Plan for the Session
Guest lecture by Eric Feron (1 hour)
Quiz on Design of Dynamic Systems (15
minutes)
Review of Reliability Improvement Case
Study -- Router Bit Life (35 minutes)
16.881 MIT
Learning Objectives
Introduce some basics of reliability
engineering
Relate reliability to robust design
Practice some advanced construction
techniques for orthogonal arrays
Introduce analysis of ordered categorical
data
Practice interpreting data from robust
design case studies
MIT 16.881
Reliability Terminology
Reliability function R(t) -- The probability
that a product will continue to meet its
specifications over a time interval
Mean Time to Failure MTTF-- The average
time Tbefore a unit fails
MTTF =
R(t)dt
Instantaneous failure rate (t)
0
(t) =Pr(System survives to t+dtSystem survives to t)
t
R(t) =e
0
( )d
16.881 MIT
Typical Behavior
Early failure period often removed by burn-in
Wear out period sometimes avoided by retirement
What will the reliability curve R(t) look like if
early failure and wear out are avoided?
Early
Wear
(t)
failure
out
Useful life
16.881
t
MIT
|
Weibull Distributions
Common in component failure probabilities
(e.g., ultimate strength of a test specimen)
Limit of the minimum of a set of
independent random variables
s
|tt
o
|
|
R(t) =e
\
.
16.881 MIT
System Reliability
Components in series (system fails when
anysubsystem fails)
R
SYST
=
R
i
i
Components in parallel (system fails only
when allsubsystems fail)
F
SYST
=
F
i
i
16.881 MIT
Router Bit Life Case Study
Printed wiring boards cut to size by a
routing operation
Router bit rotated by a spindle
Machine feeds spindle in X-Y plane
16.881 MIT
Taylor Tool Life Equation
VT
n
=C
V- cutting speed
T- time to develop a specified amount of
flank wear
Cand n- experimentally determined
constants for cutter material / workpiece
combinations
n=0.08-0.2 High speed steel
n=0.2-0.5 Carbide
16.881 MIT
Noise factors
Out of center rotation of spindle
Router bit properties
PWB material properties
Spindle speed
16.881 MIT
Noise Factor in the Inner Array
When can this be done?
Why can this work?
What are the advantages?
16.881 MIT
Quality Characteristic
We want to maximize useful life of the bit
Phadke chose to measure inches cut prior
to excessive dust production
Does this meet the guidelines for additivity?
What other choices would you suggest?
16.881 MIT
Control Factors
Suction (1 in Hg, 2 in Hg)
XY feed rate (60/min, 80/min)
In-feed (10/min, 50 in/min)
Bit type (four types)
Suction foot (solid ring, bristle brush)
Stacking height (3/16, 1/4)
Backup slot depth (0.06, 0.1)
Spindle speed (30K rpm, 40K rpm)
Do these meet the definition of a control factor?
16.881 MIT
Reqts of the Matrix Experiment
Two 4 level factors
Seven 2 level factors
Interactions
XY feed X speed
In-feed X speed
Stack height X speed
XY feed X stack height
Compute required DOF
16.881 MIT
Linear Graphs and Experimental
Design
Draw the required linear graph
Compare with standard linear graphs
Modify the standard linear graph to suit
using modification rules
Breaking a line
a c
b
a c
b
a c
b a c
b
Forming a line
16.881 MIT
Speed Assigned as an Outer Factor
Guarantees that all interactions with all
other factors can be computed
Doubles the experimental effort (for a two
level outer factor)
16.881 MIT
S/N for Fraction Defective
Average quality
L(y)
loss in passing
products
Q=k
p
1p
k
p
1-p
Signal to noise
m-
m
m+
y
=10log
10
1
p
p
(
(
"goal post" loss function
16.881 MIT
Ordered Categorical Data
The response of the system is summarized by
membership in discrete categories
The categories have an unambiguous ordering
For example, survey data often takes the form:
Poor
Fair
Good
Excellent
16.881 MIT
Router Bit Life Conclusions
Two fold increase in bit life realized while
maintaining throughput
Robust design can be used to improve life
Life as larger the better
Probability of failure as ordered categorical
Noise can be induced in inner array
A control factor can be used in an outer
array to study interactions
16.881 MIT
Next Steps
Homework #8 due 8 July
Next session Wednesday 8 July 1:05-3:55
Quiz on reliability and robust design
Mid deck active control experiment case study
Last on-campus session 9 July 1:05-3:55 -
Review session for final exam
Each student may resubmit up to three
quizzes and/or home works by Monday 13
July (grades will be averaged with the
original grades)
16.881 MIT