自回归(AR)模型是平稳信号分析的重要工具,利用峭度最大原则确定AR模型最优阶次,然后利用此AR模型对滚动轴承故障信号进行预处理,剔除可线性预测的平稳成分,得到的残余分量中理论上只包含了噪声信号和信号的非平稳部分,从而降低了后期数据分析难度.谱峭度对于非平稳信号非常敏感,它可以将非平稳信号从噪声中检测出来,因此将两者结合起来可以更有效地对滚动轴承故障进行诊断,实验结果验证了此方法的有效性.%The auto-regressive ( AR) model was an important tool for stationary signal analysis. The optimum AR model order was determined with the Maximum Kurtosis Criterion, and then this AR model was used to pre-process fault signals obtained from a rolling element bearing. As a result, it eliminated the linearly predictable stationary part and achieved the residual component only containing noise and the non-stationary part of the signal. Consequently, the difficulty of the following signal analysis was eased. Spectral kurtosis(SK) was sensitive to non-stationary signals, it could extract the non-stationary part from a noisy signal. Here, the AR model and SK were combined and used to more effectively detect faults of rolling element bearings. The effectiveness of the proposed method was verified by the test results.
展开▼
