Topic 9:

Quality and the Toyota System

1. Quality Costs
2. Statistical Process Control
3. Six Sigma
4. Just in Time Production
 Philip Crosby
• Former VP of quality control at ITT corp.
• Wrote “Quality is Free: The Art of Making Quality Certain”
• Proposed: “Zero Defects” as the goal for quality
– “Consider the AQL you would establish on the product you buy.
Would you accept an automobile that you knew in advance was 15%
defective? %5? 1%? 1/2%? How about nurses that care for newborn
babies? Would an AQL of 3% on mishandling be too rigid?”
– “Mistakes are caused by lack of knowledge and lack of attention”
 Crosby’s Quality Postures
• Uncertainty
– We don’t know why we have problems
20 with quality
18 Awakening
Cost of Quality as a % of sales

•
– It is absolutely necessary to always have
16 problems with quality
• Enlightenment
14
U ncert – Through management commitment and
12 quality improvement we are identifying
A w ake and resolving our problems
10
E nlight • Wisdom
8 W isdom
– Defect prevention is a routine part of our
operation
6 C ertain • Certainty
4 – We know why we don’t have problems
with quality
2
0
R eported A ctual
 Categories of Quality Costs
• Cost of yield loss
• Prevention costs
• cost to send your employees to
– Costs associated with preventing quality training
defects • warranty costs associated with
• Appraisal costs unplanned product repair
• cost of a new automated quality
– Costs associated with assessing testing device
quality within a productive system
• cost of rework
• Internal failure costs • loss of market share due to a
– Costs associated with losses from national product purity scandal
disposal of or fixing quality • litigation cost due to product
problems defect

• External failure costs

– Costs associated with releasing
poor quality into the demand
stream
 Rework / Elimination of Flow Units

Step 1 Test 1 Step 2 Test 2 Step 3 Test 3

Rework:
Defects can be corrected
Rework by same or other resource
Leads to variability

Step 1 Test 1 Step 2 Test 2 Step 3 Test 3

Loss of Flow units:

Defects can NOT be corrected
Leads to variability
To get X units, we have to
Step 1 Test 1 Step 2 Test 2 Step 3 Test 3
start X/y units
 Calculation of Yield Loss
• B(1-d1)(1-d2)(1-d3)…(1-dn) = m
• Thus: B=m/(1-d1)(1-d2)(1-d3)…(1-dn)
• Where:
– di = proportion of defectives generated by operation i
– n = number of operations
– m = number of finished products
– B = raw material started in process

Example:
1000 finished product needed from a flow cell
4 operations generating 2%,3%,5%,3% proportion
defective respectively.
How many units must be started in the process?
 Quality Costs

1142 1119 1086 1031

2% 3% 5% 3% 1000

23 33 55 31
 The Concept of Consistency:
Who is the Better Target Shooter?

Not just the mean is important, but also the variance

Need to look at the distribution function

 Two Types of Causes for Variation

Common Cause Variation (low level)

Common Cause Variation (high level)

Assignable Cause Variation

• Need to measure and reduce common cause variation
• Identify assignable cause variation as soon as possible
 W. Edwards Deming

• Quality is first a management

responsibility
• There are two keys to ongoing quality
improvement
– Employee training
– Reacting to process data in real time
• Variation is the disease and SPC/SQC
tools are the cure
 SPC Objectives
• Insure high quality production by reducing
and controlling process variation.
• Identify types of process variation.
– Common cause variation: small, random
forces that continually act on a process
– Special cause: variation that may be
assigned to abnormal, unpredictable forces
• Take action whenever a process is judged to
have been influenced by special causes.
 A General SPC Procedure
• Periodically select from the process a sample
of items, inspect them, and note the result.
• Because of common or special causes, the
results of every sample will vary. Determine
whether the cause of the variation is common
or special.
• Take action depending on what was
determined in step 2.

