广义频率响应函数的鲁棒建模新方法

针对基于非线性谱分析的故障诊断技术的需要,研究了非线性系统的广义频率响应函数的鲁棒在线建模方 法。首先将次元分析算法应用于Volterra级数模型的辨识,并通过对此算法的修正,得到了一种改进的Volterra系统的鲁 棒辨识算法。对其鲁棒性能进行的分析表明,改进算法的鲁棒抗噪性能被明显提高。在此基础上,提出了一种广义频率响 应函数的两步鲁棒建模新方法。仿真实验表明,该方法具有良好的鲁棒收敛性能,能够有效解决GFRF的在线建模问题。

New Robust Modeling Method of Generalized Frequency Response Functions

In allusion to the demand of the fault diagnosis technique based on nonlinear spectral analysis, robust online modeling method for generalized frequency response functions of nonlinear system is studied in this paper. Firstiy, the Volterra system can be identified by using the minor component analysis algorithm, then by modifying this algorithm, a modifying robust identification algorithm for Volterra system is obtained. The robust performance analysis for this modifying algorithm indicates that the algorithm's robust performance can be markedly improved. And then, a new two-step robust modeling method for generalized frequency response functions is presented. Simulation results indicate that the presented method can has well convergence performance, and it can realize effectively GFRF's online modeling.