Fall Technical Conference October 17 2002 Launsby Consulting 2002 Robert Launsby Co-author of three books on experimental design Co-developer of DOE Wisdom software Trained several thousand folks in industry on problem solving approaches Article “Straight Talk on DFSS”, SSFM 8/2002 www.launsby.com Download of examples using toys Launsby Consulting 2002 Agenda Model for change “Super Simulation” of air bag deployment system Trebuchets, electromagnets, antacid pills, catapults Launsby Consulting 2002 Boston Survival Skills Statement: Accents are wonderful, We all have them (we just don’t know it) South Boston (Kennedy accent) Car, Harvard, Chelmsford, Factor, Shark Data, Pizza, Cuba, peninsula “Park the car in Harvard yard” Tuna, tuner Launsby Consulting 2002 Niccolo Machiavelli “There is nothing more difficult to carry out, nor more doubtful of success, nor more dangerous to handle than to initiate a new order of things. For the reformer has enemies in all those who profit by the old order, and only lukewarm defender by all those who could profit by the new order. This luke-warmness arises from the incredulity of mankind who do no truly believe in anything new until they have had actual experience with it.” Launsby Consulting 2002 Change The application of simple statistical tools to day-to-day activities represents a huge change for many businesses Launsby Consulting 2002 Strategic Change Matrix Right Thing The world changes Done Well Stage 1 Done Poorly Continue doing old right thing Deny new reality Stage 2 Folks need practice! Stage 4 Launsby Consulting 2002 Wrong Thing Stage 3 Getting Folks Out of Stage 2 Into Stage 3 High contrast and confrontation One approach Air bag simulation using “Super Sim” Student challenge (30 minutes) Team exercise Select best settings (25 trials max) Define technology guide for each response Launsby Consulting 2002 Auto Air Bag System Gas weight Type of charge, amount of charge Launsby Consulting 2002 Orifice diameter Four factors Auto Air Bag System (cont.) Factor Range Orifice diameter 0.08 to 0.14 Gas weight 12.0 to 14.5 Prop. Type 23b, 35b Prop. Weight .85, 1.2 Launsby Consulting 2002 Auto Air Bag System (cont.) Max Pressure delivered Time to first pressure Launsby Consulting 2002 Time to 90 % max. Auto Air Bag System (cont.) 8 9 10 110 120 130 Max Pressure Time to 90 % Goal: Nominal values for all three responses 2 3 Time to First Pressure Launsby Consulting 2002 Auto Air Bag System (cont.) Response Surface**type(B)=23b,propwt(D D ( 1 c o 0.8 m 0.6 p o 0.4 s i 0.2 t 0 e12.8 ) 13 13.2 0.14 0.1292 0.1184 13.6 0.1076 0.0968 13.80.086 orif 13.4 wt Launsby Consulting 2002 Getting Folks Out of Stage 3 Into Stage 4 Trebuchet More complex DOE, PLS techniques Tools are Electromagnets great for DOE, ANOVA, regression starters, but Antacid pills don’t forget Variance reduction, simple DOE real applications Catapults Problem solving, control charts, MSA, simple DOE M&M’s for attributes MSA, C-charts Launsby Consulting 2002 Trebuchet Real Trebuchets were weapons of mass destruction in 12th century Used in TV show “Northern Exposure” to launch piano Today: Used to toss Yugo’s, pumpkins, etc, great amount of info on internet Interesting challenge for getting students beyond simple Design of Experiments, PLS Launsby Consulting 2002 Possible Trebuchet Factors Potential factors of interest Number of weights Pivot point of arm Release arm position Sling length Release bar position Launsby Consulting 2002 Possible responses Total distance Height Flight distance Trebuchet (cont.) Dependent Variable: Number Runs(N): Multiple R: Squared Multiple R: Adjusted Squared Multiple R: Standard Error of Estimate: Variable Constant Rel Bar S Length AB Distance 8 0.998703 0.997408 0.995463 2.89396 Coefficient Std Error 95% CI 156.875 -27.125 -12.375 -26.875 1.02317 1.02317 