Table of contents Partnership Processes Methods 3 rd Edition 1996
Table of contents
PartnershipProcessesMethods
3rd Edition 1996
Quality Assuranceprior to Serial Application
- Partnership- Processes- Methods
3rd Edition 1996
Verband der Automobilindustrie e.V. (VDA)
Liability exclusion
The VDA volumes are guidelines available for anyone to use. It is theresponsibility of the person using the guidelines to ensure it‘s correct use ineach specific case.
The VDA volumes take into account state of the art technology, current atthe time of issue. By using the VDA guideline, nobody is relieved of theresponsibility for their own actions. Each person acts, in this respect, at theirown risk. Liability by VDA, and those persons involved in issuing VDAguidelines, is hereby excluded.
If during the use of VDA guidelines, errors or the possibility of misinterpreta-tion are found, it is requested that these be notified to the VDA immediately,so that any possible faults can be corrected.
Referenced standards
The extracts from standards identified with their DIN number and issue dateare reproduced with the permission of the DIN Deutsches Institut fürNormung e.V. (German Institute for Standardization). The version with thelatest issue date is definitive for the use of the standard, which can beobtained from the publishers Beuth Verlag GmbH, 10772 Berlin.
Copyright
The publication, including all it‘s parts, is protected by copyright. Any useoutside of the narrow boundaries of copyright law, without the writtenpermission of the VDA, is unauthorized and illegal. This applies in particularto copying, translation, microfilming and the saving and processing inelectronic systems.
ISSN 0943-9412Copyright 1996 by
Verband der Automobilindustrie e.V. (VDA)(Association of the German Automotive Industry)D-60325 Frankfurt am Main, Lindenstraße 5
Online production:VDA-QMCD-60325 Frankfurt am Main, Lindenstraße 5
Preface to the 3rd Edition
With the 2nd Edition of the VDA Publication 4 in 1986, essential methods ofquality assurance prior to serial application were described for the first time.
Quality work in the Automotive and Supply Industry has strongly decen-tralized through the introduction of TQM (Total Quality Management) andextended to all areas of a company. Thereby the responsibility for the qualityof a product or service has been handed to the departments, which have adirect influence on the product or service provided.
Furthermore, additional methods have come into practice, which are com-prehensively described in the VDA Volume 4, Part 1, 3rd Edition usingexamples.
These methods are the precondition for effective cooperation betweensuppliers and car manufacturers; they control quality-relevant criteria in thepre-series phase (see also VDA Volume 2, Quality Assurance of Supplies inthe Automotive Industry). The phases and scopes of the methods aresummarized in an outline.
FMEA has changed in it’s overall scope and has been further developed asSystem FMEA for product and process. The significance of FMEA and thescope of this chapter 5, in connection with the comprehensiveness of thedescription, made it necessary to issue a separate VDA publication, Volume4 Part 2.
We would like to thank all the companies and their employees, for their helpwith this publication, as well as the many individual suggestions from thecircle of involved authors.
Due to the input of personally acquired „know how“ and the large scope ofthe examples described, the employees of the following companies arenamed before each chapter.
2
AUDI AG, IngolstadtBMW AG, MunichRobert Bosch GmbH, StuttgartContinental AG, HannoverDeutsche Gesellschaft für Qualität e.V, (DGQ) FrankfurtFichtel & Sachs AG, SchweinfurtFord-Werke AG, CologneGETRAG Getriebe- und Zahnradfabrik Hermann Hagenmeyer GmbH & Cie, LudwigsburgKolbenschmidt AG, NeckarsulmMercedes Benz AG, StuttgartAdam Opel AG, RüsselsheimDr. Ing. h. c. F. Porsche AG, StuttgartSiemens AG, RegensburgSteyer-Daimler-Puch AG, SteyerITT Automotive Europe GmbH, BietigheimVDO Adolf Schindling AG, BabenhausenWabco Westinghouse Fahrzeugbremsen GmbH, HannoverVolkswagen AG, WolfsburgZahnradfabrik Friedrichshafen AG, Friedrichshafen
Frankfurt/Main in May 1996
VERBAND DER AUTOMOBILINDUSTRIE E. V. (VDA)
3
Contents Page
Preface to the 3rd Edition 6
1 Introduction 7
2 Total Quality Management - TQM 9
3 Partnerships 123.1 Project Management 123.2 Simultaneous Engineering (SE) 173.3 Costs and Benefits of Preventive Quality Assurance Methods 18
4 Development processes 214.1 Introduction 214.2 Design Review 234.3 Performance Specifications 254.4 Planning of Trials 264.5 Carrying out Trials 274.5.1 Types of Inspection and Testing 284.5.2 Prototype testing - Process plan - 314.5.3 Conditions for Design Testing of Pilot Series 324.6 Documentation 324.7 Consideration of Field Experiences prior to Serial Application 334.8 Development Process (Example – Valve Spring) 344.8.1 System FMEA Product 354.8.2 Performance Specifications 354.8.3 Carrying out the Trial 364.8.4 Consideration of Results from the Field 384.9 System Audit in Development 38
5 Application of Methods (Summary) 39
4
6 Quality Function Deployment (QFD) 406.1 Description of the Method 406.2 Case Example (until end QT-I) 456.3 Quality Table II (QT II) 556.4 Quality Table III (QT III) 566.5 Quality Table IV (QT IV) 56
7 Fault Probability and Influence Analysis(Fehler-Möglichkeits und Einfluß-Analyse FMEA)- see VDA Volume 4, Part 2 System FMEA - 58
7.1 Explanation 58
8 Fault Tree Analysis (FTA) 608.1 Introduction 608.2 Purpose 638.3 Definitions (Extract from DIN 25424) 638.4 Description of the Method 658.5 Preparation of a Fault Tree 688.6 Evaluation of the Fault Tree 778.6.1 Qualitative Evaluation 778.6.2 Quantitative Evaluation 788.7 Establishing the Need for Action and Selecting Measures 798.8 Procedure (Example) 80
9 Design of Experiments (DoE) 849.1 Introduction 849.2 Problem Description and Analysis 849.2.1 Exercise and Objective 859.2.2 Taking stock of the situation 859.2.3 Target Parameters 869.2.4 Influencing Variables, Acquisition and Preparation of Data 869.2.5 Acquisition, Evaluation and Selection of Influencing
variables 869.2.6 Interactions 879.2.7 Example 879.3 Reducing the Number of Influencing variables, Selecting
Factors for Experiments 899.3.1 Reproducibility and Independence 909.3.2 Evaluation Criteria and Scale 909.3.3 Weighting of Influencing Variables 909.3.4 Effects Matrix (according to Scheffler) 91
5
9.3.5 Interactions 919.3.6 Factor Levels 929.3.7 Summary of Factor Selection in a Flow Chart 929.3.8 Example 939.4 Selecting an Experiment Strategy 949.4.1 One-Factor-Experiment 949.4.2 Complete Factorial Experiment 959.4.3 Fractional Factorial Experiment 969.4.4 Factor Search according to D. Shainin 989.4.5 Design of Experiments according to G. Taguchi 1009.4.5.1 Developing Robust Products and Processes 1009.4.5.2 Developing Robust and Sensitive Products and Processes 1029.4.6 Example 1039.5 Evaluation of the Experiment Results 1049.5.1 Presentation of the Measured Results 1049.5.2 Calculating the Effects 1079.5.3 Statistical Evaluation Procedures 1099.5.4 Example 1109.6 Computer Support 1139.7 Further Literature 114
10 Process Capability Analysis 11510.1 Introduction 11510.2 Statistical Process Control (SPC) 11610.3 Process Influences 11810.4 Process Models 12010.4.1 Process Model "A" (see Fig. 10.3) 12110.4.2 Process Model "B" (see Fig. 10.4) 12210.4.3 Process Model "C" (see Fig. 10.5) 12310.5 Controlled and Quality Capable Processes 12310.6 Process Analysis and Quality Capability 13010.6.1 Process Analysis 13010.6.2 Quality Capability Code Numbers for Production
Equipment and Processes 13410.6.2.1 Quality Capability Code Number for Continual
Characteristics 13410.6.2.2 Quality Capability Code Numbers for Counting
Characteristics 13810.6.3 Quality Capability of Measuring Equipment 139
6
10.7 Quality Control Charts (QCC) 13910.7.1 Description of QCC 13910.7.2 Purpose of QCC 14010.7.3 Types of QCC 14110.7.4 Application of QCC prior to Series Start 14210.8 Procedure (Example) 14310.8.1 Example 1 14310.8.2 Example 2 15010.8.3 Example 3 159
11 Quality-related Costs 165
12 Elementary Aids 17012.1 Introduction 17012.2 Flow Chart 17012.3 Histogram 17212.4 Check Sheets 17212.5 Quality Control Chart 17412.6 Cause/Effect Diagram (Ishikawa-Diagram) 17412.7 Pareto Diagram 17612.8 Scatter plot 178
Appendix
Flow Chart
7
1 Introduction
Innovation in car manufacturing with increasingly complex technology,increased quality awareness of the consumer and not least, the changedmarket situation, connected with inevitably raised productivity, demandappropriate quality assurance methods, especially in the product develop-ment phase prior to serial application.
To develop a new product with capable quality concepts and with a designsuitable for production, as well as qualitatively controlled production, variousmethods have been developed and practiced in car manufacturing andsupplier factories, in the last few years.
The more complex relationships and dependence on the system resulted infurther methodic quality assurance measures in the pre-series phase duringthe last few years. Especially the role of Total Quality Management (TQM)has become more predominant.
The quality department is responsible for the fundamental tasks of qualityassurance and coordinates the activities in the various technical depart-ments. These departments take over their own responsibility for the qualityassurance activities, supported by a quality representative or quality divi-sion. The following described methods were chosen so that they can beequally applied by different sized companies. With TQM as the main thread,they present a tool for securing the product quality from product design toreadiness for serial production.
It should be possible for all areas of a company, with the description of themethods, to implement and use them and to thereby economically develop,design, plan and manufacture the zero-defect-product. The procedures des-cribed are proven methods, which represent current knowledge and arebeing applied and developed further in many companies.
8
The systematic was based on the following criteria:
- The procedures described were chosen, in scope and content, sothat the momentum of a company using the methods is notrestricted.
- Differing company sizes, especially in the supply industry weretaken into consideration.
- As many concepts as possible, already in use in companies, havebeen considered, in order to provide a suggestion for all possiblyareas of interests and raise their acceptance.
- The application of the methods is explained using examples.
- This publication gives an outline of the different procedures. Moredetailed information must be obtained from specialist literature.
- The procedures not yet published in detail have been grantedmore scope.
The leading thread for application of the described methods, is the attachedchronological flow chart showing the quality assurance activities in theproduct origination phases.
With all joint product development projects between manufacturer andsupplier, the representatives of both companies should discuss the qualityobjectives and requirements at an early stage. Thereby the methods andprocedures for „Quality Assurance prior to Serial Application“, as well as thetechnical feasibility and limits shall be discussed and established.
Through the systematic application of the herein described methods, evenwith increasing complexity of products and required quality objectives, totalcosts can be minimized.
9
2 Total Quality Management - TQM
Authors: Berthold Edenhofer, Herbert Füller*), Elmar Rahm,Friedrich Scheucher, Helmut Stein
In the Sixties and early Seventies, it was ascertained that in work processesinfluenced by people, faults originate from ignorance and inattention. Qualitypromotion programs were started to motivate employees, with the aim toproduce „Zero-defect“ products. „Zero-defect-programs“ supposed to beaimed at those people at operative level, were also extended to otherfunctional areas, not directly involved in the product development process.
„Top-Down-Programs“ emerged, covering all hierarchy levels and for alltechnical departments, to improve company quality.
The term Total Quality Management (TQM) developed from this. TQM isdefined as a company-wide strategy, realized with the help of a programand coming from Top Management, with the objective that all areas workautonomously on continually improving the quality of their respectiveproducts, services and processes. The objective of this company-wide effortis to secure their companies long term future, through high customersatisfaction with the products and services.
All company activities are therefore strictly customer orientated. The stan-dard for the degree of success is customer satisfaction in relation to quality,delivery date and price.
Furthermore, a universal process orientation of all company activitiesgenerates processes of high quality and, inevitably as a result of theprocess, profitably manufactured products of high quality.
*) Moderator of the Author Team
10
Strict employee orientation creates the basis for achieving companyobjectives. High quality personnel management, a targeted, scheduledemployee qualification, delegation of responsibility, as well as a highmeasure of possible influence on the arrangement of work processesgenerates commitment and motivation with employees at levels.
Customer, process and employee orientation are assigned measurablecriteria. Thereby the setting of targets and scheduled company developmentbecome transparent.
A universal quality culture throughout the company, higher employee identi-fication and motivation, better products and services, as well as loweredcosts, are the results of the named measures.
The TQM idea effects the company structure by
- lessening the number of hierarchical levels,- transference of larger areas of responsibility,- creating smaller control loops for information flow.
One speaks of „lean-management“ when realizing these structures as partof TQM. Tools for methodical support of TQM are described in this publi-cation for all phases of product development prior to serial application.
11
Examples of quality objectives of a company are :
1. The customer must be satisfied
Quality is Trust. What quality is, decides only the customer. We supportthe customer through knowledge and ability. Customer-orientated pro-duction means, for us, consistently confirming with the requirementsprofile of the customer. Not we, but he must be satisfied with our perfor-mance. Therefore, there is no substitute for the zero-defect-product.
2. Our responsibility
Quality is valid everywhere. In our company, every employee is respon-sible for quality. This applies to all areas of our company. Thinking andacting in the company’s best interest is encouraged in all employees.We must set the example of what we expect from others.
3. Perfection
Quality has method. From development and design through to produc-tion and sales. During the lifetime of a vehicle, the products must func-tion without fault or failure. Quality already begins in the discussion withthe customer. His wishes, needs and expectations define our companyoperation.
4. Identifying with what we do
Quality begins in the head and heart. The attitude of people decidesproduct quality, not only technical aids. We have to transfer our highexpectations of quality of life onto product quality. Quality must be livedand experienced. Quality is to be continually improved, through thecourage of positive criticism and creative forward pressure. Quality is avalue of it’s own.
12
3 Partnerships
Authors: Berthold Edenhofer, Reiner Franzkowski,Herbert Füller*) , Elmar Rahm, FriedrichScheucher, Helmut Stein
3.1 Project Management
To develop and manufacture a product, that corresponds with customerrequirements and at the same time is in line with schedule and costobjectives, requires strongly organized, early cooperation of all companydepartments including the customer and supplier. This ensures optimumdevelopment, a quality production process and smooth start of the series.Consistent project management ensures a target-orientated project pro-cess.
At the point of contract receipt a project manager is assigned who coordi-nates all planning activities. This includes the assignment of a project teamand, if required, the inclusion of additional departmental representatives.
Project manager and project team are a group responsible to the companymanagement. The management decides on the product range, and basedon this, volume, costs, quality, price and schedule objectives for the indivi-dual projects are determined, using market and competitor studies andcustomer requirements.
*) Moderator of the Author Team
13
During project planning, the following three points must, at least, be takeninto account:
- The joint development of products by the car manufacturer andsupplier. Hereby, the experiences of both partners with regard to opti-mal customer satisfaction can be used.
- The planning of the project (Fig. 3.1), split into the following phases
- Definition and development- Development and engineering- Purchasing and pre-series construction- Series.
- The chronological structure of projects.
14
Project ProcessBar Chart
Parts description ________Model ________Parts Number ________Customer Number ________
Phase
Definitionand
Development
Developmentand
Planning
Purchasingand
Pre-SeriesConstruction
Series
Description
Customer RequirementsLegislationCompany objective - Product ideaMarket and competition analysisObjectives (A)Quality objectivesTechnical product studyVolume-Price-Costs frameTime frameRelease of objectives catalog (B)Quality assurance planningDesignsStyling model, interior, exteriorEstablishing performance specification ITime scheduleRelease performance specification I (C)Preplanning production and inspection technologyTechnical product description: Revision, evaluationDrawings suitab le to planningDevelopment model/prototype: preparation, evaluationQuality assurance planningDetermining initial sourcesMaster model inspectionEstablishing parts list, Beginning of parts/parts planning (costs/dates)Establishing performance specifiations IIRelease performance specifications II (D)Evaluating quality capability, Internal production/supplierPlanning of tests and test equipmentPlanning of production and production equipmentRaising of series drawingsMachine and process capability analy sisDetermining final sourcesResult of prototype/development model trialConstruction model inspectionRelease purchasing (E)Employee trainingEstablishing components inspection plansRaising component production plansTest equipment: purchasing/inspectionProduction equipment: purchasing, inspectionPre-series parts: production, purchasing, inspectionCoordination of customer equipment and documentsProduction test seriesFirst sample supplier partsRelease internal production parts (first sample test)Release to pre-series, pilot-seriesAssambly Pre-series, production or pilot lotPre-series, pilot-lot, inspection, large scale production trialsFirst sample testing for customerSingle parts series: Production, purchasing, inspectionRelease to series start (G)Production seriesResult of large scale production trial/Quality reportRelease to market introduction
1st. Year 2nd. Year 3rd. Year 4th. Year Degree of conformity in % 20 40 60 80 100
Import notification ________Release status ________Date ________
Fig
. 3.1
P
roje
ct P
roce
ss (
Bar
Cha
rt)
15
The different activities are put into a time schedule and adjusted to theproject scope.
Internal schedule coordination is the responsibility of the project manager. Insuperior networks, the content and scheduling of project plans have to becoordinated.
Inappropriate and delayed execution of project activities, especially thepostponement of set bench mark dates, leads to, besides the breaching ofdelivery dates, also usually to quality impairments and cost increases.
To achieve the quality objectives, the quality assurance methods are alsoincorporated into the project.
The activities and time periods shown in the bar chart „Quality AssuranceActivities„ (Fig. 3.2) provide a summary and their content and time framecan accordingly be adjusted for less complex projects.
16
Project ProcessQuality Assurance Activities
Parts description ________Model ________Parts Number ________Customer Number ________
Phase
Definition
and
Development
Development
and
Planning
Purchasing
and
Pre-Series
Construction
Series
Description
Objectives (A)
System FMEA Engeneering
Parts Count Method
Quality Evaluation 1 (Design Review)
Feasibility Study
QFD
Fault Tree Analysis
Release Objectives Catalog (B)
DoE Product
Design FMEA
Release Performance Specifications I (C)
Quality Evaluation 2 (Design Review)
System FMEA Process
Release Performance Specifications II (D)
Process FMEA
Process and Machine Capability Analysis
Release Purchasing (E)
Quality Evaluation 3 (Design Review)
DoE Process
Release Pre-/Pilot Series (F)
Release Series Start ( G)
System Statistical Process Control
Release Market Introduction (H)
1st. Year 2nd. Year 3rd. Year 4th. Year Degree of conformity in % 20 40 60 80 100
Import notification ________Release status ________Date ________
Fig
. 3.2
Bar
Cha
rt „
Qua
lity
Ass
uran
ce A
ctiv
ities
“ at
pro
ject
sta
rt
17
3.2 Simultaneous Engineering (SE)
Purpose
To design products in such a way, that a faultless production and qualityassured operation, and simultaneously a shortening of the developmentphase, is achieved.
