Statistical Process Control Seminar at Jireh Semiconductor. Topic Agenda

Transcription

1 Statistical Process Control Seminar at Jireh Semiconductor Instructor: John Breckline January 24, 2018 In association with BW (Ben) Marguglio, LLC Topic Agenda 2 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement 8 hours overview what/why, not how

2 John Breckline 3 Retired but Motorola: Auto Electronics, Commercial, Telecom (pagers) Nokia: MBB / Continuous Improvement Manager AT&T: MBB Transactional / Lean Projects MBB Credentials: Stat-a-Matrix (99) / Nokia (again) ASQ Certifications: CQE (87), CBB (03), CSQP (17) Certification Preparation Instructor since years in Disciplines Inspection, Sampling, Data Systems SPC, Measurement, Supplier Systems Management Six Sigma Master Black Belt Fort Worth TX Mobile: jbreckline@att.net Seminar Objectives 4 To reinforce the concepts and practices of SPC To bring a fresh perspective of the utilization of SPC To lay a foundation of statistical thinking as a basis for effective SPC To challenge those responsible for SPC to optimize the use SPC to manage production operations

3 Statistical Thinking 5 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement Statistical Thinking Map 6 Design FMEA Measurement System Analysis Process Control Define Critical Characteristics Process Capability Studies Process FMEA Process Capable No Process Improvement Yes Approval Planning Control Plan Statistical Process Control

4 Statistical Process Control (SPC) 7 Objectives & Benefits To monitor, control, and improve process performance over time by studying variation and its source. Focuses on monitoring and detecting process variation Provides on-line process feedback and control signals Distinguished between common and special causes of variation Helps improve process to perform consistently and predictably Application of Control Charts 8 Review: Basic rules for Out of Control performance: Point out of control why? Shifts why? Runs why? Others why? Batch A B C UCL=87.7 X=60.2 LCL=32.7

5 Data Use 9 Why Collect and Track Data Data Information Decision Understand Evaluate Control Predict Objective of Statistical Usage Identifying problem areas through trend analysis Assisting in identifying and alleviating risk Improving the current process Preventing problems Data Graphics 10 Data Plotting vs Recording Plotting data into a chart is used for visibility and recognition (easier to understand a visual aid than a bunch of numbers Recording data is necessary to avoid risks: No record of adjustments Subtle changes are lost Measurement errors are lost Plotting errors missed Sampling errors missed

6 Control Charts 11 Control Charts Line graphs that display a dynamic picture of process behavior Focuses attention on monitoring and detecting process variation over time Used to analyze variation in processes Distinguishes special from common causes of variation as a guide to action Theory of Variation 12 Common Cause: ~85% of all process problems are due to common cause variation. Common cause variation is predictable Special Cause: ~15% of all process problems are due to special cause variation. Special cause variation is unpredictable

7 Theory of Variation 13 Common Cause Always present Standard practices Built into the process Predictable variation Special Causes Unpredictable occurrences Significant variation Assignable cause Understand the Difference Different tools to improve / control each The Sources of Variation 14 Machine Measurement Methods Process/ System Customer Satisfaction Material Environment Man

8 Basic Statistics a Refresher 15 Terms Description Symbol Mean / Average mathematical center of the sample/population Median geographic center of the sample/population N/A Range width of distribution (largest smallest) R Standard Deviation statistical width of distribution measured by sigma increments s 99.73% Histogram Normal Curve Normal Curve 4s 3s 2s 1s 0 1s 2s 3s 4s Common and Special Causes 16 If common causes of variation dominate, the output of a process forms a distribution that is stable and predictable over time. Prediction Time Metric = Lines on Time If special causes of variation dominate, the output of a process is not stable over time and not predictable.? Prediction Time Metric = Lines on Time

9 Normal Curve Area Under the Curve 17 Normal Curve Accounts for Common Cause variation Lower natural limit (LNL) Upper natural limit (UNL) % % % % 2.14 % 2.14 % 0.13 % 0.13 % average -3*sigma average -2*sigma average -1*sigma average average +1*sigma average +2*sigma average +3*sigma Basic Statistics Distributions 18 Distributions: Not all distributions are Normal Exponential Distribution is common for one-sided specs (PPB rates) Predictive statistics require Normal Distributions (SPC is predictive) Normal Distribution not perfect Exponential Distribution Jireh?

