Professional Documents
Culture Documents
Quality Management
By Prof. N. Narayanan
In
“OPERATIONS MANAGEMENT – I”
(WMP, Term I, Part B, July – Sep, 2012)
Sn. Text
Nos. Date Topic
Reading
Core Agenda for Session
18 – Quality Ch 5: Quality 1. Case: TQM improves copier service (pp 164)
19 Management Management – Describes role played by TQM in improving
(pp 149 – service quality
179); 2. Case: Richey International’s Spies (pp 171)
3. Problem: Determining ‘Process capability
index’ at a shoe company (pp 196)
Sup 5:
Statistical 4. Problem: Control Limits for weights of oats
Process flakes boxes (pp 186 – 7)
Control (pp 5. Problem: range limits at Clinton
181 – 211) manufacturing (pp 188)
6. Case: Unisys Corp.’s costly experiment in
health care services (pp 192)
Basics in
Quality Management
Global Company Profile:
Managing Quality Provides a Competitive Advantage
Arnold Palmer Hospital (p 150 – 51)
Delivered over 10,500 babies in 2004, and
over 13,000 in 2006.
Virtually every type of quality tool is employed
Continuous improvement
Employee empowerment
Benchmarking
Just-in-time
Quality tools
1. What is Quality? (pp 152 – 56)
Different Views on ‘What is Quality?’
Failure
costs Prevention and
Failure appraisal costs
costs
• In the TQM view, the quality costs are to be not only optimized
through right trade-offs in the short term, but also brought down
to near zero levels in the long term.
Costs of Quality
Internal Failure
Prevention
Appraisal
Quality Improvement
Caselet:
The Very High Costs of Quality at Mercedes (pp 154)
1. The first efforts at revival of the Company’s performance
resulted in the diversion of attention from quality, leading, by
2003, to severe loss in quality (particularly in the suppliers’),
and hence in high quality costs:
• Series of recalls from its $50,000 E-Class sedan.
• Spate of problems in 2004 with brake control systems—680,000 cars
were recalled.
• The biggest recall in history in 2005—of 1.3 million cars with faulty
fuel pumps made by supplier, Robert Bosch.
• Many other defects arising from software problems and failed
interfaces in the complex electronic systems.
• A total of $600 million in warranty costs in a single year!
2. The costs of quality also resulted in a toll in sales, and loss of
the number one position, as luxury car maker, to BMW!
Quality and Strategy
Relationship to quality:
1.Plan:
4. Act:
Identify the
Implement the improvement and
plan (revised make
as and when a plan.
necessary).
3. Check: 2. Do:
Is the plan Test the
working? plan.
Taguchi Concepts
Three Taguchi Concepts
1. Quality loss function The purpose is to promote
quality by minimizing variances
from the target values.
2. Target-oriented quality
The purpose is to promote quality
3. Quality robustness through ‘Parameter design’ by
minimizing the effects of variances
from the target values.
Mean (Target
Value)
±3
Mean Mean
(Target
Value)
±3
Note: In a ‘six sigma process’, its ±6σ spread need not be within the
specification limits.
Process Capability & Defects
Process Capability Total Products outside two-sided
Index (Cpk) specification limits
0.25 453,255 ppm
1.50 7 ppm
Source: Quality Planning & Analysis, Juran & Gryna, Chapter 17, 3e
A ‘Six Sigma Process’, as compared to a ‘3σ Process’
A Six Sigma process, with the
variation of ±6σ lying within
A 3σ process specification limits, but with a
bias of 1.5σ
leads to a defect of
A 3σ process, with its
2,700 defects/million natural variation, ±3σ lying
(when the process centre within specification limits.
has zero bias with respect to
1.5
the target value) Upper
Lower σ
specification specification
limit limit
A ‘Six Sigma’
process can
lead to a defect of Mean Mean (Target
3.4 defects/million Value)
(as when a 6σ process ±3
centre has a bias of 1.5σ
with respect to the
target value) ±6
From a statistical viewpoint, the ‘Six Sigma’s defect level of 3.4 defects per
million’ represents a process that can have a ‘bias’ of up to 1.5σ, while
its spread of ±6σ still remains within the specification limits.