This procedure is enacted through the use of control charts

 Statistical Process Control: Control Charts
Process
Parameter
• Track process parameter over time
Upper Control Limit (UCL) - mean
- percentage defects

Center Line • Distinguish between

- common cause variation
(within control limits)
Lower Control Limit (LCL) - assignable cause variation
(outside control limits)

Time • Measure process performance:

how much common cause variation
is in the process while the process
is “in control”?
 Charting Continuous Variables
• The Xbar-R Chart: tracks the mean and
range of a variable calculated from a fixed
sample
• The Xbar-S Chart: tracks the mean and
standard deviation of a variable calculated
from a large sample
 The Xbar-R Chart
• Collect sample data by sub-group (normally containing 2 - 5 data points): record
the continuous variable under study.
• Compute the mean and range for each sub-group:

x1 + x2 + ...+ xn
x= R = xl argest − xsm allest
n
• Calculate average mean and average range
• Compute and draw control limits:

UCLR = D4 R
UCL x / LCL x = x ± A2 R
• Plot mean and range for each subgroup. LCLR = D3 R
 Parameters for Creating X-bar Charts

Number of Factor for X- Factor for Factor for Factor to

Observations bar Chart Lower Upper estimate
in Subgroup (A2) control Limit control limit Standard
(n) in R chart in R chart deviation, (d2)
(D3) (D4)
2 1.88 0 3.27 1.128
3 1.02 0 2.57 1.693
4 0.73 0 2.28 2.059
5 0.58 0 2.11 2.326
6 0.48 0 2.00 2.534
7 0.42 0.08 1.92 2.704
8 0.37 0.14 1.86 2.847
9 0.34 0.18 1.82 2.970
10 0.31 0.22 1.78 3.078
 Example of an Xbar-R Chart
Sub- Obs Obs Obs Obs Obs Mean Range
group 1 2 3 4 5
1 14.0 12.6 13.2 13.1 12.1 13.00 1.9
2 13.2 13.3 12.7 13.4 12.1 12.94 1.3 Each data point
3 13.5 12.8 13.0 12.8 12.4 12.90 1.1 is the pulling
4 13.9 12.4 13.3 13.1 13.2 13.18 1.5 force applied to
5 13.0 13.0 12.1 12.2 13.3 12.72 1.2
6 13.7 12.0 12.5 12.4 12.4 12.60 1.7
a glass strand
7 13.9 12.1 12.7 13.4 13.0 13.02 1.8 before breaking
8 13.4 13.6 13.0 12.4 13.5 13.18 1.2
9 14.4 12.4 12.2 12.4 12.5 12.78 2.2 For 5 obs.
10 13.3 12.4 12.6 12.9 12.8 12.80 0.9
. D3=0
25 13.3 12.8 13.0 12.3 12.2 12.72 1.1
D4=2.114
Total 323.50 33.80 A2=0.577
Mean 12.94 1.35
 Example (cont)
14
For this example,

Mean
13
the control
limits reduce to: 12

UCLx / LCLx = 12.94 ± (.577)1.35 Sub-group

= 13.72 & 12.16
3
UCLR = 2.114(1.35) = 2.86
LCLR = 0(1.35) = 0 Range 2

Sub-group
 The Xbar-s Chart
• Similar to Xbar-r chart except that a larger sample is taken.

• The calculation of control limits may include a sample

standard deviation as an estimate of the population standard
deviation.

• Control limits are calculated :

UCLs = B4 s
UCL x / LCL x = x ± A3 s
LCLs = B3 s
 The Statistical Meaning of Six Sigma
Process capability measure
Lower Upper
Specification Specification
Limit (LSL) Limit (USL) USL − LSL
Cp =
6σˆ
Process A
(with st. dev σ A) xσ Cp P{defect} ppm
1σ 0.33 0.317 317,000
X-3σ X-2σ X-1σ X X+1σ X+2σ X+3σ
2σ 0.67 0.0455 45,500
A A A A A

3σ

3σ 1.00 0.0027 2,700

Process B 4σ 1.33 0.0001 63
(with st. dev σ B)

5σ 1.67 0.0000006 0,6

X
6σ 2.00 2x10-9 0,00
X-6σ B
X+6σ B

• Estimate standard deviation: σˆ =R / d2

• Look at standard deviation relative to specification limits
• Don’t confuse control limits with specification limits: a process can be out of
control, yet be incapable
Control Limits and Specification Limits
• Control limits of a quality characteristic represent
natural variation in a process
• Specification limits indicate acceptable variation set
by the customer
• The process capability index is useful in comparison:

USL − LSL
Cp =
6σ
• The capability index may be adjusted to to consider
how well the process is “centered” within the limits

C p k = C p (1 − k )
K=2 |design target - process average | / specification range
 Process Capability Example

