基于支持向量机的综合传动装置磨损状态研究

Study on Wear State of Power-shift Steering Transmission System Based on Support Vector Machine

结合光谱油液分析技术,运用支持向量机对某履带车辆综合传动装置磨损状态进行研究。建立了一种多输出最小二乘支持向量回归算法,并将其应用到综合传动装置的光谱油液分析数据的预测研究中。采用交叉验证方法,讨论了回归算法中参数的选取问题,并将磨损元素预测值与试验值进行了对比分析。结果表明,该方法在较短里程(4000km以内)具有较高的准确率,可以用于综合传动装置磨损状态的预测研究。

The wear state of power-shift steering transmission （PSST） system was studied with support vector machine method combining spectrometric oil analysis technology.A method of multiple output least squares support vector regression was developed.The spectrometric oil analysis data were studied using multiple output least squares support vector regression.The selection of regression parameters was studied using cross validation method.The predictive values of wear elements were compared with the actual values.The result shows that this method has better prediction accuracy for wear elements in relatively short-distance miles （no more than 4 000 km）,it can be used for studying wear state and preventing faults of PSST system.