10 Central Limit Theorem 19 Concept The sample mean become normally distributed as sample size increases The spread of the sample means are less than the spread of the individuals of the sample Value Non-normal distributions can be assessed using normal distribution statistics and tools (X-bar-R, etc.) Most inferential statistical tools assume normality of data Confidence Intervals SPC / Control Charts Performance vs. Requirements 20 Requirements Customer Specifications Internal Operating Control Levels Compare Performance to Requirements LSL USL O

11 The Loss Function inverted normal distribution curve Rejected Product Taguchi Loss Function Acceptable Product Rejected Product 21 Best Output Lower Specification Limit (LSL) Nominal Upper Specification Limit (USL) Key Learnings

12 Critical Characteristics 23 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement Key Characteristics 24 What is Important to End Customer to Manufacturers to Jireh Operations to Supplier Operations Identifying Characteristics Design FMEA Early Supplier Involvement Process FMEA White / Black Diamond Designation

13 Key Characteristics 25 Product: Process: Key Process Characteristics 26 Mother Earth Materials Man.. Measurement Machine Methods Mother Earth Materials Man.. Measurement Machine Methods

14 Measurement / Process Capability 27 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement Measurement 28 Measurement Systems Analysis (MSA) Gage Repeatability & Reproducibility (GR&R) Variable GR&R Understand capability of measurement Percent of Contribution Variance Percent of Study Std Dev Percent of Tolerance (P/T Ratio) Effect of sample selection <10% acceptable 10-30% marginal >30% high risk Attribute R&R Understand capability of inspection

15 Gage R&R Terms 29 Repeatability The ability of a single instrument to generate accurate measurements consistently same operator, same part, same time Reproducibility The ability of a system to generate accurate measurements consistently different operators or times (accuracy/precision) P/T Ratio Precision to Tolerance how much measurement error is related to specification Critical to product acceptance & capability studies (decision to Spec) R&R Contribution Measurement error related to process / characteristic variation Important to process control (SPC) and process improvement (DOE) Variable Gage R&R 30 2 total Observed = Variability 2 product Product Variability 2 measurement system + Measurement Variability Measurement Variability - Additive - Product Variability Observed values average

16 Why care about Process Capability? 31 Satisfy our customers Understand our processes Prioritize areas in need of quality improvement (variation reduction) activities Verify that process improvements are successful Track improvements over time Give us information to set realistic tolerances Help us identify how to cost products Select the best qualified supplier We can improve our company Bottom Line Process Capability Roadmap 32 Select Characteristic Study Scope Short-term - Long-term Is Data Normal? No Improve / Transform OK Understand Confidence Interval (based on sample size) Multiple Sources? Yes Segregate/ Understand Improve Poor Measurement Capability? N/A Cpk / Ppk Acceptable? No Improve Process OK OK Improve No Is Process Stable? Yes Maintain the Gain

17 Process Capability 33 Long Term vs. Short Term Variation Fishbone diagram Mother Earth Materials Man 5 Min Variation Short Term 10 Hr Variation Long Term Output Measurement Machine Methods 6 Mo Variation Performance Stability & Normality 34 Stability Without stability, cannot use data as a predictor of future Special Causes create instability removal for study? Normality: Calculations are based on Normal Distribution Further from normal less accurate totally invalid Calculation Tests & Probability Plotting Stability Question Normality Question

18 Process Capability 35 How it s measured Cp capability index Pp performance index Cpk ratio to the smallest value (USL/LSL) Ppk ratio to the smallest value (USL/LSL) Process Capability 36 Process Capability rule of thumb: Cpk > 1.50 Process is Six Sigma Cpk > 1.00 All Product Meets Requirements Cpk = 1.00 Most Product Meets Requirements Cpk < 1.00 Some Product Does Not Meet Requirements Ppk typically 0.33 less than Cpk What are Jireh standards for Cpk? What is difference between Cpk & Ppk in Jireh? Note: Above calculates take in 1.5 s shift of mean over time

19 Process Capability 37 Objective of Process Capability Analysis Determine how the natural process limits compare with the specification range Depending on the comparison to standard: Study Outcome Do nothing Change the specs Center the process Reduce process variability -5- Accept the losses Key Learnings

20 SPC: Attribute & Variable Data 39 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement Data Types 40 Type Description Model Symbol Attribute Discrete counted Variable Continuous measured What About %

21 Data Collection 41 Attribute Data Counted: How many How often What kind Discrete Convert Attribute to Variable Measure attribute (length/area) Create a scale: Likert Check Sheets for Attribute data NOT check-lists Defects / errors by time frame farmer s count Quick visual analysis sum by column or row Attribute Distributions 42 Binomial Distribution Defectives number of pieces unacceptable Poisson Distribution Defects errors to a single piece

