The Oregon Wood Innovation Center at Oregon State University

QC Tools

Pareto Analysis

Flowcharts

Cause and effects diagrams

SPC

2

The Role of Quality (pre- WWII)

3

Make ProductRaw Materials

Production

Inspect Product

QAShip product

rework

RejectDid we forget

someone here?

Oops...

4

Modern Quality Management

Customer

ExpectationsTranslation

Quality

Control

Voice of the customer Voice of the processProcess specifications

5

Introduction to SPC

7

8

3.04

3.041

3.042

3.043

3.044

3.045

3.046

3.047

3.048

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

X-b

ar

UCL LCL Average X bar-bar

What is SPC? An Example

Day Shift

9

3.04

3.041

3.042

3.043

3.044

3.045

3.046

3.047

3.048

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

X-b

ar


Swing Shift

10

3.043.0413.0423.0433.0443.0453.0463.0473.048

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

X-b

ar


11

3.04

3.041

3.042

3.043

3.044

3.045

3.046

3.047

3.048

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

X-b

ar


What the Customer Sees

Reducing reject rate from

30% to 20% = $120,000 per year

Origins of SPC

1920-30’s: SPC invented as economical means to control quality in manufacturing

1940-50’s: During WWII, Bell Labs personnel train armament manufacturers to use SPC U.S. economy intact, SPC “disappears”

Japanese devastated; W. Edwards Deming arrives

1960-70’s: Quality techniques blossom in Japan

1980’s: world begins to recognize “Made in Japan” as a sign of quality. In U.S., automotive and electronics industry lead “quality

revolution”

12

Origins of SPC

1990’s: SPC recognized by most U.S. industries as a vital QC tool In wood products industry, principal use is size control in

sawmills

2000’s: Real-time SPC “Black box” SPC?

13

Philosophy of SPC

Variation is the enemy and it is inevitable

inherent, common, random

special, assignable

Real improvement comes from defect

prevention, not defect detection - “You can’t

inspect quality into a product”

Prevent defects by monitoring, controlling &

reducing variation

Continuous process improvement to reduce

variation

14

If SPC is the answer,

what was the question?

What is the distribution of process output?

centering, range or “spread”, likelihood of an

extreme value

Is the process capable of meeting customer

expectations?

What is causing the variability?

When is it reasonable to “get tough” with

employees?

15

If SPC is the answer,

what was the question?

Can we afford to minimize the variability?

Over time, how can we be sure the process

hasn’t changed?

When should we “tinker” with the process and

when should we leave it alone?

16

SPC is only a tool

It complements, not replaces, existing knowledge

Use of SPC leads to

increased profits

increased productivity

increased employee involvement in the process

increased morale

New tools require change. Change fear, resistance,...

17

How & Why SPC Works

SPC in a “nutshell”

18

SPC in a “nutshell”:

Describe the distribution of process output

Estimate the limits within which the process

operates under “normal” conditions

Determine if the process is “stable”

Determine if the process is “capable”

19

Neither accurate,

nor precise

Precise, but

not accurate

Precise and

accurate

Distribution of process output

20

Moisture Content (%)

21

Creating Histograms

Using Excel

22

HistogramsHands-On

Open histogram1.xls

Enter in column E 2.5

2.55

2.6

2.65

Etc. to 3.05

Click on ‘Data’, ‘Data Analysis’, ‘Histogram’

Input range = C4:C128

Bin Range = E4:E15

Click ‘Chart Output’

23

HistogramsHands-On

Open histogram2.xls

24



Estimate the limits within which the process

operates under “normal” conditions



25

Estimating Process Parameters:

Central tendency - arithmetic mean (average)

Dispersion - range or standard deviation

26

Estimating Process

Parameters:

Central tendency

sample average

(an estimate of the population mean, µ)

Dispersion or “spread”

sample range

sample standard deviation

(both are used to estimate the population standard deviation, )

27

Important Concepts:

Population – “the whole enchilada”

Parameter – a specific characteristic of the

population

Sample – a subset of the population, used to

estimate certain population parameters

Statistic* - a value derived from the sample,

an estimate of a parameter

* The word “statistics” has 2 meanings

1) the science of variation

2) estimates of population parameters

28


Process centering

n

i

i

n

xX

1

An estimate of the population mean, µ

mean or average,

29

Why isn’t the average good

enough?

