Design for Data Analysis for Engineers 1 MPD 575 DFX Cohort 8 November 29, 2007 Developed by: Tjuana Buford Professor: Jonathan Weaver.

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

11

MPD 575 DFXCohort 8

November 29, 2007Developed by: Tjuana

BufordProfessor: Jonathan Weaver

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

22

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Examples Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

33

Introduction to DFDA (Design for Data Analysis)

Definitions for Data Analysis:

Data analysis is the process of looking at and summarizing data with the intent to extract useful information and develop conclusions.1

Confirmatory data analysis is based on confirming or falsifying existing hypotheses. 1

Exploratory data analysis is based on discovering new features in the data1

Data analysis – a procedure that prepares a data model for implementation as a non-redundant, flexible, and adaptable database2

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

44

Introduction to DFDA Cont.

Design for DFDA relates to the following Design for X Modules:

Design for Testability

Design for Serviceability

Design for Robustness

Design for Reliability

Design for Reuse

Design for Failure

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

55

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Example Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

66

Define DFDA for Engineers

Testing falls within Systems Engineering Fundamentals, which creates a need for analysis of data. Falling within the generic, right hand side of the SEF “V”, verification using data analysis may include:

Analysis of large or small amounts of dataVarying data formats from multiple sourcesData from multiple engineers within a departmentData from several engineering departmentsReport generationResult retention capability

Engineers are required to meet requirements through testing,

but who designs the test and analysis?

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

77

Define DFDA for Engineers Cont.

Tests and analysis are designed by Test engineers, R&D engineers, Technical Specialists, Software Engineers, and many more!

"Just as a requirement specifies the functional performance to be delivered (not how it is to be designed), a Design Verification Method defines what the test or analysis must deliver, not how the test or analysis is to be designed. “3 (emphasis added)

How good is your Analysis?

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

88

Define DFDA for Engineers Cont.Test Engineers who create analysis procedures need to have both engineering and computer science knowledge to be able to move beyond Microsoft Excel as a method of data analysis. Past experience with strictly IT personnel creating or choosing data analysis tools has resulted in tools which do not stand up to engineering requirements, for example:

Allowing final results to be alteredMisinterpretation of engineering theoryVarious forms of analysis documentation throughout companyNo version controlNo means of sharing (process, analysis, results) easily with other engineersCost to local engineering departments for design, and maintenance of tools

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

99

Define DFDA for Engineers Cont.According to Broy (2006)4, The typical electrical engineer lacks the following:

Do not know enough about softwareDo not understand the software processesDo not understand project management

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1010

Define DFDA for Engineers Cont.

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1111

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Examples Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1212

DFDA and SEF (Systems Engineering Fundamentals)

FabricationFabrication//

VerificatioVerificationn

FabricationFabrication//

VerificatioVerificationn

SubsysteSubsystem m

VerificaVerificationtion

SubsysteSubsystem m

VerificaVerificationtion

SystemSystemVerificatVerificat

ionion

SystemSystemVerificatVerificat

ionion

CustomerCustomerSatisfactioSatisfactio

nn

IntegratIntegratee

IntegratIntegratee

ComponentComponentDesignDesign

ComponentComponentDesignDesign

Ver.Ver.

Req.Req.

VerificationVerificationRequirementsRequirements

Customer Customer Focused Focused FeedbackFeedback

Verification using data analysis can occur at any level

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1313

DFDA and SEF (Systems Engineering Fundamentals)

FabricationFabrication//

VerificatioVerificationn

FabricationFabrication//

VerificatioVerificationn

SubsysteSubsystem m

VerificaVerificationtion

SubsysteSubsystem m

VerificaVerificationtion

SystemSystemVerificatVerificat

ionion

SystemSystemVerificatVerificat

ionion

CustomerCustomerSatisfactioSatisfactio

nn

IntegratIntegratee

IntegratIntegratee

ComponentComponentDesignDesign

ComponentComponentDesignDesign

Ver.Ver.

Req.Req.

