Multi-objective Optimization Of Wheel Cross-Sectional Shapes Based On Surrogate Model

Urban rail transit brings convenience to people’s lives, but also brings vibration and noise pollution problems which can’t be ignored in the urban environment. And the wheel noise is the main part of the vibration and noise, so, reducing the wheel vibration and noise is an effective measure to reduce urban rail vehicle noise pollution. In this paper, we take a subway S-shaped spokes wheel as study object, introducing the NURBS curve to establish wheel cross-sectional shapes parametric geometric model, taking the vibration level difference as the evaluation index of reducing wheel vibration and noise, conducting multi-objective optimization of wheel cross-sectional shapes on the base of the surrogate model, getting the corresponding optimization program, and conducting the verification of noise and intensity. The main research contents of this paper are are as follows:(1) The wheel cross-sectional shape parametric geometric model is established according to the Ministry of Railways standard TB453-85, using NURBS spline principle. The wheel cross-sectional shape is changed by modifying control vertices and weight factors. This thesis used MAT LAB to program, generating wheel cross-sectional shapes.(2) This research takes the weight factors of the wheel’s web and the tread、hub connections fillet as design variables, and takes wheel modal、quality、vibration level difference as a response, using Optimal Latin Hypercube experimental design method to establish the surrogate model of wheels moda、quality、vibration level difference, then conducts the error analysis of the surrogate model.(3) This paper takes the weight factors of the wheel’s web and the tread、hub connections fillet as design variables, the wheel’s first8natural frequencies as constraints, the wheel’s quality、vibration level difference as objective function and establishes the mathematical model of multi-objective optimization.(4) On the basis of the Surrogate model, this paper uses traditional optimization method-Linear Weighting Method of preference-based search after the first decision to optimize, and gets the optimization program to verify the noise and intensity.(5) On the basis of the Surrogate model, this paper uses multi-objective genetic algorithm NSGA-Ⅱ of non-preference-based decision after the first search to optimize to get Pareto optimal solution set, and selects the appropriate program according to the preference to verify the noise and intensity. The results show that optimization program has played a significant effect of reducing vibration and noise. The maximum amplitude of the noise reduction is12dB, and the wheel’s strength meets the requirements.