Target 3.4”3.2

3.7

4.2

2.8

3.1

X = 3.4

We need a way to quantify the “spread”

in the data

X = 3.40

3.43

3.35

3.43

3.41

3.38

30


Process “spread”:

standard deviation,

range,

1

1

2

n

Xx

s

n

i

i

Both can be used to estimate the population

standard deviation, σ

XXRminmax

31

Standard Deviation

3.2

3.7

4.2

2.8

3.1

How can we express the spread of

these numbers with a single value?

Asked another way - How “far away”

are these numbers from their average?

2.8 3.1 3.2 3.7 4.2

3.4

32

Calculating Standard Deviation

- 3.4 = (-0.2)2 = 0.04

- 3.4 = (+0.3)2 = 0.09

- 3.4 = (+0.8)2 = 0.64

- 3.4 = (-0.6)2 = 0.36

- 3.4 = (-0.3)2 = 0.09

X = 3.4 sum = 1.22

1.22

5 - 1= 0.305

55.0305.0 s

1

1

2

n

Xx

s

n

i

i

3.2

3.7

4.2

2.8

3.1

33

Let’s try it using Excel

Start Excel

Enter the following data:

Calculate the mean and standard deviation

for each set of data

34

3.2

3.7

4.2

2.8

3.1

3.43

3.35

3.43

3.41

3.38

Do 2 numbers really tell us much?

Wouldn’t it be nice, if knowing only the

average and standard deviation, we could

determine the limits within which the process

should operate?

For example, given an average of 2.76%, and a

standard deviation of 0.102%, we might say that

over 99% of the product will have a MC between

2.45 and 3.07%.

35

Determining the “limits”:The Normal Distribution

Fre

quency

µ µ + 3µ - 3

2

2

2

)(

2

1)(

x

exF

36

ExampleF

requency

69.7 77.8

Adult males -

Average height (µ) = 69.7 inches

Standard deviation () = 2.7 inches

61.6

Source: University of Guelph, Ontario

37

72.4 75.164.3 67.0

So what?

38

By knowing the average and standard

deviation, and that the numbers are

approximately normally distributed, we know

everything we need to know about the

distribution of process output - it is

predictable...

IF- we can somehow be sure that

the process is stable

39

ExampleF

requency

2.76 3.07

Moisture Content -

Average = 2.76%

Standard deviation = 0.102

2.45

40



Estimate the limits within which the process operates under “normal” conditions



41

Is the process “stable?”

or

in SPC lingo, is the process

in statistical control?

42

Causes of Variation Random, chance, constant, common,

unknown causes

the “rhythm” of the process

Assignable, special causes

something has changed

43

Examples of things that may be assignable

causes of variation:

machine troubles (damaged saw teeth, plugged

blowpipe, etc.)

faulty measuring device

operator overcontrol

worker fatigue

drastic changes in raw material

undocumented change in procedures

Assignable Causes of Variation:

44

“Statistical Control”

“A process is described as in control

when a stable system of chance

causes seems to be operating.”

(Grant & Leavenworth, Statistical Quality Control)

45

The affect of “control” on the process

In control - The process is

reasonably predictable.

It is likely to be unprofitable to

search for assignable causes of

variation.

Statistically rare occurrences

are indications of lack of control

Out of control – Assignable

causes of variation must be

present.

A little investigation is

suggested to find and eliminate

those causes

46

How about a little poker?

47

Either this guy’s cheatin’ or I’ve just witnessed a

statistical anomaly!

48

A statistically rare event -

An outcome with a very small probability of

occurrence (“the odds are against it”)

In manufacturing, an outcome that is

“statistically rare” usually signals the

presence of assignable causes of variation

How can you know what is statistically rare

and what is commonplace?

SPC

49

What is a “statistically rare

occurrence?”

