Synthesis of experimental designs of maneuvering captive-model tests with a large number of factors

Abstract

An approach to synthesizing D-optimized experimental designs for an arbitrary number of factors was developed and tested on a third-order polynomial regression model with 5–8 factors. Three options were envisaged for the internal optimization procedure: an exhaustive search, a quasirandom search with the help of the Sobol sequences, and a genetic algorithm. The calculations performed have shown the pronounced superiority of the variant involving a genetic algorithm. Captive-model tests with a catamaran model with varying Froude number, drift angle, rate of yaw, sinkage, trim, and heel are presented as an example of the practical synthesis of the experimental design. The linear regression model constructed is a third-order 5-factor polynomial with respect to all factors except the Froude number. The influence of the latter is accounted for by representing the polynomial’s regression coefficients as functions of the Froude number represented as a truncated Fourier series with a linear term added.