基于DBN重载货车闸瓦磨耗量预测研究

Wear Volume Prediction of Brake Shoes Used in Heavy Freight Wagon Based on DBN

针对铁路货车闸瓦磨损严重导致闸瓦更换成本居高不下的问题,基于深度信念网络(DBN)拟合了铁路重载货车闸瓦磨损量的分布规律,实现了对铁路重载货车不同位置的闸瓦磨损厚度预测。其预测结果与基于支持向量回归机(SVR)结果进行对比,平均预测精度提高了6.2%达到91.8%。当运行里程数较少时(2万公里、4万公里、8万公里),闸瓦的磨损厚度受磨合时间、闸瓦安装等因素影响相对严重,随着运行里程数的增加预测精度显著增加,当里程数为10万公里、12万公里以及14万公里时预测精度分别为93.9%,91.5%,95.4%。深度学习算法的应用提高了智能性和准确性,对闸瓦磨损厚度的预测可为闸瓦更换提供参考依据,适当延长闸瓦服役寿命,减少闸瓦更换的经济成本。

In view of the high cost of replacing brake shoes in heavy freight cars caused by serious wear,the wear thickness of brake shoes is predicted at different positions of heavy freight cars based on the wear distribution of brake shoes fitted by Deep Belief Network(DBN).The prediction results are compared with that by support vector regression machine(SVR),which shows the average accuracy is improved by 6.2%to 91.8%.When the running mileage is less(20,000 km,40,000 km,80,000 km),the wear thickness of brake shoes is relatively severely affected by running-in time,installation and other factors.The prediction accuracy increases significantly with the increase of running mileage.When the mileage is 100,000 km,120,000 km and 140,000 km,the prediction accuracy is 93.9%,91.5%and 95.4%respectively.The application of a deep learning algorithm improves intelligence and accuracy.The predicted wear thickness can be an important reference to change brake shoes,which service lives can be extended.That will reduce the cost of replacing brake shoes.