(Not in the book)
Six Sigma
– a structured approach to process improvement
1. Reliability 7. Credibility
2. Responsiveness 8. Security
3. Competence 9. Understanding/
4. Access knowing the
customer, all in
5. Courtesy addition to the
6. Communication 10.Tangibles
Service Recovery Strategy
Standard
What is
Organization Standard
Inspected
Olive Garden Busboy Serves water and
Restaurant bread within 1 minute
Busboy Clears all entrée items
and crumbs prior to
dessert
Waiter Knows and suggest
specials, desserts
Part B
Process B
Process A
Spread of a
LSL process is
USL
indicative of
its capability
LCLa UCLa
Process B
A process that is
aligned closer to
the desired
target is likely to
LCL UCL be more capable
b
Offset of the process center
from the target value
LSL USL
Thus, any measure of ‘process capability’ must be based on both the
‘process spread’, and any ‘offset’ that may be present in the process center.
Process Capability Ratio
Upper Specification Limit – Lower Specification Limit
Cp = 6
213 – 207
= = 1.938
6(.516)
Process Capability Ratio – An Example
(Continued)
LSL USL
3 3
Upper Lower
Cpk = minimum of Specification – x , x – Specification
Limit Limit
1. A capable process must have a Cpk of at least 1.0
2. A ‘capable process’ is not necessarily in the center
of the specification2, but it falls within the
specification limit at both extremes.
1
‘Process Capability Index’ as a measure also takes into account (unlike ‘Process Capability’)
any bias in the process mean (i.e., process being not in the centre of the specifications).
2
A process with Cpk value above 1.0 will be ‘capable’, in spite of the bias in the process.
Process Capability Index
of an ‘Insole cutting machine’ (pp 196)
(.251) – .250
Cpk = minimum of ,
(3).0005
Process Capability Index
This process
Cpk = zero has bias, but
in spite of it,
it is just
‘capable’
because of
Cpk = between 0 and 1 low
standard
deviation, σ.
or
Cpk = 1
Cpk > 1
LSL USL
Process Capability & Defects
Process Capability Total Products outside two-sided
Index (Cpk) specification limits
0.25 453,255 ppm
1.50 7 ppm
Source: Quality Planning & Analysis, Juran & Gryna, Chapter 17, 3e
A 3σ Process
– is a process with its ±3σ spread within the specification limits.
Mean (Target
Value)
±3
Mean Mean
(Target
Value)
±3
Note: In a ‘six sigma process’, its ±6σ spread need not be within the
specification limits.
A ‘Six Sigma Process’, as compared to a ‘3σ Process’
A Six Sigma process, with the
variation of ±6σ lying within
A 3σ process specification limits, but with a
bias of 1.5σ
leads to a defect of
A 3σ process, with its
2,700 defects/million natural variation, ±3σ lying
(when the process centre within specification limits.
has zero bias with respect to
1.5
the target value) Upper
Lower σ
specification specification
limit limit
A ‘Six Sigma’
process can
lead to a defect of Mean Mean (Target
3.4 defects/million Value)
(as when a 6σ process ±3
centre has a bias of 1.5σ
with respect to the
target value) ±6
From a statistical viewpoint, the ‘Six Sigma’s defect level of 3.4 defects per
million’ represents a process that can have a ‘bias’ of up to 1.5σ, while
its spread of ±6σ still remains within the specification limits.
(Not in the book)
Six Sigma Quality – Process Requirements
(Not in the book)
Process Average
The UCL and the UCL reflect the capability of the process
in attaining the ‘target value’
When a Process is “Out of Statistical Control”
• A process is said to be “out of statistical control” (with
respect to a particular ‘control chart’ drawn for the
process), when it has some ‘assignable cause/s’ of variation
in the process, making the sample statistic from the
process, plotted on the control chart, give a signal of such a
possible presence.
• This is not the same as the process being ‘out of control’,
since,
• Even if a process is ‘capable’ in adhering to the specification
limits, it can still be ‘out of statistical control’ with respect to
the control chart in use’, and
• A process in ‘statistical control’ (as indicated by ‘control
charts’ in use) may not necessarily meet the design
specifications, and hence may be ‘not capable’.
