Stuart Kozola, Technical Marketing
Ascension Vizinho-Coutry, Application Engineering Group
LMCS 2008
Journe nationale pour la modlisation et la simulation 0D/1D
17th April 2008
2008 The MathWorks, Inc.
Reliability Analysis and Robust Design
Using MATLAB Products
Agenda
Overview of reliability analysis and robust design
Overview of MATLAB
Demonstration
Summary of products shown
Additional resources
Question and answer
2
Benefits of Reliability Analysis and
Robust Design
Gain more knowledge about design
Understand failure rates / mechanisms
Sensitivity to changes in design parameters
Capture variation in design process and operation
Better designs
Improved performance
Reduced variability
Improved quality
Reduce development cost/effort
Fewer prototypes
Faster design iterations
Shorter development cycle
3
What is a Reliable and Robust Design?
Requirement
Performance Measure, f(x)
Reliability probability of
satisfying a requirement
Probability of
failure, 1-p
Probability of
success, p
Optimal
Not reliable
Not robust
Robustness ability to
handle variation without loss
of performance
Requirement
Initial
(mean target)
Design Requirement
f(x)
Robust
Design
Not optimal
Reliable
Robust
f(x)
xo xr
Design Variable, x
Engineering Design Workflow
Modify Design
Parameters
Initial
Design
Parameters
NO
F(X)
Model or
Prototype
Objectives
Achieved?
YES
Final
Design
OPTIMIZATION PROCESS
Design process can be performed:
Optimization benefits include:
Manually (trial-and-error or iteratively)
Finding better (optimal) designs
Automatically using optimization techniques
Speeding up design evaluation
F(X) = Goal or min/max F(x)
By changing X
Subject to Design Requirements
Reducing labor/effort
Finding non-intuitive designs
MATLAB Provides the Foundation for
Engineering Analyses
The leading environment for
technical computing
Customizable
Numeric computation
Data analysis and visualization
The de facto industry-standard,
high-level programming language
for algorithm development
Toolboxes for statistics, optimization,
symbolic math, signal and image
processing, and other areas
Foundation of the
MathWorks product family
Demo: Automotive Suspension System
Design
Task: To design the suspension system for a new luxury car model
Solution Approach:
Model
Traditional
Approach
Reliability
and
Robust Design
Design Optimization Traditional
Approach
Minimize passenger vertical and rotational acceleration
Z&&,&&
Design Variables:
front/rear springs
front/rear shock absorbers
kf, kr
cf, cr
Constraints:
Level car
Available parts
Low natural frequency
Required damping ratio
Modify
[kf, kr, cf, cr]
Initial
Design
Parameters
[kf,kr,cf,cr]
NO
Z&&, &&
[kf kr cf cr]
Car mass = constant
Minimum
Found?
YES
Optimal
Design
Design Optimization Reliable and
Robust Design Approach
Minimize passenger vertical and rotational acceleration
Z&&,&&
Design Variables:
front/rear springs
front/rear shock absorbers
kf, kr
cf, cr
Constraints:
Level car
Available parts
Low natural frequency
Required damping ratio range
Reliability constraint on strut
Uncertainty in vehicle mass
Modify
[kf, kr, cf, cr]
Initial
Design
Parameters
[kf,kr,cf,cr]
NO
Z&&, &&
[kf kr cf cr]
Car mass =
Minimum
Found?
YES
Optimal
Design
Distribution
Optimization Toolbox
Graphical user interface and
command line functions for:
Linear and nonlinear programming
Quadratic programming
Nonlinear least squares and nonlinear
equations
Multi-objective optimization
Binary integer programming
Customizable algorithm options
Standard and large-scale
algorithms
Output diagnostics
10
Our example
function mycost = myCostFcn(x,simParms)
%% Extract suspension variables (called by Simulink Model)
kf = x(1); cf = x(2);
kr = x(3); cr = x(4);
struct2var(simParms);
%% Run Simulink Model
simTime = [0 8];
sim('mldemo_suspnfast.mdl',simTime); % returns Zdotdot, thetadotdot
%% Compute total acceleration
rf = 0.5*Lf;
% location of front passengers (m)
rr = 0.9*Lr;
% location of rear passengers (m)
totalAccel = (Zdotdot + rf * thetadotdot).^2 + ...
(Zdotdot - rr * thetadotdot).^2;
%% Return a single (scalar) value
mycost = sum(totalAccel);
function [c,ceq] = mynonlcon(x,simParms)
%% Extract suspension variables
kf = x(1); cf = x(2);
kr = x(3); cr = x(4);
struct2var(simParms);
%% Define desired damping ration range
cupper = 0.5;
% upper limit for damping coefficients
clower = 0.3;
% lower limit for damping coefficients
%% Define mass distribution on tires
Mf = Mb*Lr/(Lf+Lr)/2;
Mr = Mb*Lf/(Lf+Lr)/2;
% Inequality constraints c <= 0
c = [sqrt(kf/Mf)/(2*pi)-2;...
sqrt(kr/Mr)/(2*pi)-2;...
cf/(2*sqrt(kf*Mf))-cupper;...
front
clower-cf/(2*sqrt(xkf*Mf));...
cr/(2*sqrt(kr*Mf))-cupper;...
clower-cr/(2*sqrt(kr*Mf))];
% Equality constraints ceq = 0
% fn <= 2 Hz for front
% fn <= 2 Hz for rear
% damping ratio for
% damping ratio for rear
11
Genetic Algorithm and Direct Search Toolbox
Graphical user interface and
command line functions for:
Genetic algorithm solver
Single objective
Multi-objective with Pareto
front
Direct search solver
Simulated annealing solver
Useful for problems not easily
addressed with Optimization
Toolbox:
Discontinuous
Highly nonlinear
Stochastic
Discrete or custom data types
Undefined derivatives
12
Statistics Toolbox
Statistics Toolbox provides interactive
and command line tools for:
Data collection and management
Descriptive statistics
Multivariate statistics
Probability distribution fitting and
modeling
Hypothesis testing
Analysis of variance/covariance
Linear and nonlinear modeling
Visualization
Statistical Process Control
Design of Experiments
13
Simulink for Modeling and Simulation
Block-diagram environment for modeling,
simulating, and analyzing dynamic
systems
Add-on tools extend functionality
Application specific (e.g. Simulink
Control Design, Signal Processing
Blockset)
Physical modeling (e.g. SimMechanics,
SimPowerSystems, SimHydraulics)
And more
Simulink specific optimization tools
Simulink Parameter Estimation
Simulink Response Optimization
14
Summary
Models are not perfect and operate under uncertainty
Accounting for uncertainty in upfront design process
improves design performance and quality
MATLAB , Statistics, and Optimization products
provide tools needed for reliability analysis and robust
design
Capture and model uncertainty
Find optimal and robust design
15
Links to Digest Articles of Interest
for Reliable/Robust Design
- Designing for Reliability and
Robustness
- Improving an Engine Cooling Fan
Using Design for Six Sigma
Techniques
- Using Statistics to Analyze
Uncertainty in System Models
16
Contact Information
Ascension Vizinho-Coutry
avizinho@mathworks.fr
Tl: 01 41 14 88 30
Christophe Bouveur
Christophe.Bouveur@mathworks.fr
Tl: 01 41 14 87 79
Loic Noury
Loic.Noury@mathworks.fr
Tl: 01 41 14 87 85
17
2008 The MathWorks, Inc.
Questions?