Aims
- To use all resources to their optimum- To design products for quality assured production- To improve application of production technologies- To achieve cost effective production- To shorten the development phase.
Realization
At the earliest possible stage, all expert departments, especially planningand production, cooperate with the development and design department, inorder to achieve the optimal qualitative and cost-effective result in productdesign. Performance specifications are established according to require-ments, so that the quality and operational requirements can be realized bythe developer.
The requirements set for production processes, are to be taken into accountright from the beginning of development.
At the same time, decisions about internal production or purchasing are tobe made, so that the processes for internal production procedures are such,that the expected quality and costs are met and for purchasing, the sameresults are provided.
The development time is shortened through partial simultaneous operationsin the respective phases.
18
3.3 Costs and Benefits of Preventive Quality Assurance Methods
The difference between the costs and benefits of preventive qualityassurance methods, is the measure of economy of the method to beapplied. (Fig. 3.3).
Cost
Benefit
Co
st a
nd
Ben
efit
, no
min
ated
Product Development Time Series Start
Cost-/BenefitEquilibrium
Start ofprofitability
Fig. 3.3 Economics of the application of preventive QA methods
Therefore, the costs and benefits aspects are described, and theiroccurrence over the product life-span examined in the following.
Costs
The costs of applying of methods for preventive quality assurance mainlyconsists in the time the employees spend on the increased systematic.Expenses for material, traveling, etc. form another, smaller part.
As preventive quality assurance methods are involved here, the main costsoccur in the early stage of product design and development. The peak costsnormally lie in the first third of the design and development phase of aproduct. The costs decrease very rapidly during the course of productionand process development. (See cost curve in Fig. 3.3)
19
The most important influencing factors to minimize the costs are fullydeveloped, effective method concepts, suitable equipment for methodsupport and clear integration of the methods into the operational processes.
The named cost aspects are easily established. It is more difficult, however,to examine the benefits in advance.
Benefits
The benefits of applying methods for preventive quality assurance consistsof being able to prevent error costs, i.e. costs resulting from defects, asearly as possible. Benefit aspects are as easily named as cost aspects.
Benefit aspects are
- less design costs to eliminate design defects- less costs for more targeted trial runs- less planning costs to eliminate planning defects in production
processes- less disruptions in the processes- less rejects, reworks and lower warranty expenditures and
improved image compared to competitors.
Defect transmission from the performance specifications over productdevelopment, process planning, production to the final customer productmeans multiplication of the defect costs. This relationship is also known asthrough the so-called Decimal Rule (see Fig. 3.4).
20
FMEA SPC
Developing Purchasingand andPlanning Production 100,-
10,- 1,- 10,-
DM
per
fau
lt
Devolpment Pilot Series Cusomer
Fig. 3.4 Decimal Rule
The basic principle of the Decimal Rule states, that whilst it costs only a fewpennies to change a few lines during the design phase, having to alter therunning production process costs many times more.
The most expensive type of defect transmission occurs, when the conse-quences of a development defect are only identified when the product isused by the customer.
As preventive quality assurance methods prevent potential problems, thegained benefit does not occur at the time the method is applied, but in thefuture. The sooner a defect is identified, the longer one has to wait for thebenefit, however, it also means, that according to the decimal rule thebenefit is greater.
As prevented defects normally cannot be proven, economic contemplationof a theoretical nature, e.g. what percentage of defects can be prevented byusing preventive quality assurance methods, are limited in their meaning-fulness. It is only possible to identify the shown, future strategic benefit,apply the preventive quality assurance methods and discover the monetarybenefit after the „waiting period„.
21
4 Development processes
Authors: Turecht Beltz*), Helmut Höfter,Dr. Hans Schlemm, Dr. Walter Weiblen,Werner Wielan, Dr. Uwe Wittkowski*)
4.1 Introduction
Technical products, such as vehicles and their components are, in additionto the following described preventive quality assurance methods, mainlydeveloped to proven basic methods.
This includes internal works, national and international design guidelines,standards and calculation methods, describing state of the art technology.
Trials are necessary at those points, where technical characteristics cannot,or only inadequately, be determined by theoretical procedures (e.g.calculations) during development.
The foundation for a quality product is laid with the confirmation or improve-ment of concepts through experiments after design. Targeted application oftrials also saves time and money. The planned quality shall be reached withthe smallest possible scope of trail. For these reasons, the differing trailsteps must be exactly planned into the development process and monitored.
Fundamental requirements and procedures:
In the individual development phases of a vehicle, aggregate or componentconstruction, tests and trails are carried out (Fig. 4.1), the results of whichdemonstrate the achieved design and quality status.
*) Moderator of the Author Team
22
Setting theobjective
Trial productionmeans
Production preparation
Development
Pre-development
A
B
C
ProductionRelease
Pre-develop-ment
ReleaseDevelopment
- Technical release- Customer release
Purchasingrelease
ReleasePre-series/Pilot Series
ReleaseSeries start
Developmentmodel, prototype
Construction modelProduction
trialseries
First sampling
- Pre-series- Pilot Series
Fig
. 4.1
Pro
ject
pro
cess
pla
n w
ith th
e de
velo
pmen
t pha
ses
of a
veh
icle
23
At least one comparable form of a project process plan with „milestones“and „check points“ should be available at car manufacturers and thesuppliers of differing types of sub-systems. In the mutual interest of thecustomer and supplier, the trials and their results are to be repeatedlycoordinated, with regards to time and content, by the respective projectmanagement.
The possible method is described in the following paragraphs.
4.2 Design Review
Purpose
Problem prevention in the earliest possible development phase.
Objectives
- Discovering the best solution to achieving the development goals
- Establishing the individual requirements
- Smallest deviations (tolerances) with regards to safety and relia-bility during production
- Economically meeting the requirements and delivery dates.
Performance
- The concept review takes places after determination of theconcept but prior to setting the tolerances. The review determineswhich Hardware and Software properties have to be included.
- The design review is to ensure, that the given design can bemanufactured within acceptable tolerances under the foreseenproduction conditions. The product and process specifications aredetermined.
24
Final Design Review
- Final design review is the formal hand-over of the design docu-ments of a product to the production department. Simultaneously,concept and design status are „frozen“.
Method
When developing complex systems, various methods must be used. Animportant aid, thereby, is a functioning structure and process organization,in which development projects can be controlled and monitored using pro-ject process plans. These three named reviews must be listed following theindividual process phases in the project process plan (Fig. 4.1).
The method is described in the following.
Basic requirements :
- planning of all meetings- deciding on the participants (interdisciplinary combination)- ensuring that all meetings are minuted including process plan,
checklists, time requirements, measures and recommendations- establishing the criteria to judge and decide, especially for safety and
reliability.
Handling of problem points:
- suitability of the product- fulfilling the performance specification requirements- quantified instructions to the design sub-groups- experiments which prove the current status- maintainability- parts-standardization, where possible- safety analysis for production equipment and personnel- tolerances and producibility.
- Strains on the product- during production- during transport- in use.
25
4.3 Performance Specifications
Performance specifications are an important basis for trials. From theperformance specification of the vehicle, performance specifications for theindividual systems and components are derived. Only in this way can theybe tested separately from one another.
The trial shall ensure and prove, that the systems fulfil the performancespecification requirements.
These are essentially:
- Set targets, e.g. regarding improvement of energy consumption,emissions, noise levels, recycling capability
- functionality, suitability for use, customer acceptance, maintainabilityand reparability
- safety and reliability- legal requirements, regulations, rules, standards.
For some of these requirements, there are established trial procedures.They are either legally stipulated (e.g. as for type tests for approvals) orestablished through internal company rules (e.g. endurance test runprograms).
Additionally, the performance specification must include criteria for thebreakdown of the aggregate or component and for their acceptable degreeof deterioration.
For the development and/or for the delivery of vehicle components, it isimperative to record testing procedures agreed between the purchaser andthe supplier in the performance specification. Thereby, it should beinvestigated (also for the purpose of cost reduction), if the test proceduresof different purchasers for the same part, could be standardized.
Especially with regard to product liability, it is important during deliverynegotiations to describe the requirements of the delivery subject exactly andcomprehensively. The recipient has the obligation of informing the supplierof the requirements.
26
4.4 Planning of Trials
Performance specification requirements and results from
- System FMEA for the product and Design FMEA for componentsare used during the planning of trials.
The System FMEA described in chapter 7 for products examines functionsand possible functional failures at different levels (Hierarchical System, Fig.4.2). This examination goes down to the level of the components. TheDesign FMEA is raised for this.
Vehicle
Drive
Engine
Valve
Valve spring
Fig. 4.2 Hierarchical System
27
With System FMEA, system weak spots become apparent. With the help ofDesign FMEA design weak spots on components are discovered andevaluated. In many cases, through an optimization of the design, trials canthen be avoided. Components with residual high risks require exactsafeguarding through trial testing.
When planning trials, the following must be decided on:
- trial necessity- trial procedure- scope of the trail- trial methods- measuring equipment and it’s calibration- processing the trail data.
The planning also involves ensuring trial focal points concentrate on newcomponents or new technologies, corresponding to the degree of technicalinnovation in the product. Previous trial scopes must be critically examinedand, if necessary, reduced.
Corresponding to the hierarchical system „vehicle“, planning and testingmust take place, parallel in time or overlapping for the vehicle, it’saggregates and components.
4.5 Carrying out Trials
The trial performance must always satisfy the opposing requirements of „asearly as possible“ and „largely close to serial production“. An acceptablecompromise is to be found and established, possibly also with follow-upactivities at a later date.
For the first trial, only design models or prototypes on all levels of thesystem are available and form an initial focal point of the trail.
28
The next trial step is carried out with pilot production parts from vehicles,aggregates and components in close-to-serial design.
The trial results are immediately documented. They are evaluated andincluded in the design review.
With growing technical maturity, further improvement potentials emerge.These are also to be evaluated and documented.
The trials are to be confirmed with serial products.
4.5.1 Types of Inspection and Testing
The tests can be divided into two groups:
- Material testing and- Functional testing of components, aggregates and vehicles.
Material testing
Testing parameters for metals are, for example:
- chemical composition- microstructure- hardness- yield strength and- corrosion resistance.
Testing parameters for synthetics, textiles and paints are, for example:
- hardness- extension- impact resistance- tear resistance- wear resistance
29
- temperature stability- chemical stability- light fastness and- resistance to atmospheric corrosion.
Testing of Components, Aggregates and Vehicles
Company specific tests are procedures, which are determined in companiesthrough their technical knowledge, experience, as well as their technicalpossibilities.
Furthermore, there are tests under the framework of approvals for releaseinto public traffic. These tests are carried out according to legal require-ments and decrees. National regulations are often based on internationalguidelines.
In the Federal Republic of Germany, as a Member of the „EuropeanCommunity“, these types of tests are carried out according to EC Guidelines(Fig. 4.3).
The limit values or regulations for the construction of vehicles, as well astest procedures contained in these standards and guidelines are obligatoryfor these legally required tests to approve vehicles for release into publictraffic.
30
72/245 Ignition noise and electromagnetic compatibility
71/127 Rear mirror 77/649 Field of visibility 77/541 Safety belts and restraint systems
76/115 Anchoring of the safety belts
74/60 Parts in passenger space
78/932 Headrests
70/222 Application of the registration plates
76/760 Rear registration plate lighting
77/540 Park light
77/539 Reverse light
77/538 Rear fog light
76/757 Rear lights
76/758 Clearance lights and SBBR-lights
70/157 Noise level, exhaust system
70/156 Overall type approval
78/549 Wheel covers
74/483 Protruding outer edges
71/320 Brake system
70/221 Fuel tank, undercarriage
70/220 Emissions
70/387 Door locks and hinges
76/769 Hazardous materials67/548
74/408 Seats and their anchoring
75/443 Reverse gear, speedometer
74/297 Steering system behavior during accidents
70/311 Steering system
76/759 Direction indicators
70/388 Horn
74/61 Security device against unauthorized use
78/316 Switches, control lights and indicators
77/389 Towing arrangement
76/756 Headlights and
76/761 lighting mounts
76/762 Front fog lights
80/1269 Engine performance
80/1268 Fuel consumption
76/114 Chassis number, plates
78/317 Screen de-ice/ de-mist blowers78/548 Heating
78/318 Wind shield wipers and washer
FE-Vehicle Regulations
EG-Guidelines Construction regulations Safety and Environment / Passenger vehicles Fig
. 4.3
EC
-Gui
delin
es,
Con
stru
ctio
n re
gula
tions
Saf
ety
and
Env
iron-
men
t – P
asse
nger
veh
icle
, Sta
tus
03.1
992
31
4.5.2 Prototype testing- Process plan -
Checking product prototypes for compliance to the specifications. Checkingdesign models and prototypes for compliance with the quality objectives withthe help of a detailed assessment program. Analysis and documentation ofthe results including the evaluation of nonconformities. Eliminatingnonconformities including monitoring.
The trials are to be arranged so that they are representative for the requiredcharacteristics and application conditions. That means, for example, thatsimulation or accelerated procedures must be chosen so that, amongstothers driving conditions, environmental influences and mechanical agingprocesses correspond with the application requirements.
For the check of the body, for example, the following testing types orscopes, on different building stages of the car body prototype, are carriedout:
- Car body stability: endurance fatigue test (hydraulic ram teststand, moving doors and covers).
- Car body safety: e.g. tensile and impact tests. Car shellsfitted with, for example, safety belts, panelboard, seats, seat belt height adjustment orsplit rear seat.
- Crash testing: Frontal, rear and side impact test, roll-overtests.
- Endurance test: e.g. driving over 100.000 km run orcorrosion test.
- Partial car body: Examinations on partial car bodies e.g. forepart to B-column with built-in engine, gearetc.
32
4.5.3 Conditions for Design Testing of Pilot Series
Construction status: Pilot production, i.e. construction after com-pleted development under serial productionconditions with positive sampled parts.
Construction date: Early enough before series start, so thatcorrective actions for found nonconformitiescan be defined and implemented prior toseries start.
Testing : Testing is carried out under conditions rele-vant to the customer. It includes all quality-and cost-related criteria that come intoquestion, for example:
- functional safety- reliability- comfort- charm (subjective impression)- optics, acoustics- economic viability- legal requirements, environmental protec- tion requirements.
Evaluation of the results: The evaluation of the test results is carriedout jointly by the production, development,purchasing and sales departments. Com-pliance with the quality objectives is thecondition for the release to start serial pro-duction.
4.6 Documentation
A compilation of the tests carried out, the results (specified/actual values)and evaluations is to be performed. In cases of product liability, a completedocumentation of the testing and development phase is helpful.
33
4.7 Consideration of Field Experiences prior to Serial Application
Passenger vehicles today are designed for an average life expectancy of 10to 15 years. Of these, the customer drives only about 2.000 to 3.000 hours.The rest of the time, there are no wear effects on the components, whichresult from driving. Internal strains (e.g. from weight), as well as environ-mental influences (e.g. temperature, humidity, air pollution), do however,lead to aging processes, which can be the causes for system failures (con-tact problems in electrical and electronic equipment, embrittlement of syn-thetic parts, sealing failures).
In addition, the entire spectrum of loads and strains which occur duringcustomer use cannot be fully covered by trials during development, for timeand cost reasons. With the performance of e.g. fleet trials, trials at thecustomer’s own test circuit, field trials by selected customers, such as taxidrivers, the post office or forwarding companies, and with the evaluation ofthe feedback from customers, further knowledge to help improve qualitywhen revising the serial product (Model updating), or for subsequent pro-duct generations, can be gained.
When planning, performing and evaluating field trials, following aspectsshould be taken into account:
- The objective of field trials is to be clarified in detail and the results tobe expected must stand in adequate relation to the generally highexpenditure.
- The field trial must include all relevant cases of customer applicationin adequate relation (motorway, town traffic, country roads, dirt tracks,maneuvering etc.). Thereby, the applications should not be carriedout in blocks behind one another (e.g. 100.000 km motorway, then10.000 km town traffic) but instead should be varied.
34
- Field trials without continuous monitoring (e.g. through classificationequipment, tachometers, recording of all relevant events andsurrounding conditions) often have only limited meaningfulness dueto insufficient information to identify the failure causes.
- With regard to the aspect of damage or failure causes, as well as thelife expectancy, intensive feedback to experimental trials of the com-ponents and subassemblies must be established, to continuallyimprove simulation techniques in the development phase and corre-lation of the results gained with the behavior in customer use.
If products, in particular with regard to late occurring breakdowns, are to beoptimized, customer service departments must raise a questionnairefocussed on the respective problem area, which makes use of all availableinformation about the special application conditions of the failed system.
4.8 Development Process (Example – Valve Spring)
The position of the component „valve spring“ can be seen in the system„Vehicle“ (Fig. 4.2). The development process is illustrated in Fig. 4.5.
35
4.8.1 System FMEA Product
Possible failure cause, valve spring, with effects on the engine or vehicleare represented on the system FMEA-form (Fig. 4.4).
Type/Model/Production/Charge:
System number/System element:
Function/Task
Vehicle
Drive
Possible effectof the fault
B
Poor performance
Vehicle breakdown
10
10
Type/Model/Production/Charge: Engine
Possible effectof the fault
B
to high wear on the Spring power Wrong con-valve spring retainer/ too high structivecam/tappets layout of the valve spring
Engine damage
10
10
Fault-Probability-
X System-FMEA Product
System number/System element:
Function/Task Valve
Possible effectof the fault
Possible causeof the fault
B = evaluation number A = evaluation number for for the significance the occurrence probabilityV = responsibility T = date for the completion
Fig. 4.4 System FMEA Form 96 / Risk analysis
4.8.2 Performance Specifications
During the constructive design of the valve spring, forces (from operatingspeed range, cam profile, valve acceleration) and dimensions (necessarywire strength, no. of turns, available fitting room), as well as the materialspecifications were established.
36
Functional Testing to fulfill the Requirements on the Component
Requirements on the component - valve spring:
- dimensional accuracy- spring power = f (Way) corresponding to the requirements- material testing.
Functional Testing to fulfill the Requirements on the Component in theEngine Assembly
- spring loads- spring resonance outside operating speed range.
Endurance Testing in Component Test-stand
Pulse tests (e.g. n = 10 springs with N = 2x106 alterations of load withoutfailure).
Endurance Testing in Engine Test-Stand / Vehicle Endurance Run
- Spring power / lengths after endurance testing in engines still withinestablished tolerances
- No unauthorized on the valve pinion components, as far as a valvespring influence comes into question
- No valve spring fracture.
