utdallas.edu/~metin 1 quality management chapter 8

1utdallas.edu/~metin

Quality Management

Chapter 8


Learning Goals

Statistical Process Control X-bar, R-bar, p charts Process variability vs. Process specifications Yields/Reworks and their impact on costs Just-in-time philosophy


Steer Support for the Scooter


Steer Support Specifications

Go-no-go gauge


0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

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

R

79.9

79.91

79.92

79.93

79.94

79.95

79.96

79.97

79.98

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

X-b

ar

Control Charts


Statistical Process Control (SPC)

SPC: Statistical evaluation of the output of a process during production/service

The Control Process– Define

– Measure

– Compare to a standard

– Evaluate

– Take corrective action

– Evaluate corrective action


Not just the mean is important, but also the variance

Need to look at the distribution function

The Concept of Consistency:Who is the Better Target Shooter?


Statistical Process Control

CapabilityAnalysis

ConformanceAnalysis

Investigate forAssignable Cause

EliminateAssignable 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 occurredInvestigate for assignable cause• Find “Root Cause(s)” of Potential Loss of Statistical ControlEliminate assignable cause• Need Corrective Action To Move Forward


Statistical Process Control

Shewhart’s classification of variability: – Common (random) cause

– assignable cause

Variations and Control– Random variation: Natural variations in the output of

process, created by countless minor factors» temperature, humidity variations, traffic delays.

– Assignable variation: A variation whose source can be identified. This source is generally a major factor

» tool failure, absenteeism


Common Cause Variation (low level)

Common Cause Variation (high level)

Assignable Cause Variation

Two Types of Causes for Variation


Mean and Variance

Given a population of numbers, how to compute the mean and the variance?

deviation Standard

)(Variance

Mean

},...,,{Population

1

2

2

1

21

N

x

N

x

xxx

N

ii

N

ii

N


Sample for Efficiency and Stability

From a large population of goods or services (random if possible) a sample is drawn. – Example sample: Midterm grades of OPRE6302 students

whose last name starts with letter R {60, 64, 72, 86}, with letter S {54, 60}

» Sample size= n» Sample average or sample mean= » Sample range= R» Standard deviation of sample means=

x

population theofdeviation Standard: where n

x


Sampling Distribution

Sampling distribution

Variability of the average scores of people with last name R and S

Process distribution

Variability of the scores for the entire class

Mean

Sampling distribution is the distribution of sample means.

Grouping reduces the variability.


Normal Distribution

Mean

95.44%

99.74%

x

at x. cdf normal )1,_,,(normdist:functions lstatistica Excel

at x. pdf normal )0,_,,(normdist:functions lstatistica Excel

devstmeanx

devstmeanx

normdist(x,.,.,0)

Probab

normdist(x,.,.,1)


Cumulative Normal Density

)_,,(norminv :prob""at cdf offunction Inverse

)1,_,,(normdist:at x (cdf)function Cumulative

:functions lstatistica Excel

devstmeanprob

devstmeanx

0

1

x

normdist(x,mean,st_dev,1)

prob

norminv(prob,mean,st_dev)


Normal Probabilities: Example If temperature inside a firing oven has a normal

distribution with mean 200 oC and standard deviation of 40 oC, what is the probability that

– The temperature is lower than 220 oC=normdist(220,200,40,1)

– The temperature is between 190 oC and 220oC=normdist(220,200,40,1)-normdist(190,200,40,1)


Control Limits

Samplingdistribution

Processdistribution

Mean

LCLLowercontrol

limit

UCLUppercontrol

limit

Process is in control if sample mean is between control limits. These limits have nothing to do with product specifications!


Setting Control Limits:Hypothesis Testing Framework

Null hypothesis: Process is in control Alternative hypothesis: Process is out of control Alpha=P(Type I error)=P(reject the null when it is true)=

P(out of control when in control) Beta=P(Type II error)=P(accept the null when it is false)

P(in control when out of control)

If LCL decreases and UCL increases, we accept the null more easily. What happens to – Alpha?– Beta?

Not possible to target alpha and beta simultaneously, – Control charts target a desired level of Alpha.


Type I Error=Alpha

Mean

LCL UCL

/2 /2

Probabilityof Type I error

st_dev)mean,/2,-norminv(1UCL

st_dev)mean,/2,norminv(LCL

The textbook uses Type I error=1-99.74%=0.0026=0.26%.

