Surrogates and Approximations in Engineering Design
Surrogates and Approximations in Engineering Design.
In the context of engineering design, often the functional objective and the design constraints are approximated by connecting the design variables and the responses of interest at few points in the design space. Since these are approximations of the original functions, they are called surrogates and are widely used in design studies. This course will focus on introducing such surrogates – how to build, evaluate and use them in design. Surrogates discussed will include polynomial regression, kriging, and radial basis function while Design of Experiments discussions will include Latin hypercube sampling and Hammersley sequence.
What Will I Learn?
- -Introduction, physical versus computational experiments, introduction to engineering optimization, need for surrogates in optimization
- -Sampling plans, Latin squares, Latin hypercubes sampling, stratification, Orthogonal arrays, Hammersley sequences
- -Surrogates: Polynomial Regression, Radial basis function, Kriging
- -Using surrogates in design space exploration and exploitation, infill criteria, adaptive sampling