基于改进局部均值分解与复合多尺度气泡熵融合的直流电机故障诊断

Fault Diagnosis of DC Motors Based on the Fusion of Improved Local Mean Decomposition and Composite Multiscale Bubble Entropy

摘要:

直流电机出厂故障检测的核心挑战在于电机运行时声音信号中微弱故障特征的精准提取，此类信号常受机械振动、电磁噪声及环境干扰的影响，导致传统方法易受模态混叠与频带交叠干扰；针对这一难题，提出了一种基于改进型局部均值分解与复合多尺度气泡熵融合的直流电机故障诊断方法；并采用粒子群PSO优化极限学习机ELM进行分类建模；实验结果表明，所提出的方法能够有效地区分电机的正常状态、转子轴弯曲、叶片断裂以及轴承故障状态，在特定的320组测试样本中仅仅1组测试出错；因此，该方法为直流电机的故障诊断提供了一种新的技术途径，有助于提高电机的可靠性和运行效率。

Abstract:

The core challenge in the factory fault testing of DC motors lies in the accurate extraction of weak fault features from operational acoustic signals. These signals are often contaminated by mechanical vibrations, electromagnetic noise, and environmental interference, leading traditional methods to suffer from mode aliasing and frequency band overlap disturbances. To address this challenge, this study proposes a novel DC motor fault diagnosis method based on improved Local Mean Decomposition (LMD) and Composite Multiscale Bubble Entropy (CMBE) fusion. Additionally, a Particle Swarm Optimization (PSO)-enhanced Extreme Learning Machine (ELM) is employed for classification modeling. Experimental results demonstrate that the proposed method effectively distinguishes motor states, including normal operation, rotor shaft bending, blade fracture, and bearing faults. In a specific test involving 320 samples, only 1 misclassification occurred. Consequently, this approach offers a new technical pathway for DC motor fault diagnosis, significantly enhancing motor reliability and operational efficiency.