Sampling distribution


Time

ProcessParameter

Upper Control Limit (UCL)

Lower Control Limit (LCL)

Center Line

• Track process parameter over time - mean - percentage defects

• Distinguish between - common cause variation (within control limits) - assignable cause variation (outside control limits)

• Measure process performance: how much common cause variation is in the process while the process is “in control”?

Statistical Process Control: Control Charts


Control Chart

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

UCL

LCL

Sample number

Mean

Out ofcontrol

Normal variationdue to chance

Abnormal variationdue to assignable sources

Abnormal variationdue to assignable sources


Observations from Sample Distribution

Sample number

UCL

LCL

1 2 3 4


Number of Observations

in Sample Sample size (n)

Factor for X-bar Chart

(A2)

Factor for Lower

control Limit in R chart

(D3)

Factor for Upper

control limit in R chart

(D4)

Factor to estimate Standard

deviation, (d2)

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

Parameters for computing UCL and LCLthe Table method


Period x1 x2 x3 x4 x5 Mean Range

1 1.7 1.7 3.7 3.6 2.8 2.7 2 2 2.7 2.3 1.8 3 2.1 2.38 1.2 3 2.1 2.7 4.5 3.5 2.9 3.14 2.4 4 1.2 3.1 7.5 6.1 3 4.18 6.3 5 4.4 2 3.3 4.5 1.4 3.12 3.1 6 2.8 3.6 4.5 5.2 2.1 3.64 3.1 7 3.9 2.8 3.5 3.5 3.1 3.36 1.1 8 16.5 3.6 2.1 4.2 3.3 5.94 14.4 9 2.6 2.1 3 3.5 2.1 2.66 1.4

10 1.9 4.3 1.8 2.9 2.1 2.6 2.5 11 3.9 3 1.7 2.1 5.1 3.16 3.4 12 3.5 8.4 4.3 1.8 5.4 4.68 6.6 13 29.9 1.9 7 6.5 2.8 9.62 28 14 1.9 2.7 9 3.7 7.9 5.04 7.1 15 1.5 2.4 5.1 2.5 10.9 4.48 9.4 16 3.6 4.3 2.1 5.2 1.3 3.3 3.9 17 3.5 1.7 5.1 1.8 3.2 3.06 3.4 18 2.8 5.8 3.1 8 4.3 4.8 5.2 19 2.1 3.2 2.2 2 1 2.1 2.2 20 3.7 1.7 3.8 1.2 3.6 2.8 2.6 21 2.1 2 17.1 3 3.3 5.5 15.1 22 3 2.6 1.4 1.7 1.8 2.1 1.6 23 12.8 2.4 2.4 3 3.3 4.78 10.4 24 2.3 1.6 1.8 5 1.5 2.44 3.5 25 3.8 1.1 2.5 4.5 3.6 3.1 3.4 26 2.3 1.8 1.7 11.2 4.9 4.38 9.5 27 2 6.7 1.8 6.3 1.6 3.68 5.1

Average

3.81

5.85

n

xxxX n

...21

},...,min{

},...,max{

21

21

n

n

xxx

xxxR

• Collect samples over time

• Compute the mean:

• Compute the range:

as a proxy for the variance

• Average across all periods - average mean - average range

• Normally distributed

The X-bar Chart: Application to Call Center


• Define control limits

• Constants are taken from a table

• Identify assignable causes: - point over UCL - point below LCL - many (6) points on one side of center

• In this case: - problems in period 13 - new operator was assigned

0

2

4

6

8

10

12

1 3 5 7 9 11 13 15 17 19 21 23 25 27

UCL=X +A2 ×R=3.81+0.58*5.85=7.19

LCL=X -A2 ×R=3.81-0.58*5.85=0.41

CSR 1 CSR 2 CSR 3 CSR 4 CSR 5 mean 2.95 3.23 7.63 3.08 4.26 st-dev 0.96 2.36 7.33 1.87 4.41

Control Charts: The X-bar ChartThe Table method


Range Control Chart

ranges sample of average theof multipleA

ranges sample of average theof multipleA

3

4

RDLCL

RDUCL

Multipliers D4 and D3 depend on n and are available in Table 8.2.

EX: In the last five years, the range of GMAT scores of incoming PhD class is 88, 64, 102, 70, 74. If each class has 6 students, what are UCL and LCL for GMAT ranges?