4.8.3 Carrying out the Trial
The trial is carried out according to the shown process (see Fig. 4.5), toprove the compliance with the requirements as given in the performancespecifications (see Pt 4.8.2).
The successful testing of the described functions of the component is theprecondition for the start of the endurance trial.
37
The subsequent endurance trial is carried out in parallel on pulse teststands, engine test stands and during vehicle endurance runs. If defects arefound, the above named tests or trials have to be repeated with optimizedparts.
Engine project start Start of TestingValve spring
Valvespringrelease
Enginepilotstart
Engine serial begin
Definition of the requirements
Constructive design
Test model purchasing
Material testing
Functional testing
Pulse testing
Engine test stand endurance test
Vehicle endurance test
Fig. 4.5 Development process „Valve spring“
When necessary optimization measures, or new definition of the require-ments during the entire engine development, make one or more trials runsnecessary, it is possible that some trial results (e.g. vehicle endurance test)may only be available after start of serial production. Here attention shouldbe paid, that a wide enough basis from other trials (e.g. pulse tests, enginetest stand endurance tests) are available, to ensure the long term durabilityof the valve spring.
38
4.8.4 Consideration of Results from the Field
As with the valve spring example, even small deviations of material qualitycan result in high damage (total engine failure), respective customer irrita-tion and high warranty costs, an exact observation of field failures becomesvery significant, in order to be able to react as early as possible.
4.9 System Audit in Development
To verify if all necessary quality assurance elements are implemented indevelopment, the performance of a system audit according to VDA,Volume 6, Part 1, „Quality System Audit“ is recommended.
The quality element, described therein under Element 08 - Design Control(Product development), encompasses the quality assurance duties indesign and development.
39
5A
pp
lication
of M
etho
ds (S
um
mary)
In the following table, the application phases of the m
ethods, the involvedand responsible departm
ents are shown.
Method
Total Quality Management
Project Management
Simultaneous Engineering
Design Review
Quality FunctionDeployment
Fault Probability andInfluence Analysis(FMEA)
Fault Tree Analysis
Design of Experiments
Process CapabilityAnalysis
Process Control
Quality related costs
Elementary aids
Page
9
12
17
23
40
58 (andVolume
4/2)
60
84
115
116
165
170
Application in Phase
all
Release Pre-developmentuntil Release of First SeriesDelivery
Target until Series ReleaseCustomer
Release PerformanceSpecification 1 (Concept)Release PerformánceSpecifications 2(Design/Engineering)
Target until ReleaseMaterial Purchasing
System-FMEA-ProductStart Development untilRelease PlanningSystem FMEA ProcessRelease planning until releaseMaterial Purchasing
all
For Products:Target until Series StartFor Production:Release Material Purchasinguntil End of Running Period
Release Material Purchasinguntil Release to Series Start
Release to Series Start untilProduct Running Period
all
all
Involved Department
all
Development, Engineering,Planning,QA, Sales,Purchasing
Development, Engineering,Planning,QA, Sales,Purchasing
Development, Engineering,Planning,QA, Sales,Purchasing
Sales, Development, Engineering,Design, Production, Dispostion
Planning, Development, Engineering, QualityPlanning, Development,Engineering, Quality, Production
Corresponding expert teammembers
Planning, Trial, Engineering
Planning, Quality, Production
Quality, Production
Quality
all
Responsible forexecution
Company managementResponsible Coordinator
Project Manager
Project Manager
Project Manager(Team representativeDevelopment/Design)
Sales (Development)
Development teamrepresentativeProduction planningteam representative
Team representatvie(team member
Trial team represen-tativeProduct planning teamrepresentative
Product planning(Production)
Production
Controlling (reportingalso Q-Management)
all
Guidelines for application inthe VDA Volume 6, 2nd Edition
01.1 to 01.6Inclusion of all chapter contents
07.208.1, 08.709.1, 09.7
08.2 and 08.3
08.2
08.2, 14.2, 18.2, 22.3
08.2, 18.2, 22.3
08.2, 18.2, 22.3
14.1, 14.2
14.1, 15.1, 12.1, 12.2, 12.5,20.1
05.1 to 05.4
15.6, 18.3, 22.4, 22.5, 26.6
40
6 Quality Function Deployment (QFD)
Authors: Herbert Füller,Dr. Martin Kaminski*),H. Kumpfmüller, Armin Schemion
6.1 Description of the Method
The tasks of quality assurance have over the course of time changed fromdefect identification through inspection in production to defect preventionthrough risk minimization in development and planning. The continuation ofthis development to the start of product development, shows, that qualityassurance must begin already at the concept stage (Fig. 6.1).
Product Origin
END
Concept
Design
ProductionPlanning
Production
ProductDevelopment
QualityAssurance
Def
ect
prev
entio
nD
efec
tba
ttle
Fig. 6.1 Product Development and Quality Assurance
*)
Moderator of the Author Team
41
Quality Function Deployment (QFD) is a comprehensive planning andcommunication system, which helps to coordinate all resources of thecompany, to develop, produce and market the product and services that areexpected by the customer, so that an improvement of the companyperformance through increased competitiveness is achieved.
As an aid QFD uses quality tables, which consist of several matrix fields anddue to their shape are called „House of Quality„.
42
internal companycharacteristics(HOW)
Competitionanalysis of
Customer requirem.
Customerrequirements
(WHAT ) 1 2 3 45
1
2
3
4
5
6
7
8
9
Difficulties in reaching the target( 1= easy, 5 = difficult)
Test methods, current, futurespecification
Today 5
Aim 4
Competitor 3
2
1
Critical, internal company characteristics
ROOF
Strongl. pos.
Positive
Negative
Strongl. neg.
MATRIX WEIGHTS
Strong 9
Medium 3
Weak 1
Customer
X
XXX
. .
Line
num
ber
Prio
rity
Com
plai
nts,
war
rant
y ca
ses,
sale
s ar
gum
ents
Crit
ical
cus
tom
er r
equi
rem
ent
Tod
ay
Com
petit
ion
Tar
get
1 2 3 4 5
Crit
ical
inte
rnal
ch
arac
teris
tic
Com
pany
Customer
Fig. 6.2 House of QualitySource: American Supplier Institute, N.Y.
With the help of these quality tables, the „voice of the customer„ istranslated into the „company language„, whereby the answers to thequestions „WHAT?„ and „HOW?„ are related to each other (Fig 6.3). This isdone in four phases (Fig. 6.4).
43
Customer
WHAT?
HO
W?
Com
pany
Customer
Fig. 6.3 Voice of the customer and language of the company
Pro
duct
Par
ts
Pro
cess
Wor
kin
stru
ktio
n
Process
Parts
Product
Customer
Fig. 6.4 Continuity from customer to production
The first quality table (QT-l) is drawn up in the concept phase; it convertsthe customer requirements into internal company characteristics andtechnical terms. These internal company characteristics are transformedinto parts characteristics during the development phase using a secondquality table (QT-II). These then again form the basis for determining theprocess parameters during production planning, with the help of a thirdquality table (QT-III). During the standardization phase, the fourth qualitytable (QT-IV) is established, which derives standards for the individualworking steps as well as, servicing requirements and necessary trainingmeasures from the process parameters.
44
QFD can be applied to products and services, for new developments duringthe concept phase, for decisions about new product generations and forfurther development of existing products fit to market requirements.
It is worth investing the time for QFD in the long run, as the following advan-tages arise
- customer orientation,- transparency and- team work.
The application of the QFD method requires, that each department andevery company employee understand themselves to be an „Internalcustomer„ and at the same time „Internal supplier„. As a „customer„ hereceives an input, performs his own work process and transfers the outputas a „supplier„ to his „customer„. In the sense of this net product chain, notonly the end user is a customer, but also each organizational unit within acompany.
QFD promises success, when compliance with all customer requirements isa company-wide target. Accordingly, the QFD team has to represent theinterests of the following internal and external customers:
- Car manufacturers (internal or external)- Suppliers (external or internal)- End users (external)- Authorities (external)- Company employees (internal) from the following departments
- Sales- Service- Development- Engineering- Material planning- Planning- Logistics- Production- Controlling.
45
Coordination is taken over by an independent body of the company, e.g. anemployee of the quality department.
Whilst in QT-l phase every team member must provide his input, in the QT-II phase, the representatives of external customers, sales and servicedepartments are not required. In the QT-III and QT-IV phase, thedevelopment department also no longer needs to be represented.
6.2 Case Example (until end QT-I)
In the following, the concept for an ice scraper using QFD is developed.This example aims to clarify the individual phases and steps of the method,however it does not claim to be exhaustive.
When preparing the quality table QT-l, one proceeds as follows (Fig. 6.5):
1) Obtaining and weighting the „voice of the customer„ (customerrequirements)
2) Competition analysis of the „voice of the customer„ from a customer’sperspective
3) Defining complaints, warranty cases and sales arguments
4) Establishing critical customer requirements
5) Determining internal company characteristics and ...
5a) ::: their interaction with customer requirements
6) Competition analysis of the internal company characteristics from acompany’s perspective
7) Compiling test methods, current and future specifications Evaluatingthe difficulties of achieving the objective
8) „In-House-Test“ to establish the critical internal company characte-ristics
9) Correlating the critical internal company characteristics with all others.
10) Correlation of the critical internal company characteristics with allothers.
46
10
5
1 5a 2,3,4
6,7,8 9
Fig. 6.5 Steps in the „House of Quality“
1st Step: Establishing and weighting of the „voice of the customer„(customer requirements)
In the first step, the QFD team compiles the whishes, requirements andneeds of all external and internal customers. Hereby, expressed and notexpressed (silently assumed) customer expectations, as well as unexpectedpositive properties (innovations) are taken into account and are weighted inthree categories (9: very important, 3: important, 1: less important).
The aspects; functionality, reliability, producibility, environmental conditionsand protection are to be dealt with in the „House of Quality“. Small costs orhigh profit are, in contrast, universally valid requirements, which should beexamined separately; thereby, the completed „House of Quality„ – estab-lished without financial criteria – serves as a decision aid, as it makes thepossibilities and effects of cost savings or reductions transparent.
2nd Step: Competition analysis of the „voice of the customer„ from thecustomer‘s perspective
The compliance with the customer requirements established in the first step,is evaluated in comparison to the competition from the customerperspective, in the second step. Each individual requirement is therebyawarded a grade from 1 (= very bad compliance) up to 5 (= very goodcompliance) for the company and it’s best competitor. Based on thiscompetition analysis, the target for the company is set for each requirementusing the same grading system.
47
Thereby, three profiles emerge line by line in the corresponding space of the„House of Quality„: company actual-value, best competitor and companytargets.
3rd Step: Defining complaints, warranty cases and sales arguments
As a third step, available complaints, warranty cases and sales argumentsare compiled regarding each customer requirement. A code letter, forexample, (C=Complaint, W=Warranty case, S=Sales argument) and theline number is to be filled into the space provided in the „House of Quality„.The complaints, warranty cases and sales arguments are then listed in aseparate table under the respective code letter and line number. „B7„therefore would be e.g. a complaint, which refers to the customerrequirement in line 7 of the „House of Quality“ (see Fig. 6.6).
4th Step: Establishing critical customer requirements
Based on the established evaluations from the customer’s perspective, thecritical customer requirements are marked with a star in the fourth step.
The first four steps are shown in Fig. 6.6.
5th Step: Determining the internal company characteristics and theirinterrelation with customer requirements
In the fifth step, for all customer requirements at least one internal companycharacteristic must comply. To gather ideas and a logical structure, fishbonediagrams and fault tree structures may be useful. The established internalcompany characteristics are assigned to the columns of the „House ofQuality„. This results in a matrix between the customer requirements in therows and internal company characteristics in the columns. The fields of thismatrix are completed based on the question: „Can the customerrequirement of the corresponding line be met through the internal companycharacteristic of the respective column?„.
48
internal companycharacteristics (HOW)
Customer requirements(WHAT)
no risk ofinjury
easyhandling
Weighting
Strong 5Medium3Weak 1
Compet. Analysisof customerrequirements
1 2 3 4 5
must not damage vehicle 1 3
easily producable 2 3
easy to stow away (take out) 3 3
matching the vehicle 4 1
many functions 5 9
during production 6 1
during use 7 9
function at every temperature 8 3
fingers must not freeze 9 9 B9, V9
no physical effort 10 3
must clean pane 11 9 G11, V1
Ice
scra
per
Line
num
ber
Prio
rity
Com
plai
nts
war
rant
y ca
ses
sale
s ar
gum
ents
Tod
ayC
ompe
titio
n
Tar
get
Crit
ical
cut
omer
req
uire
men
t
Weighting
Strong 5Medium 3Weak 1
Fig. 6.6 1st Step to 4th Step
Table of complaints, warranty cases and sales arguments:
B9: scraped-off ice always falls onto hands V9: scraped-off ice cannot fall onto hands G11: ice scraper does not scrape along the complete length, pane
immediately ices over V11: ice particles are removed immediately (possibly with a squeegee).
The completed matrix shows the significance of the individual internal com-pany characteristics to the compliance of the customer requests (Fig. 6.7).
very good =good =not so good =not at all = no entry.
49
Phase I - Product Planning
Internal companycharacteristics (HOW)
Customer requirements(WHAT)
Column numbermust not damage the vehicle
easy to produce
easy to store (take out) everywherematches the car
many functionsduring production
during use
functions at any temperaturefingers shall not freeze
no strenuous effort
must clear the pane
no risk ofinjury
easy handling
1
2
34
5
6
7
89
10
11
1 2 3 4 5 6 7 8 9
Line
num
ber
Prio
rity
Har
dnes
s
Sur
face
fini
sh
Tem
pera
ture
sta
bilit
y
Mat
eria
l com
posi
tion
Fle
xibi
lity
Saf
e ag
ains
t fra
ctur
e
Fra
ctur
e ty
pe
The
rmal
cap
acity
Ope
ratin
g m
ater
ials
Deg
ree
of a
utom
atio
n
Com
plex
ity
Seq
uenc
e of
ope
ratio
ns
Qua
lific
atio
n
Tra
inin
g
Ext
erio
r co
ntou
r
Col
orin
g
Num
ber
of fu
nctio
ns
?? H
eat s
ourc
e
Han
dlin
g in
put
Shi
fting
pow
er
10 11 12 13 14 15 16 17 18 19 20
Com
plai
nts,
war
rant
y ca
ses,
sale
s ar
gum
ents
Ice
scra
per
Tod
ayC
ompe
titor
Tar
get
1 2 3 4 5
1 2 3 4 5 6 7 8 9
Competition Analysisof customer
requirements
Rating:
StrongMediumWeak
Material SelectionProductionprocedure
Form/Design
Han
dlin
gpo
wer
?
??
?
B.V9
G.V11
Fig 6.7
5th S
tep
50
6th Step: Competition Analysis of internal company characteristicsfrom a company perspective
Simultaneously to the procedure in the second step, the internal companycharacteristics are evaluated in a competition analysis in the sixth step.Here again, three profiles emerge applying the grades 1 (very badcompliance) to 5 (very good compliance): Own company today, bestcompetitor, objectives.
7th Step: Compiling test methods, current and future specifications
In the seventh step, test methods, as well as current and future specifi-cations, are compiled for all internal company characteristics. Therefore, T1,T2, T3 etc. for test methods, H1, H2, H3 etc. for current specifications, andZ1, Z2, Z3 etc. for future specifications have to be filled in the columnprovided in the "House of Quality". The individual definitions are described inmore detail under their respective coding in the attachment (see Fig. 6.8).
51
Phase I - Product Planning
Internal companycharacteristics (HOW)
Customer requirements(WHAT)
Column number 1 2 3 4 5 6 7 8 9
Har
dnes
s
Sur
face
fini
sh
Tem
pera
ture
sta
bilit
y
Mat
eria
l com
posi
tion
Fle
xibi
lity
Saf
e ag
ains
t fra
ctur
e
Fra
ctur
e ty
pe
The
rmal
cap
acity
Ope
ratin
g m
ater
ials
Deg
ree
of a
utom
atio
n
Com
plex
ity
Seq
uenc
e of
ope
ratio
ns
Qua
lific
atio
n
Tra
inin
g
Ext
erio
r co
ntou
r
Col
orin
g
Num
ber
of fu
nctio
ns
?? H
eat s
ourc
e
Han
dlin
g in
put
Shi
fting
pow
er
10 11 12 13 14 15 16 17 18 19 201 2 3 4 5 6 7 8 9
Material SelectionProductionprocedure
Form/Design
Han
dlin
gpo
wer
Test methods, current and futurespecifications
Difficulty to achieve the target(1 = easy, 5 = difficult)
Competition analysisof internal companycharacteristics
TodayCompetitorTarget
5
4
3
2
1T
,H,Z
1
T,H
,Z2
T,H
,Z3
T,H
,Z4
T,H
,Z5
T,H
,Z6
T,H
,Z7
T,H
,Z8
T,H
,Z9
T,H
,Z10
T,H
,Z1
1
T,H
,Z12
T,H
,Z13
T,H
,Z14
T,H
,Z15
T,H
,Z16
T,H
,Z17
T,H
,Z18
T,H
,Z19
T,H
,Z20
1 4 2 5 4 3 1 5 1 2 4 2 3 2 4 1 4 5 3 3
List of the test methods, current and future specifications:
T1:
Hardness
test, H
1: P
lastic indentation
hardness 84 N
/mm
² R
ubberhardness S
hore A70, Z
1: =H
1, etc.
Fig. 6.8
6th S
tep to 8 th S
tep
52
8th Step: Evaluating the difficulties of achieving the objective
In the eighth step, the difficulty for each internal company characteristic toreach the respective target is evaluated using the steps 1 (easy) to 5 (diffi-cult). The steps 6 to 8 are shown in Fig. 6.8.
9th Step: „In House-Test“ to establish the critical internal companycharacteristics
In the ninth step (Fig. 6.9), one examines the almost fully completed "Houseof Quality". Here one has to see, whether the evaluation from thecustomer’s perspective, in the horizontal, harmonize with the grading fromthe company’s perspective, in the vertical. If that is not the case, appropriatecorrections have to be made in the "House of Quality". Subsequently, thecritical internal company characteristics are marked with a star.
