基于小波核最小二乘支持向量机的齿轮磨损预测

Gear Wear Prediction Based on the Least Square Wavelet Kernel Support Vector Machine

针对齿轮在磨损过程中的磨损程度,可以用振动信号来表征,并通过对磨损过程中振动信号的预测来实现磨损预测,提出了一种基于小波核的支持向量机磨损预测算法。首先,分析了最小二乘小波在磨损预测中建模方法,其核函数采用小波核,改善了系统非线性性能;然后用量子行为粒子群优化算法(QPSO)优化SVM参数,具有较快的搜索速度并保持了时间序列的特征。验证实验中用齿轮箱振动信号的统计指标表征齿轮磨损状态。实验结果表明,该预测方法能够有效地进行齿轮磨损预测。

Aimed at the wear process of gear wear, vibration signals were able used to characterize wear intensity, wear prediction was able achieved by the prediction of vibration signals, a wear prediction algorithm based on wavelet kernel support vector machine （ SVM） was proposed. Firstly, the least square wavelet modeling method in the wear prediction was analyzed, and wavelet kernel was used as the kernel function to improve the nonlinear performance of the system. Then the SVM parameters were optimized by the Quan-tum-behaved Particle Swarm Optimization （ QPSO） , and the system possessed faster searching speed and maintained the characteristics of time series. The statistical indicators of the vibration of gear box were used in validation experiment to characterize gear wear intensi-ty. The results of the experiment show that the prediction method can effectively predict gear wear.