079.6*0 159.279.6*2

.0D ,2D 6,nFor .6.795/)74701026488(

34

34

RDLCLRDUCL

R

Are the GMAT ranges in control?


0

2

4

6

8

10

12

1 3 5 7 9 11 13 15 17 19 21 23 25 27

0

5

10

15

20

25

30

1 3 5 7 9 11 13 15 17 19 21 23 25 27

X-B

ar

R

Control Charts: X-bar Chart and R-bar ChartFor the Call Center


X-bar and Range Charts: Which?

UCL

LCL

UCL

LCL

R-chart

x-Chart Detects shift

Does notdetect shift

(process mean is shifting upward)

SamplingDistribution


X-bar and Range Charts: Which?

UCL

LCL

LCL

R-chart Reveals increase

x-Chart

UCL

Does notreveal increase

(process variability is increasing)SamplingDistribution


• Compute the standard deviation of the sample averages• stdev(2.7, 2.38, 3.14, 4.18, 3.12, 3.64, 3.36, 5.94, 2.66, 2.6, 3.16, 4.68, 9.62, 5.04, 4.48, 3.3, 3.06, 4.8, 2.1, 2.8, 5.5, 2.1, 4.78, 2.44, 3.1, 4.38, 3.68)=1.5687

• Use type I error of 1-0.9974

Control Charts: The X-bar ChartThe Direct method

8.531.5687)9987,3.81,norminv(0.

st_dev)mean,/2,-norminv(1UCL

-0.911.5687)0013,3.81,norminv(0.

st_dev)mean,/2,norminv(LCL

0.0026


Tolerances/Specifications– Requirements of the design or customers

Process variability– Natural variability in a process

– Variance of the measurements coming from the process

Process capability– Process variability relative to specification

– Capability=Process specifications / Process variability

Process CapabilityLet us Tie Tolerances and Variability


Process Capability: Specification limits are not control chart limits

LowerSpecification

UpperSpecification

Process variability matches specifications

LowerSpecification

UpperSpecification

Process variability well within specifications

LowerSpecification

UpperSpecification

Process variability exceeds specifications

Sampling Distribution

is used


Process Capability Ratio

When the process is centered, process capability ratio

A capable process has large Cp.

Example: The standard deviation, of sample averages of the midterm 1 scores obtained by students whose last names start with R, has been 7. The SOM requires the scores not to differ by more than 50% in an exam. That is the highest score can be at most 50 points above the lowest score. Suppose that the scores are centered, what is the process capability ratio?Answer: 50/42

X

pC6

levelion specificatLower - levelion specificatUpper


Processmean

Lowerspecification

Upperspecification

+/- 3 Sigma

+/- 6 Sigma

3 Sigma and 6 Sigma Quality


• Estimate standard deviation:• Or use the direct method with the excel function stdev()• Look at standard deviation relative to specification limits

= R / d 2

3

Upper Specification (USL)

LowerSpecification (LSL)

X-3A X-2A X-1AX X+1A

X+2 X+3A

X-6BX X+6B

Process A(with st. dev A)

Process B(with st. dev B)

x Cp P{defect}

1 0.33 0.317

2 0.67 0.0455

3 1.00 0.0027

4 1.33 0.0001

5 1.67 0.0000006

6 2.00 2x10-9

The Statistical Meaning of Six Sigma


Use of p-Charts

p=proportion defective, assumed to be known When observations can be placed into two categories.

– Good or bad

– Pass or fail

– Operate or don’t operate

– Go or no-go gauge


• Estimate average defect percentage

• Estimate Standard Deviation

• Define control limits

1 300 18 0.0602 300 15 0.0503 300 18 0.0604 300 6 0.0205 300 20 0.0676 300 16 0.0537 300 16 0.0538 300 19 0.0639 300 20 0.067

10 300 16 0.05311 300 10 0.03312 300 14 0.04713 300 21 0.07014 300 13 0.04315 300 13 0.04316 300 13 0.04317 300 17 0.05718 300 17 0.05719 300 21 0.07020 300 18 0.06021 300 16 0.05322 300 14 0.04723 300 33 0.11024 300 46 0.15325 300 10 0.03326 300 12 0.04027 300 13 0.04328 300 18 0.06029 300 19 0.06330 300 14 0.047

p =0.052

SizeSample

pp )1( = =0.013

pUCL= + 3

pLCL= - 3 =0.091=0.014

Period n defects p

Attribute Based Control Charts: The p-chart


0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0.160

0.180

13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Attribute Based Control Charts: The p-chart


