Software, Electronic and Mechanical Systems Laboratory
The increasing sophistication of methods and software tools for modeling and simulation enabled significant changes in the way engineers approach design problems. The progress in development of new simulation and analysis methods has resulted in a progression of standard practices for using the tools. Our practice has moved from discrete numerical experiments, to design optimization, to design optimization with sensitivity analysis. In this talk we describe a new standard practice for the the future: design for robust performance. This practice differs from optimal design in two significant ways. First one doesn't seek the optimal design; rather, one seeks a good enough design. Second, one seeks a design that is insensitive to variability in conditions that affect performance.
Following a brief discussion of general features of robust design, I will give an example of robust design of a diffractive optical element. The goal of this work was to provide a device to create a uniform intensity profile from a non-uniform laser beam. The added complication is that the shape of the laser beam varies randomly over time. A constructive approach to the design will be discussed as well as recent efforts to find a solution using optimization.