基于Mamdani模糊神经网络的相敏轨道电路故障诊断方法研究

摘要

Addressing the uncertainty and fuzziness of 25 Hz phase-sensitive track circuit faults, this paper proposes an improved track circuit fault diagnosis method based on Mamdani fuzzy neural network. The adaptive-momentum BP learning method is used to train and optimize model parameters with the derivation process and the setting of initial value of system parameters is discussed. Simulation results show that under the same experimental conditions, the improved method reduces the training error and effectively improves the stability and convergence speed of the learning process. Application of this method to 25 Hz phase-sensitive track circuit faults diagnosis is of high feasibility.%针对25 Hz相敏轨道电路故障的不确定性与模糊性，提出一种基于Mamdani模糊神经网络的轨道电路故障智能诊断改进方法。采用自适应动量BP学习调整法对模型参数进行训练优化，给出推导过程，并讨论系统参数初始值的设定。仿真实验表明，在相同实验条件下改进方法降低了训练误差，并有效地提高了诊断学习过程的稳定性与收敛速度，对25 Hz轨道电路故障进行智能模糊诊断是可行的。