Inspection

Where/When » Raw materials» Finished products

» Before a costly operation, PhD comp. exam before candidacy

» Before an irreversible process, firing pottery

» Before a covering process, painting, assembly

Centralized vs. On-Site, my friend checks quality at cruise lines

Inputs Transformation Outputs

Acceptancesampling

Processcontrol

Acceptancesampling


ProcessStep

Bottleneck

Based on labor andmaterial cost

MarketEnd ofProcess

Defectdetected

Defectoccurred Defect

detectedDefectdetected

Cost of defect

$$ $

Based on salesprice (incl. Margin)

Recall, reputation,warranty costs

Defectdetected

Discovery of Defects and the Costs

CPSC, Segway LLC Announce Voluntary Recall to Upgrade Software on Segway™ Human TransportersThe following product safety recall was conducted by the firm in cooperation with the CPSC. Name of Product: Segway Human Transporter (HT) Units: Approximately 6,000

Recall Alert

U.S. Consumer Product Safety CommissionOffice of Information and Public AffairsWashington, DC 20207September 26, 2003


Examples of Inspection Points

Type ofbusiness

Inspectionpoints

Characteristics

Fast Food CashierCounter areaEating areaBuildingKitchen

AccuracyAppearance, productivityCleanlinessAppearanceHealth regulations

Hotel/motel Parking lotAccountingBuildingMain desk

Safe, well lightedAccuracy, timelinessAppearance, safetyWaiting times

Supermarket CashiersDeliveries

Accuracy, courtesyQuality, quantity


The Concept of Yields

90% 80% 90% 100% 90%

Line Yield: 0.9 x 0.8 x 0.9 x 1 x 0.9

Yield of Resource = rate Flow

resource the atcorrectly processed units of rate Flow

Yield of Process = rate Flow

correctly processed units of rate Flow


Rework / Elimination of Flow Units

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

Rework



Rework: Defects can be corrected Same or other resource Leads to variability Examples: - Readmission to Intensive Care Unit

Loss of Flow units: Defects can NOT be corrected Leads to variability To get X units, we have to start X/y units Examples: - Interviewing - Semiconductor fab


Why Having a Process is so Important:Two Examples of Rare-Event Failures

Case 1: Process does not matter in most cases• Airport security• Safety elements (e.g. seat-belts)

Case 2: Process has built-in rework loops• Double-checking

1 problem every 10,000 units

99% correct

“Bad” outcome happens with probability (1-0.99)3

Good

Bad

99% 99%

99%

1%

1% 1%

Learning should be driven by process deviations, not by defects

“Bad” outcome only happens Every 100*10,000 units


Rare events are not so rare:Chances of a Jetliner Crash due to Engine Icing

Engine flameout due to crystalline icing: Engine stops for 30-90 secs and hopefully starts again.

Suppose 150 single engine flameouts over 1990-

2005 and 15 dual engine flameouts over 2002-2005. What are the annualized single and dual engine flameouts?10=150/15 and 5=15/3

Let N be the total number of widebody jetliners flying through a storm per year. Assume that engines ice independently to compute N.Set Prob(2 engine icing)=Prob(1 engine icing)2

(5/N)=(10/N)2 which gives N=20

There are 1200 widebody jetliners worldwide. It is safe to assume that each flies once a day. Suppose that there are 2 storms on their path every day, which gives us about M=700 widebody jetliner and storm encounter very year. How can we explain M=700 > N=20? The engines do not ice independently. With M=700, Prob(1 engine icing)=10/700=1.42% and Prob(2 engine icing)=5/700=0.71%. Because of dependence Prob(2 engine icing) >> Prob(1 engine icing) 2 .

Unjustifiable independence leads to underestimation of the failure probabilities in operations, finance, engineering, flood control, etc.


Just-in-Time Philosophy

Pull the operations rather than pushing them– Inventory reduction

– JIT Utopia» 0-setup time

» 0-non value added operations

» 0-defects

Discover and reduce process variability


Push vs Pull System What instigates the movement of the work in the system?

