1 Success depends upon the ability to measure performance. Rule #1: A process is only as good as the ability to reliably measure. Rule #2: A process is only as good as the ability to repeat. Gordy Skattum, CQE 2 VIII-2 It is impossible for us to improve our processes if our gaging system cannot discriminate between parts or if we cannot repeat our measurement values. Every day we ask “Show me the data” - yet we rarely ask is the data accurate and how do you know? 3 VIII-2 What is a measurement system used for? Product Control ◦ Detection, conformance to a design specification Process Control ◦ Prevention, real-time control, assessing a feature to its natural process variation 4 VIII-2 AIAG – Automotive Industry Action Group ◦ Collaboration between “Big 3” to create one set of guidelines for all suppliers MSA Reference Manual – Measurement System Analysis ◦ Introduction to MSA ◦ Covers normally occurring measurement situations ◦ Developed to meet specific needs of the automotive industry Gage Repeatability and Reproducibility ◦ A study to understand the within-system and between-system variation in a measurement system ◦ A comparison of standard deviation 5 VIII-2 Discrimination ◦ In selecting or analyzing a measurement system, we are concerned about the system’s discrimination, or the capability of the system to detect and faithfully indicate even small changes of the measured characteristic - also known as resolution. ◦ The smallest readable unit 100ths graduation decimal rule 6 VIII-4 Yes Accurate (on target) No Precise (low variation) Yes No 7 VIII-4 Bias ◦ The difference between the observed average of measurements and the reference value. The reference value, also known as the accepted reference value or master value, is a value that serves as an agreed-upon reference for the measured values. Bias is measured as “accuracy” or as “accuracy shift.” Choices for addressing bias error ◦ Calibrate the gage; adjust, correct, or apply an offset ◦ Change the system (instrument, condition, masters, …) Bias average reading true value 8 VIII-4 Stability ◦ Stability (or drift) is the total variation in the measurements obtained with a measurement system on the same master or parts when measuring a single characteristic over an extended time period. ◦ The change in bias over time. Stability Time 2 Time 1 9 VIII-4 Linearity ◦ Linearity is the bias over the operating range of a measurement system. This, along with bias, is checked as part of the calibration procedure. Choices for addressing linearity ◦ Calibrate, adjust the gage or build offset table ◦ Change the system (condition, masters, …) 10 VIII-4 Repeatability (EV) ◦ Repeatability is the variation in measurements obtained with one measurement instrument when used several times by one appraiser while measuring the identical characteristics on the same part. Includes all within-system variation. Repeatability 11 VIII-4 Reproducibility (AV) ◦ Reproducibility is the variation in the average of the measurements made by different appraisers using the same measuring instrument when measuring the identical characteristics on the same part. Includes all between-system variation. Operator B Operator C Operator A Repeatability 12 VIII-4 The sensitivity of a measurement system to detect process variation Rules for determining effective resolution 1. Count the number of “0” plot points on the process range control chart. If >25%, then the gage lacks effective resolution 2. Count the gage discrimination levels between the UCL and the LCL of the process average control chart. If <5 levels, gage lacks effective resolution 13 VIII-4 Individuals Chart 6 Levels between UCL-LCL? UCL=5.2633 4 CEN=2.16 2 0 LCL=-0.9433 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 -2 Moving R Chart Number of zeros? 5 4 UCL=3.8115 3 2 CEN=1.1667 1 0 LCL=0.0 -1 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 14 VIII-4 A capability measure Typically we compare ◦ Gage R&R to tolerance ◦ Gage R&R to process variation Three levels of results ◦ 0-10% ◦ 10-30% ◦ +30% 15 VIII-5 If you have a poor measurement system… Difficult/impossible to make process improvements Causes quality / cost / delivery / responsiveness problems False alarm signals, increases process variation, loss of process stability Improperly calculated control limits Can make your processes worse! 