【摘要】 轮轨接触点位置是衡量轮轨接触状态、评价列车运行安全性的重要参数，其测量受到电桥信号中高次谐波的干扰。针对此问题，提出一种基于BiGRU神经网络模型的轮轨力与接触点位置测量方法。通过测力轮对有限元仿真分析，发现垂向力作用于轮对踏面不同横向位置时，辐板应变发生相应改变，据此可以对轮轨接触点位置进行检测。根据Fourier级数理论对电桥的组桥方案进行设计，使电桥信号以一次谐波为主，消除偶数次和奇数次中的3、5次谐波，并对解析法的计算原理展开分析。对BiGRU神经网络模型进行结构设计，以测量电桥信号为输入进行轮轨力与接触点位置的计算。结果表明，相比于方程解析法，本文方法可以消除电桥信号中高次谐波的干扰，有效提高轮轨力与接触点位置计算精度。更多还原

【Abstract】 The wheel-rail contact point location is a critical parameter for assessing the state of wheel-rail interaction and evaluating the operational safety of trains. Its measurement is influenced by the presence of higher-order harmonics in the Wheatstone bridge signal. To address this issue, this study introduces a measurement method for wheel-rail forces and contact point location based on the BiGRU neural network model. Finite element simulation analysis of the instrumented wheelset revealed that the strain in the spokes changes correspondingly when a vertical force is applied to different lateral positions on the wheel tread, enabling the detection of the wheel-rail contact point location. The Wheatstone bridge configuration is designed according to Fourier series theory, prioritizing the fundamental harmonic and eliminating the even order and 3rd, and 5th harmonics. The paper also analyzes the computational principles of the analytical method. The BiGRU neural network model is structured to compute the wheel-rail force and contact point location using the measured bridge signal as input. The findings indicate that the proposed method can mitigate the interference of higher-order harmonics in the bridge signal, significantly enhancing the precision of the wheel-rail force and contact point location calculations.更多还原