基于支持向量机的机械设备状态趋势预测研究

Research on Condition Trend Prediction of Mechanical Equipment Based on Support Vector Machine

提出了用支持向量机对机械设备状态趋势进行预测的新方法,构造了相应的支持向量回归机,并分别用仿真数据和实际数据对其性能进行了验证.将该支持向量回归机应用于某机组振动信号的预测,采用径向基核函数和合适的参数,使该向量回归机对振动量峰峰值的单步预测误差小于2%,24步预测误差小于5%,表明该算法对机械设备的运行状态趋势具有较好的预测能力.

A new method of condition trend prediction of mechanical equipment based on support vector machine was presented and the support vector regression machine was constructed. Both simulation data and actual data were used to validate the performance of this regression machine. The support vector regression machine was applied to the trend prediction of the vibration signal from machine sets. The single-step prediction error for peak-peak value of the vibration signal is less than 2% and the 24 steps prediction error is less that 5% with radial basis function (RBF) kernel function and proper parameters. These results show that the support vector regression machine has excellent performance of condition trend prediction for mechanical equipment.