Obtaining a sample value that is “very far away”

from the average is a highly unlikely event

In SPC, we usually define “very far away” as 3

µ µ + 3µ - 3

99.73%

50

Determining the state of control:The Control Chart Graphical view of the process over time

- central tendency - trends

- spread - changes

Average = X

UCL

LCL

Time51

Time

LCL

UCL

How the control chart works

X

52

Control Charts:

Two different broad categories:

variables measurement data

attributes fraction defective (go/no-go data)

defect counts

53

Determining the state of control:Interpreting the Control Chart

X

UCL

LCL

Time

X

UCL

LCL

Time

54



Estimate the limits within which the process operates under “normal” conditions



55

Is the process capable of

meeting specifications?

Process Capability Analysis:

Process capability - Measuring process potential by comparing the specification width to the variation of the process

or

How “wide” are the specs relative to the natural spread of the process?

56

It’s up, it’s… good???

23 ft. 4 inches

57

What are the “specs?”

23.3011.65

58

A visual look at process variation

µLSL USL(Target) 59

Case Example

XYZ Forest Products Inc.

60

Case Example intro

XYZ Forest Products Inc.

Producers of wood handles for push

brooms

Multiple species, sizes, and styles

Orders started falling off;

complaints/rejects increased

Complaints about a variety of quality issues

Some customers stated competition had

better quality

So where to begin???

61

Case Example intro

Quality improvement team visited customers,

examined ‘scrap & rework’ bins

What to do next?

62

Pareto Analysis &

Checksheets

Where should we focus?

What’s the most important/costly quality problem?

63

Checksheets

Record historical data or compile info as it

occurs

Observe patterns and trends

Stratification? – collect data separately for

shift, operator, etc.

Can drive the need for precise & consistent

definitions of nonconformities (or events)

64

NonconformitiesEstablishing precise & consistent definitions

A simple experiment:

Have several different people examine product

and tally # of nonconformities by category

Compare differences between inspectors

Develop a standard set (actual product

and/or photos) for each nonconformity

65

Checksheets

66

Pareto Charts

Based on Pareto principle

a.k.a., the “80:20 rule”

“the vital few vs. the trivial many”

It’s all about prioritizing!

Bar chart – similar to a histogram

Graphical method to view checksheet information

Data plotted in descending order

Include cost data

67

Pareto Chart - example

Category Frequency Cost/item

Size out of spec. 194 $0.12

Loose knots 18 $0.27

Raised grain 4 $0.07

Dents 3 $0.14

Stain/rot 31 $0.27

Fuzzy grain 105 $0.08

Splits 11 $0.14

Machine tear-out 61 $0.18

Burn marks 44 $0.08

Oil/grease marks 2 $0.06

68

Flowcharts

Where in the process might the

problems be occurring?

69

Flowcharts

Where (specific processes) might the

problems be occurring?

What quality characteristic should we

measure?

How does the process actually work (vs.

ideal)?

70

Flowcharts

1. Determine start & stop points

2. List major steps in process

3. Put steps in order

4. Draw flowchart

5. Test chart for accuracy & completeness

6. Look for opportunities to improve process

(remove non value-added activities)

71

Flowchart

Symbols

72

73

74

FlowchartingHands-On

Break into teams of 4 and draw a chart for:

Stacking lumber

Grading lumber

Kiln operation

Customer complaint resolution process

75

Cause & Effect Diagrams

What are some potential causes of the

problem?

76

Cause & Effect Diagrams(a.k.a. Ishikawa or ‘fishbone’ diagrams)

Purpose – organize & display interrelationships

of various theories of root cause of a problem

First step in root cause analysis

Enables clarification of thinking about potential

causes

Helps to avoid ‘pet theory’; can aid consensus-

building

Often done via brainstorming and/or w/

checksheets & flowcharts

77


Process is often more important than product

Types

Dispersion analysis: cause categories

Process classification: process steps as

categories

Customer needs – ‘head’ is need; ‘bones’ are

specifications, components, features, etc.

78


Effect

EnvironmentOperator

Machine Methods Materials

79

Cause & Effect DiagramsHands-On

Groups of 4 – the problem (effect) is:

Size out of specification

Machine tear-out on handles

Inconsistent color on units

Internal burning/checking

Surface checking

?

80

C&E Diagrams

81

Acknowledgements

SUSTAINABLE

The Oregon Wood Innovation Center at Oregon State University

Documents