时变参数模型及其在非平稳振动分析中的应用

TIME-VARYING AUTOREGRESSIVE MODEL AND ITS APPLICATION TO NONSTATIONARY VIBRATION SIGNAL ANALYSIS

对时变参数模型TVAR(Tim e-varying Autoregressive Model)进行了研究,并将其应用于转子实验台非平稳振动信号的分析。TVAR是模型参数随信号统计特性而变化的变参数AR(Autoregressive Model)模型,适用于分析非平稳信号。利用TVAR对调频仿真信号进行分析并与典型时频分析方法进行比较,结果表明TVAR具有时频分辨率高、无交叉干扰项以及对噪声不敏感等优点。基于TVAR分析了转子实验台正常及故障工况下连续变速过程中采集的振动信号,实验表明TVAR能够有效地分析非平稳振动信号,并具有较强的信号特征提取能力,为非平稳工况下转子故障诊断及模态分析等提供了一种有效的分析方法。

Time-varying autoregressive model(TVAR) is investigated and applied to analyze the signals collected from a rotation machine test rig under nonstationary conditions.TVAR is an improved autoregressive model with coefficients evolving with signal statistical characteristics.The performances in time-frequency analysis are compared between TVAR and some traditional methods by analyzing some frequency modulation(FM)signals.It is shown that TVAR has high resolutions,no cross terms and is insensitive to noises,etc.Using TVAR nonstationary signals collected in the continuously varying speed process under normal or fault states are analyzed.The results show that TVAR excels at disposing nonstationary signals and has a superior feature extracting ability;TVAR is an effective method for fault diagnosis and modal analysis under nonstationary conditions.