LSSGB Lean Six Sigma Green Belt: DMAIC, Process Improvement

Posted: March 4, 2011 | Author: Snapjudge | Filed under: Management | Tags: 6Sigma, A3, Analyze, Batch, Cert,Certifications, Change, CTQ, Cycle Time, Define, Design, DFSS, DMADV, DMAIC, Exams, FMEA, Green Belt, IDOV, Improve,Improvements, Inputs, IPO, KPIV, Lead Time, Lean, LSSGB, MAP, Measure, Methods, Notes, Optimize, Outputs, Process,Process Time, SIPOC, Six Sigma, Tests, VOC, VSM, X, Y |Leave a comment

1 0 Rate This Lean is all about o o o o eliminating waste Discover the Root Cause Determine the Flow Speed Identify Bottlenecks to Increase Speed

Six Sigma is all about o o o Variation Eliminate the wide range Reduce the Deviation Deliverables

VOP & VOC SIPOC / VSM

Error Xs and Ys Confidence in Data MSA

Root Causes Why are we having the issues?

Kaizen Plan Project Plan: What/When/How/Who (Resources/Timeline)

Audits Control Plan

JDI

1 Person & Less than a Week

Kaizen

Team & Two Weeks

10 Steps for Process Improvement

1. 2. 3. 4. 5. 6. 7. 8. 9. Champion starts A3 Map Current Process Validate Measurement System Collect & Organize I/O Data Screen for Key Input Variables (KPIVs) Identify Root Causes and Solutions to achieve Objectives from Step I Improve KPIVs to achieve Objectives from Step I Establish Standard Work Instructions and Control Plan Control KPIVs with or without Statistical Process Control (SPC)

10. Validate that the Project Charter has been achieved and Celebrate

Tollgate Reviews
Has the Project been clearly defined? Is the Measuring system adequate?

Kaizen = Lean Office Implementation Model

1. Select Process to Improve 2. Draw the current State Map 3. Find Wastes Select Lean Office Tools or Other Improvement Ideas 4. Draw the Future State Map & Action Plan 5. Achieve the Future State & Standardize

8 Wastes
Defects Over-work Waiting Non-utilized People Transportation Inventory in Excess Motion Extra Processing

Lean Office Waste Elimination Tools

A3 Report 3P = Production, Preparation, Process IT Supply Chain Management Kanban Pull Systems Quality @ the Source Poka Yoke & Jidoka Continuous (Cellular) Flow Standardized Work Instruction Sheet (SWIS) Cross Training & Team Work 5S & Visual Office Value Stream Mapping

What is Six Sigma?

Statistical Measure of Variation Target of 3.4 defects per million operations GE added Define phase

DMAIC
Basic premise is that sources of Variation can be

1. 2. 3.

Identified Quantified Eliminated or Controlled Define the business opportunity Measure the Process current state Analyze Determine the Root Cause (or Y = f(x) Improve Eliminate waste & variation Control Evidence of sustained results

Process Focus for 6 Sigma

a. Inputs: Independent, Cause, Control b. Output: Dependent on above Input, Effect, Monitor Y = f(X)

Thoughts
Kaizen Kanban Efficiency & Effectiveness Central Tendency Theorem sample Size: Good Rule of Thumb is 30 How many DPMO in 1,000,000? 3.4 Define your Process with cost and/or Time Data Points Accuracy vs. Precision Attribute (Counting data) vs. Variable (Measured Continuous) Defect vs. defective

Define Phase
Purpose

Identify Process and/or Product to be improved Identify, Clients, their needs Define Performance standards Establish Objectives Quantify gaps between outputs & requirements

CTC Critical to Customer

VOC ==> CTQ := Critical to Customer Characteristics 1. 2. 3. CTC Critical to Cost CTQ Critical to Quality CTD Critical to Delivery

VSM Value Stream Mapping

1. 2. 3. 4. Document current situation Identify waste & design in future state Draw the future state Prepare Implementation Plan Information Flow vs. Product Flow Process Cycle Time vs. Waiting Time Lean ==> VSM 6 Sigma ==> SIPOC

IPO
Types: 1. 2. 3. 4. N Noise C Controls controllable Inputs (Knob Variables) Data S Standard Operating Procedures

5.

