Abstract:
Xiao Tang
and S. Krishnamurty, “Performance Estimation and Robust Design Decisions,”Design
EngineeringTechnical Conference ’2000 ,DAC-14239
Decision based engineering
design research community addresses the notion of design as a decision
making process. The quintessence of the decision based design approaches
is to first model the design situation that results in a mapping from the
design option space to the performance attribute space, and then to obtain
the designer’s preference information regarding the tradeoffs among the
attributes in order to construct a system utility function on the attributes
representing the overall desirability of any combination of alternatives.
It can be argued from this perspective that the essence of design decision-making
lies in understanding the impact of uncertainty on utility. This paper
deals with two major issues critical to the development and implementation
of a decision-based robust design, namely, representation of design performance
under conditions of uncertainty and the development of a robust design
decision model.
In this paper, a reliability-based robust optimal design decision model
is presented to handle a primary source of uncertainty associated with
performance estimation, namely, the consequence of the unpredictability
of the model parameters or variables. Specifically, this paper presents
a computationally efficient procedure for accurate estimation of performance
variance using a novel Surround Point Method (SPM) and discusses its incorporation
into a decision-based robust design framework. Using Monte Carlo simulation
(MCS) results in constructing an optimal design of outer array in a design
of experiments (DOE) setup, this approach aims to mimic MCS-like results
throughout the design space without requiring the need to run computationally
intensive simulations at each design instance. Its application in design
optimization eliminates the need for surrogate response surface models,
which can be lead to significant errors in variance estimation. Preliminary
results appear to indicate that the SPM-based uncertainty estimation method
can offer the best promise in achieving an optimal balance between computational
complexity and design-scenario independence. As such, it can be expected
to be a viable and applicable probability estimation tool in generic engineering
design, and particularly useful in highly nonlinear configuration design
with many design variables. Towards its implementation in a design scenario,
this paper explicitly introduces the concept of robustness criteria and
the development of quality utility for use in design decisions. These issues
are discussed in the context of engineering design decision-making with
the aid of two case studies and the results are discussed.
Keywords: Engineering
Design, Robust Design, Design of Experiments, Expected Utility Theory