The method for the optimal experiment design
Natalija Pozniak, Leonidas L. Sakalauskas · 2019
Purpose – is to develop the Bayesian method of optimal engineering design by a series of experiments, aiming to manage experimental resources in a rational economic way. Research methodology – is based on modelling of experimental data by Gaussian random fields (GRF) and using matri-ces of fractional Euclidean distances. Next, the P-algorithm for the planning of the experiment series is created in order to optimize the values of the response surface. Findings – the application of the developed method in engineering design enable us to create plans for the experiment se-ries in order to create new functional products and processes managing experimental resources in a rational economic way. Research limitations – the creation of the plans of the experiment series can require a large amount of computer time re-lated to the application of the Monte Carlo procedure in order to ensure the optimality of created plans. However, this limitation can be avoided using distributed computing tools. Practical implications – the created method helps engineers to seek solutions to experimental problems, considering the economic viability of each potential solution along with the technical aspects. Originality/Value – in creating functional products and processes engineers are using the experimental design process, which usually is highly iterative. The developed approach enables us to design the experimental series inflexible way, de-creasing the number of required experiments and avoiding of rather expensive methods such as factorial experiments, steepest descent, etc., usually applied for experimental design in engineering practice