【摘要】 为了寻求一种更加有效的列车车轮扁疤故障分析算法,提出一种通过轮轨噪声来确定车轮扁疤严重程度的检测方法。该方法将遗传算法与小波神经网络相结合,同时为了避免出现局部极小值,加速学习速度,在小波神经网络中增加了动量模型;在搜寻小波神经网络隐含层链接权值之前,使用遗传算法进行计算以优化小波神经网络结构;硬件只需2组麦克风阵列以及2个速度感应器就可以提供实时结果,成本远低于我国现有的检测方法。对不同列车车速下的轮轨信号进行了实时测试,结果表明:与传统神经网络、小波神经网络和遗传算法相比,该方法的检测准确率最多分别提高了16%、11%和3%,并且收敛最快。更多还原

【Abstract】 A novel strategy is proposed to provide a more effective wheel flat scar fault detection algorithm by means of wheel/rail noise.In this strategy,genetic algorithm is combined with a wavelet neural network,and a momentum model is added into the genetic wavelet neural network to avoid the local minimum and to accelerate the learning speed.Before searching the hidden-layer weights of the network,the structure of the network is optimized by genetic algorithm.This strategy requires only two groups of microphone arrays and two speed sensors for real-time measurements.Consequently,the cost is much lower than that of the existing detection methods in China.The proposed strategy has been applied to the real-time detection of train wheel/rail signals at different speeds.Numerical results reveal that the proposed strategy is of the fastest convergence,and its detection accuracy increases at most by 16%,10%,and 3%,respectively,compared with the conventional neural network,wavelet neural network,and genetic algorithm.更多还原