X Critical Inputs: Statistically proven to impact the Process Y(s) : Y = f(X)

Metrics
FY: Final Yield = Units Passed / Units Submitted - doesnt take into account the Hidden Factory = defects + Rework TPY: Throughput Yield = Units thru the Process, Right the 1st time / Units submitted RTPY: Rolled Throughput Yield NY: Normalized Yield

Defect Metrics
DPU: Defects per Unit = Number of Defects (D) / (U) Total Units TDPU: Total Defects per Unit Opportunity counting TOP: Total Number of Opportunities = Units * Opportunities

Defects per Opportunity

DPO = OP * U = Number of opportunities per unit * Number of Units per characteristic DPO = D / TOP = Number of Defects / Total Number of opportunities per Characteristic

DPMO
DPMO: Defects per Million Opportunities = PPM DPMO = [Total Defects * 1,000,000] / [Total Units * Total Opportunities per Unit] DPMO = [Defects * 1,000,000] / Total Opportunities DPMO = DPO * 1, 000, 0000

Defects vs. Defectives

Single Instance of non-conformance to a Customer Requirement Wrong Issues Some Problems Unit Containing 1 or more defects

Always considered a Fail on a Pass/Fail Scale Non-functional Unit Useless

Measure Phase
Deliverables
Process Map Measurement system Analyses on the Ys (Low Hanging Fruit) Action Items assigned to Team

Data Types
1. a. b. 2. a. b. c. d. Maths Variable Attribute or or Continuous Discrete

Measurement Model Nominal Averages Ordinal Rank Interval Range

Measures
Mean Average Median Midpoint Mode Most frequent

Measure Key Tools

Pareto Histograms

Measurement system Analysis DPMO & Sigma

Measurement Uncertainty Sources of Variations: 6M Control

Materials Methods Machines Man/Woman Measurement Mother Nature

Normal Distribution?
Poisson - Skewed-left data: Median is between Mode & Mean, with Mean on Left Skewed-right data: Median is between Mode & Mean, with Mean on Right Bi-Modal Distribution: Multi-Modal Means could be same; But the Standard Deviation differ Narrow Distribution Standard Normal Curve: Mean = 0; Std Deviation = 1

Measures of Spread
Accuracy & Precision Variance: Accuracy describes centering Stability: Precision describes spread Range = Max Min Deviation = (X mu): Distance between a data point & Mean

MinTab
Normal Probability Plot Normally distributed Data will appear on the Plot as a Straight Line If P-Value > 0.05, the data is Normal

Normality Test: Cubic vs. Quadratic Quadratic Behavior: P-Value Positive Skew P-Value < 0.005 Negative Skew P-Value < 0.005

Measurement system Analysis

Precision
Repeatability Do People agree with Themselves Reproducibility Do People agree Among themselves

Accuracy
1. 2. a. b. 3. 4. a. b. Stability Linearity = Slope * Process Variation % Linearity = Slope * 100 Resolution: Precision Scaling system Bias constant Bias Variable Bias

Analyze
1. 2. Screen for KPIV Identify Root Causes & Solutions to achieve objectives for A3 :: Making sure of the

assumptions from above step are verifiable here 3. Brainstorm

4. 5. 6. 7. 8. 9.

Affinity Diagram Cause and Effect Matrix Fishbone 5 Whys Data Mining Scatter / Correlation

10. Regression 11. Data Transformation 12. FMEA 13. Hypothesis Testing for Means 14. Graphical techniques 15. Chi-Square 16. ANOVA 17. SPC for Analysis Statistical Process Control 18. Lean Analysis Spaghetti Map 19. Video Process Visual analysis 20. OEE

Lean Six Sigma Analysis Model

a. 1. 2. 3. 4. 5. 6. Process Map Inputs, VSM, Outputs Data Tables: Causes & their associated Counts Pareto Cause & Effect Diagram Cause & Effect Matrix Quantify & Prioritize Normal? Statistical & Graphing Techniques Root Cause or Causes

Analyze Phase Deliverables

Prioritized list of Potential sources of Variation Variation Component Studies Measurement Analysis on the Xs Data Collected to Validate Sources Graphical & Statistical analysis of data P-value establishing level of significance & probability Correlation & Regression analysis to determine variable relationships Reduced list of Potential KPIV that affect the output Updated Control charts, process map & FMEA