53
Phase I - Product Planning
Internal companycharacteristics (HOW)
Customer requirements(WHAT)
Column number
must not damage the vehicle
easy to produce
easy to store /take out) everywhere
matches the car
many functions
during production
during use
functions at any temperature
fingers shal not freeze
no strenuous effort
must clear the pane
1 2 3 4 5 6 7 8 9
Line
num
ber
Prio
rity
Har
dnes
s
Sur
face
fini
sh
Te
mpe
ratu
re s
tabi
lity
Mat
eria
l com
posi
tion
Fle
xibi
lity
Saf
e ag
ains
t fra
ctur
e
Fra
ctur
e ty
pe
The
rmal
cap
acity
Ope
ratin
g m
ater
ials
Deg
ree
of a
utom
atio
n
Com
plex
ity
Seq
uenc
e of
ope
ratio
ns
Qua
lific
atio
n
Tra
inin
g
Ext
erio
r co
ntou
r
Col
orin
g
Num
ber
of fu
nctio
ns
?? H
eat s
ourc
e
Han
dlin
g in
put
Shi
fting
pow
er
10 11 12 13 14 15 16 17 18 19 201 2 3 4 5 6 7 8 9
Material SelectionProductionprocedure
Form/Design
Han
dlin
gpo
wer
Test methods, current and futurespecifications
Difficulty to achieve the target(1 = easy, 5 = difficult)
Competition analysisof internal companycharacteristics
TodayCompetitorTarget
5
43
2
1
T,H
,Z1
T,H
,Z2
T,H
,Z3
T,H
,Z4
T,H
,Z5
T,H
,Z6
T,H
,Z7
T,H
,Z8
T,H
,Z9
T,H
,Z10
T,H
,Z1
1
T,H
,Z12
T,H
,Z13
T,H
,Z14
T,H
,Z15
T,H
,Z16
T,H
,Z17
T,H
,Z18
T,H
,Z19
T,H
,Z20
1 4 2 5 4 3 1 5 1 2 4 2 3 2 4 1 4 5 3 3
no risk of injury
easy handling
Ice
scra
per
Com
plai
nts,
war
rant
y ca
ses,
sale
s ar
gum
ents
B.V9
G.V11
To
day
Com
petit
or
Ta
rget
1 2 3 4 5
Competition Ana-lysis of customerrequirements
Crit
ical
cus
tom
er r
equi
rem
ents
1
2
3
4
5
6
7
8
9
10
11
??
Fig. 6.9
9th S
tep
5410th
Ste
p:
Cor
rela
ting
the
criti
cal i
nter
nal c
ompa
ny c
hara
cter
istic
s w
ithal
l oth
ers
Phase I - Product Planning
Internal companycharacteristics (HOW)
Customer requirements(WHAT)
Column number
must not damage the vehicle
easy to produce
easy to store /take out) everywhere
matches the car
many functions
during production
during use
functions at any temperature
fingers shal not freeze
no strenuous effort
must clear the pane
1 2 3 4 5 6 7 8 9 Li
ne n
umbe
r
Prio
rity
Har
dnes
s
Sur
face
fini
sh
Tem
pera
ture
sta
bilit
y
Mat
eria
l com
posi
tion
Fle
xibi
lity
Saf
e ag
ains
t fra
ctur
e
Fra
ctur
e ty
pe
The
rmal
cap
acity
Ope
ratin
g m
ater
ials
Deg
ree
of a
utom
atio
n
Com
plex
ity
Seq
uenc
e of
ope
ratio
ns
Qua
lific
atio
n
Tra
inin
g
Ext
erio
r co
ntou
r
Col
orin
g
Num
ber
of fu
nctio
ns
?? H
eat s
ourc
e
Han
dlin
g in
put
Shi
fting
pow
er
10 11 12 13 14 15 16 17 18 19 201 2 3 4 5 6 7 8 9
Material SelectionProductionprocedure
Form/Design
Han
dlin
gpo
wer
Test methods, current and futurespecifications
Difficulty to achieve the target(1 = easy, 5 = difficult)
Competition analysisof internal companycharacteristics
TodayCompetitorTarget
5
432
1
T,H
,Z1
T,H
,Z2
T,H
,Z3
T,H
,Z4
T,H
,Z5
T,H
,Z6
T,H
,Z7
T,H
,Z8
T,H
,Z9
T,H
,Z10
T,H
,Z11
T,H
,Z12
T,H
,Z13
T,H
,Z14
T,H
,Z15
T,H
,Z16
T,H
,Z17
T,H
,Z18
T,H
,Z19
T,H
,Z20
1 4 2 5 4 3 1 5 1 2 4 2 3 2 4 1 4 5 3 3
no risk of injury
easy handling
Ice
scra
per
Com
plai
nts,
war
rant
y ca
ses,
sale
s ar
gum
ents
B.V9
G.V11
Tod
ay
Com
petit
or
Tar
get
1 2 3 4 5
Competition Ana-lysis of customerrequirements
Crit
ical
cus
tom
er r
equi
rem
ents
1
2
3
4
5
6
7
8
9
10
11
??
??
?
xx
x?
Fig
. 6.1
010
th s
tep,
fully
com
plet
ed „
Hou
se o
f Qua
lity“
55
This ends the QT-I phase. Parallel to the processing of the other QFDphases, the results gained in the QT-I phase are used for further activities,such as risk analysis (FMEA, FTA), experiment planning (DoE) and SPCpreparation (Fig. 6.11).
QFD
RiskAnalysis
DoE SPC
Criticalcharacteristics
Openinterrelationsn
Customer relev antcharacteristics
Fig. 6.11 QFD as a basis for further methods
6.3 Quality Table II (QT II)
Components or Parts Planning (Phase 2)
The critical internal company product characteristics gained in quality table I(QT 1), are basis for the further procedure in phase 2.
The product characteristics are now the "WHAT" requirements and the"HOW" interpretations (constructive solutions) have to be worked out accor-dingly.
56
The evaluation of this "WHAT-HOW" table is carried out as described instep 5 of phase 1.
Critical characteristics for components and parts identified here, are furtherdealt with in phase 3.
6.4 Quality Table III (QT III)
Process Planning (phase 3)
The critical components and parts characteristics are taken up again as"WHAT" requirements and the "HOW" interpretations (process solutions)are worked out in the quality tables.
The evaluation of this "WHAT-HOW" quality table is carried out asdescribed in step 5 of phase 1.
Critical processes identified here, are further dealt with in phase 4.
6.5 Quality Table IV (QT IV)
Work instruction planning (phase 4)
The last quality table to be drawn up in this series contains the criticalprocesses as "WHAT" requirements. For this, the "HOW" interpretations(work instructions) are worked out. The evaluation of this quality table iscarried out as in step 5 of phase 1.
57
Literature
(1) Rolf Blank: Quality consciousness starts at the customer!(Lecture manuscript, Kodak AG)
(2) Jürgen Ebeling: Quality on new paths (BMW Brochure)
58
7 Fault Probability and Influence Analysis(Fehler-Möglichkeits und Einfluß-Analyse FMEA)- see VDA Volume 4, Part 2 System FMEA -
Authors: Turecht Beltz, Berthold Edenhofer*), Jürgen Eilers,Andreas Krause, Guido Laschet, Hans-JoachimPfeufer, Detlef Gaide*) , Dietmar Zander
7.1 Explanation
System FMEA is a risk analysis system accompanying development andplanning integrated into the expert departments. System FMEA is an impor-tant methodical tool, in order to enable early identification and prevention ofpotential defects, especially in new concepts.
The flow chart of quality assurance activities shows, at what point SystemFMEA is necessary as a method of TQM during development and planningand when it's presents is required by the quality assurance system (seeattached Flow Chart).
The procedure during the analysis and evaluation of the defects isdescribed in VDA Volume 4, Quality Assurance prior to Serial Application, inthe 2nd edition of 1986.
It has been found, that the existing practice has disadvantages.
These disadvantages made a further development towards SystemFMEA Product and System FMEA Process necessary. Due to the spe-cial significance of System FMEA and the comprehensive descriptionof it's contents, a separate VDA Volume 4, Part 2 is published whichexclusively deals with this subject.
*) Moderators of the Audit Team
59
Note
The FMEA forms issued after the VDA Volume 4, 2nd Edition of 1986 remainvalid. For newly created System FMEA's, new forms are to be used.
60
8 Fault Tree Analysis (FTA)
Authors: Dr. Dieter Keller, Friedrich Scheucher*),Manfred Scholz
8.1 Introduction
Risk analyses mainly serve the timely identification and elimination of weakspots, as well as the comparison of alternative systems. The task of theseprocedures is, to establish the probability of system failures occurring and toshow their potential effects, for the evaluation of a damage or theproceedings of an accident. Thus, the risk is a function of both theprobability and the resulting consequences. Therefore, two parameters arealways of interest in risk analysis:
- the frequency of occurrence and- the system failure effects.
Fig. 8.1 provides an outline of the most common technical riskanalyses.
*)
Moderator of the Author Team
61
FMEA
FMECA
Sequence ofEventsAnalysis
Fault Tree Analysis
InductiveAnalysis
Failure Modes-Analysis
DeductiveAnalysis
RiskAnalysis
Parts-Count-Method
Parts-Stress-Method
Mark-off-Procedure
Failure RateAnalysis
Sy stem
Analy sisState
Fig 8.1 The most common Technical Risk Analyses
62
The first safety/risk analyses (USA 1950) were limited to the examination ofthe various failure modes of components/groups of a system and to failureeffects of the respective failure mode. For the evaluation of failure effectsand occurrence probabilities, four categories were formed heuristically:
Failure Effects Failure Occurrence Probability
No effects
Minor effects
Critical effects
Catastrophic effects
very unlikely(1 occurence/107 operating hrs)
unlikely(1 occurrence/105-107 operatinghrs)
quite likely(1 occurrence/104-105 operatinghrs)
likely(1 occurrence/104 operating hrs)
However, it quickly became apparent, that an exclusive failure mode andfailure effects analysis was difficult to carry out with the increasingcomplexity of equipment and systems, and unsuitable for a quantitativereliability analysis. Based on the knowledge of reliability theory andBooleanalgebra, the engineers of the Bell Telephone Laboratories (H.Watson, 1961) succeeded in demonstrating the failure conduct of steeringsystems in a Boolean Model with logical symbols. Fault Tree Analysis wasborn!
Especially in the last ten years, FTA has constantly been refined and iscurrently probably the most common analysis method for the evaluation ofsafety and reliability of large complex systems.
63
8.2 Purpose
With the aid of Fault Tree Analysis, the logical connections of componentsand partial system failures, which lead to an undesired event, areestablished and graphically presented. The purpose of the analysis is todetect not only the failure causes, but also their functional relationships.
With Fault Tree Analysis one is able
- to identify all possible failures, as well as failure combinations andtheir causes, which lead to an undesired event (top event),
- to present particularly critical events or event combinations (e.g.failures which lead to the top event),
- to calculate reliability parameters, such as e.g. occurrence probabili-ties of the top event or system availability when necessary,
- to achieve objective evaluation criteria for system concepts and- to gain a clear documentation of the failure mechanisms and their
functional relationships.
Fault Tree Analysis is a method with a very universal application. It can beused preventively, as well as identify the causes of already existingproblems.
8.3 Definitions(Extract from DIN 25424)
Examination Unit
The subject of an examination, which depending on type and scope, isdetermined by th examiner according to functional and constructive aspects.Examination units are e.g. systems, components and function elements.
64
System
A system is a summary of technical and organizational means for the self-contained fulfillment of a task. One has to differentiate between a technicalsystem (Hardware System) and a functional system. Corresponding to thediffering functions of a technical system, one or more functional systemsexist.
Component
A component is the lowest examination unit of a technical system. Eachcomponent is assigned one or more function elements.
Function Element
A function element is the lowest examination unit of a functional system. Itmay only describe an elementary function such as e.g. switching, turning,blocking etc.
Failure
Failure of a technical unit arises, when an acceptable deviation (tolerance)from a performance target of this technical unit is exceeded. In a functionalsystem, such a failure represents a loss of function elements, which isdescribed in the FTA as a failure of the function element.
Fault
Unacceptable deviation of a characteristic.
Failure Mode
The various component failure possibilities are described as failure modes.
65
Undesired Event (top event)
The top event (Fault Tree Start) is the failure of the examined functionalsystem. This can be caused by various failure combinations.
Failure Combinations
A failure combination is the simultaneous existence of function elementfailures, which lead to a top event. The smallest failure combinations arecombinations of failures, which contain at least the number of failuresnecessary to cause the top event.
8.4 Description of the Method
Besides the presentation of the functional structure, the developed modelalso allows quantitative representation of the system failure reaction to beexpected. The structural and methodical modeling is carried out usingBoolean Algebra. The FTA can generally be divided into three scopes ofwork:
- Presentation of the cause/effect relationship- Establishing reliability parameters for the basic events- Calculation of reliability parameters.
The starting point is a system, consisting of n components, which can be inexactly two conditions (intact/defect). The system itself can also be in thesetwo conditions. Each node (component/system condition) can be in exactlyone of these conditions. A function is given for each node which specifiesthe dependence on the conditions of the predecessors. The conditions of allnodes without predecessor are described as an independent variable of thefault tree (Fig. 8.2). This lowest level is also described as base event or faulttree leaf.
66
O (Output)
Base Event
Fig. 8.2 Node without predecessor
The components correspond to nodes without predecessors (Leaves of thefault tree), while the system status is represented by a node withoutsuccessor (top events) (Fig. 8.3).
Top Event
I (Input)
Fig. 8.3 Node without successor
The function, which determines the status of a node through its pre-decessors, is a Boolean Function. It is described with a basic function, theso-called gates (see fig. 8.4).
Successor
Gate
Predecessor X Predecessor Y
Fig. 8.4 Boolean Function
67
When developing fault trees, component failures (inputs) are classified intothree categories: Primary failure, secondary failure and command failure.
A primary failure is a component failure, which occurs under acceptableapplication conditions. The cause for a primary failure lies in theconstruction or material characteristics of the component itself.
A secondary failure describes a component failure, which is caused byunacceptable external influences, such as e.g. environmental conditions,application conditions or influence of other system components.
Command failure are caused by human operating errors or misuse.
The actual fault tree now consists of graphical symbols for the abovementioned inputs and their connections. These connections which representlogical relationships, determine, according to characteristic rules, an outputfrom their inputs which is described in binary:
„0“ (intact)„1“ (defect).
For the (1 of 2)-evaluation, the so-called OR-Gate, one of two inputsignals (I1 or I2) is sufficient, to receive the respective output signal (O).Thereby, the output signal not occurring is more unlikely then with a single-channel installation. Fig. 8.5 illustrates the (1 of 2)-evaluation with thecorresponding function table:
I1 I2 O
> = 1 1 1
1 0 1
I1 I2 0 1 1
0 0 0
Fig. 8.5 (1 of 2)-Evaluation
68
Is, e.g., the braking of a hoisting winch initiated by an OR-connection,braking is also possible if an input signal fails. However, operationaldisturbances through faulty initiation of the brakes have to be accepted.
For the (2- of2)-evaluation (AND-Gate) two input signals have to be givenat the same time, in order to obtain the output signal. In this case, theoccurrence of output O becomes more unlikely then with a single-channelinstallation. Figure 8.6 shows the (2 of 2)-evaluation.
O I1 I2 O
& 1 1 1
1 0 0
I1 I2 0 1 0
0 0 0
Fig. 8.6 (2 of 2)-Evaluation
If an actuation occurs by a (2 of 2)-evaluation, an unacceptable start in caseof a fault can be avoided. However, increased operational disturbancesthrough blocking of the actuation have to be expected.
The diagrams for inputs and connections are explained in DIN 25424(Part 19). The summary of gates is shown in DIN 25424 (Part 2).
8.5 Preparation of a Fault Tree
With the help of illustrative examples, the development of fault trees withfundamental AND/OR-connections is explained. Figure 8.7 shows a simpleelectric circuit with power supply, two switches and an engine.
69
System A is in a defined initial state, in which switch 1 and switch 2 areclosed. The top event is chosen to be „Engine cannot be turned off„.
EPower Source Engine
Switch 1 Switch 2
Fig. 8.7 Description of System A
The corresponding fault tree in Figure 8.8 illustrates the two basiccomponent failures for this situation.
Switch 1 cannot be openedSwitch 2 cannot be opened.
Assuming, that in this defined system the cabling and plugs do not influencethe failure, the top event only occurs, when both component failures takeplace (AND-Gate).
70
Engine cannotbe turned off
& Gate
Switch 1cannot beopened
Switch 2cannot beopened
Fig. 8.8 Fault Tree of System A with an AND-connection
Figure 8.9, on the other hand, shows a typical OR-connection. In this case,another top event is assumed for the system: „Engine does not start„.The above-mentioned assumptions again apply here. The figure illustrates,that the engine does not start, if the engine itself fails, external influences,not taken into account, exist or the required power supply is notappropriately fitted.
Engine doesnot start
≥ 1
Switch 1cannot beopened
Secondary failureEngine
Switch 2cannot beopened
Fig. 8.9 Fault Tree of System A with an OR-connection
71
The engine failure hereby is a primary failure, the external influencingfactors, secondary failures and the inappropriate power supply a commandfailure (e.g. through an operation error). In general, primary failures are notfurther developed. If the FTA is used as an analysis to determine the failurecause for existing problems, then a complete listing of the causes shoulddefinitely be aimed for; thus the primary failures are also to be analyzedfurther.
The branches of the command or secondary failures are to be developedfurther to the base events. Figure 8.10 shows by example the fault tree withdeveloped secondary failure of the engine.
Engine doesnot start
≥ 1
Primary FailureEngine
Secondary FailureEngine
Command FailureEngine
≥ 1
External Influences
Failure of otherSystem Components:
Switch 1 and 2,Power Source
≥ 1
Blocking of theenginedue to
dirt
Fracture of enginehousing due to
too hightemperaturesor vibrations
Fig. 8.10 Fault Tree of System A with developed Secondary Failure
72
In order to be able to transfer even highly complex technical systems intorealistic models, the following operational steps for the fault tree preparationare to be carried out:
Operational steps for theFault Tree Preparation
1st Step
System analysis
2nd Step
Definition of the top events and the failurecriteria
3rd Step
Determination of reliability parametersand time intervals
4th Step
Determination of the failure modes of thecomponents
5th Step
Preparation of the Fault Tree
Fig. 8.11 Operational Steps for the Fault Tree Preparation
73
1st Step: System Analysis
The preparation of the fault tree presupposes exact knowledge of thefunctional processes of the normal functioning system. With the help ofsystem analysis, the action of the system, under consideration of its inter-faces with it‘s environment, is to be made transparent.
- System Functions / System Requirements
To clearly define the system function, all required functions are shown andrelated to the functional elements (system elements). Hereby, the perfor-mance targets and tolerances of each of the system functions are to betaken into account. For this, the usual technical documents, e.g. signaland/or current conduction plans, performance specifications and designsdrawings are needed. To illustrate system networks and interfaceinfluences, function block diagrams (FBD) are used. This graphical, two-dimensional presentation within a verbal, one-dimensional approach, asfollowed in specifications, is highly superior, as soon as the functionalprocesses are no longer only sequential.
- Environmental Conditions
The system must fulfill the required functions under influence ofenvironmental conditions, not affected by the technical system itself, in thevarious operational phases. Environmental influences, as well as physicaland chemical characteristics of the system elements are to be considered.