22 Attribute Control Charts 43 Defect: is each nonconformance to the specified acceptance criteria Defective: is an item (unit) in a sample that has one or more nonconformance to the acceptance criteria A defective contains at least one defect Defectives Control Charts np Plots number of non-conforming units (needs fixed n) p Plots proportion of non-conforming units (n varies) Defects Control Charts c Plots number of defects (requires fixed n) u Plots number of defects per inspection unit (n varies) Attribute Charts 44 Sample Size Varies Sample Size Fixed Defectives (binomial) p Chart (most common) % or portion defective np Chart Number of defectives Defects (Poisson) u Chart Average number of defects c Chart Actual number of defects

23 Human Error Modes Omission 2. Excessive / insufficient repetition 3. Wrong order 4. Early / late execution 5. Execution of restricted work 6. Incorrect selection 7. Incorrect counting 8. Misrecognition 9. Failing to sense danger 10. Incorrect holding 11. Incorrect positioning 12. Incorrect orientation 13. Incorrect motion 14. Improper holding 15. Inaccurate motion 16. Insufficient avoidance BW (Ben) Marguglio, LLC Variable Data SPC Terms 46 Xbar-R (Xbar = average of readings; R = range) n sample size X or x reading (data) Xbar average of readings Xbarbar average of averages R range of data points Rbar average of ranges S or s standard deviation NOTE: UCL/LCL are boundaries for 99.73% of the data population

24 The Control Chart 47 UCL PLOT POINT Measure AVG 99.73% LCL Time Remember: Control limits are determined by the process average values... NOT SPECIFICATIONS NOTE: Process spread is equal to a +/ 3 sigma from the mean Using Variable Data Charts 48 Variable Charts Constructing a chart Historic Data: minimum 20 data-sets (time increments) Reference table values Calculate / post UCL / LCL Visualize Zones (+/ 3 std deviation) Plot data points Interpret Immediately

25 Variables Control Charts 49 X-Bar and R Chart (2 charts) Plot (1) subgroup average Plot (2) subgroup range X-Bar and S Chart (2 charts) Plot (1) subgroup average Plot (2) subgroup standard deviation typically when n>9 ImR/ XmR Chart (individuals moving range 2 charts) Plot (1) individual value Plot (2) difference between 2 consecutive individuals Key Learnings

26 Operational SPC 51 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement Control Plan / SPC 52 Based on Capability & Stability Understand source of input variation Understand frequency of input variation SPC: Run-time, Set-up, Special Event Control Plan Document who, what, when, sample Out of Control Action Plan

27 Control Plan 53 Components of a Control Plan Process Step to be controlled (critical to process/output) Variable to control (input or output) Specification / acceptable management limits Measurement Method Control Method SPC / other monitoring Sample Size / Sample Frequency OCAP Out-of-Control Action Plan Who to contact initiate responsibility for action What action to take (investigative or prescribed) What documentation (SOP or records) Key Process Characteristics 54 Per YOUR determination of Key Process Inputs Mother Earth Materials Man.. Measurement Machine Methods

28 Types of Control Charts 55 Variable Data Xbar R Xbar S X MR Median Attribute Data p Chart np Chart c Chart u Chart Effective SPC Antenna Assembly Long Term Process Capability % X Tool Wear Direction Suggested Monitoring 100% 50% Nom- 0% -50% Upper Spec Locating Slot Width Upper Crush Rib Lower Crush Rib Snap Height X X X X X X Assembly Features X -100% X X X Lower Spec X X -150%

29 Where Should I Control? 57 Set up controls at: Changes of authority Start of significant, irreversible activity After critical quality feature created Vital few process inputs (Xs) Significant cost decision points Service gates Set-up operations Material Introduction Use your process flow chart to identify control points SPC Sampling - Rational Subgrouping 58 Subgrouping Select SPC samples in a way that makes each subgroup as homogeneous (same) as possible A sampling process that reflects the actual process Sources of Variability Time-to-Time (lot-to-lot) Within Piece Between Pieces M T W T F S S M T W T F S S M T W T F S S M T W T F S S

30 Key Learnings Interpretation of SPC Charts 60 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement

31 Statistical Thinking Map 61 Design FMEA Measurement System Analysis Process Control Define Critical Characteristics Process Capability Studies Process FMEA Process Capable No Process Improvement Yes Approval Planning Control Plan Statistical Process Control Control Chart Interpretation 62 Control Charts Average and control limits are a function of past history common cause variation If process remains stable, only common cause variation is existent in the process Special Cause conditions create Out-of-Control signals on the control chart Four general rules: 1. Average: Out-of-Control Range: In-Control 2. Average: In-Control Range: Out-of-Control 3. Average: Out-of-Control Range: Out-of-Control 4. Average: In-Control Range: In-Control 5+ Specific Rules (sigma zones)