Cause and Effect

Products
1. 2. 3. 4. 5. 6. Man or Woman Measurement Method Materials Machines Mother Nature

4Ps Control Transactional

1. 2. 3. 4. 5. 6. People Policy & Procedure Supplies Place Measurement Environment

Fishbone

Cause: Products or Transactional (Above) Effect - Problem Statement Fishbone Ishikawa Root cause Finding Ideas: Systematically list all Categorizing Layers of possibilities; Xs Controllable C (Knowledge) Procedural P (People, Systems) Noise N (External, Uncontrollable)

Affinity Diagram
1. 2. 3. 4. 5. Generate Ideas Spoken data, opinions, issues Display Ideas < 15 items of information? > Do NOT use it Sort Ideas into Groups Root Causes & Pain Points Create Header Cards Draw Finished Diagram

Cause & Effect Matrix

1. 2. 3. 4. 5. 6. 7. Identify Client Requirements Outputs Assign Priority to each Output using a ranking scale List all Inputs Correlate Inputs to Outputs using a ranking scale Cross Multiply correlation values with Output Priorities Add across Rows for each Process Step or Input Sort Inputs by Highest Total

FMEA Failure Mode & Effect Analysis

Risk Analysis

Tool to analyze & Quantify the risks associated with NEW design

Form Components
1. a. 2. a. b. c. 3. a. b. c. 4. a. 5. a. b. c. 6. a. b. c. 7. a. 8. Process Step/Input What is the Process Step? Potential Failure Mode What can go wrong with the Process Step? Process defects Historical bugs Prior issues data Potential Failure Effects What is the Effect on the Outputs? VSM Process Map Ys SEV Severity How Bad? Class Control Procedural Noise Potential Causes What are the Causes? From Brainstorming & Causes and Effects diagram/matrix 5 Whys OCC Occurrence How Often? Current Controls

a. b. c. d. 9. a. b.

How can this be found? SPC Statistical Process Control Mistake Proofing Poka Yoke Detection after Failure DET Determine How well? Probability of control detecting the failure mode

10. RPN a. b. c. Risk Priority Number SEV * OCC * DET it is used in Pareto

11. Actions Recommended a. b. What can be done? Responsible Person & Date

Types
Design FMEA; Process FMEA; Similar to Fishbone / Cause & Effect Matrix System FMEA; Equipment FMEA

Approach?
1. 2. 3. 4. Identify Potential failure modes and their effects Prioritize risks associated with specific causes Identify ways of eliminating or reducing the specific causes Document a plan to prevent the failures

DFSS Design for Six Sigma

DMADV Design, Measure, Analyze, Design, Verify QFD Quality Function Deployment IDOV Identify, Design, Optimize, Verify

Graphs, Charts, Plots, Maps

Good Graph? variety of Data Selection of

Variables Graph Range Histogram Box Plot Summarize data abt

Shape Dispersion Center of the Data Spot Outliers Minimum of 10 observations One of the variables shd be Discrete or categorical & Other to be Continuous. Time Series Plot Multi-Vari Chart What is the largest source of Variation?

Data Mining
Classification Class Clustering Similar Group Data Regression Prediction Association Rules

Correlation
Is it directly proportional or inversely related to each other? Correlation Coefficient always has a value between -1 & +1 X Independent variable; Y Dependent variable r Coefficient of Determination: Variability |r| > 0.80 > Relationship is significant |r| < 0.20 > Relationship is NOT significant Correlation Coefficient is HIGH & p-value is LOW Reject the NULL hypothesis, there is a
2

correlation Scatter Plot

Regression
Prediction Equations Forecasting: Linear; Quadratic, Exponential, Cubic Correlation tells us the Strength of the Relationship; Not a numerical relationship r How the value varies? :: Whether it varies or NOT? [Correlation] r How dependent is X to Y? :: Whether you could be confident of r?
2

Modeling
1. are Normality Assumptions satisfied?

2. 3.

are Equal Variance Assumptions satisfied? are Independence Assumptions satisfied?