- Dependence and Reaction
Here, the system is to be examined with regards to the following criteria:
- Cooperation of the system elements to create the system functions,- Reaction of the system to environmental conditions,- Reaction of the system in case of internal failures and failure of
required aid sources (power supply, services).
74
2nd Step: Definition of the top event and failure criteria
The meaningfulness of FTA depends on the description of the top event andrelated boundary conditions. When establishing the top event, twofundamentally different approaches are available:
- Preventive Approach
If the FTA is carried out from a preventive point of view, the definition of thetop events arises from the noncompliance of functions or requirements.When defining the top events, relevant safety, as well as comfort productrequirements may be considered.
- Corrective Approach
Here, an occurred failure or system failure is defined as a top event.
3rd Step: Determining reliability parameters and time intervals
When evaluating a quantitative fault tree, one differentiates between thefailure probability over a defined time period and the failure at a randompoint in time.
If one wants to differentiate such quantitative statements for the top event,then appropriate data on the base events is required.
4th Step: Determining the failure mode of the components
After the system analysis and definition of the top event, all failure modes ofthe components which are to be considered for the fault tree model, are tobe differentiated. For a detailed FTA, it is normally not sufficient, to useundifferentiated failures of components as base events. On the contrary,differing failure modes of a component (e.g. contacts of a relay do notopen/close) can
75
have completely different effects on the top event, so that they cannot besubtotaled under a base event, but have to be inserted at different areas inthe fault tree. When determining the reliability parameters for base events,an additional difficulty is created due to the fact, that failure probabilities ofcomponents are known, but not for individual failure modes. Some databooks contain information on the failure modes of certain components.
If no quantitative information about the failure modes is available, one can atleast refer to a worst-case scenario, by using the total failure probability of acomponent as the estimation of the highest probability of its individual failuremodes. It is often useful, to carry out a failure mode analysis (e.g. FMEA)prior to the fault tree analysis, to gather „material„, so that possible failuremodes are known in advance.
5th Step: Fault Tree Preparation
In order to obtain comparable results during the development of fault treesand to limit the arbitrariness of the person preparing them in a sensiblemanner, it is useful to establish a general procedure. Figure 8.11 explainssuch a scheme in more detail.
Starting point for preparation of a fault tree is the identification of the topevent for the system to be examined.
Firstly, it is assessed, whether the described top event can be categorizedas a failure of a system element alone. If this is the case, normally an OR-connection follows with a maximum of three inputs "Primary Failure","Secondary Failure" and "Command Failure". Otherwise, the failures, thatindividually or together result in the top event, are identified. These failuresare entered into commentary squares, logically connected and furtherdeveloped like a top event. Each recorded failure result results in a separate"Fault Tree Mast".
76
As mentioned earlier, primary failures are not developed further, exceptwhen the FTA is used with existing problem fields as a pure cause analysis.If secondary and command failures relate to things which are not part of theexamined system, they are also not further developed. Otherwise, theoperational sequence as for a top events follows. In contrast to a primaryfailure, secondary and command failure do not necessarily have to bepresent. Once the fault tree mast is worked through, one moves on to thenext failure. When no failure is left, the preparation of the fault tree iscompleted.
Selection oftop event
Can the selected eventbe reduced to failure
of hust one component?
Entry of componentfailure into the fault tree
Apply a connection(generally OR) with max.3 inputs: Primary failure, secondary failure and
command failure
Are secondary andcommand failures to befurther developed under
the framework of theassessment?
START
END
Selection of an eventstill to be developed
Are there any branchesleft which end with events
requiring furtherdevelopment?
Sidentification of all failureevents which can directlycause the top event,individually or as a group.
These “cause” events areentered into the fault treeand connected by gates.
Selection of one of the“cause” events
No
Yes
Yes
YesNo
Fig.8.12 Scheme for the Preparation of Fault Trees
77
8.6 Evaluation of the Fault Tree
After preparation of the fault tree, qualitative and/or qualitative statementsabout the system failure reaction can be gained depending on the exercise.For relatively simple fault trees, the evaluation can be carried out manually;complex trees, however, require the use of EDP.
8.6.1 Qualitative Evaluation
FTA is a complete procedure, i.e. if applied consistently, principally all eventcombinations leading to the top event, can be found. It’s limits are thereforenot inherent in the procedure, but in the expertise and care of the user. Theresult of a fault tree analysis is only realistic and therefore meaningful, in sofar as it is possible to portray the system and it's functional failure behavioras a causal chain of effects.
- Critical Paths
Even without using input data (such as e.g. failure rates) for the occurrenceof certain function failures, a fault tree can already provide qualitativeinformation regarding system reliability. Especially when applying FTA as apure cause finding instrument, the developed cause/effect structure helps tosystematically recognize critical paths. A critical path is defined as thebranch, in which the stated component failures are not covered bypreventive system or test mechanisms. Every open result is filled into a list,sorted according to respective area and forwarded to the relevant expertdepartments for clarification. From the results gained, current risks of thepotential system weak spots can be approximately be differentiated.
78
- Minimal Cut Sets
Another possibility of making qualitative statements is, to examine thesystem using the method of the minimal cut set after individual and multiplefailures.
A minimal intersection, also called minimal cut set or critical set, is aminimal combination of system components, the failure of which leads tothe top event. From these combinations, a qualitative statement on theweakest branches of the fault tree can be given.
If the top event is initiated by a minimal intersection with one element in asafety system, a system modification may be required. Reliability can beincreased by integrating a redundancy, and the critical minimal intersectioneliminated.
8.6.2 Quantitative Evaluation
Besides the stated qualitative statements about a system, the quantitativeevaluations of fault trees are of most interest. With the help of reliabilityparameters (failure probabilities or non-availability of the system), theexpected frequency of the top event or the non-availability of a systemfunction occuring can e.g. be calculated. The required occurrencefrequencies are deducted either from observations (e.g. laboratory test orfield experience) or from relevant data sources (e.g. MIL-HDBK 217 F).
- Occurrence Probability of the Top Event
The precondition for the calculation of the probability of the top eventoccurring is the corresponding quantification of all base events.
This calculation can principally be carried out by hand; the procedures to beapplied are described in DIN 25424, Part 2. When somewhat complex faulttrees are concerned, random and also systematic calculation errors canhardly be avoided, so that the use of a suitable calculation program is, inmost cases, practically indispensable.
79
- Sensitivity Analyses
The calculation of the probability of the top event occuring allows decisionmakers to make a global judgement, whether this probability is acceptableor not. If not, one has to question which base events mainly cause theunacceptable occurrence, and if eliminating measures are possible.
This is the purpose of sensitivity analyses. They enable the identification ofthe base events and minimal cut sets, which prevailingly influence theoccurrence probability.
The importance of base events and cut sets is described by so-calledimportant parameters. By calculating the importance for all base events orminimal cut sets, a quantitative sequence of importance can be acquired.Thereby, one obtains indications which measures can most effectivelyminimize the occurrence probability of the top event.
8.7 Establishing the Need for Action and Selecting Measures
In order to determine, whether action is required, the results of the fault treeanalysis, regarding the failure probability of the top event, have to be com-pared to the qualitative and quantitative requirements. If the resulting valuesdeviate from the specified values, action is necessary.
In this case, measures to achieve the specified values have to be taken.Hereby, it is necessary to define the weak spots responsible for the non-compliance with the specified values.
80
- Recommended Measures for Improvement:
To eliminate the above defined weak spots, appropriate measures aresuggested. In order to select the best of these measures, the followingaspects have to be considered and evaluated:
- Feasibility- Costs- Time schedule.
Following this, the required measure is determined, whereby it has to beclearly defined, who is responsible for the implementation and what timeschedule has to be met for the application of the measure.
- Check of Success
In order to prove and secure the success of the measure taken, a new cal-culation of the fault tree has to be carried out. If this calculation showsresults which comply with the specified requirements, achieving theobjective is ensured.
8.8 Procedure(Example)
The following simple example of an electronic circuit (see Fig. 8.13) shallillustrate the structure of a fault tree. A detailed qualitative or quantitativeevaluation is abandoned at this point; furthermore the example does notclaim to be exhaustive.
The process as shown in Fig. 8.11 is followed.
81
1st Step System Analysis
The required system function is defined as follows:
VCC
LED
RC
C
E
C
E
TR1
TR2
B
B
RB2
RB1
u1
Fig 8.13 Electronic Circuit (Principle illustration)
The light-emitting diode LED must light in time with the control voltage u1.
The environmental conditions are as given in the specification.
The fault tree is to be prepared under the assumption, that the failure modeof the transistors is described as a short-circuit or interruption between theemitter and the collector. The principle of the circuit is shown in Figure 8.12.
2nd Step Definition of the top event
The top event is:
LED does not illuminate in time.
The top event thereby includes the complete failure or permanent operationof the LED, as well as frequency shifting between control voltage and theLED.
82
Elements of the circuit:
u1 = control voltageVcc = supply voltageLED = light-emitting diodeRC = resistorTR1 = transistor 1TR2 = transistor 2RB1 = resistor for limiting base current TR1
RB2 = resistor for limiting base current TR2.
3rd Step: (Reliability) does not apply
4th Step: Determining the failure modes
5th Step: Establishing the fault tree (see Fig. 8.14)
Figure 8.14 shows the fault tree structure for the above mentioned topevent. Depending on the exercise, only the depth of the analysis can bespecified.
The generated fault tree already allows a quantitative system evaluation:The system complexity can be significantly reduced and the systemreliability increased through concept changes. The degree of complexity isreflected in the existence of failures of equal components (Transistor 1and 2, resistor 1 and 2), which, independent of each other, respectively leadto the top event.
In this example, a targeted component reduction (concentration or replace-ment of equal components) means functional minimization and at the sametime reduction of risk potentials.
83
LED does not
illuminate in time
Wrong Polarity LED
Damages LED
Cold sold.joint
Prim. failure LED
Poor soldering joints
Short circuit LED
Interruption LED
LED burns outShort circuit ResistanceRC for limiting
collector voltage
LED defect
Faulty supply
Faultycontrol voltage
InterruptionResistor RB1 for
limiting basecurrent
Prim. failure RB1(Interruption)
Damage RB1
Cold sold.joint
High ohm value RB1through ageing
Transistor TR1 isnot triggered
Prim. failure RB2(Interruption)
Damage RB2
Cold sold.joint
High ohm value tRB2through ageing
Transistor TR2 isnot triggered
InterruptionResistor RB2 for
limiting basecurrent
Cold sold.jointInterruption RC
Prim. failure RC
Damage RC
No voltage on the LED
Interruption/Punch-through transist.
TR1
Prim. failure TR1(Interruption)
Wrong Polarity
Cold sold.joint
MechanicalDamage TR1Damage TR1
Short circuit RB1
Interruption/Punch-through transist.
TR2
Prim. failure TR2(Interruption)
Wrong polarity
Cold sold.joint
MechanicalDamage TR2Damage TR2
Short circuit RB2
Prim. failure TR1(Short circuit)
&Prim. failure TR2
(Short circuit)
>=1
>=1
>=1
>=1
>=1
>=1
>=1
>=1
>=1
>=1
>=1
Fig. 8.14 Fault Tree Structure of Electronic Circuit
84
9 Design of Experiments (DoE)
Authors: Rainer Franzkowski*), Dr. Johannes Krottmaier,Hartmut Nowack, Dr. Elmar Rach, Erich Wald
9.1 Introduction
If reliable, producible and marketable products are to be released intoproduction at series start, it becomes necessary, prior to series start, to testdesign and product alternatives in the development phase. Thereby, theeffects of changed parameters have to be examined and evaluated inpractical experiments.
One effective method, to examine the effects of changes in influencingvariables (factors) on the properties of target parameters, is statisticalplanning of experiments. Thereby, design or product alternatives areexamined for their combined effect and influence on the target parameters.It is advantageous that the scope of the experiment and information depth,prior to the start of the experiment, are exactly determined and that theresult is covered statistically.
The following presented possibilities for planning and carrying out of experi-ments can be applied at any phase of production planning, productionpreparation, development and production.
9.2 Problem Description and Analysis
It is necessary, when describing and analyzing problems, to workaccurately. This initially costs time. This time can, however, very quickly beregained, as hereby only really necessary experiments have to be carriedout.
*) Moderator of thhe Author Team
85
This often goes so far, that already after problem analysis, the desiredresult, as described in the formulation of the exercise, is available.
9.2.1 Exercise and Objective
The exercise formulation and objective must show, if an improvement, arelative or absolute optimum is to be achieved, or if a different objectiveexists. For this purpose, quality characteristics and evaluation criteria are tobe established, based on which, the achievement of the objective can beassessed.
9.2.2 Taking stock of the situation
Product:
An exact product description should contain, besides a drawing or sketch,complete details about the functions to be fulfilled. These can be preparedanalog to the FMEA with the help of forms.
Process:
Under a process one understands a production process, as well as afunctional process. The process can be represented using a flow chart |1| ora process plan. If hierarchical orders or processes with mutualdependencies are to be shown, then block diagrams |1| are also suitable.
Environment:
When describing the environment, boundary conditions, as well as alreadyrecognized or suspected disturbance factors, important for problem solvingare to be cited. In practice, these are mostly not controllable systeminfluences.
86
9.2.3 Target Parameters
An essential step in problem analysis is setting of the target parameter(s).Target parameters are directly or indirectly measurable physicalparameters, which are suitable to assess the achievement of the objectivein the experiment.
9.2.4 Influencing Variables, Acquisition and Preparation of Data
For this step, data acquisition across all departments is necessary. Dataand information from current production, as well as results from previousexaminations are included in this.
The information about the actual situation is structured, whereby a largeselection of simple and proven tools are available to use:
- Failure list, failure location diagram |1|- Pareto Analysis (also called ABC-Analysis) |1|- Graphic illustration (Line-diagram, Bar chart, Pie chart) |1|- Stratification, Histogram, Correlation diagrams, Box Plots, |1| Proba-
bility grid |2|.
For new developments, theoretical knowledge or knowledge derived fromthe results of simulations are to be considered.
9.2.5 Acquisition, Evaluation and Selection of Influencing variables
For the acquisition of possible influencing variables, brainstorming is carriedout. Possible aids when collecting and structuring suspected influencingvariables are:
- Meta-plan technique |3|- Cause/Effect-Diagram
(also known as Fishbone- or Ishikawa-Diagram) |1|- Fault tree diagram. |4|
87
The determined influencing variables are subsequently sorted and listed asgroups. One differentiates
- independent, controllable and impressionable influencing variablese.g. technical or physical parameters, like dimensions, pressure,number of revolutions,
- hardly or not impressionable influencing variables e.g. surroundingtemperature, humidity, customer behavior, application profile.
9.2.6 Interactions
It is necessary to consider possible interactions. An interaction betweeninfluencing variables exists, when the effect of one influencing variable,depends on the level at which the other influencing variable(s) is (are) set.
9.2.7 Example
As an example, to explain and illustrate the procedure in the framework ofdesign of experiments, a demonstration experiment is discussed. It dealswith the fitting bearing of a drive shaft of an automatic gear, illustrated in(Fig. 9.1).
Fig 9.1 Fitting bearing of a drive shaft of an automatic gear, ∅ 31 mm
88
Taking stock and discussing the initial situation, resulted in the objective ofthe examination being the improvement of the surface finish of the bearing.As target parameters, to allow assessment of this process, the followingwere established:
1. Surface Roughness Rz,2. Chip shape.
The surface roughness RZ on 31 mm diameter was measured according toDIN 4768, see Fig. 9.2. The assessment of the chip shape was carried outagainst the modified standard series INFOS (according to W. König andW. Eversheim 1977).
Reference distance (total measured distance)
Fig 9.2: Determining the surface roughness RZ acc. to DIN 4768
The contact length is divided into five equal, individual measuring lengths afterleaving out the run on and run off section. The mean surface roughness RZ isdefined as an arithmetical mean value of the surface roughness of the fiveadjacent individual measuring lengths. The surface roughness Rmax is thelargest roughness of the five measurements.
The result of extensive brainstorming about the possible influencingvariables of the surface finish is shown in the Cause/Effect Diagram inFig. 9.3. Interactions between most of the influencing variables listed can beexpected.
89
Tool
Anglesetting
Cuttingshape groove
Toolmaterial
CeramicsPCD HSS
CBN tungstencarbidecoated
tungstencarbide
Positioning-Tool/Production piece
Cut depth
Lubricant
with withoutcompo-sition
littlemuch
contami-nation
Machine
A
C
F
G
groundfloor
floor
Installation
Cuttingspeed
B
D
HCuttingedge radius
Positive/negativecutting edge geometry
Material/production piece
Carbon contentEHardness (HRC)
Primary structureAlloying elements
Pretreatmentproduction piece Fixing of the
production piece
Lot size
R (µm)Chip form
Z
time
shift
Time of day
Lighting
Temperature
Attention
Man/Environment
Employee
unskilled
skilled
Rate of feed
?
Fig 9.3 Cause/Effect Diagram for the target parameters surface rough-ness and chip shape of a fitting bearing of a drive in an auto-matic gear
9.3 Reducing the Number of Influencing variables, SelectingFactors for Experiments
The step of reducing the number of possible influencing variables,determined during the problem analysis, to a manageable amount for theexperiment, is usefully performed by the working group, which carried outthe problem analysis. Thereby it is guaranteed, that all practical and techni-cal expertise has an influence on the factor selection. In connection with themethod presented here, Scheffler |6| is especially referenced.
90
9.3.1 Reproducibility and Independence
When setting the selected influencing variables, they must be able to beadjusted to determined levels with reproducible accuracy. If the setting ofone influencing variable is changed, it must not have an effect on the settingof other influencing variables.
9.3.2 Evaluation Criteria and Scale
Depending on the problem, evaluation criteria for the selection of theinfluencing variables are established and assigned with a suitable multistagescale. Each favorable case is highly rated; the more unfavorable therelation to the evaluation criteria, the smaller the rating.
For example:
Effort of setting the influencing variable1 high ... 10 low
Suspected variable influence1 low ... 10 high
Costs when altering the influencing variable1 high ... 10 low
The multiplication of the means determined for each evaluation criteria,results in a figure indicating the priority of an influencing variable. Theinfluencing variables are sorted according to their priority.
9.3.3 Weighting of Influencing Variables
The evaluation and weighting of influencing variables are to be carried outby all team members together. Thereby, a mean rating for each evaluationcriteria of an influencing variable is firstly produced. Then all mean ratings ofan influencing variable are multiplied with one another.
91
The result is a figure indicating the priority of this influencing variable.Finally, the influencing variables are sorted according to their priority andthereby weighted.
9.3.4 Effects Matrix (according to Scheffler)
Under an effects matrix, one understands the tabular illustration of thechanges of the target parameter(s) when there is variation in the influencingvariables. Curves and symbols characterize the suspected or knownchange.
