Process Improvement

1
Process Improvement

• (Continued)

2
Introduction to Design of Experiments (DOE)

• Quickly optimize processes
• Reduce development time
• Reduce manufacturing costs
• Reduce scrap and rework
• Increase throughput
• Improve product quality
• Make products/processes more robust
• Reduce need for control charting

3
Definition of DOE

• Experimental designs are specific collections of
  trials run so the information content about a
  multi-variable process is maximized. With
  response-surface experimental designs, the goal
  is to put this information into a picture of the
  process.
• J. Stuart Hunter -

4
Important Contributions From Different Approaches
5
Process Knowledge

• If what we know about our processes cant be
  expresses in numbers, we dont know much about
  them.
• If we dont know much about them, we cant
  control them.
• If we cant control them, we cant compete.
• Motorola University -

6
Old Philosophy of Quality

• Quality is based on conformance to specifications

Loss due to scrap rework
Loss due to scrap rework
LSL
USL
7
New Philosophy of Quality
L3
L3 gt L2 gt L1
L2
L1
LSL
USL
Target
8
Taguchis Concept

• DESIGN quality into the product and process.
• Design the PRODUCT to be least sensitive to
  variations rather than trying to control the
  factors.
• Design the product so that its performance
  parameters are CLOSEST TO THE TARGET.
• Minimize costs within quality constraints rather
  than maximize quality within cost constraints.

9
Quality Effort by Activity
Development
Design
Manufacturing
Solve Problems
10
Taguchis Quadratic Loss Function
L0
LSL
USL
Target
L1 k(y1 - T)2
11
Example

• Let V(out) 115 Vdc
• y V(out) m 115Vdc
• LD(50) 115 /- 20 Vdc (Consumers Tolerance)
• Repair Cost 100
• L(y) k(y - 115) 2
• k L(y)/(y - m) 2 100/20 2
• k 0.25
• If V(out) 110 Vdc
• L(110) 0.25(110 - 115)2 6.25

12
Example
Suppose adjustment costs 2.00, i.e., the cost to
rework. When should a unit be reworked? L(y)
0.25 (y - 115)2 2.00 0.25 (y - 115)2 8
(y - 115)2 y 115 /- 8 0.5 y 115 /-
2.83
13
Basis of The Taguchi Method

• System Design
• Parameter Design
• Tolerance Design

14
What is Designed Experiments?

• Purposeful changes of the inputs (factors) to a
  process in order to observe corresponding changes
  in the output (responses).

15
Strategies for Experimentation

• Screening
• Modeling (Characterization)
• Sensitivity
• Optimization
• Robust (parameter) Design
• Tolerance Design

16
Objectives of an Experimental Design

• Obtain maximum information using minimum
  resources.
• Determine which factors shift average response,
  which shift variability, which have no effect.
• Find factor settings that optimize the response
  and minimize the cost.
• Build empirical models relating the response of
  interest to input factors

17
Methods of Experimentation

• Full Factorials
• Fractional Factorials
• Plackett-Burman
• Latin Square
• Hadamard Matrices
• Foldover Designs
• Box-Behnken Designs
• D-Optimal Designs
• Taguchi Designs

18
Full Factorial Experiments

• Advantages
• Tests all factors at all levels
• Evaluates all main effects
• Evaluates all interactions

19
Full Factorial Experiments

• Disadvantages
• Large number of runs
• Large number of samples
• Takes long time to run
• Expensive

20
Fractional Factorial Experiments

• Advantages
• Fewer runs
• Faster to complete
• Fewer Samples
• Less costly

21
Fractional Factorial Experiments

• Disadvantages
• Cannot test all factors at all levels
• Cannot evaluate all main effects
• Cannot evaluates all interactions

22
Factorial Versus Taguchi
23
Taguchi Designs

• Orthogonal Arrays
• Screening Designs
• Robust Designs
• Minimal Runs

24
Layout for Taguchi L-8
25
Example