32 Out of Control 63 Out of Control Special causes of variation are present Is detected by either having any point outside the control limits or by unnatural patterns 7 Rules to interpret control charts Interpreting a Control Chart A lack of control is indicated whenever a single point falls outside the control limits. UCL Measure AVE LCL Time

33 Interpreting a Control Chart out of 5 points in zone B (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time Interpreting a Control Chart out of 3 points in zone A (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time

34 Interpreting a Control Chart or more consecutive points on one side of centerline (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time Interpreting a Control Chart A trend of 6 or more consecutive points increasing or decreasing (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time

35 Interpreting a Control Chart Stratification, 15 or more points in zone C (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time Interpreting a Control Chart Mixture or systematic variation (A) UCL (B) Measure (C) (C) (B) (A) AVE LCL Time

36 Out-of-Control Action Plan Plan (OCAP) (OCAP) Start Activators Checkpoints Terminators Yes Yes 71 No Yes No Yes No Yes No Yes No End Activators - out-of-control decision rules Checkpoints - list of possible assignable causes Terminators - corrective actions Actions: Who to contact initiate responsibility for action What action to take (investigative or prescribed) What documentation (SOP or records) Engagement Renewal 72 SPC / Stats Review Critical Characteristics Product or Process Measurement Capability Process Capability SPC: Attribute / Variable Establishing SPC in Operations SPC Interpretation Renewed Engagement

37 The Loss Function inverted normal distribution curve Rejected Product Taguchi Loss Function Acceptable Product Rejected Product 73 Best Output Lower Specification Limit (LSL) Nominal Upper Specification Limit (USL) Multi-Vari Studies 74 Studying Relationships Between suspect input variables and the output variables of a process while in normal operation Passive Data Collection Process is monitored in its natural state without intervention or can be done with historical data Full Range of Variation To allow inputs to vary so we can observe their effects on the outputs Inherent Issues To identify process problems and limitations

38 Data Collection 75 The identified inputs (Xs) are monitored in concert with the critical outputs (Ys) Relate variation in the inputs to variation in outputs Xs Ys Multi-Vari Study - Graphical Day Main Effects Plot for Thickness Data Means Shift Product Opr Exp Lot Code Color Booth SolCode TempCode RHCode SpeedCode Mean Fri Mon Sat Sun Thu Tue Wed AX100 B2000 VT687 XR71 High Low Medium 18721A 19982D 22110A 24431C 25501D 26621C 26792A 27032B Black Blue Brown Green Orange Red (min, Q1) 2 - (Q1, med) 3 - (med, Q3) 4 - (Q3, max) 1 - (min, Q1) 2 - (Q1, med) 3 - (med, Q3) 4 - (Q3, max) 1 - (min, Q1) 2 - (Q1, med) 3 - (med, Q3) 4 - (Q3, max) 1 - (min, Q1) 2 - (Q1, med) 3 - (med, Q3) 4 - (Q3, max)

39 1.53 Multi-Vari Study Example Multi-Vari Chart for A by Trial - Part Range of measurement all inspectors Oper single part Trial A Master Inspector measurement Part 4 5 SPC Checklist Activity 78 Team Activity Form Teams of 4-5 Using current SPC Procedure & Process Specs Create a Checklist for daily/weekly review of SPC Activity: 30 minutes to complete Report-Out 5 minutes per team of line items Debrief: collective sharing of team Checklists Similarities & Differences Value of SPC Checklist?

40 Key Learnings Overall Feedback Key Learnings Cpk based individual; SPC on X-bar chart Data interpretation action Stable chart as the foundation for predictable future Clear steps to set up SPC Interpretation of WE rules Fully understand the key parameters then create specific chart to monitor OCAP design review to come up with good procedure Daily review and in some cases to reset SPC limits Charting SPC on a paper a good exercise Multi-variable correlation to generate actions to improve SPC Customer always sees individual not average of the sample; this reminds us to look at and address SPC more carefully Interpretation immediately; looking at SPC chart real-time Process capability, stability, normality review before SPC setup Daily review checklist item exercise/ocap for continuous improvement 80

41 Overall Feedback Key Learnings 81 For your attention and active participation! John

42 John Breckline 83