USL=10
10 − 9.5
LSL=9.5 Cp = = 4.167
6(.02)
 = .02
9.5 10.0

K=2 |9.75 - 9.95| / .5 = .8

C pk = 4.167 (1 − .8) = .8334

 PC Example (cont)

USL=10
10 − 9.5
LSL=9.5 Cp = = 4.167
6(.02)
 = .02
9.5 10.0

K=2 |9.75 - 9.79| / .5 = .16

C pk = 4.167 (1 − .16) = 3.917

 Charting Discrete Attributes
• Charts that track the number of units
defective
– P Chart: fraction of a sample that is defective
given different sample sizes
– NP Chart: fraction of a sample that is
defective given constant sample sizes
Attribute Based Control Charts: The p-chart
Period n defects p
1 300 18 0.060
2
3
300
300
15
18
0.050
0.060
• Estimate average defect percentage
4 300 6 0.020
5
6
300
300
20
16
0.067
0.053
p =0.052
7 300 16 0.053
8 300 19 0.063 • Estimate Standard Deviation
9 300 20 0.067
10
11
300
300
16
10
0.053
0.033 p (1 − p )
12 300 14 0.047 σˆ = Sample Size
=0.013
13 300 21 0.070
14 300 13 0.043
15 300 13 0.043
16 300 13 0.043
17 300 17 0.057 • Define control limits
18 300 17 0.057

UCL= p + 3σˆ =0.091

19 300 21 0.070
20 300 18 0.060
21 300 16 0.053
22
23
300
300
14
33
0.047
0.110
LCL= p- 3σˆ =0.014
24 300 46 0.153
25 300 10 0.033
26
27
300
300
12
13
0.040
0.043
• Divide time into:
28 300 18 0.060 - calibration period (capability analysis)
29 300 19 0.063
30 300 14 0.047 - conformance analysis
Attribute Based Control Charts: The p-chart

0.180
0.160
0.140
0.120
0.100
0.080
0.060
0.040
0.020
0.000
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
 Example of a P Chart
Sub-group Sub-group Number of Percent UCL LCL
Number Size (n) Defectives Defective
(np) (np/n)100
1 115 15 13.0 18.8 1.8
2 220 18 8.2 16.5 4.1
3 210 23 11.0 16.6 4.0
4 65 5 7.7 21.6 0.0
5 220 18 8.2 16.5 4.1
6 255 15 5.9 16.0 4.6
.
Quantities of light
Total 5925 610 10.3 bulbs are tested to
see if they function

Note: control limits calculated assuming z=3

 Example of a P Chart (cont)
For this example, the control .0924 3
.103 ± 3 = .103 ± (.304)
limits reduce to: n n
Percent Defective

25
20
15 UCL
10 p
5 LCL

Sub-group
 The NP Chart
• Similar to the P Chart except assumes
constant sample size
• Calculation of the control limits must be
performed only once

Center line = pn

UCL / LCL = p n ± z p n(1 − p )

 Discrete Attributes (cont)
• Charts that track the number of defects in
one or more units
– U Chart: defects in a variable sized sample
volume
– C Chart: defects in a fixed sized sample
 The U Chart
• Collect sample data: for each sample record the
number of units sampled (n) and the number of
defects (c)
• Compute the number of defects per unit for each
sample sub-group: (u = c/n)
• Calculate the mean defects per unit: Σc
u=
Σn
• Compute and draw control limits
• Plot u
u
UCL / LCL = u ± z
n
 The C Chart
• Similar to the U Chart except assumes
constant sample size
• Calculation of the control limits must be
performed only once

Center line = C
UCL / LCL = C ± z C
 Example of a C Chart
Sub-group Number of Sub-group Number of
Number Defects Number Defects
1 7 11 6
2 5 12 3
3 3 13 2 In this example,
4 4 14 7 a data point
5 3 15 2
6 8 16 4 represents the
7 2 17 7 number of rips
8 3 18 4 found in 5 yards
9 4 19 2
10 3 20 3 of nylon fabric

Total 82

Note: control limits calculated assuming z=3

 Example of a C Chart
For this example, the control C = 4.1
limits reduce to: UCL / LCL = 4.1 ± 3 4.1

