Development of a parsimonious GA-NN ensemble model with a case study for Charpy impact energy prediction

摘要: A parsimonious genetic algorithm guided neural network ensemble modelling strategy is presented. Each candidate model to participate in the structurally selected using a algorithm. This provides an effective route improve performance of individual models as compared more traditional approaches, whereby structure through some trial-and-error methods or heuristics. The developed this paper highly efficient and requires very little extra computation for developing model, thus overcoming one major known obstacles model. key techniques behind implementation include formulation fitness function, generation qualified models, well specific definitions assemble strategies. case study presented which exploits complex industrial data set relating Charpy impact energy heat-treated steels, was provided by Tata Steel Europe. Modelling results show significant improvement over previously same set.