Y
Y
1
2
X X X1 2 3 4
X
? ? ?
Effect known
Effect known and not linear
Effect suspected
Effect not known
X1 - Xn = Influencing variablesY1 ... Yn = Target parameters
Fig. 9.4 Effects Matrix according to Scheffler (Blick27.drw/Designer)
9.3.5 Interactions
For the selection of a suitable and economical experiment plan it isespecially important to gain prior information about possible interactionsbetween the influencing variables. If actual interactions are not taken intoaccount when establishing the experiment plan, the results of theexperiment can lead to incorrect statements.
92
9.3.6 Factor Levels
Those influencing variables, which have been taken into account in theexperiment planning, are called factors. In the simplest case, one assumestwo factor levels. Both levels of the individual factor are chosen at „suitable“intervals from one another based on technical considerations, boundaryconditions, as well as experimental feasibility. For qualitative factors, theadjustment inaccuracy must be negligibly small compared to the levelintervals.
9.3.7 Summary of Factor Selection in a Flow Chart
Influencing variablereproducubleindependent
?
No
Influencing variable evaluationevaluation criteria and
evaluation scale
Weighting of the influencing variabletrough mathematical connections with
Raise effects of the targetparameters factors
Acquiring prior information aboutpossible interactions between
the factors
Determine the factors levelsand test their consistency
Not suitable forstatistical experimentplanning
1
2
3
4
5
6
Yes
Fig 9.5 Flow Chart of Factor Selection and Levels
93
9.3.8 Example
Adequate basic knowledge about the principle effects of the diverseinfluencing variables are available for the cutting process. With the help ofthis prior knowledge, the following eight essential influencing variables fromFig. 9.3 were established as factors for the examination:
A = Cooling lubricantB = Cutting speedC = Cutting depthD = Rate of feedE = Material / carbon contentF = Cutting angleG = Chip form grooveH = Cutting edge radius
In order to achieve as simple an experiment plan as possible, each of thesefactors was examined at two levels. With the help of known experience withthis process, the team set the following factor levels:
Factor Levels
Factor - +
ABCDEFGH
no100 m/min
1 mm0,2 mm/hMaterial 2
45°small
0,8 mm
yes150 m/min
2 mm0,3 mm/hMaterial 1
75°large
1,2 mm
Fig. 9.6 Factors and Factor Levels of the Cutting Process.
94
In as far as knowledge or suspicions about the effects of these factors onthe target parameters were available, they were illustrated in the effectsmatrix of Fig. 9.7.
Targetparameter
largeSurface roughness small
favorable
Chip form
unfavorable
Influencingfactors
ACoolinglubricant
- +
BCuttingspeed
- +
CCuttingdepth
- +
DRate
offeed
- +
EProduc-
tion piecematerial- +
FAnglesetting
- +
GChipform
groove- +
HCuttingedgeradius- +
Fig, 9.7 Effects matrix of the known and suspected effects of the eightselected factors on the target parameter.
9.4 Selecting an Experiment Strategy
Some frequently used experiment plans are briefly described. Theyrepresent only a small extract from known experiment strategies. Moredetailed information is, for example, given by Juran |7|.
9.4.1 One-Factor-Experiment
Assessment of the effects of a quantitative or qualitative factor on one ormore target parameters. The factor is adjusted to two or more levels. Theassessment is carried out the same number of times at all levels (n > 1).
The effects of influencing variables, which are not included in theexperiment plan, are to be eliminated. This can be done either by keepingthem constant during the experiment or, in that the test pieces are randomlyassigned to the levels.
95
The one-factor-experiment provides statements about the effect of thisfactor on one or more target parameters under strictly set conditions.Furthermore, the residual or experiment variation can be estimated.
Typical application: Assessment of the effect of a factor.
9.4.2 Complete Factorial Experiment
Assessment of the effect of several quantitative or qualitative factors on oneor more target parameters. Each factor is set to two or more levels. Theassessment is carried out on all possible factor level combinations, thesame number of times (n possibly >).
The effects of influencing variables, which are not included in theexperiment plan, are to be eliminated. This can be done, either by keepingthem constant during the experiment or, in that the test pieces are randomlyassigned to the factor level combinations.
The complete factorial experiment provides statements about the effects ofthe examined factors on the target parameter(s) and about the interactionsbetween the examined factors. Furthermore, the residual or experimentvariation can be estimated.
Typical application:
Examination of the effects of a small number of factors, if interactionsbetween them can be expected or cannot be excluded.
96
Example:
Examination of the effects of four factors A, B, C, and D on one targetparameter when the factors are each set to two levels ' - ' and ' + '. Name ofthis experiment plan: 24.
No. A B C D
12345678910111213141516
-+-+-+-+-+-+-+-+
--++--++--++--++
----++++----++++
--------++++++++
This experiment gives statements about:
- the main effects of A, B, C and D- the two-factor interactions AB, AC, AD, BC, BD and CD- the three-factor interactions ABC, ABD, ACD and BCD and- the four-factor interaction ABCD.
Further literature: E. Scheffler |6|.
9.4.3 Fractional Factorial Experiment
Assessment of the effects of several quantitative or qualitative factors onone or more target parameters. Each factor is set to two or more levels. Theassessment is carried out with a selected part of the possible factor levelcombinations, the same number of times.
97
The effects of influencing variables, which are not included in theexperiment plan, are to be eliminated. This can be done, either by keepingthem constant or, in that the test pieces are randomly assigned to theselected factor level combinations.
The fractional factorial experiment provides statements about the effects ofthe examined factors on the target parameter(s) and about the interactionsbetween the examined factors. Thereby, however, depending on theselection of the factor level combinations, commixtures occur. Interactioneffects are mixed with themselves and also with main effects. The degree ofcommixture depends on the chosen experiment plan.
Typical application:
Examination of the effects of a large number of factors, when at least onepart of the possible interactions is demonstrably not available or can bejustifiably excluded. Frequently it is assumed, that higher interactions, i.e.interactions between more than two factors, can be ignored.
Example:
Examination of the effects of four factors A, B, C and D on one targetparameter when the factors are each set to two levels ' - ' and ' + '.Performing the experiment with half of the possible factor levelcombinations of the complete factorial experiment.
Name of this experiment plan: (1/2 . 24 =) 24-1.
No. A B C D
12345678
-++-+--+
-+-+-+-+
--++--++
----++++
This experiment gives statements about the following effects commixed withone another:
98
A + BCD,B + ACD,C + ABD,D + ABC,AB + CD,AC + BD,AD + BC.
This experiment plan is therefore usable, when at least one of the fourfactors does not show any interaction with the other factors and when,beyond this, higher interactions can justifiably be ignored.
Further literature: E. Scheffler |6|.
9.4.4 Factor Search according to D. Shainin
Identifying those factors, which have the strongest influence on the targetparameter(s).
The most important influencing variables are selected for the examinationsand sorted, from a technological point of view, according to their suspectedsignificance:
A, B, C, D, E, . . . . .
Subsequently each of these factors is assigned with two levels, one level ' +', from which the more favorable effect on the target parameter is expected,and one level ' - ', from which the less favorable effect on the targetparameter, is expected, from a technological point of view.
In the first experiment phase, two experiments are carried out the samenumber of times (at least twice). In the first experiment, all factors are set tothe level ' + ', and in the second experiment, all factors are set to the level ' -'. If the difference between both settings is significant, one can move ontothe second experiment phase.
99
Is the difference between both settings not significant, the cause must bedetermined. Either important influencing variables were overlooked and nottaken up as factors in the examination, or the levels ' + ' and ' - ' werepartially wrongly assigned. In a further brainstorming session by the experts,the influencing variables are newly discussed and the setting of the levels ischecked with the help of one-factor-experiments.
In the second experiment phase, the following four experiment results arecompared to one another:
A - B - , C - , D - , E - , . . .A + B - , C - , D - , E - , . . .A - B +, C +, D +, E +, . . .A + B +, C +, D +, E +, . . .
Depending on how strong the effect of A is compared to the effect of theremaining factors, A can be classed as the crucial factor, as one of variousessential factors, or as an unimportant factor. The same examination is thencarried out with the other factors B, C, etc., until all essential factors arefound.
This method identifies from a large number of factors, those factors whichhave the strongest influence on the target parameters(s). Furthermore, theresidual or experiment variation can be estimated.
Typical application:
As pre-assessment, when it is expected, that very few factors have a strongeffect on the target parameter(s), whilst the majority of the remaining factorsonly show a small effect on the target parameters (Pareto principle). Sub-sequently a complete factorial experiment can be carried out with the identi-fied crucial factors.
Further literature : K. Bhote |8|.
100
9.4.5 Design of Experiments according to G. Taguchi
Developing products and process which are robust with regard toparameters of the application profile. It is a development tool to achievefollowing goals:
- Development and improvement of products and processes, whichshall be robust with regard to parameters of the application profile(robust design).
- Development and improvement of products and processes, whichshall be robust with regard to parameters of the application profileand simultaneously react sensitively to one or more selectedinfluencing variables (dynamic characteristics).
Taguchi uses traditional experiment plans for his procedure. For two-leveledexperiment plans, however, he uses the numbers '1' and '2' instead of ' - 'and ' + ' to identify the levels.
9.4.5.1 Developing Robust Products and Processes
If an application profile, with a temperature range from -10°C to +40°C isgiven, as illustrated in Fig. 9.8, and if constant product properties within thisarea are required, then material A2 is preferable to material A1. Whendeveloping robust products and processes, product-specific factors, i.e.factors which determine the product properties, as well as process-specificfactors and those specific to the application profile, are to be considered.The latter are called disturbing factors.
101
Tensile strength
A1
A2
-10 °C Temperature 40 °C
Fig. 9.8 Tensile strength of two materials in one temperature range
Product-specific factors and disturbing factors are assigned to different,independent experiment plans, see Fig. 9.9. Each factor level combinationof the product- specific factors is examined with all factor level combinationsof the disturbing factors.
Disturbing parameter matrix
Factor A B C D E F G
1 1 1 1 1 1 1 12 1 1 1 2 2 2 23 1 2 2 1 1 2 24 1 2 2 2 2 1 1 Results5 2 1 2 1 2 1 26 2 1 2 2 1 2 17 2 2 1 1 2 2 18 2 2 1 2 1 1 2C
hara
cter
istic
s co
mbi
natio
n
xyz
111
122
212
221
Fig 9.9: Schematic experiment structure for product-specific factorsand disturbing parameters to develop robust products andprocesses
102
9.4.5.2 Developing Robust and Sensitive Products and Processes
The steering of a vehicle should be designed, so that it
- responds to changes in the steering angle by the driver (signal para-meter) well and evenly, over a wide area,
- reacts irrespective, as far as possible, of external conditions(disturbing parameter), i.e. that constant steering conditions aremaintained under very differing road conditions.
Measuring procedures should be designed, so that they
- react sensitively to changes in the parameter to be measured (signalparameter) and
- irrespective of the external conditions (disturbing parameter) deliverreproducible results.
The objective of development is therefore a product or measuringprocedure, that on the one hand, reacts sensitively to an input parameter, aso-called variable factor (e.g. steering angle), on the other hand, however,does not react to parameters of the application profile (e.g. road conditions).From Fig 9.10 it is apparent, that the factor types can be extended by afurther group, namely the variable factors.
Signal parameter matrix
S S1 2
Factor A B C D E F G
1 1 1 1 1 1 1 12 1 1 1 2 2 2 23 1 2 2 1 1 2 24 1 2 2 2 2 1 1 Results Results5 2 1 2 1 2 1 26 2 1 2 2 1 2 17 2 2 1 1 2 2 18 2 2 1 2 1 1 2C
hara
cter
istic
s co
mbi
natio
n
xyz
111
122
212
221
111
122
212
221
Dis
turb
ing
para
met
er m
atrix
Dis
turb
ing
para
met
er m
atrix
Fig 9.10 Schematic experiment structure to develop robust andsensitive products and processes.
Further literature: M. Phadke |9|.
103
9.4.6 Example
In the example of the „examination of the surface finish of a fitting bearing“eight factors at two levels each are to be included in the experiment plan,see Paragraph 8.3.8. In this case, therefore, factorial experiments with twolevels for each factor can be used. To keep the expenditure of theexperiment (i.e. no. of different partial experiments to be carried out) small,a fractional factorial experiment was chosen, in which the main effects (A,B, C, etc.) are not commixed with two-factor-interactions (AB, AC, BC, etc.),but are only commixed with higher interactions, whilst the two-factor-interactions are commixed with each other and with higher interactions. Inthis case, one speaks of an experiment of resolution IV. These types ofexperiments with low expenditure, have the property, to be able to estimatethe main effects in good approximation. On the other hand, the price thatone must pay for the low experiment expenditure, is the fact that
interactions cannot be established1.
As the experiment contains a total of 16 partial experiments, which cannotbe carried out in one stage with differing settings and for unchangedboundary conditions, an experiment in 4 blocks was planned, which werecarried out on four different days:
1 These experiment plans are therefore readily used as a start to an examination. One initially
determines which factors have a significant main effect, in order to, subsequently, ifnecessary, carry out experiments with these (usually few) factors, which then also allowinteractions to be determined.
104
BlockPart.
experi-ment
Setting of the factors
A B C D E F G H2 + - - - - - + +
I 3 - + - - + + - +14 + - + + - - + -15 - + + + - + - -5 - - + - + + + -
II 8 + + + - + - - -9 - - - + - + + +12 + + - + - - - + Matrix (5)
6 + - + - - + - +III 7 - + + - - - + +
10 + - - + + + - -11 - + - + + - + -1 - - - - - - - -
IV 4 + + - - - + + -13 - - + + + - - +16 + + + + + + + +
Sample size per partial experiment: n = 10
9.5 Evaluation of the Experiment Results
The evaluation of experiments isvda-t explained, in example, with the helpof factorial experiment plans with K factors at two levels each (2k).
9.5.1 Presentation of the Measured Results
The simplest, trivial case k = 1 corresponds to a one-factor-experiment, inwhich the relationship between one target parameter y and only oneinfluencing factor A is examined. The relationship is illustrated in Fig. 9.11.
The change of the target pearameter y when changing from A- to A+ iscalled the effect of factor A. The size of the effect is dependent on thechoice of the settings A- and A+.
105
The above fundamental observations be transferred onto complete andfractional experiments with two or more influencing variables.
y-
y-2
y-1
A- A+ A
Effect (A)
Fig 9.11 Graphic illustration of an effect with a one-factor-experiment
With a two-factor-experiment, the factors A and B are changedcorresponding to the following matrix, to two levels:
No. A B y
1 - - y1
2 + - y2 Matrix (6)
3 - + y3
4 + + y4
In the column y, the means y1, ..., y
4 of the results of the four experiment
rows are shown. They can be illustrated in the following form:
106
A- A+ A
B
B+
B-
y4
y-1 y-
2
y-3
-
Fig. 9.12 Graphic illustration of the results of a two-factor-experiment
A- A+ A
y-3
-y4
y-2
y-1
y-
B+
B-
Fig. 9.13 Graphic illustration of the results of a two-factor-experimentwith the factors A and B
This form of illustration is also usable, when one (or more) of the examinedfactors is not a quantitative, adjustable variable, but instead a qualitativevariable with fixed levels (e.g. Material 1 - Material 2).
Naturally in this case, an interpolation of intermediate values is not sensible.
107
9.5.2 Calculating the Effects
The effect of a factor indicates the change of the target parameter y whenthe setting is changed from level – to level +, averaged over the settings ofall remaining factors. Of course, the effect depends on the explicit choice oflevels.
The graphical determination of the effects is shown in Fig. 9.14 to 9.16 forthe example of the two-factor-experiment.
In as far as the factors remain additive, the course of the two lines inFig. 9.13. is parallel. If the effect of one factor is, however, dependent on thesetting (level) of another, then an interaction between these two factorsexists, they are not additive.
The evaluation matrix of the two-factor-experiment contains, besides thecolumns for the factors A and B, a column AB for the interaction of thesefactors.
No. A B AB y
1 - - + y1
2 + - - y2 Matrix (7)
3 - + - y3
4 + + + y4
108
A- A+ A
y-3
-y4
y-2
y-1
y-
B+
B-
Effect (A)
B- B+ B
y-3
-y4
y-2
y-1
y-
B+
B-
Effect (B)
A- A+ A
y-3
-y4
y-2
y-1
y-
B+
B-Effect (A)at B -
Effect (A)at B +
Effect(AB) = (Effect(A) at B+ - Effect(A) at B-)/2
Figs. 9.14,9.15, 9.16: Graphic illustration of the main effects A and B,as well as the interaction effect AB
109
The effect of the factor X is calculated as the difference from the mean of allresults y, at which X has the level + and the mean of all results, at which Xhas the level -. This calculation is valid simultaneously for interactions andcan generally be used for orthogonal experiment plans with K factors.
For the examined example, the following is therefore valid:
Effect (A) =y y2 4
2
+
Effect (B) =y y3 4
2
+-
y y1 2
2
+(8)
Effect (AB) =y y1 4
2
+-
y y2 3
2
+
With fractional factorial experiment plans, commixtures of the factors withinteractions can occur. The effects of the commixed values can then nolonger be calculated separately.
9.5.3 Statistical Evaluation Procedures
Due to the repetitions of the experiment, it is useful to supplement theevaluation matrix on the right with corresponding columns.
No. A B AB Results y sI
1 - - + y11, ...,y1m y1 s1
2 + - - y21, ...,y2m y2 s2 Matrix (9)
3 - + - y31, ...,y3m y3 s3
4 + + + y41, ...,y4m y4 s4
110
The statistical values established, allow the significance of the calculatedeffects to be judged through comparison with the experiment variation.
For further considerations or examinations, one limits oneself to thosefactors which show significant effects and/or interaction effects.
The performance of the significance test is usefully carried out with the aidof computer support.
9.5.4 Example
The experiment was carried out, as determined in Matrix (5). Only theresults for the surface roughness are described in detail in the following. foreach of these partial experiments, the mean x and standard deviation swere determined. A tabular representation of these results is not given here.The graphic illustration in Fig. 9.17 is more clear.
25
20
15
10
5
0
xin µm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Part. exp. no.
x = 14,36 µm=
Fig. 9.17 Mean x of the 16 partial experiments
111
In this picture one can recognize clear differences between the results of thedifferent partial experiments, caused by the setting of the factors.
The desired information about the effects of the different factors is gainedthrough the investigation of the corresponding effects, see Paragraph 8.5.2and 8.5.3. The calculation of the individual effects and their significance isnot applied in this case.
The results are summarized in Fig. 9.18. Here the effect D = 4,5 µm means,that the surface roughness increases by 4,5 µm, when the rate of feed israised from the lower level = 0,2 mm to the upper level = 0,3 mm.
