多维非平稳时间序列在机床主轴故障诊断中的应用

Application of Multidimensional Non-stationary Time Series on Fault Diagnosis of Spindle

采用ARIMA模型将数控机床主轴故障初期的非平稳时间序列转化成标准平稳时间序列,然后利用多维自回归(AR)模型进行数据处理与趋势预测,并分析了基于多维自回归序列参数估计的Yule-Walker算法以及FPE阶次判定准则。实测数据的计算结果表明:经稳态处理后的多维AR时序模型能够很好地拟合数控机床主轴故障模型,预测的精度符合要求。

The non-stationary time series, obtained during spindle running in ill conditions, was transformed into standard normal stationary time series using ARIMA model, and the AR model was used in data processing and forecasting. Yule-Walker algorithms and rank confirmation with final prediction error （FPE） were analyzed based on the multidimensional autoregressive parameter estimation. The experimental results show that stabilized multidimensional AR model can simulate spindle fault model, and meet the needs of forecasting.