Research on surrogate models and optimization of equipment dynamics for complex systems

In this paper, we use tracked vehicles as an example to study complex systems and surrogate model building method. Due to the high cost and long development cycle associated with the traditional development approach, involving static design and experimental verification, there is an urgent need to explore an effective dynamic optimization design method to enhance vibration reduction performance. This paper introduces a methodology for constructing a model of complex equipment surrogate, aimed at streamlining subsequent parameter optimization processes. The dynamics model of the tracked vehicle is utilized, and the root-mean-square value of plumb vertical acceleration at the center of mass is selected as the dynamic performance index. The impact of structure parameters on dynamic performance is analyzed, and a range of design parameters are determined. The optimal Latin hypercube experimental design method is employed to generate sample data. A radial basis function neural network, optimized by particle swarm optimization, is used to construct a dynamic surrogate model for the tracked vehicle. The surrogate model constructed in this paper greatly reduces the time cost in optimizing complex equipment parameters. In the past, parameter optimization required conducting numerous simulations to collect data, with each simulation taking about 700 s. However, the surrogate model introduced in this paper only requires similar to 100 s for completion. This significant reduction in time not only saves costs but also maintains a high level of accuracy.