In Push systems, work release is based on downstream demand forecasts– Keeps inventory to meet actual demand

– Acts proactively» e.g. Making generic job application resumes today (e.g.: exempli gratia)

In Pull systems, work release is based on actual demand or the actual status of the downstream customers– May cause long delivery lead times

– Acts reactively» e.g. Making a specific resume for a company after talking to the recruiter


Push/Pull View of Supply Chains

Procurement,Manufacturing andReplenishment cycles

Customer OrderCycle

CustomerOrder ArrivesPush-Pull boundary

PUSH PROCESSES PULL PROCESSES


Kanban

Direction of production flow

upstream downstream

Kanban

Kanban

Kanban

Authorize productionof next unit

Pull Process with Kanban Cards


Bro

wse

rer

ror

Number ofdefects

Cause of DefectAbsolute Number Percentage Cumulative

Browser error 43 0.39 0.39

Order number out of sequence 29 0.26 0.65

Product shipped, but credit card not billed 16 0.15 0.80

Order entry mistake 11 0.10 0.90

Product shipped to billing address 8 0.07 0.97

Wrong model shipped 3 0.03 1.00

Total 110O

rder

nu

mb

er o

ut

off

seq

uen

ce

Pro

du

ct s

hip

ped

, bu

tcr

edit

car

d n

ot

bill

ed

Ord

er e

ntr

ym

ista

ke

Pro

du

ct s

hip

ped

to

bill

ing

ad

dre

ss

Wro

ng

mo

del

ship

ped

100

50

Cumulativepercents ofdefects

100

75

50

25

Pareto Principle or 20-80 rule


•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”

Reduce Variability in the ProcessTaguchi: Even Small Deviations are Quality Losses

LowerSpecification Limit

Target value

QualityLoss

High

LowPerformance Metric Target

value

QualityLoss

Performance Metric, x

UpperSpecification Limit

Traditional view of Quality loss Taguchi’s view of Quality loss

Performance Metric


• Double-checking (see Toshiba)• Fool-proofing, Poka yoke (see Toyota)

• Computer plugs• Set the watch 5 mins ahead

• Process recipe (see Brownie)• Recipes help standardize

Accommodate Residual (Common) Variability Through Robust Design


Materials

MachinesSpecifications /information

People

Vise positionset incorrectly

Machine toolcoordinates set incorrectly

Vice position shiftedduring production

Part clampingsurfaces corrupted

Part incorrectlypositioned in clamp

Clamping force toohigh or too low

Cutting tool worn

Dimensions incorrectlyspecified in drawing

Dimension incorrectly coded In machine tool program

Material too soft

Extrusion stock undersized

Extrusion dieundersized

Extrusion ratetoo high

Extrusion temperaturetoo high

Error in measuring height

Steer support height deviates from specification

Ishikawa (Fishbone) Diagram


Summary

Statistical Process Control X-bar, R-bar, p charts Process variability vs. Process specifications Yields/Reworks and their impact on costs Just-in-time philosophy


Jesica Santillan died after a bungled heart-lung transplant in 2003. In an operation Feb. 7, Jesica was mistakenly given organs of the wrong blood type.

Her blood type was 0 Rh+. Organs come from A Rh- blood type.

Her body rejected the organs, and a matching transplant about two weeks later came too late to save her. She died Feb. 22 at Duke University Medical Center.

Line of Causes leading to the mismatch• On-call surgeon on Feb 7 in charge of pediatric heart transplants,

James Jaggers, did not take home the list of blood typesLater stated, "Unfortunately, in this case, human errors were made during the process. I hope that we, and others, can learn from this tragic mistake."

• Coordinator initially misspelled Jesica’s name• Once UNOS (United Network for Organ Sharing) identified Jesica,

no further check on blood type• Little confidence in information system / data quality• Pediatric nurse did not double check• Harvest-surgeon did not know blood type

Process Failure in Healthcare: The Case of Jesica Santillan


- As a result of this tragic event, it is clear to us at Duke that we need to have more robust processes internally and a better understanding of the responsibilities of all partners involved in the organ procurement process.

William Fulkerson, M.D., CEO of Duke University Hospital.

- We didn’t have enough checks. Ralph Snyderman, Duke University Hospital

Jesica is not the first death in organ transplantation because of blood type mismatch.