D
E
H G A
B F
Rate of feed
Material Cutter radiusChip form groove
Cooling lubricant
Cutting speed
Angle settingCut depth C
66
4
2
0
-2
-4
-6
Significant
Notsignificant
Significant
Effect ofRoughnessin µm
Fig. 9.18 Effects of all factors, ordered to their absolute size and signifi-cance
At this point, it has to be pointed out again, that this experiment plan onlyallows the evaluation of main effects, whilst statements about interactionsare not possible.
Analog to the surface roughness, the chip form was evaluated. The resultsis shown in Fig. 9.19. The chip form is better evaluated the smaller the chipbasic number.
112
G
C
H
F
A
D
E
Chip form groove
Cut depth
Cut edge radius
Angle setting
Cooling lubricant
Rate of feed
Significant
Notsignificant
Significant
Effectof thechip basicnumber
2,0
1,5
1,0
0,5
0,0
-0,5
Material
B
Cutting speed
Fig. 9.19 Effect of all factors, ordered according to their absolute valuesand significance
The best possible settings from the point of view of both target parametersroughness and chip form can be read from the Figs. 9.18 and 9.19 asfollows:
Factor
A B C D E F G H
Roughness + (+) (-) - + (+) + +
Chip form - (+) - (+) (-) + - -
Thereby the symbols in brackets represent non-significant effects. Theprimary objective here is low surface roughness with an acceptable chipform for the production. One can recognize that the influencing variableseffect the two target parameters differently, and a compromise must befound.
A follow up experiment with the optimal setting from the point of view of thesurface roughness
A B C D E F G H
+ + - - + + + +
113
resulted in excellent (low) surface roughness, but in a chip form unaccep-table for production. A modified setting, therefore, had to be found for amore favorable chip form which effects the surface roughness as little aspossible. This is obviously possible with the help of factor G, which has arelatively low effect on the surface roughness, but on the other hand, hasthe largest influence on the chip form. The following modified setting wastested:
A B C D E F G H
+ + - - + + - +
The experiment result delivered approximately the same surface roughnesswith an essentially improved and chip form acceptable for production.
9.6 Computer Support
There is a multitude of Software on offer to support the user of statisticalexperiment planning. Therein are independent, complete programs, as wellas some, that are an integral part of a set up modular, comprehensivestatistics Software. The palette of sub-menus, offered as standard oroptions to these programs, is correspondingly varied. The majority of theseprograms support the statistical and graphical evaluation of experiments,however, not the planning phase.
114
9.7 Further Literature
|1| Juran Institute: Quality Improvement Tools
|2| DGQ-Publication 18-19: Forms with Probability Grids
|3| K. Nagel: Success, Oldenbourg-Verlag
|4| DIN 25 424: Fault Tree Analysis
|5| S. Häffner: Statistical Experiment Planning in Company Practice,Company Experiment: Turning, FHT Esslingen, Expert Field: Preci-sion Mechanics, 1992
|6| E. Scheffler: Introduction to the Practice of Statistical ExperimentPlanning, 2nd Edition VEB Deutscher Verlag für Grundstoffindustrie,1984
|7| J. Juran: Quality Control Handbook, McGraw Hill Book Company,New York, 1988
|8| K. R. Bhote: Quality – The Way to the Top of the World, Institute forQuality Management, Großbottwar, 1990 (Translation of World ClassQuality, American Management Association)
|9| M. S. Phadke: Quality Engineering Using Robust Design (in Germantranslation), Gfmt
|10| G. Box, W. Hunter, J. Hunter: Statistics for Experimenters, Wiley &Sons, New York, 1978
|11| Retzlaff, Rust, Waibel: Statistical Experiment Planning, VerlagChemie, Weinheim, 1978
|12| D. Wheeler: Understanding Industrial Experimentation, StatisticalProcess Controls, Inc., Knoxville, 1988
|13| E. Spenhoff: Process Safety through Statistical Experiment Planningin Research, Development and Production, gfmt-Verlags KG, 1991
|14| B. Gimpel: Quality Optimization of Production Processes, VDI-VerlagDüsseldorf, 1991
115
10 Process Capability Analysis
Authors: Manfred Marteloc, Jürgen Jacob, Karl-HeinzSchleiß, Heinrich Ziesler*)
10.1 Introduction
In order to produce in series with high quality and productivity, productionprocesses have to be planned and installed in time. Verification of theprocess capability is hereby an important condition, without which it isimpossible long term, to offer a high quality product to the customer with aneconomically viable input.
Verification of the process capability is carried out under the framework ofStatistical Process Control (SPC).
The process capability established as the result of SPC shall not onlydescribe the status present during the examination period. It is much moreimportant, that such a result can also be used to make a prediction regar-ding the future quality situation.
An important condition for the successful use of SPC is a careful processanalysis. A statement on the process capability is only admissible, if thisanalysis shows the process to be controlled, i.e. if the distribution parame-ters of the characteristics do not change in practice or change only withinknown terms or limits. If, on the other hand, a process is not controlled, nostatement on process capability can be made.
The process with its influencing factors and its possibility of process controlwith the help of set values of the process and product characteristics, isillustrated in Fig. 10.1.
*) Moderator of the Author Team
116
Measureson a process
Process characteristics
Quality information
Product
ProcessManMachineMaterialMethodMedium/Environment
Product characteristics
Fig. 10.1 Process Control Cycle
10.2 Statistical Process Control (SPC)
The name SPC (Statistical Process Control) describes a procedure basedon statistical methods for the technical evaluation of production processesand products. In general, SPC deals with data acquisition, the analysis ofestablished characteristic values, the calculation of process capabilityresults and the assessment of the effectiveness of introduced measures.
Measure
Data acquisition
Analysis
Evaluationof results
SPC(Statistical Process Control)
Fig. 10.2 SPC Steps
117
The above cycle starting with the data acquisition, does not represent aunique process finished with the introduction of measures. On the contrary,in connection with the leading motive of constant quality improvement, thispicture should be regarded as a continuous cycle.
Data Acquisition:
Besides technical knowledge on the process to be examined, SPC requiresthe recording of characteristic values of the produced units and theirevaluation using quality control charts. Process control is only possible withthe help of this data.
Analysis:
A quality capability analysis of the process to be examined, can only becarried out on the basis of available technical knowledge and findings fromthe quality control chart.
Result Evaluation:
In order to obtain information about the quality capability of the process tobe examined, one compares the process variation with the limit values,under consideration of the process status. To simplify understanding, thisresult is expressed as a quality capability code number (capability index).
Measure:
In order to evaluate the effectiveness of an introduced measure, qualitycontrol charts are used as with data acquisition. They fulfil the task ofmaintaining and, whenever possible, improving the achieved quality status.By using this procedure, the cycle demonstrated in Fig. 10.2 is not onlycompleted, but at the same time, the conditions to restart the cycle toimprove quality are established.
Maintaining a quality control chart to illustrate and monitor the process overtime, is therefore an essential basis, whereby however, only a process-orientated quality control chart can show whether the process is actuallycontrolled. To verify the process capability, the characteristic behavior is
118
examined and evaluated over a time axis. If the process is controlled, thecurrent status can be monitored using a quality control chart, and changesto the initial state identified with statistical probability corresponding to theapplied control chart type. However, especially with the process analysis,further analysis methods may become necessary in addition to the qualitycontrol chart.
10.3 Process Influences
Every process is subjected to various influencing factors, which, as a sum oftheir effects, cause the distribution of the values of the characteristics on theproduct. Based on a process-related quality control card, two types ofinfluencing factors can initially be differentiated:
• random influences• systematic influences.
Random influences are the cause of deviations, produced by the chosenprocedure, from the process mean. The deviation of a single characteristic’svalue from the mean is not predictable, but statistical procedures allow,based on a suitable distribution model, the quality capability of the processto be established.
Systematic influences are the cause for additionally deviations occurring inthe process.
Using a process-related quality control chart it is possible, to differentiaterandom and systematic influences from one another. It is thereforesuggested, to carry out this examination with the help of Shewhart*) – Quality Control Charts (e.g.: x /R-, x /s-chart). Going from the acquiredvalues of the characteristics, the intervention limits are establishedaccording to pure
*) W.A.Shewhart: Economic Control of Quality of Manufactured Product, 1931
119
statistical methods with the Shewhart- Quality Control Chart and theprocess-related intervention limits of the control card resulting from thecalculation, make the differentiation of random and systematic influencespossible.
The general aim is to eliminate all systematic influences, in order if possible,to reach a process status, where only random influences are present. Asmentioned earlier, one should initially try to describe, and, if technicallyfeasible and economically viable, eliminate the systematic influences underconsideration of all quality and economic criteria. However, as the Shewart -Quality Control Chart, proposed for the first step, is based on pure statisticalcriteria for the differentiation of random and systematic influences, it issensible, to bring technical knowledge of the process into the examination.
In the second step it is permissible, based on an available examinationresult, to divide the systematic influences into two types:
• predictable systematic influences• disturbance factors.
Predictable systematic influences are systematic influences that can bedescribed by physical regularities and empirical examinations.
Disturbance factors can neither be described by statistical nor process regu-larities and must therefore be eliminated in all circumstances.
If the result of the first examination allows the occurring systematic distribu-tion to described as a predictive systematic distribution, then a correspon-dingly adjusted quality control chart may be applied to monitor and controlthis process. It has to be considered though, that this procedure may onlybe
120
chosen, when systematic influences can be described as predictable syste-matic influences through technical knowledge of the process, physical regu-larities, empirical examinations or by long term observations.
To describe and explain the influences on the process, the quality controlchart, the central element and tool in SPC, has already been used. It is pro-posed, to initially base all analyses on the use of quality control cards. Thisproposal should, however, not prevent the use of other or supplementaryanalysis procedures, if those have proven to be more suitable for a certainprocess.
Examples for the effects of random influences:
• Temperature fluctuations of a casting tool in warm operating state• Positioning accuracy of a machining tool when positioned repeatedly
Examples for the effects of unpredictable systematic influences:
• Temperature changes during the warm-up phase of a casting tool• Wear of machining tools• Tool change of non-adjustable tools.
Examples for the effects of disturbing factors:
• Cooling system failure during a casting process• Tool fracture on machining equipment.
10.4 Process Models
Time response and distribution of the characteristic values are the mainproperties, which are used for the assessment of the quality capability of aprocess. A statement on the process capability is only possible, if thetimeresponse has been examined, if the confirmation that the process iscontrolled is available, and if the distribution of the characteristic’s values
121
characteristics can be sufficiently described by a statistical model ortechnical knowledge of the process. If one is aiming to obtain a precisemathematical description, various options are available. Whilst, on the onehand, it is certainly desirable to find a process model that describes reality indetail, on the other hand, it has to be considered, whether the increasedeffort of finding an exact description is really necessary, or if a simpledescription would suffice.
In many cases, the normal distribution is sufficient as a distribution modelfor determining the process capability.
On the basis of normal distribution, three differing process models aredemonstrated in the following, through which the main process typesavailable in practice are sufficiently well described.
10.4.1 Process Model "A" (see Fig. 10.3)
• The current distribution of the characteristic’s values at the time tcorresponds to a normal distribution with the standard deviation σ (t)and the mean µ (t).
• The process standard deviation σ (t) is constant.• The process mean µ (t) is constant.
Fig. 10.3 Example: A process with dimensional controlRecommended QCC: Shewhart – Quality Control Chart
122
The process model A is equally valid for all other distribution forms, if thecondition of the process mean µ (t) being constant is given. All calculations,however, then have to be carried out on the basis of the respectivedistribution model.
The two following process models B and C describe processes, where theprocess mean µ (t) subject to variations.
These two models should be regarded as marginal cases of the possibilitiesarising in such processes. For most technical applications within SPC thecomplete process result is able to be sufficiently described by the normaldistribution, in cases of extreme variations of the mean over time, however,one should check, whether the process model C would not be more suitableto describe the process.
10.4.2 Process Model "B" (see Fig. 10.4)
• The current distribution of the values of the characteristics at time tcorresponds to a normal distribution with the standard deviation σ (t)and the mean µ (t).
• The process standard deviation σ (t) is constant.• The process means µ (t) scatter in a normal distribution around a
long term constant mean value.
Fig. 10.4 Example: A process dependent on non-adjustable tools, re-commended QCC: Shewhart – Quality Control Chart withextended limits
123
10.4.3 Process Model "C" (see Fig. 10.5)
• The current distribution of the values of the characteristics at time tcorresponds to a normal distribution with the standard deviation σ (t)and the mean µ (t).
• The process standard deviation σ (t) is constant.• The process mean µ (t) changes according to known regularities.
Fig. 10.5 Example: A process, subject to tool wear with relatively lowinternal and extreme mean value variation
Recommended QCC:
Supposition Quality Control Charts.
However, these supposition quality control charts must comply with the con-cept of continuous quality improvement, i.e. the calculation of theintervention limits must be orientated to the real process events and notexclusively to limit values.
10.5 Controlled and Quality Capable Processes
One speaks of a controlled process, when the process behavior (time res-ponse and the distribution of the characteristic values) virtually does notchange or only changes in a known manner or within known limits (e.g.when the standard deviation and the process mean over time for a normaldistribution are constant, or only vary within known limits). Only a controlledprocess allows a statement on the process capability.
124
A controlled process can therefore be described using statistical methods,whereby a statement not only on the current but also on the future qualitycapability to be expected is made possible. One has to consider, however,that the parameters of the population, estimated from the random sampleknown values, are connected with a confidence interval.
Besides the word „controlled„, one also uses the synonyms „stable„ or„statistically under control„ to describe a stable process.
One describes a process as capable – or to be more precise - qualitycapable -, when it is proven to be capable of producing units which complywith quality requirements and when values of characteristics lie withinspecified limits. For a process, which is in accordance with the model ofnormal distribution, this means, that the standard deviation of the values ofthe characteristics from the population, defined as ± 4 σ , must definitely liewithin the specified limits. Through additional agreements betweenmanufacturer and customer, however, other or supplementary conditionscan be determined.
Whether a process can be regarded as quality capable depends on thelimits of the characteristics and any additional agreements betweencustomer and manufacturer. Often, quality capability code numbers areused, to describe this relation (see also Para. 10.6.2).
A very important point that must be taken into account is the fact, that astatement on the quality capability (see Fig. 10.6) is only valid, if it has beenproven in advance, that the process to be assessed is actually controlled.Therefore, the evaluation of the distribution as well as the mean valuebehavior of the random samples is required.
125
Process capability analysis
Assess if the processis controlled
Statement about thequality capability
is permissible
Statement about thequality capabilityis not permissible
Yes No
Fig. 10.6 Flowchart to determine quality capability
The following Figure 10.7 shows exemplary differing process results.
126
Process result Controlled Quality capable
Yes Yes
Yes,if
systematicinflueces
arepredictable
Yes,if
systematicinflueces
arepredictable
Yes,if
systematicinflueces
arepredictable
Yes
Yes
No
Yes No
nostatementpossible
No
No No
Case G
Case F
Case E
Case D
Case C
Case B
Case A
Fig. 10.7 Control and Quality Capability of Processes
127
Case A: This process is controlled and quality capable. This statusshould always be aimed for, as such a process offers thesafest and most reliable basis for a high quality niveau.
Case B: This process is controlled with a normally distributed meanµ (t). Through the additional deviation of the mean incontrast to Case A, the standard deviation of the populationis also larger than in Case A. The aim therefore is, toidentify the cause for this additional deviation, in order to beable to eliminate it. If this is not possible, the process is tobe monitored with the help of the Shewhart Chart withextended limits.
Case C: This process can also be regarded as controlled, if the addi-tional deviation as in Case B can be described by statisticalregularities or technical conditions. However, here again it isimportant, to establish the cause for this additionaldeviation, in order to be able to eliminate it wheneverpossible. If this is not possible as a result of an examination,the process is to be monitored with the help of anappropriately adjusted quality control chart which complieswith the described technical status. When calculating thequality capability code number the additional deviation mustdefinitely be taken into account.
Case D: For this case, it is required that the additional deviationwhich leads to large changes of the mean over time, isexamined. Through improvement measures, at least CaseC has to be achieved. As long as Case D exists, a sortingcheck has to be carried out.
Case E: For this case, to examine and, whenever possible, toreduce the random distribution influences and to aim forCase A. As long as Case E exists, a sorting check has to becarried out.
Case F: In this case, it is likely to be misjudgment to categorize theprocess as quality capable. However, as stationary condi-tions can neither be proven for the distribution nor position,this process must be regarded as uncontrolled, and a state-ment on the quality capability is impossible. As long as the
128
process status F exists, a sorting check is appropriate, orother suitable measures have to be applied, which enable areliable statement.
Case G: This process in not controlled, and a statement on the quali-ty capability is not possible. Therefore, fundamentalchanges and improvements to the process are required. Aslong as the process status G exists, a sorting check has tobe carried out.
As described in the Cases A to G, distribution form of the values of thecharacteristics and time response of the examined process play animportant part in the decision on process capability. Depending on theprocess capability, a statement on the measures connected to theapplication of SPC, is possible. As a basis for this calculation, the standarddeviation of the process is determined to be 6σ.
The following summary (see Fig. 10.8) illustrates, which additionalmeasures beyond the application of SPC, depending on the qualitycapability of the process, are required. Using the example of normaldistribution, the various situations are demonstrated.
129
Tolerance6σ
Fault-freeProduction
SPCApplica-tion
AdditionalMeasures
Tolerance
Stand. Dev.= 6 s
> 1,33 uncompli-cated
yes Continuos qualityimprovement underconsideration ofeconomic factors
> 1,33 problema-tic
yes Optimize processposition (center)
1 to 1,33 problema-tic
yes Reduce processdistribution
< 1 not possible yes 100 % Sorting checkand parallel processimprovement required
Fig. 10.8 Process Capability and additionally required Measures
If the characteristics are not normally distributed, then this has to be takeninto account respectively.
If, by using reliable statistical analyses, it is proven, that a normaldistribution is not applicable and a different distribution pattern betterdescribes the process result, then the quality capability should always beestablished based on the distribution form which best describes the presentprocess status.
130
Even in such a calculation, based on other distribution rules, a distributionrange that is comparable to the 6 σ standard deviation of the normaldistribution, is always compared to the limit values.
10.6 Process Analysis and Quality Capability
When a production process is installed, several quality capability exami-nations are normally required. The procedure described hereafter, servesas an orientation and can be applied for most processes. The selection andnumber of examinations, are to be determined and carried out, appropriateto the process scope.
10.6.1 Process Analysis
The examinations necessary for process analysis can be divided into twoareas:
• Process analysis prior to series start• Process analysis after series start.
In the following description of the principal procedure, the rules of normaldistribution are used as a basis. If it becomes evident, that this assumptionis not valid, then the respective distribution form is to be applied.
Graphical and calculation methods, such as those below, serve the processanalysis:
• Quality Control Chart• Probability Grid• Analysis of Variance.
