Deep-Learning-Coupled Numerical Optimization Method for Designing Geometric Structure and Insertion-Withdrawal Force of Press-Fit Connector

Abstract

The press-fit connector is a typical plug-and-play solderless connection, and it is widely used in signal transmission in fields such as communication and automotive devices. This paper focuses on inverse designing and optimization of geometric structure, as well as insertion-withdrawal forces of press-fit connector using artificial neural network (ANN)-assisted optimization method. The ANN model is established to approximate the relationship between geometric parameters and insertion-withdrawal forces, of which hyper-parameters of neural network are optimized to improve model performance. Two numerical methods are proposed for inverse designing structural parameters (Model-I) and multi-objective optimization of insertion-withdrawal forces (Model-II) of press-fit connector. In Model-I, a method for inverse designing structure parameters is established, of which an ANN model is coupled with single-objective optimization algorithm. The objective function is established, the inverse problem is solved, and effectiveness is verified. In Model-II, a multi-objective optimization method is proposed, of which an ANN model is coupled with genetic algorithm. The Pareto solution sets of insertion-withdrawal forces are obtained, and results are analyzed. The established ANN-coupled numerical optimization methods are beneficial for improving the design efficiency, and enhancing the connection reliability of the press-fit connector.