VerificationVerificationRequirementsRequirements

Customer Customer Focused Focused FeedbackFeedback

Analysis and test results are recorded in the DVP&R as the Verification plan is

executed.(actual results vs.

targets)

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1414

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Examples Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1515

Example 1: My manager asked me to develop a regression model for some data from a new process, what are the basic steps I should follow?5

Input Data

Output Data

Process The Data

Process The Data

“On a fundamental level, all computer programs do the same thing”6

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1616

Example 1 Cont.The Basic StepsIn the form of aFlowchart.

StartStart

Select form of model based on current data or results from prior model.

Select form of model based on current data or results from prior model.

Design New

Experiment

Design New

Experiment

New Data needed to fit model?

New Data needed to fit model?

CollectNewData

CollectNewData

Fit model using parameter

estimation method suggested by data and /or process

knowledge

Fit model using parameter

estimation method suggested by data and /or process

knowledge

Validate model to assess its adequacy

Validate model to assess its adequacy

New model

describes data well?

New model

describes data well?

EndEnd

YesYes

NoNo

YesYes

NoNo

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1717

Example 2I use Excel to analyze data collected from highway driving studies. It takes me 20 hours to analyze approximately 22,000 data points, and create a report. What can I use to speed up this process?

*Fictitious Data

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1818

Example 2 Cont.There are many software packages on the market to analyze data including:

Excel Macros/Data Analysis Toolkit

MATLAB/Simulink

Minitab

Labview

NumPy and SciPy for Python (Freeware)

Software choices have to take into account cost, functionality, ability to generate reports, customer service and support.

Use trial software offers to help test software packages.

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

1919

Example 2 Cont.Most Engineers that have a multitude of data to analyze will eventually move away from the most basic products to products that make their time more productive.

Some issues with software packages:Some issues with software packages:

Excel – Okay for small projects but for large projects, major version changes from Microsoft can kill macros and Excel functions you depend on.

MATLAB/Simulink – Large online community of users, support available online, by phone, and in person (for a fee). Ability to handles large amounts of Data. Each additional toolbox costs more money. Has report generator capability.

Labview – Widely used in the automotive industry.

Python – Customizable to individual needs. Free to use. Growing in popularity.

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2020

Example 2 Cont.General Comparison of Numerical Analysis Software7

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2121

Example 2 Cont.Operating System Compatibility7

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2222

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Examples Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2323

Summary

Use data collection software to collect data as close as possible to the software format that will be used to analyze the data

Make sure the software package chosen has the functions you need

Software should be able to process data from different sources

The goal is consistent results

Reach out to other engineers and developers

Participate in the data analysis community

Ford Employees - Contact Tjuana Buford (Core Developer) – [email protected] for access to the Data Processing tool presented in example 3.

The Death tool is used by 400+ engineers within Ford PD

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2424

Introduction to DFDA (Design for Data Analysis)

Define DFDA for Engineers DFDA and SEF (Systems Engineering Fundamentals)

Examples Summary References

Design for Data Design for Data Analysis for Analysis for EngineersEngineers

2525

References1. http://en.wikipedia.org/wiki/Data_analysis

2. http://asgard.kent.edu/systems1/dataanal/tsld005.htm

3. Ford Motor Company – Systems Engineering Fundamentals Reference Guide (2005) Ford Confidential

4. Broy, M., Pretschner, A., Salzmann, C., and Stauner, T., "Software-Intensive Systems in the Automotive Domain: Challenges for Research and Education," 2004-01-1780, SAE World Congress, Detroit, Michigan 2006

5. NIST/SEMATECH e-Handbook of Statistical Methods, http://www.itl.nist.gov/div898/handbook/, 2007.

6. Bronson, Gary J., “C for engineers and scientists an introduction to programming”, West Publishing company, St. Paul, MN 1993

7. http://en.wikipedia.org/wiki/Comparison_of_numerical_analysis_software

8. DEATH Design developed by Erwin Peters 2002, Ford Motor Company Confidential. Copyright 2002-2007 All Rights Reserved