Taking Advantage of
‘Central Limit Theorem’
In Developing Control Charts
• Regardless of the distribution of the population, the
distribution of sample means, drawn from the
population will tend to follow a normal curve
• In case of ‘Control Chart for Variables’, the sample
statistic, ‘mean’ can be taken to follow normal
distribution with:
1. The mean of the sampling distribution x=
(x) will be the same as the population
mean
2. The standard deviation of the
sampling distribution (x) will equal
the population standard deviation x =
n
() divided by the square root of
the sample size, n
Sampling Distribution
Sampling
distribution of
the sample
means
Process
distribution of
the population
x=
(mean)
Variables
Measures dimensions such as weight, speed,
height, or strength
Falls within an acceptable range
Variables Attributes
Characteristics that Defect-related
can take any real characteristics
value
Classify products as
May be in whole or either good or bad,
in fractional or count defects
numbers
Categorical or
Continuous random discrete random
variables variables
Measurement Methods
• Variable Based
– Detailed observation of the characteristic (such as
length, diameter, weight)
– measurement will be expensive and more time
consuming but will provide a wealth of information about
the process
• Attribute Based
– simple clustering of the characteristic into a few
categories (such as good or bad)
– Two frequently used attribute measures are:
• Proportion of defects (denoted as p)
• Number of defects (denoted as c)
– measurements are easy to make, quick & less
expensive, but reveal very little information about the
process
8. Control Charts for Variables
Control Charts for Variables
Control Chart
for sample of 9 Variation due
Out of to assignable
boxes control causes
17 = UCL
15 = LCL
| | | | | | | | | | | | Variation due
1 2 3 4 5 6 7 8 9 10 11 12 to assignable
Out of causes
Sample number control
Setting Chart Limits
For x-Charts when we don’t know
1
American Society for Testing Materials, 1951
Setting Control Limits – An Example S2 (pp 187)
UCLx = x + A2R
= 12 + (.577)(.25)
= 12 + .144
= 12.144 ounces
From Table
S6.1
Setting Control Limits
Process average x = 12 ounces
Average range R = .25
Sample size n = 5
= 12 – .144
= 11.856 ounces
R – Chart
where
R = average range of the samples
D3 and D4 = control chart factors
from Table S6.1
Setting ‘Range Limits’ for Variables (p188)
Average range R = 5.3 pounds
Sample size n = 5
From Table S6.1 D4 = 2.115, D3 = 0
= (2.115)(5.3)
= 11.2 pounds Mean = 5.3
UCL
(x-chart detects
x-chart shift in central
tendency)
LCL
UCL
(R-chart does not
R-chart detect change in
mean)
LCL
Mean and Range Charts
Sample Sample
(b) Sample 3 4
Sample 2
These 1
sampling (Sampling mean is
distributions constant but
result in the dispersion is
charts below increasing)
UCL
(x-chart does not
x-chart detect the increase
in dispersion)
LCL
UCL
(R-chart detects
R-chart increase in
dispersion)
LCL
9. Control Charts for Attributes
Control Charts for Attributes
1. For variables that are categorical
• Good/bad, yes/no, acceptable/unacceptable
2. Measurement is typically counting defectives
3. Charts may measure
• Percent defective (p-chart): Control charts for
fraction defectives are effective when a piece
considered defective even if there is a single defect.
• Number of defects in each piece (c-chart):
Control charts for defectives are effective when a
large number of defectives can occur in each piece,
but the actual number that do occur is relatively
small.
Control Limits for p-Charts
Proportion defectives p, of the population (as
representing fraction defectives), will have a
binomial distribution, but applying the Central
Limit Theorem allows us to assume a normal
distribution for the ‘sample statistics’.
p(1 – p)
UCLp = p + z^p ^p =
n
LCLp = p – z^p
where p = mean of the ‘fraction defectives’, based
on all the samples taken
z= number of standard deviations
^ p = estimate of the standard deviation of the
‘fraction defectives’, based on all the samples taken
n = sample size
(Note: The calculation of the standard deviation reflects the binomial distribution of the population.)
Insurance Records at Mosier data Systems (pp 191 – 192)
p-Chart for Data Entry
Sample Number Fraction Sample Number Fraction
Number of Errors Defective Number of Errors Defective
1 6 .06 11 6 .06
2 5 .05 12 1 .01
3 0 .00 13 8 .08
4 1 .01 14 7 .07
5 4 .04 15 5 .05
6 2 .02 16 4 .04
7 5 .05 17 11 .11
8 3 .03 18 3 .03
9 3 .03 19 0 .00
10 2 .02 20 4 .04
Total = 80
80 (.04)(1 - .04)
p = (100)(20) = .04 p^ = = .02
100
p-Chart for Data Entry
UCLp = p + z^p = .04 + 3(.02) = .10
LCLp = p – z
^ p = .04 – 3(.02) = 01
.11 –
.10 – UCLp = 0.10
.09 –
.08 –
Fraction defective
.07 –
.06 –
.05 –
.04 –
.03 – p = 0.04
.02 –
.01 –
.00 –
| | | | | | | | | | LCLp = 0.00
2 4 6 8 10 12 14 16 18 20
Sample number
1
LCL taken as zero, when the calculation yields negative value.
