基于自适应FMD的滚动轴承复合故障诊断

Composite Fault Diagnosis of Rolling Bearings Based on Adaptive FMD

摘要:

针对滚动轴承复合故障信号具有非平稳、非线性且伴有背景噪声的特性,导致故障特征提取困难,难以实现诊断的问题,提出一种基于自适应特征模态分解(FMD)的滚动轴承复合故障诊断方法；为了克服传统FMD算法需要人工设置关键参数的问题,提出自适应FMD:通过改进传统算法的截止条件,计算算法运行过程中得到的每一组模态的平均相关峭度(CK),选取平均CK最大的模态组,用来自适应得到参数组合滤波器个数K和滤波器长度L下的最优模态个数n；结合网格搜索法搜索K和L,实现自适应FMD关键参数[K,L,n]的选取；依据时频域相关系数与峭度准则对分解后的信号进行筛选重构；对重构信号进行包络解调,提取故障特征频率,完成故障诊断；分别使用仿真信号和帕德博恩大学轴承数据集对所提的方法进行验证,实验结果表明,所提的方法可以有效提取出滚动轴承复合故障特征,在噪声环境下可以很好的完成诊断任务,且相比于其他方法拥有更好的故障特征提取能力。

Abstract:

To address the difficulties in fault feature extraction and diagnosis caused by the nonstationary, nonlinear, and noisy characteristics of composite fault signals in rolling bearings, a composite fault diagnosis method based on adaptive Feature Mode Decomposition (FMD) is proposed. To overcome the need for manual setting of key parameters in traditional FMD algorithms, an adaptive FMD method is introduced. By improving the stopping criterion of the conventional algorithm, the average Correlated Kurtosis (CK) of each mode group obtained during the decomposition process is calculated. The mode group with the highest average CK is selected to adaptively determine the optimal number of modes n for a given combination of filter number K and filter length L. A grid search method is then employed to optimize K and L, enabling adaptive selection of the key parameters [K,L,n]. The decomposed modes are subsequently screened and reconstructed using a criterion based on time-frequency domain correlation coefficients and kurtosis. The reconstructed signal is then envelope-demodulated to extract the fault characteristic frequencies for diagnosis. The proposed method is validated using both simulated signals and the Paderborn University bearing dataset. Experimental results show that this method can effectively extract the composite fault features of rolling bearings, can complete the diagnosis task well in a noisy environment, and has better fault feature extraction capabilities than other methods.