16 VIII-5 Since the purpose of the analysis of a measurement system is to understand the systems variation, the use of graphical tools is very important. Personal investigations have unveiled many powerful gage analysis software packages. • SPCXL (Sigma Zone) • Minitab • All will deliver identical results* • SPCXL is easy to use and inexpensive. • Minitab is a complete statistical analysis package which requires a lot of training. 17 VIII-7 The four standard methods for analyzing measurement systems are: • Range method (short form) • Average and Range (long form) • ANOVA (Analysis of Variance) • Attribute gage study (both short and long form methods) Each method has its advantages and disadvantages as well as limitations. Refer to the AIAG MSA Reference Manual (Measurement Systems Analysis) for additional information. 18 VIII-7 G AG E R & R S TU D Y E Q U IP M E N T : D AT E: OPERATOR A NAME: S A M P L E (n ) 1 S T T R IA L -B 2 N D T R IA L -C 3 R D T R IA L -D RANGE OPERATOR B NAME: 1 S T T R IA L - F 2 N D T R IA L - G 3 R D T R IA L - H RANGE OPERATOR C NAME: 1 S T T R IA L -J 2 N D T R IA L - K 3 R D T R IA L -L RANGE 1 2 3 4 5 6 7 8 9 10 TO TA LS S U M O F B ,C ,D S U M O F F ,G ,H R bar A AVE. X BAR A S U M O F J ,K ,L R bar B AVE. X BAR B R bar C AVE. X BAR C R bar A R bar B R bar C # T R IA L S D4 SUM 2 3 .2 7 AVE. R BAR 3 2 .5 8 (A V E . R B A R ) * (D 4 ) = U C Lr M A X. X B A R M IN . X B A R A LL R A N G E V A LU E S O V E R U C L A R E R E C A LC U LA TE D X B A R D IF F . TO R E D U C E TO A V E . R A N G E V A LU E R E P E A T A B IL IT Y -E Q U IP M E N T V A R IA T IO N (E V ) EV = AVE. R BAR * AD D IT IO N AL IN F O R M AT IO N k1 EV = n N O . T R IA L S (m ) 2 3 k1 4 .5 6 3 .0 5 EV = m R E P R O D U C IB IL IT Y - A P P R A IS E R V A R IA T IO N (A V ) AV = S Q R T O F [(X B A R D IF F .)*(k 2 )]^2 -[(E V ^2 )/(n *m )] k2 AV = OPERATORS 2 3 k2 3 .6 5 2 .7 n = NUMBER OF PARTS m = N U M B E R O F T R IA L S R E P E A T A B IL IT Y A N D R E P R O D U C IB IL IT Y R&R = S Q R T O F [(E V )^2 + (A V )^2 ] T o le ra n c e USL LS L % G R R to l G R R to l a t 5 .1 5 R&R = d iffe rn c e = G R R to l a t 6 (c o n s ta n t va lu e s ro u n d e d ) www.jimakers.com/downloads/Basic Quality Tools.xlsx 23 VIII-8 Following are examples of gage analysis charts one would find using SPCXL. M S A D a ta T e m p la te F o r A ttrib u te d a ta e n te r A fo r A c c e p t a n d R fo r R e je c t D a te : P a rt T yp e : D e s c rip tio n : 3 /4 /2 0 0 4 USL: LSL: P a rt # 1 2 3 4 5 6 7 8 9 10 R e fe re n c e 1 .0 0 .6 R ep 1 0 .6 5 1 0 .8 5 0 .8 5 0 .5 5 1 0 .9 5 0 .8 5 1 0 .6 O p e ra to r 1 R ep 2 R ep 3 0 .6 0 .6 2 1 0 .9 5 0 .8 0 .8 0 .9 5 0 .9 0 .4 5 0 .5 1 1 0 .9 5 0 .9 5 0 .8 0 .8 5 1 1 0 .7 0 .6 5 R ep 1 0 .5 5 1 .0 5 0 .8 0 .8 0 .4 1 0 .9 5 0 .7 5 1 0 .5 5 O p e ra to r 2 R ep 2 R ep 3 0 .5 5 0 .5 0 .9 5 0 .9 5 0 .7 5 0 .7 5 0 .7 5 0 .8 5 0 .4 0 .4 5 1 .0 5 1 0 .9 0 .9 0 .7 0 .7 5 0 .9 5 0 .9 5 0 .5 0 .5 5 R ep 1 0 .5 1 .0 5 0 .8 0 .8 0 .4 5 1 0 .9 5 0 .8 1 .0 5 0 .8 5 O p e ra to r 3 R ep 2 R ep 3 0 .5 5 0 .5 1 1 0 .8 0 .8 5 0 .8 0 .8 0 .5 0 .4 5 1 .0 5 1 0 .9 5 0 .9 0 .8 0 .8 1 .0 5 1 0 .8 0 .8 5 M S A X b a rR M e th o d R e s u lts S o u rc e T o ta l M e a s u re m e n t (G a g e ) R e p e a ta b ility R e p ro d u c ib ility P ro d u c t (P a rt-to -P a rt) T o ta l V a ria n c e 0 .0 0 1 8 2 4 4 9 0 .0 0 0 8 7 2 2 2 0 .0 0 0 9 5 2 2 7 0 .0 2 9 9 1 3 7 7 0 .0 3 1 7 3 8 2 6 USL LSL P re c is io n to T o le ra n c e R a tio P re c is io n to T o ta l R a tio R e s o lu tio n 1 0 .6 0 .6 4 0 7 1 0 1 2 0 .2 3 9 7 6 1 1 4 5 .7 B IA S A N A L Y S IS R e fe re n c e N o t A va ila b le B ia s S ta n d a rd D e v ia tio n % C o n trib u tio n 0 .0 4 2 7 1 4 0 0 8 5 .7 5 % 0 .0 2 9 5 3 3 3 7 3 2 .7 5 % 0 .0 3 0 8 5 8 8 1 3 3 .0 0 % 0 .1 7 2 9 5 5 9 7 5 9 4 .2 5 % 0 .1 7 8 1 5 2 3 3 8 1 0 0 .0 0 % 24 VIII-8 M S A - R a n g e C h a rt 0 .1 4 0 .1 2 P a rt R a n g e 0 .1 O p e ra to r 1 0 .0 8 O p e ra to r 2 O p e ra to r 3 U C L = .1 2 9 C e n te r = .0 5 0 .0 6 LCL = . 