p-Chart for Data Entry
UCLp = p + z^p = .04 + 3(.02) = .10
Possible
LCLp = p - z^p = .04 – 3(.02) = 0 causes
assignable
present
.11 –
.10 – UCLp = 0.10
.09 –
.08 –
Fraction defective
.07 –
.06 –
.05 –
.04 –
.03 – p = 0.04
.02 –
.01 –
.00 –
| | | | | | | | | | LCLp = 0.00
2 4 6 8 10 12 14 16 18 20
Sample number
Control Limits for c-Charts
The number of defects per unit of the
population, c, will have a Poisson distribution 1,
but applying the Central Limit Theorem allows
us to assume a normal distribution for the
sample statistics
UCLc = c + 3 c LCLc = c – 3 c ^
^
^
14 – UCLc = 13.35
UCLc = c + 3 c 12 –
Number defective
10 –
=6+3 6 8 –
= 13.35 6 –
4 – c= 6
2 –
LCLc = c – 3 c 0 –
LCLc = 0
=6–3 6 | | | | | | | | |
= 01 1 2 3 4 5 6 7 8 9
Day
1
LCL taken as zero when the calculation yields negative value.
Managerial Use of Control Charts
Which Control Chart to Use
Variables Data
Target
Target
Target
Target
Target
Target
75 –
50 –
Return whole shipment
0 10 20 30 40 50 60 70 80 90 100
% Defective in Lot
75 –
As sample size of a
sampling plan increases,
Probability of its OCC will more and
Acceptance more be closer to the
50 –
‘perfect’ OC curve, but
such a plan will become
25 – costlier bec
10 –
= 0.10 | | | | | | | | | Percent
0 0– 1 2 3 4 5 6 7 8 defective
AQL LTPD
Consumer’s risk
of acceptance of Good Indifference
bad lot with lots zone Bad lots
percent defects
above LTPD
Producer’s and Consumer’s Risks
(as related to an ‘Operating Characteristics Curve’)
Producer's risk ()
Probability of rejecting a good lot (also
known as type I error – "false positive": the error
of rejecting a null hypothesis when it is actually true )
n = 50, c = 1
n = 100, c = 2
Average Outgoing Quality
Hour
Defect 1 2 3 4 5 6 7 8
A /// / / / / /// /
B // / / / // ///
C / // // ////
Absenteeism
Seven Tools of TQM
3) Cause-and-Effect Diagram: A tool that identifies
process elements (causes) that might effect an
outcome
Causes
Materials Methods
Effect
Manpower Machinery
Cause-and-Effect Diagrams
Material Method
(ball) (shooting process)
Grain/Feel Aiming point
(grip)
Size of ball
Air pressure Bend knees
Hand position
Balance
Lopsidedness
Follow-through
Missed
free-throws
Training Rim size
Machine
Manpower (hoop &
(shooter) backboard)
Seven Tools of TQM
Percent
A B C D E
Pareto Charts
Data for October
70 – – 100
– 93
60 – – 88
50 – 54
– 72
Frequency (number)
Cumulative percent
40 –
30 –
Number of
20 –
occurrences
10 –
0 – 12
4 3 2
MRI Flowchart
1. Physician schedules MRI 7. If unsatisfactory, repeat
2. Patient taken to MRI 8. Patient taken back to room
3. Patient signs in 9. MRI read by radiologist
4. Patient is prepped 10. MRI report transferred to
5. Technician carries out MRI physician
6. Technician inspects film 11. Patient and physician discuss
8
80%
1 2 3 4 5 6 7 11
9 10
20%
Seven Tools of TQM
6) Histogram: A distribution showing the frequency
of occurrences of a variable
Distribution
Frequency
Target value
Time
An SPC Chart
10%
| | | | | | | | |
1 2 3 4 5 6 7 8 9
Lower control limit
Game number
Statistical Process Control (SPC) Chart
Uses statistics and control charts to tell
when to take corrective action
Drives process improvement
Four key steps
Measure the process
When a change is indicated, find the
assignable cause
Eliminate or incorporate the cause
Restart the revised process
The End: Any Questions?