1.02317 1.02317 ± 2.84133 ± 2.84133 ± 2.84133 ± 2.84133 Tolerance T 1 1 1 P(2 Tail) 153.323 -26.511 -12.095 -26.266 Model looks good! Launsby Consulting 2002 0 0 0 0 Trebuchet (cont.) S Contour Plot 3 100 2.4 L e 1.8 n 1.2 g t 0.6 h 0 140 Prediction is 165 2 Actual value from trial was 194 Launsby Consulting 2002 120 180 160 3 4 Rel Bar Distance 5 Trebuchet (cont.) Eventually fit special cubic model to obtain useful predictions for flight distance with previous factor ranges Note: you can download details from my website Launsby Consulting 2002 Electromagnets Some possible factors Diameter of wire Possible Responses Number of turns Gauss Amount of current Type of core Distance of gauss meter from core Launsby Consulting 2002 Distance from coil to move compass needle Electromagnets (cont.) Main Effects 40 30 f i e l d 20 s t r 10 0 green(-) gold(+) 1.5(-) 3(+) 0(-) 1(+) wire(A) voltage(B) dist(C) Factors Launsby Consulting 2002 10(-) 20(+) turns(D) Electromagnets (cont.) Contour Plot**w ire(A)=green,dist(C)=0.000 3 20 2.7 18 v o 2.4 l t a g 2.1 e 16 14 1.8 1.5 10 12 12 14 16 turns field str Launsby Consulting 2002 18 20 Antacid pills Scatte r Plot 200 Factors: Water temp, amount of cola t i m e Response: 100 Dissolve time 0 1(-) 1.5 w ate r(A) Factors 2(+) Experiment demonstrates interaction, nonlinearity, and variance reduction Launsby Consulting 2002 Antacid pills (cont.) Inte ractions 120 Interaction tested to be significant 100 80 t i m e 60 40 20 acid(+) acid(0) acid(-) 0 1(-) Launsby Consulting 2002 1.5 w ate r(A) Factors 2(+) Catapult Possible factors: pull back angle, “pin on pole”, hook position, cup position, stop position, ball type, etc. Possible responses: total distance, height, flight distance. Positives: simple linear model provides useful predictions Other uses: Control charts (older rubber band will demonstrate shift in mean), problem solving Launsby Consulting 2002 Catapult (cont.) Main Effe cts 140 116 92 d i s t 68 44 20 150(-) 186(+) Angle (A) Launsby Consulting 2002 -1(-) 1(+) Band(B) Factors -1(-) 1(+) Pe g(C) Catapult (cont.) Inte ractions 200 Band(+) Pe g(+) Pe g(+) d i s t 100 Pe g(-) Pe g(-) Band(-) 0 150(-) 186(+) Angle (A) Launsby Consulting 2002 150(-) 186(+) Angle (A) Factors -1(-) 1(+) Band(B) Catapult (cont.) Contour Plot**Peg(C)=-1.00000 1 120 100 0.6 80 B a n d 0.2 60 -0.2 -0.6 -1 150 40 20 157.2 164.4 171.6 Angle dist Launsby Consulting 2002 178.8 186 Paper Helicopter Numerous potential factors including Wing length Wing width Number of weights Type of weight Paper weight Body length Launsby Consulting 2002 Paper Helicopter Launsby Consulting 2002 Dependent Variable: Number Runs(N): Multiple R: Squared Multiple R: Adjusted Squared Multiple R: Standard Error of Estimate: Variable Coefficient Constant 1.65896 Wing Width(A) 0.186528 Wing Length(B) 0.420556 Weight(C) -0.290972 Body Length(D) -0.0811806 Paper(E) -0.181181 AB -0.00777778 AC 0.0419444 AD -0.0390972 AE -0.0203472 BC -0.0827778 BD -0.05625 BE -0.04625 CD 0.0696528 CE 0.0709028 DE -0.011875 Wing Length**2 0.0766667 Time 36 0.99464 0.989308 0.980304 0.0733144 Std Error P(2 Tail) 0.0399463 0 0.0145436 0 0.0157747 0 0.0145436 0 0.0131981 0 0.0131981 0 0.0157747 0.628 0.0145436 0.009 0.0131981 0.008 0.0131981 0.14 0.0157747 0 0.015873 0.002 0.015873 0.009 0.0131981 0 0.0131981 0 0.015873 0.464 0.0473242 0.122 Paper Helicopter Response Surface**Wing Width(A)=2.40000,Weight(C)=1.40 1 D ( T i m e ) 0.8 0.6 0.4 0.2 0 2 1.8 1.6 Paper 1.4 1.2 1 5 Launsby Consulting 2002 5.8 6.6 7.4 8.2 Wing Length 9 M&M’s Simple to use Simple to define defect Useful for attributes R&R study and Ccharts You can sample (eat) some of the product Launsby Consulting 2002 Summary Please visit my website to download case studies and find listing of what is required for each experiment www.launsby.com Launsby Consulting 2002