Process Failure in Healthcare: The Case of Jesica Santillan


Step 1: Define and map processes - Jaggers had probably forgotten the list with blood groups 20 times before - Persons involved in the process did not double-check,

everybody checked sometimes - Learning is triggered following deaths / process deviations are ignored

Step 2: Reduce variability - quality of data (initially misspelled the name)

Step 3: Robust Design - color coding between patient card / box holding the organ - information system with no manual work-around - let the technology help

RFID tagged patients: Tag includes blood type and other infoElectronic medicine box: Alarming for the obsolete medicine

The Three Steps in the Case of Jesica


1. Management Responsibility2. Quality System3. Contract review4. Design control5. Document control6. Purchasing / Supplier evaluation7. Handling of customer supplied material8. Products must be traceable9. Process control10. Inspection and testing

11. Inspection, Measuring, Test Equipment12. Records of inspections and tests13. Control of nonconforming products14. Corrective action15. Handling, storage, packaging, delivery16. Quality records17. Internal quality audits18. Training19. Servicing20. Statistical techniques

Examples: “The design process shall be planned”, “production processes shall be defined and planned”

How do you get to a Six Sigma Process? Do Things Consistently (ISO 9000)


Zero InventoriesZero DefectsFlexibility / Zero set-upsZero breakdownsZero handling / non value added

Just-in-time Production• Kanban• Classical Push• “Real” Just-in-timeMixed ProductionSet-up reduction

AutonomationCompetence and TrainingContinuous ImprovementQuality at the source

Organization MethodsPrinciples

The System of Lean Production

Pardon the French, caricatures are from Citroen.


• Avoid unnecessary inventory• To be seen more as an ideal• To types of (bad) inventory: a. resulting from defects / rework b. absence of a smooth process flow• Remember the other costs of inventory (capital, flow time)

Inve

ntor

y in

pro

cess

Buffer argument:“Increase inventory”

Toyota argument:“Decrease inventory”

Principles of Lean Production: Zero Inventory and Zero Defects


71

2345

68

ITAT=7*1 minute

3

1

2

4

ITAT=2*1 minute

Good unit

Defective unit

ITAT: Information Turnaround Time


• Flexible machines with short set-ups• Allows production in small lots• Real time with demand• Large variety

• Maximize uptime• Without inventory, any breakdown will put production to an end• preventive maintenance

Avoid Non-value-added activities,specifically rework and set-ups

Principles of Lean Production: Zero Set-ups, Zero NVA and Zero Breakdowns


Push: make to forecast Pull: Synchronized production

Pull: Kanban

• Visual way to implement a pull system• Amount of WIP is determined by number of cards

• Kanban = Sign board • Work needs to be authorized by demand

• Classical MRP way• Based on forecasts• Push, not pull• Still applicable for low cost parts

• Part produced for specific order (at supplier)• shipped right to assembly• real-time synchronization• for large parts (seat)• inspected at source

Methods of Lean Production: Just-in-time


InventoryInventoryInventory

CycleInventory

Production with large batches

End ofMonth

Beginning ofMonth

Cycle


End ofMonth

Beginning ofMonth

Cycle


End ofMonth

Beginning ofMonth

Cycle


End ofMonth

Beginning ofMonth

Inventory

End of

Month

Beginning of

Month

Produce Sedan

Produce Station wagon

End of

Month

Beginning of

Month

Produce Sedan


End of

Month

Beginning of

Month

Produce Sedan


End of

Month

Beginning of

Month

Production with small batches

Methods of Lean Production: Mixed Production and Set-up reduction


• Create local decision making rather than pure focus on execution• Use machines / tools, but avoid the lights-off factory

• Automation with a human touch

• Cross training of workers• Develop problem solving skills

Organization of Lean Production: Autonomation and Training


• Solve the problems where they occur - this is where the knowledge is - this is the cheapest place

• Traditional: inspect and rework at the end of the process

• Once problem is detected, send alarm and potentially stop the production

Own Process Next Process End of Line FinalInspection

End User

$ $ $ $ $

• Rework• Reschedule

• very minor • minor delay

• Significant Rework• Delayed delivery• Overhead

• Warranty cost• recalls• reputation• overhead

Defect found

Defect fixed

Organization of Lean Production: Continuous Improvement and Quality-at-the-source

utdallas.edu/~metin 1 quality management chapter 8

Documents

variation slide

control process

dev slide

distribution of sample

sample mean

absenteeism slide

scooter slide

gauge slide