lecture12

# lecture12 - http/ocw.mit.edu MIT OpenCourseWare 2.830J...

This preview shows pages 1–15. Sign up to view the full content.

MIT OpenCourseWare ____________ http://ocw.mit.edu 2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303) Spring 2008 For information about citing these materials or our Terms of Use, visit: ________________ http://ocw.mit.edu/terms .

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 1 M anufacturing Control of Manufacturing Processes Subject 2.830/6.780/ESD.63 Spring 2008 Lecture #12 Full Factorial Models March 20, 2008
2.830J/6.780J/ESD.63J 2 M anufacturing Outline • Modelling “Effects” from Multiple Inputs • ANOVA on Effects • Linear and Quadratic Models • Model Coefficient Calculation – Regression (General Approach) – Contrasts (for Factorial Designs)

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 3 M anufacturing What Is the Effect? • What is the relationship between Hold Time and Dimension? I.M. Hold Time Change 2.025 2.030 2.035 2.040 2.045 2.050 2.055 2.060 0.5 1 1.5 2 2.5 Hold Time Dimension
2.830J/6.780J/ESD.63J 4 M anufacturing But Wait… There’s More! IM Run data 2.015 2.020 2.025 2.030 2.035 2.040 2.045 2.050 2.055 2.060 111111111111111222222222222222333333333333333444444444444444 Levels changes Velocity low high low high Hold time low low high high Do the two inputs interact? Dimension

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 5 M anufacturing Model Form • Linear • Quadratic • Exponential • General Polynomial? • Interactions What data needed to decide and/or estimate parameters of different model forms?
2.830J/6.780J/ESD.63J 6 M anufacturing Multiple Input/Treatment Models • In general k inputs – If 2 levels for each ….. 2 k combinations – If 3 levels for each …… 3 k combinations • Why use more than one input? – More than one output – Change mean and variance • Process Robustness • Optimization of Quality Loss

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 7 M anufacturing linear term interaction term i = input index k = total number of inputs mean higher order terms (model form error) A General Linear Model for k inputs η=β 0 + β i i = 1 k x i + ij x i i = 1 k j = 1 j < i k x j + h . o . t . + ε residual error
2.830J/6.780J/ESD.63J 8 M anufacturing Two Input Model η=β 0 + β 1 x 1 + β 2 x 2 + β 12 x 1 x 2 + h . o . t . + ε 4 coefficients to determine How many data points (factors, levels) are needed to uniquely identify?

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 9 M anufacturing Consider a One Input Case I.M. Hold Time Change 2.025 2.030 2.035 2.040 2.045 2.050 2.055 2.060 0.5 1 1.5 2 2.5 Hold Time Dimension
2.830J/6.780J/ESD.63J 10 M anufacturing Linear One Input Example Linear model η=β 0 + β 1 x η = 1 2 X = 1 x 1 x + β = 0 1 ε = 0 (for means) Assume 2 levels x , x + = X + ε Since X is square and ε= = X 1 With 1 trial at each level we get:

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 11 M anufacturing Linear Model with Replicates • Line will no longer intersect specific points • What is “best fit? I.M. Hold Time Change 2.025 2.030 2.035 2.040 2.045 2.050 2.055 2.060 0.5 1 1.5 2 2.5 Hold Time Dimension
2.830J/6.780J/ESD.63J 12 M anufacturing Minimum Error Line Fits • Define squared error for data for a given β o and 1 • Find o and 1 that lead to minimum of the sum of all e 2 • OR - Solve the matrix equation to get = ( X T X ) 1 X T η where X is a non-square matrix of all inputs for all replicates and η is the vector of all trial outputs e 2 = ( η− ( 0 − β 1 x )) 2

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2.830J/6.780J/ESD.63J 13 M anufacturing Aside η = X β + ε ε = η X squared error J = ε T = η X () T X ( ) The minimum value of J is then found by the vector partial derivative: J ∂β = 0 =− 2
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 09/24/2010 for the course MECHE 2.830J taught by Professor Davidhardt during the Spring '08 term at MIT.

### Page1 / 63

lecture12 - http/ocw.mit.edu MIT OpenCourseWare 2.830J...

This preview shows document pages 1 - 15. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online