Terminology
Response Variable: Output Y Dependent Variable Uncontrolled Regressor Variable: Input X Independent Variable Noise Variables Regression Equation - Prediction Equation Residuals - Diff between predicted response values & observed response values SS Sum of Squares DF Degrees of Freedom Multiple Regression: 2 or more Xs; 1 Y

Residuals
be Normally Distributed be Independent of each other data must be time-ordered (Residuals vs. Order Graph) have a Constant Variance (# of residuals above & below line equally spread) Residuals vs. Fitted Values: Shd be randomly scattered with NO patterns Residuals vs. Order of the Data: Shd show no trends & approx the same # of points above &

below the line

Value Stream Analysis Lean analytical Tools

1. 2. 3. 4. a. b. c. Identify waste in the Current State Map Define the Objectives for the Future State Map Answer the transition questions for Process Improvement List all Lean six sigma solutions until Future state goals can be achieved Which processes can be eliminated and how? Which processes can be combined and how? How can the slowest processes be improved?

d. e. f. 5.

Where to apply continuous flow? How to improve processes to achieve Future State How can different process be harmonized? Draw Lean Solutions on Current state Map

Reduce Lead Time & Improve Delivery

Total cycle Time = TCT = Lead Time = LT = PT + WT Process Time + Waiting Time

Overall Employee Effectiveness (OEE)

measure of performance Availability: %age of time available to work

Work Time Absenteeism & Time not in workstation Productivity: %age of time work is done at Expected Rate

Avail time Interruptions, unnecessary meetings & speed losses Quality: Efficient time Time producing errors & work not to standard Overall Equipment Effectiveness

Hypothesis Testing
If we have 2 or more factors/variables use Chi-factor or ANOVA Identify sources of Variability using Historical or Current Data

Types
Passive Historic Active A Modification is made to a process & then sample data is obtained

Methodologies
Hypothesis Tests of Mean Hypothesis Tests of Proportion Hypothesis Tests of Range Works for diff in Means, Variations, Proportions, Std deviations Inferential Stats Small samples are used to answer questions There is always a chance that the conclusion obtained is WRONG Hypotheses are statements abt Population parameters, not abt Sample statistics

Forming a Hypothesis
1. 2. Burden of proof rests with H0 Assume H0 to be true until proven otherwise

Null vs. alternate Hypotheses

1. a. b. c. d. e. 2. a. b. If p is LOW, then NULL H0 must GO < 0.05 Alternate is good Null Hypothesis NOT valid Less likely p< If p is HIGH, > 0.05 > H0 is Good Null Hypothesis is likely true p>

Significance Level
Level of Error: How confidence? There is an % chance that we are wrong when we say that Alternate Hypothesis is valid. Reasonable doubt =

H0 assumed to be true Type I Error or Producers Risk Risk of finding a difference when there really isnt one Risk Type II Error Consumers Risk Risk of NOT finding a difference when there really is one Statistical Power 1 Chance of rejecting the null hypothesis when the NULL hypothesis is false Likelihood of detecting an effect when an effect is present Ability to detect that the NULL hypothesis is False Two Tailed or Double sided tests: / 2 2 sample t-s 2 sample means One Tailed or Single Sided Tests 1-sample t Are we improving the system? < 0 Is it better the new way? Just use

Hypothesis Tests of Mean

When Dependent variable (Response) is Continuous Independent Variable (Factor) is discrete Methods 1 Sample Z Sample Mean & Target value when population std is known :: Single Mean

Comparison 1 Sample t Sample Mean & Target value when population std is NOT known

2 Sample t Means obtained from two independent samples Paired t Means obtained from paired samples

When Hypothesis test is Appropriate These hypotheses deal with Mean values Only one factor for examination one given characteristic Data follows a Normal distribution? p > 0.05 Whether sample size is good? > 30

Chi Square Test ::

No need to be Normal Tests the hypothesis of Independence between two variables Used for Attribute data sample size must be large enough to have each cell populated by 5 or more observations

(occurrences)

Contingency Tables
Null Hypothesis Classifications are Independent: H0 :p1 = p2 = p3 = = pn Ha at least one p is different Used for Frequency (Count) type data

Analysis of Variance ANOVA

Hypothesis test for Means Uses 2 Components of Variance Within Variance NO Change :: Signal :: Between Variance After a Change :: NOISE p-value of the ANOVA test < 0.05 > REJECT the null hypothesis; Alt: At least 1 is diff

SPC

Control Chart
Run chart with confidence intervals calculated Random Variation Region UCL, LCL, Mean Common Cause Variation vs. Special (Non-Random) :: Outliers Forced by External Factors