131
Process analysisprior to series start
Process analysisafter series start
Series start
Process capability: prior to series start after series start
Short termcapabilityanalysis
Primary processcapability analysis
Long term processcapability analysis
ProcessManMachineMaterialMethodMedium
Continuousquality improvement
Time
Short termcapability
Minimum scope 50 parts
Preliminaryprocess capability
Minimum scope 100 parts.To use the requires control chart,at least 20 individual samplesare necessary.
Long termprocess capability
Appropriate long time period undernormal series conditions, by ensuring,that all influencing factors take effect.Standard value: at least 20 productionsdays.
Fig. 10.9 Process Analysis
All types of capability analysis are principally based on the same procedure,and the calculation of the quality capability code number is carried outaccording to identical calculation modalities.
Differences only exist in the number of analyzed parts, the period of analysisand the degree of conformity depending on the extent to which the finalseries conditions were realized at the time of analysis.
132
Depending on whether the quality capability analysis was carried out prior toor after the series start, a division into three types of analysis takes place.
• Process capability analysis prior to series start- Short term capability analysis- Preliminary process capability analysis
• Process capability analysis after series startLong term capability analysis.
The three types of analysis are explained in more detail in the following:
Short term capability analysis
If the required minimum number of parts for a preliminary process capabilityanalysis is not available, a short term capability analysis can be carried out.This type of analysis is mostly applied to inspections of productionequipment and machines at the manufacturer, and is therefore calledmachine capability analysis. In this analysis, mainly those influencesderiving from the production equipment will take effect, while the influencesfrom material, man, method, and mean/environment should be keptconstant. The result of the short term capability analysis is a preliminarystatement regarding the suitability of the production equipment.
Normally, 50 production parts manufactured in sequence are divided into 10samples of 5 parts each and are recorded in the chronological order of theirextraction, in order to identify, for example, any possible trends. As with only50 parts produced in sequence, the actual operational distribution range isoften not fully covered, the result of a short term capability analysis has tobe assessed with respect to the future process situation. Furthermore, the50 measurement values are used during the evaluation of the distribution.
Preliminary capability analysis
To carry out this analysis, a minimum of 20 samples with at least 3 parts areextracted at equal time intervals. However, to increase the meaningfulnessof this analysis, one should try to take 25 samples with 5 parts each. Basedon
133
a quality control chart, it is assessed whether the process is controlled. Afterconfirmation, that the process to be examined is controlled, the processmodel and distribution are identified. In addition to a process-relatedShewhart-Quality Control Chart, probability grids and analysis of variancecan be applied.
To enable a meaningful first estimation of the long term process capabilityto be expected, the final series conditions should already be realized and alldistribution influences effective, as a precondition for this analysis. Theprocedure for assessing the process capability prior to series start should,however, also be incorporated into the assessment of machines andproduction equipment. An appropriate comment next to the quality capabilitycode number established in this analysis, should point out the boundaryconditions (which points presently still deviate from the later final seriesconditions).
Long term capability analysis
This analysis serves to assess the quality capability under real process con-ditions. Such an analysis has to cover a longer time period suited to the pro-cess, so that all factors influencing the distribution, can show their effects.
To secure the results of the trial run, a long term observation of the processis absolutely necessary. The procedure is the same as for the preliminaryprocess capability analysis, and the quality control chart from the trial run iscontinued. An observation period of 20 production days should be appliedas the normal procedure.
134
10.6.2 Quality Capability Code Numbers for Production Equipmentand Processes
To be able to make a statement on the quality capability of productionequipment (machines) and processes in a quantitative manner, qualitycapability code numbers (capability indices) are established. Whendetermining quality capability code numbers, the boundary conditions underwhich the process ran, are to be documented.
By stating quality capability code numbers, information in a simple form isavailable on the quality capability of a process. For continuous characte-ristics, the most important and commonly used quality capability code num-bers are described and presented in a summary in the following.
It is sensible, to coordinate the individual calculation procedures, as well asthe minimum requirements between the customer and supplier.
10.6.2.1 Quality Capability Code Number for Continual Characteristics
For normally distributed characteristics, the quality capability code number isestablished by comparing the tolerance of a characteristic with the standarddeviation of the process described by ±3σ. For particular quality capabilitycode numbers the process status is also taken into account.
The indices Cm and Cp result from the comparison of the tolerance with thestandard deviation of the process described by ±3σ and are therefore ameasure for the potential quality capability. The indices Cmk and Cpk havethe same basis as Cm and Cp , but in addition, the process status is takeninto account (see Fig. 10.10).
135
Short termprocess capability
Preliminaryprocess capability
Long termprocess capability
Quality capability code number for Process potential Process capability
LLV ULV
Standard deviation
Tolerance Tolerance
Standard deviation
LLV ULV
Target parameterMiddle position
Fig. 10.10 Quality Capability Code Number depending on the type ofanalysis
Quality capability code numbers state, to what extent the produced unitscomply with the set requirements of a certain characteristic. These require-ments are determined by the upper limit (or maximum) value ULV and thelower limits (or minimum) values LLV. The tolerance
T = ULV - LLV
is defined by these two limit values.
A comparison of the values of characteristics with the limit values is notusefully carried out on each individual characteristic‘s values separately, butwith the totality of the characteristic‘s values. i.e. with the distribution whichbest describes the overall distribution of the characteristic‘s values.
136
This distribution of the totality of the characteristic‘s values is a goodapproximation to a normal distribution for the process models "A" and "B"described in paragraph 10.4. Even for the process model "C", a goodestimation of the process capability can often be made based on the normaldistribution.
Under condition, that the totality of the characteristic‘s values can, in goodapproximation, be described by a normal distribution with the mean µtot andthe standard deviation stot , the following code numbers are used.
The quality capability code number
T 6 stot
compares the tolerance T with the standard deviation of the values of thecharacteristics described by 6 stot . In order to categorize a process asquality capable, it is necessary, that the standard deviation is always smallerthan the tolerance.
The above mentioned relationship is the basis for determining the potentialprocess capability, however, on its own, does not yet present a sufficientdescription of the process capability.
Another requirement is namely, that the process is well centered, and there-fore the process mean µtot has to be included in the calculation.
As explanation, two situations of a production with T = 8 stot are presented(see Fig. 10.11).
137
Mean µtot centered Mean µtot not centered
ULLLLV
T = tolerance
ULLLLV
T = tolerance
Specified value
tottot = Specified value
tot tot6 = stand. dev. 6 = stand. dev.
oo
µ µ
Fig. 10.11 Calculation basis for the quality capability code number
Through the calculation of the quality capability code number, based on theformulas below, one can describe, how well the process complies to thequality requirements compared to the upper limit value (ULV) or lower limitvalue (LLV):
Cpu = ULV tot
tot
− µσ3
Cpl = µσ
tot
tot
LLV−3
To evaluate a process one uses the more critical, i.e. the smaller of the twovalues as the quality capability code number Cpk.
Cpk = Smallest of the two values Cpl and Cpu
Only for processes that correspond to the model "A", stot = s, i.e. is equal tothe standard deviation of the current process distribution. In all other cases,is stot > s.
138
It has to be ensured, that the process to be assessed is controlled, and thatthe process distribution range, on which the calculation is based, corres-ponds to the proven process model. Only under consideration of bothcriteria, does one receive realistic values for the quality capability codenumber Cp and Cpk.
As the mean µ and the standard deviation s only enter the calculation of thecapability code number as estimate values, it has to be considered that thecapability code numbers are therefore also only estimate values (Cp, Cpk),which are subject to chance variation.
One obtains, for example, the following 99 % confidence levels for Cp:
(The 99% confidence level for Cp is the area, which covers the true value ofCpk with a 99 % probability.)
for 1 sample of the scope n = 50 0,75 ⋅ Cp ≤ Cp ≤ 1,26 ⋅ Cp
for 25 samples of the scope n = 5 0,82 ⋅ Cp ≤ Cp ≤ 1,18 ⋅ Cp
for 50 samples of the scope n = 5 0,87 ⋅ Cp ≤ Cp ≤ 1,13 ⋅ Cp
This means, that quality capability code numbers are subject to inaccuracy.The confidence level for the quality capability code numbers Cpu, Cpl and Cpk
are even slightly larger than that for the quality capability code number Cp.The same applies to other process types and quality capability codenumbers for counting characteristics.
10.6.2.2 Quality Capability Code Numbers for Counting Characte-ristics
The calculation of quality capability code numbers for counting characte-ristics is not recommended. To describe the quality capability of the coun-ting characteristics it is more sensible, to directly state the established rejectrate as a percentage value.
139
10.6.3 Quality Capability of Measuring Equipment
The measuring equipment and devices used in capability analysis have tobe capable of fulfilling the required measuring accuracy resulting from theapplication purpose.
Definitions on the investigation procedure and results (accuracy, linearity,stability, repetitive and comparative precision) are provided in DIN 55350,Part 13.
Methods for determining the measuring uncertainty are described inDIN 1319, Part 3 and Part 4.
10.7 Quality Control Charts (QCC)
10.7.1 Description of QCC
A quality control chart is a form, in which the results of examinations of arunning number of random samples, are graphically illustrated across a timeaxis. To judge if the examined process is controlled, the results entered inthe quality control chart are compared with process-related interventionlimits (see Fig. 10.12).
Upper Intervention Limit UIL
Lower Intervention Limit UIL
Middle line
Me
an
Va
lue
x
(mm
)
27,55
25,50
27,45
Time06 07 08 09 10 11 12 13 14 15 16 17 18
Fig. 10.12 Example of a Quality Control Chart with Intervention Limits
140
Process-related intervention limits are, with quality control charts, the basisfor determining if one has to intervene in the process or not. Theintervention limits entered in the QCC must, however, in no case bemistaken for the limit values of a characteristic.
In contrast to the limit values determined by the engineer, process- relatedintervention limits have the great advantage, that the process compared tothem is proven capable of fulfilling the quality requirements connected withthese intervention limits (see Fig. 10.13).
Upper Intervention Limit UIL
Lower Intervention Limit UIL
Middle line
Me
an
Va
lue
x
(mm
)
27,55
25,50
27,45
Time06 07 08 09 10 11 12 13 14 15 16 17 18
Upper Warning Limit LWL
Lower Warning Limit LWL
Fig. 10.13 Example of a Quality Control Chart with Warning and Interven-tion Limits
10.7.2 Purpose of QCC
The QCC is a tool for analysis, evaluation and control of processes. For thesubjects of process capability and statistical process control (SPC), theQCC is the central element, that allows, with relatively low expenditure, ahigh quality and productivity standard to be reached and ensured.
The QCC fulfills the task, through appropriate signals, of indicating that aprocess can no longer be classed as controlled. This information thenmakes it possible for purposeful measures to improve quality to beimmediately implemented.
141
10.7.3 Types of QCC
Under the leading idea of process control and the continual quality improve-ment, two types of QCC come mainly into use:
• Shewhart - QCC• Supposition - QCC
The intervention limits of the QCC are determined with the help of the para-meters σ and µ, which are estimated through the trial run.
With the Shewhart – QCC the intervention limits are determined based onthe mean and under exclusive consideration of the chance variation, foundin the trial run, established with the standard deviation σ of the momentaryvariation (Process Model "A").
For Shewhart – QCC with extended intervention limits, the same basisapplies initially. The intervention limits of the chart are, however, calculatedon the basis of the total standard deviation σtot and adjusted accordingly(Process Model "B").
The application of Supposition Quality Control Charts must also refer to theprocess, and intervention limits must not orientate themselves exclusively tolimit values. One must always strive to keep the intervention limits of theposition chart as close as possible, and to only allow extensions over theShewhart model, when technological conditions or economic aspects makesuch a procedure necessary.
With Supposition Quality Control Charts, the intervention limits are initiallyestablished based on the standard deviation of the individual randomsamples. In the position chart, it is however allowable, to expand the inter-vention limits by the amount of a justified and for this process approvedadditional mean fluctuation (Trend) (Process Model "C"). However, in orderto do justice to the concept of continual quality improvement, one mustalways aim to keep this extension to a minimum.
142
10.7.4 Application of QCC prior to Series Start
In all phases of the process, even as early as possible before series start,the QCC should be used to record results. Here the QCC serves as asimple basis for the determination of a preliminary quality capability codenumber. Supplementary statistical analyses, required for certain processes,are not made superfluous by this.
10.7.5 Application of QCC in Serial Production
The basic objective of any production must be to work with qualitycompetent processes and to continually improve these. This is possible withthe help of QCC techniques. The production department evaluates andinfluences it’s process based on the entries in the QCC. Recognizeddisturbances – e.g.: exceeding of the intervention limits, - give rise tosuspicion of systematic process changes and are reason to intervene in theprocess.
When using Shewhart – QCC to monitor the process status, one shouldalways aim to cover the specified value (target parameter) with the processmean. Furthermore, it should constantly be attempted to reduce the distri-bution. If this is achieved, the intervention limits must be adjusted accor-dingly.
This procedure is based on the consideration, that within the framework ofgiven production conditions, the „actual value“ should deviate as little aspossible from the „specified value“, in order to keep the total economic lossto a minimum.
Continual process improvement is, however, not to be expected due to theformalism of the quality control chart. Only measures on the process,through information and disturbance notifications from the quality controlchart, can improve quality capability.
143
10.8 Procedure (Example)
The examples are split as follows:
Example1: Determining short term capability.
Example 2: Preliminary process capability analysis with subsequentcontinuation of the process and determination of the longterm process capability.
Example 3: Long term process capability analysis
In example 1, the calculation is based on the chance variation (1-α=99 %).
In examples 2 and 3, the calculations are based on three times the standarddeviation of the corresponding random sample known value.
Both calculation bases can be used for all types of process capability evalu-ation and are interchangeable.
The different calculation bases can lead to slightly differing results. Thesedifferences are, however, without significance for the practical applicationand can therefore be ignored.
The examples are illustrated in detail; in order to be able to follow theseindividually, it is necessary to study the more in-depth expert literature.
10.8.1 Example 1
Deterimining short term capability
This type of analysis is, for example, used for the assessment andinspection of new machines and production equipment, when, as in thepresent case, the necessary number of parts for a preliminary processcapability analysis is too low. In such a case, when one would like to assessonly the production equipment itself and not the entire process, it has to beensured as a basis, that all influences, apart from those caused by themachine to be assesed, are mostly eliminated or kept constant over theentire assessment period.
144
In this example, the procedure for short term capability assessment, toevaluate and inspect a sanding machine used to machine the exterior of acylinder, is described.
Operation: Machining of the exteriorCharacteristic: DiameterSpecified size: 100,00 mmUpper limit value (ULV): 100,020 mmLower limit value (LLV): 99,980 mmNumber and sample size: 50 parts manufactured in sequence and
extracted in samples of 5 pieces
The results of the parts produced in sequence are recorded and grouped infives.
1St Sample 2nd Sample 3rd Sample 4th Sample 5th Sample
99,996 100,000 100,002 100,003 99,999
99,997 100,001 100,002 100,000 100,001
99,998 100,002 100,003 100,003 100,000
100,000 99,995 100,002 100,002 100,002
100,002 99,999 99,997 99,992 99,999
6th Sample 7th Sample 8th Sample 9th Sample 10th Sample
100,002 100,000 100,003 99,997 99,996
99,994 100,001 99,997 100,004 100,006
99,998 100,002 100,000 100,000 99,999
99,998 100,000 99,998 99,999 99,997
100,002 100,000 99,998 100,004 100,000
145
Based on the available readings x and s for each of the 5er samples aredetermined and subsequently entered in the quality control chart.
Equation: x = sum of the sample valuesnumber of sample values
= 1
1nx j
j
n
=∑
s =1
12
1nx xj
j
n
−−∑
=( )
Numeric results for the mean and standard deviations
Sample No. x s
1 99,9986 0,00241
2 99,9994 0,00270
3 100,0012 0,00239
4 100,0000 0,00464
5 100,0002 0,00130
6 99,9988 0,00335
7 100,0006 0,00089
8 99,9992 0,00239
9 100,0008 0,00311
10 99,9996 0,00391
Subsequently the calculation of the intervention limits is made, which arethen also entered in the quality control chart, to judge if the process is con-trolled.
In this example, a quality control chart is used, in which the interventionlimits were calculated based on the chance variation (1-α=99 %).
146
In all cases, the calculation of the intervention limits must be based on thereadings taken from the process to be assessed. Therefore, initially the total
mean x and the estimated value $σ for the standard deviation must bedetermined.
Total mean $µ = χ = 1
1kx i
i
k
=∑
Estimated value for the standard deviation $σ = s 2
= s
ki k
k2
=∑
In this example the following numeric values result:$µ = 99,99984 mm$σ = 0,002910 mm
The calculation of the intervention limits UIL = Upper Intervention LimitLIL = Lower Intervention Limit
is carried out to the following equationsfor the mean value chart: OEG = $µ + AE⋅ $σ
UEG = $µ - AE⋅ $σ
for the standard deviation chart OEG = BOEG⋅ $σUEG = BUEG⋅ $σ
AE, BUIL and BLIL are constants dependent on the sample size, which in thisexample of a 5er sample, take the following values:
AE, = 1,152 BUIL = 1,927 and BLIL = 0,227
Results of the calculation:
Mean value chart Standard deviation chart
Upper Intervention Limit = 100,0032 mm Upper Intervention Limit = 0,00561mmLower Intervention Limit = 99,9965 mm Lower Intervention Limit= 9,00066 mm
147
Quality Control Chart
Sample No.
Sample No.
Based on the now available quality control chart it can be assessed, if theprocess can be classed as controlled.
The process examined here is controlled.
It is assessed, by illustrating in a probabilty grid or with another correspon-ding numeric test for normal distribution (e.g. Chi², Shapiro- Wilk), if thevalues for the characteristics corrspond in approximation to a normaldistribution.
148
Probability Gird
The entries in a probability grid allow a further calculation to determine thequality capabiltiy code number according to the normal distribution.
The test according to Shapiro-Wilk and Chi²-Test confirm that the valuesfound, for this example, are approximately normally distributed.
These tests for normal distribution are not dealt with here and should becarried out with corresponding analysis software to save time and effort.
A good summary the process results is also given through the illustration asa histogram and the presentation of the individual values.
149
Histogram Illustration of the individuall values
The calculation of the quality capability code number can now be carried outbased on the normal distribution.
Equations: $C m =T
6 $σ
$C mo =OGW − $
$
µσ3
, $C mu = $
$µ
σ− U GW
3
$C mk = smallest of the two values $C mo und $C mu
Calculation: $C m =0,040
6 ⋅ 0,002910 = 0,040
0,01746 = 2,29
$C mo =100,020 - 99,99984
3 ⋅ 0,002910 = 0,020160,00873 = 2,31
$C mu =99,99984 - 99,980
3 ⋅ 0,002910 = 0,019840,00873 = 2,27
Result: $C m = 2,29
$C mk = 2,27