Computer experiments for circuit optimisation using symmetric Latin hypercube designs

Abstract

Computer experiments play an important role in a product development process. When the input-output relationship in a computer experiment is known through a transfer function equation, the experimental runs are computationally inexpensive to carry out. The objective of developing a meta-model in such situations is mainly optimisation as prediction is easily achieved using the transfer function. This is in contrast to computer experiments where experimental runs are expensive and the main objective of developing meta-models is prediction and/or optimisation. In this paper, we provide an optimisation procedure that uses meta-models developed from computer experiments using symmetric Latin hypercube designs. This approach is demonstrated using transfer function of a filter circuit.