Charts Control
X-s Chart : Variable, n > 10 X-R Chart : Variable, 1 < n < 10 IMR Chart : Variable, n == 1 np Chart : Attribute, Defectives, Constant Sample Size YES Number of defectives per lot p Chart : Attribute, Defectives, Constant Sample Size NO % defective per subgroup c Chart : Attribute, Defects, Constant Opportunity YES Number of defects per subgroup u Chart : Attribute, Defects, Constant Opportunity NO Number of defects per unit

Nelson Tests for Special causes Control

Any point outside control limits 6 consecutive points increasing or decreasing

Takeaways
Information regarding a sample is called Statistics Information regarding the Population is called Parameters

Disadvantage of Box-Cox transformation is the difficulty in reversing it. Why do you need Normal Data? Yes

t-tests ANOVA Control Charts Capability NO

Pareto Fishbone 5 Whys Run Charts Agreement Analysis Bar Charts Variation is inversely proportional to Standard Deviation Poka Yoka Error Proofing Devices T-tests measure the significance of differences between means Increase Sample Size OR Decrease Predictability Power What is p-value: Level of significance Error = 0.05 ==> 5% ==> Confidence Interval :: 95% Rejected H0 when we should NOT have rejected 10% Confidence Interval :: 90% Failed to reject H0 when we should have rejected

Validations Questions to Ask when ANALYZE

How good is your Measuring System

How Reliable is the Data? What is getting Measured? Is it Normal? Is Sample representative of the Population? Find out basic statistics like Mean, Standard deviation

Improve
Agenda
v Visual Office Cellular Flow Layout Links Manual & Automated operations Maximize Value-added Content Minimize Waste SWIS 5S Pull Systems Visual Displays (IDs) Ideas Board v Cross Training v Quality at the Source (Poka Yoke) v Basic DOE v Kaizen v Action Plan

Visual Office
Following will be quickly accessible: 1. Team Goals

2. Team Performance 3. Useful Information 4. Ideas 5. Work Instruction

SWIS Standard Work Instruction Sheets

v Sequence of activities that each worker performs to complete one cycle

Standard Operating Procedures

1. Document 2. Explain 3. Display 4. Show 5. Be Shown 6. Audit 7. Habit

5S
1. 2. 3. 4. 5. Sort Set in Order Shine Standardize Sustain

What is it?
1. Clean & Organize the Work Area 2. Maintain a safe and clean work environment 3. Document Team processes 4. Standardize Team processes 5. Continuously Improve Team Processes 6. Safety

Pull Systems
Lead Time = Wait Time + Process Time o Order Point

o o o o o o

Two Bin Supermarket Kanban Min / Max Breadman Vendor-Managed

Order Points Pull Locations

1. Supermarkets 2. Buffers 3. Heijunka 1. Setting the pace 2. Context Switching 3. Asymmetric Multi-processing

Supermarkets
Order Point 1. { Adjusted Average Daily Demand (AADD) * 2. Replenishment Lead Time (LT) } 3. Safety 1. {AADD * LT] * %Error Order Quantity o For Purchased Part or Raw Materials +

o AADD * Frequency of Supply o For Manufactured Items

o Machines: AADD * EPE Interval o Manual Process: Order Point

Kanban Implementation Model/Create Pull System

1. Never send defective products downstream 2. Downstream process draws from upstream processes only what is required by the Kanban 3. Upstream process produces items in exact quantity and sequence indicated by the Kanban 4. Never produce or convey parts without a Kanban 5. Attach Kanban to the actual container of parts

6. Use Kanban to fine tune the production system

Error Proofing Principles

Error: Nonconformity BEFORE it passes to another area Defect: An error that HAS BEEN passed to another area 1. Elimination 2. Replacement 3. Facilitation 4. Detection 5. Mitigation

DOE Methodology
1. Define the Practical Problem 2. Establish the experimental Objective 3. Select the OUTPUT (Response) Variables 4. Select the INPUT (Independent) Variables 5. Choose the levels for the input variables 6. Select the experimental design 7. Execute the experiment and collect data 8. Analyze the data from the designed experiment and draw statistical conclusions 9. Draw practical solutions 10. Replicate or validate the experimental results 11. Implement Solutions

