Adaptive sequential experimentation based on revised simplex search

This paper presents an effective experimentation strategy for expensive industrial experiments. In these experiments, there is no prior knowledge about the behaviour of the system under testing and the cost of running experiments is high and the total testing budget is limited. Adaptive sequential experimentation strategy is needed which is able to explore and reach the best design space quickly. Our experimentation strategy is based on revised simplex search method that combines the advantages of simplex search method and adaptive one-factor-at-a-time method and utilises response surface modelling to predict the results of subsequent experiments. Six experimental datasets are used to test our new adaptive sequential experimentation strategy. The results show that our strategy is able to explore the neighbourhood of the best design space and reach the best experimental point significantly faster than the existing approach.