The wheel tread wear of heavy haul freight car in operation leads to shortened wheel turning period, reduced operation life, and poor train operation performance. In addition, wheel rail wear is a complex non-linear problem that integrates multiple disciplines. Thus, using a single physical or mathematical model to accurately describe and predict it is difficult. How to establish a model that could accurately predict wheel tread wear is an urgent problem and challenge that needs to be solved. In this paper, a tread wear prediction and optimization method based on chaotic quantum particle swarm optimization (CQPSO)-optimized derived extreme learning machine (DELM), namely CQPSO-DELM, is proposed to overcome this problem. First, an extreme learning machine model with derivative characteristics is proposed (DELM). Next, the chaos algorithm is introduced into the quantum particle swarm optimization algorithm to optimize the parameters of DELM. Then, through the CQPSO-DELM prediction model, the vehicle dynamics model simulates the maximum wheel tread wear under different test parameters to train and predict. Results show that the error performance index of the CQPSO-DELM prediction model is smaller than that of other algorithms. Thus, it could better reflect the influence of different parameters on the value of wheel tread wear. CQPSO is used to optimize the tread coordinates to obtain a wheel profile with low wear. The optimized wheel profile is fitted and reconstructed by the cubic non-uniform rational B-spline (NURBS) theory, and the optimized wear value of the tread is compared with the original wear value. The optimized wear value is less than the original wear value, thus verifying the effectiveness of the optimization model. The CQPSO-DELM model proposed in this paper could predict the wear value of different working conditions and tree shapes and solve the problem that different operating conditions and complex environment could have a considerable effect on the prediction of tread wear value. The optimization of wheel tread and the wear prediction of different tread shapes are realized from the angle of artificial intelligence for the first time.

运营中的重载货车车轮踏面磨损，导致车轮转动周期缩短，使用寿命降低，列车运行性能不佳。此外，轮轨磨损是一个综合多学科的复杂非线性问题。因此，使用单一的物理或数学模型来准确地描述和预测它是困难的。如何建立一个能够准确预测车轮胎面磨损的模型是一个亟待解决的问题和挑战。在本文中，提出了一种基于混沌量子粒子群优化（CQPSO）优化推导极限学习机（DELM）的胎面磨损预测和优化方法，即CQPSO-DELM，以克服该问题。首先，提出了一种具有衍生特征的极限学习机模型（DELM）。下一个，在量子粒子群优化算法中引入混沌算法来优化DELM的参数。然后，通过CQPSO-DELM预测模型，车辆动力学模型模拟不同测试参数下的最大车轮花纹磨损进行训练和预测。结果表明，CQPSO-DELM预测模型的误差性能指标小于其他算法。因此，它可以更好地反映不同参数对车轮踏面磨损值的影响。CQPSO 用于优化胎面坐标以获得低磨损的车轮轮廓。通过三次非均匀有理B样条(NURBS)理论拟合和重构优化后的轮廓，并将优化后的胎面磨损值与原始磨损值进行比较。优化后的磨损值小于原始磨损值，从而验证了优化模型的有效性。本文提出的CQPSO-DELM模型可以预测不同工况和树形的磨损值，解决了不同工况和复杂环境对胎面磨损值预测影响较大的问题。首次从人工智能的角度实现了轮毂胎面的优化和不同胎面形状的磨损预测。本文提出的CQPSO-DELM模型可以预测不同工况和树形的磨损值，解决了不同工况和复杂环境对胎面磨损值预测影响较大的问题。首次从人工智能的角度实现了轮毂胎面的优化和不同胎面形状的磨损预测。本文提出的CQPSO-DELM模型可以预测不同工况和树形的磨损值，解决了不同工况和复杂环境对胎面磨损值预测影响较大的问题。首次从人工智能的角度实现了轮毂胎面的优化和不同胎面形状的磨损预测。