Global stochastic comprehensive sensitivity analysis based on robustness grouping and improved Pelican algorithm-optimized radial basis function neural network

Abstract Modeling analysis is one of the important means to analyze practical engineering, and as technology continues to evolve, various models are getting closer and closer to reality, while at the same time, there are more and more parameters in the models. It is important to analyze the impact of these parameters on the project to assist engineers in making plans or decisions. Sensitivity analysis (SA) can describe the effect of changes in these parameters on the model. However, complex models often have dozens or even hundreds of parameters, and most current SA methods struggle to deal reliably and effectively with these high-dimensional problems. In addition, it is difficult to obtain the sensitivity of continuous points in the parameter space with traditional SA methods. Therefore, this paper proposes a method that combines adaptive grouping and an improved pelican optimization algorithm for an optimal radial basis function (IPOA-RBF) agent model to solve these problems. Firstly, a clustering grouping method considering grouping robustness is established to obtain objective and stable parameter grouping results in high-dimensional SA. Secondly, a proxy model based on radial basis function neural network and an IPOA are proposed to capture the logic of the proxy model to obtain the parameter sensitivity of continuous points in the parameter space. Finally, the superiority and applicability of this method is verified using an arch dam simulation model.