Adaptive Elliptic Basis Function Model Construction Strategy Based on Particle Swarm Optimization

To address the problems of poor generality and difficulty in obtaining opti-mal parameter combinations for existing surrogate models, this paper pro-poses a construction strategy for an adaptive Elliptic Basis Function neural network model based on particle swarm optimization (PSO-EBF). The K-means clustering algorithm is introduced in the EBF model to improve the efficiency and diversity of model construction. On this basis, the PSO algorithm is introduced to take the EBF model parameters: width coefficient, number of kernel functions, smoothing coefficient, etc. as design variables, and take the model determination coefficient \({R}^{2}\) as the optimization objective, and automatically search for the optimal parameter combination of the model. The numerical experimental results show that the PSO-EBF model constructed in this paper can effectively improve the efficiency of surrogate model construction. Compared with ISight commercial software, the accuracy of the PSO-EBF model is higher than that of the traditional EBF model, which proves that this adaptive model can effectively solve the complexity and blindness of surrogate model construction process and improve the efficiency of model construction.