10 UCL
Defectives

5 C

Sub-group
We assume the process is in an
“in control” state when:
• Points are within the control limits
• Consecutive groups of points do not take a particular form.
– Runs on one side of the central line (7 out of 7, 10 out of 11,
or 12 out of 14)
– Trends of a continued rise or fall of points (7 out of 7)
– Periodicity or same pattern repeated over equal interval
– Hugging the central line (most points within the center half of
the control zone)
– Hugging the control limits (2 out of 3, 3 out of 7, or 4 out of
10 points within the outer 1/3 zone)
 Statistical Process Control
Capability Conformance
Analysis Analysis

Eliminate Investigate for

Assignable Cause Assignable Cause

Capability analysis
• What is the currently "inherent" capability of my process when it is "in control"?

Conformance analysis
• SPC charts identify when control has likely been lost and assignable cause
variation has occurred

Investigate for assignable cause

• Find “Root Cause(s)” of Potential Loss of Statistical Control

Eliminate or replicate assignable cause

• Need Corrective Action To Move Forward
How do you get a Six Sigma Process?
Step 1: Do Things Consistently
ISO 9000 can be very helpful
Step 2: Reduce Variability in the Process
Taguchi: Even small deviations are quality losses.
It is not enough to look at “Good” vs “Bad”
Outcomes.
Only looking at good vs bad wastes
opportunities for learning; especially as
failures become rare (closer to six sigma) you
need to learn from the “near misses”

Step 3: Accommodate Residual Variability Through Robust

Design

Double-checking and Fool-proofing

A Systems View
of Total Quality
Management

PLANNING
CONTINUOUS MANAGEMENT
IMPROVEMENT COMMITMENT
& LEADERSHIP

CUSTOMER EMPOWERMENT MGT BY FACT FOCUS

TRAINING
EMPLOYEE ANALYTICAL
INVOLVEMENT PROCESS
THINKING
 Toyota Production System

• Pillars:
1. just-in-time, and
2. autonomation, or automation with a human touch

• Practices:
– setup reduction (SMED)
– worker training
– vendor relations
– quality control
– foolproofing (baka-yoke)
– many others
 JIT Implementation
• Adopt goal to eliminate all forms of waste
• Improve workplace cleanliness and order
• Promote flow manufacturing
• Level production requirements
• Improve and standardize all process steps
 The Seven Zeros
• Zero Defects: To avoid delays due to defects. (Quality at the
source)
• Zero (Excess) Lot Size: To avoid “waiting inventory” delays.
(Usually stated as a lot size of one.)
• Zero Setups: To minimize setup delay and facilitate small lot sizes.
• Zero Breakdowns: To avoid stopping tightly coupled line.
• Zero (Excess) Handling: To promote flow of parts.
• Zero Lead Time: To ensure rapid replenishment of parts (very
close to the core of the zero inventories objective).
• Zero Surging: Necessary in system without WIP buffers.
Cross Training and Plant Layout
• Cross Training:
– Adds flexibility to inherently inflexible system
– Allows capacity to float to smooth flow
– Reduces boredom
– Fosters appreciation for overall picture
– Increase potential for idea generation
• Plant Layout:
– Promote flow with little WIP
– Facilitate workers staffing multiple machines
– U-shaped cells
• Maximum visibility
• Minimum walking
• Flexible in number of workers
• Facilitates monitoring of work entering and leaving cell
• Workers can conveniently cooperate to smooth flow and
address problems
U-Shaped Manufacturing Cell

Inbound Stock Outbound Stock

 Kanban
• Definition: A “kanban” is a sign-board or card in Japanese and
is the name of the flow control system developed by Toyota.
• Role:
Kanban is a tool for realizing just-in-time. For this tool to work
fairly well, the production process must be managed to flow as
much as possible. This is really the basic condition. Other
important conditions are leveling production as much as possible
and always working in accordance with standard work methods.
• – Ohno 1988
• Push vs. Pull: Kanban is a “pull system”
– Push systems schedule releases
– Pull systems authorize releases
 One-Card Kanban
Outbound Outbound
stockpoint Completed parts with cards stockpoint
enter outbound stockpoint.

Production When stock is Production

cards removed, place card authorizes
production card start of work.
in hold box.
 The Lessons of JIT

– The production environment itself is a control

– Operational details matter strategically

– Controlling WIP is important

– Speed and flexibility are important assets

– Quality can come first

– Continual improvement is a condition for survival