Design of Experiments
Nomenclature for Factorial Experiments
2
k

Level

Factor

Measurements

Input Variables

Fractional Factorial / Tagucci Factorial What are the INPUTS being manipulated in this design How many RUNS are there in this experiment

Experimental Design
RCBR Randomization

 Center Points Used for investigating curvature Blocking Used to account for Noise Variables Repetitions

Two Level Sample Size

Based on the Number of Factors (Xs) find the best DOE Approach K <= 4 - Full Factorial K = 5 - Fractional 6 <= k <= 11 - Taguchi/Plackett-Burman

Kaizen
Finish it in less than 4 hours

Steps for the Kaizan Event

1. Decide on the type of Kaizen 2. Pre-Kaizen Activities 3. Select the Right Team 4. Train 5. Run the Kaizen Event 6. Celebrate

Pre-Kaizen Activities
Start a Kaizen A3 by filling olut the Define Box Draw a Current State Value Stream Map or IPO Gather Objectives related data for the VSM or IPO Study the current process in detail Go to the Gemba (Real Floor) Collect data yourself if needed Determine how long the event will be Schedule production to allow the resources to be available

Steps to Improvement Implementation

1. Brainstorm possible solutions 2. Prioritize Solutions 1. Based on Difficulty 2. Impact on KPIV 3. Cost 4. Time 3. Develop the Action Plan 4. Validate Improvements

Control
Detection Feedback Correction

Agenda
1. SPC 2. Visual Controls Responding to visual displays 1. Verbal Instructions 2. Written Instructions 3. SWIS 4. SPC 5. Poka Yoke (Mistake Proofing) 6. DFSS 3. Control Plan 4. Sustaining & Celebrating 5. Optimize Improvements

Deliverables
Statistically validates optimum process operation conditions Systematically experiment with KPIVs Determine optimum conditions of KPIVs to result in optimum KPOV to meet CTQ Proposed optimum operating conditions and solution

 Proposed operating details Continue the Lean Journey A new & improved process performance baseline and capability Validated significance of improvements Confirmation of control plan and standards

Statistical Process Control

Sets limits of variation Provides Trigger for Action Process Behavior Charts Muda Waste Mura Variation Muri Overburden

SPC Purpose
Causes of Variation 1. Common Cause 1. Natural Variability 2. Random Variation Region 3. UCL to LCL 4. +3 till -3 5. 99.73% Area 1. Special Cause 1. Unnatural Variability (6 Ms) 2. Non-Random Variation region

Control Charts
1. C 1. Track Number of defects per subgroup of Units 2. Sample Size is Constant 2. U 1. Track the Number of Defects per Unit 2. Sample Size is Variable

3. nP a. 1. 1. Sample Size is Constant 2. Sample Size is usually > 50 1. P 1. Track the fraction of defective units 2. Sample Size is Variable 3. Sample Size is usually > 50 Track the Number of Defective Units per subgroup

Visual Control Correction ARCI

Accountable 1 person Responsible Consultant Informed

Control Plan
v Written summary describing systems used for monitoring/controlling process or product variation v Document all control methods v Design FMEAs v Design Reviews v Defect Prevention Methods

Training Plan
1. Module 2. Who 3. Schedule 4. Who will be trained 5. Trainers 6. Final Location

Documentation Plan
1. Process Maps/Flowcharts 2. Procedures (SOPs)

3. FMEA 4. Monitoring Plan 5. Response Plan 6. Systems & Structures

Acronyms
5S CPK Capability Meeting Specifications: process capability index COPQ Costs of Poor Quality CPK Capability (Symptoms to 6 Sigma) CTC Critical to Customer CTC Critical to Cost CTD Critical to Delivery CTQ Critical to Quality CTS Critical To Satisfaction CTX Critical to Quality DFSS Design for Six Sigma FMEA JIT KDD Knowledge Discovery in Databases KPIV Key Process Input Variables LEI Lean Engineering Institute LSL Specification Limit PIV Process Input Variables POV Process Output Variables QFD RPN Risk Priority Number

SL Specification Limit SPC SWIS Standard Work Instruction Sheets TRIZ USL Specification Limit VOB Voice of Business VOC Voice of Client/Customer VOP Voice of Process VSM Value Stream Mapping Z = Sigma