基于PSO-VMD的滚动轴承故障诊断方法
Fault Diagnosis Method of Rolling Bearing Based on PSO-VMD

滚动轴承的振动信号特征提取具有挑战性，变分模态分解(Variational Mode Decomposition, VMD)可以有效地提取特征，但参数的选择需要仔细考虑。本文提出了一种创新方法，结合粒子群优化(Particle Swarm Optimization, PSO)算法，为VMD提供自适应参数优化，以提升其性能。首先，通过PSO算法自适应地确定VMD中的模态数量K和二次惩罚因子α，并使用包络熵作为目标函数。然后，使用优化的参数，通过VMD方法对故障信号进行分解，得到多个本征模态函数(Intrinsic Mode Functions, IMFs)。随后，采用特征融合系数(Feature Fusion Coefficient, FIC)选择最优的IMF，再通过包络分析提取故障特征频率。使用西储大学(Case Western Reserve University)数据集进行的实验结果表明，该方法可以有效解决VMD参数选择对故障诊断的影响，成功识别滚动轴承故障。与基于峭度指数的传统方法相比，该方法在选择最优IMF时具有更高的准确性和稳定性。
Rolling bearing vibration signal feature extraction is challenging, but variational mode decomposition (VMD) effectively addresses this issue. While VMD excels in feature extraction, parameter and sensitive mode selection require careful attention. This paper proposes an innovative approach that enhances VMD with an adaptive parameter optimization technique using Particle Swarm Optimi- zation (PSO). Firstly, the modal number K and quadratic penalty factor α in VMD are adaptively determined by the PSO algorithm, utilizing envelope entropy as the objective function. Subsequently, the optimized parameters are applied to decompose the fault signal using the VMD method, resulting in multiple intrinsic mode functions (IMFs). The feature fusion coefficient (FIC) is then employed to select the optimal IMF, followed by envelope analysis for feature frequency extraction. Experimental results using the Case Western Reserve University data set demonstrate the effectiveness of this method in mitigating the influence of VMD parameter selection on fault diagnosis. Compared with traditional methods based on kurtosis index, this approach exhibits higher accuracy and stability in selecting the optimal IMF for rolling bearing fault identification.