0 .0 4 0 .0 2 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 P a rt N u m b e r M S A - X b a r C h a rt 1 .2 1 0 .8 P a rt A v e ra g e O p e ra to r 1 O p e ra to r 2 O p e ra to r 3 0 .6 U C L = .8 5 7 C e n te r = .8 0 6 L C L = .7 5 4 0 .4 0 .2 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 P a rt N u m b e r 25 VIII-8 The total amount of variance in the gage and in the process total gage 2 2 2 process where , total gage R from control chart d2 from sample size R d2 from gage study from # repeated measuremen ts 26 VIII-14 Most processes are designed to meet the customer specification Because we are using all of our tolerance, we’re forced to keep the process exactly centered. If the process shifts at all, nonconforming parts will be produced Lower Limit Target Upper Limit 27 VIII-19 Using 75% or less of a tolerance will allow processes to shift slightly without producing any defects. The goal is to improve your process in order to use the least amount of tolerance possible ◦ Reduce the opportunity to produce defects ◦ Reduce the cost of the process 28 VIII-19 Defines the width of the process distribution Cp is calculated by dividing the tolerance zone width by the width of the +/- 3 sigma distribution This Cp number (or index) tells how many times the distribution will fit into the tolerance zone A Cp of at least 1.33 is desired Cp Tolerance 3 USL LSL 6 * * Which standard deviation do I use? 29 VIII-25 If a process uses 50% of a tolerance zone, the Cp value would be 2.0 If a process uses 100% of the tolerance zone, the Cp value would be 1.0 If a process uses 200% of the tolerance zone, the Cp value would be 0.5 30 VIII-25 Process capability as a percentage of tolerance The inverse of the calculations for Cp Divide the width of the +/- 3 sigma distribution by the width of the tolerance zone A CR of no more than .75 is desired CR 3 Tolerance 6 * USL LSL * Which standard deviation do I use? 31 VIII-25 If a processes Cp = 1.0 the CR = 100% If a processes Cp = 2.0 the CR = 50% If a processes Cp = .5 the CR = 200% neat 32 VIII-25 Takes into account not only the spread of the distribution, but also the location of it as well A Cpk of at least 1.33 is desired Calculating Cpk: USL Mean Mean LSL Cpk min , 3 * 3 * * Which standard deviation do I use? Cpk = Cp - a “Penalty” for offcenter distributions! 33 VIII-26 If a process uses 100% of a tolerance zone, Cp = 1.0 If the distribution is not centered, the Cpk <1.0 Cpk = 1.0 Cpk <1.0 34 VIII-26 If a process uses 1/2 of the tolerance zone, the Cpk = 2.0 If the process is not centered, the Cpk value would be <2.0 LSL Target USL Cpk = 2.0 LSL Target USL Cpk <2.0 this stuff is so awesome 35 VIII-26 Low Speed Limit High Speed Limit MEAN 1 65 Cp = USL - LSL 6 6 CR = USL - LSL 70 75 * * * MEAN - LSL Cpk = 3 Min Cpk = USL - MEAN 3 36 VIII-26 1st Shape – control chart Stabilize process Am I in control? 2nd Spread – Cp Reduce variation Cp>1.33? 3rd Location – Cpk Center process Cpk>1.33? Control then capability 37 VIII-26 We can determine next steps to improve the process by comparing the Cp and Cpk numbers. For example: High Cp, high Cpk… Process is centered (accurate) and capable (precise). No improvements are needed. High Cp, low Cpk… Process is capable (precise) but not centered (accurate). Improvements should shift the process mean to match the target. Low Cp, low Cpk… Process is not centered (accurate), and variation must be reduced to be precise. 38 VIII-26 USL = 1.505 LSL = 1.500 = .00045 CR = Cp = USL = .507 LSL = .506 = .00006 CR = Cp = USL = 2800 PPH LSL = 2700 PPH Xbar = 2750 PPH = 12.5PPH CR = Cp = Cpk = USL = 750 Mhz LSL = 735 Mhz Xbar = 740 Mhz = 1.333Mhz CR = Cp = Cpk = USL = 1.503 LSL = 1.500 Xbar = 1.501 = .00045 CR = Cp = Cpk = USL = .251 LSL = .250 Xbar = .250 = .00015 CR = Cp = Cpk = 39 VIII-26