Transmission system of gearbox has a very complex structure and its faulty vibration signals always include strong noise. Feature extraction of weak signals from the background of strong noise is a difficult problem in vibration signal processing. Sparse decomposition method can extract the weak signal features adaptively under strong noise background, but it needs large computer time consuming in searching for the optimal matching atoms. To speed up the matching process for the optimal atoms, the signal sparse decomposition algorithm using genetic algorithm to optimize matching pursuit and tracking is proposed. Results show that the optimized algorithm can greatly reduce the computation time for searching for the optimal atom parameters in matching and tracking algorithm, and the main feature of gear's faulty vibration signal is the modulation phenomenon which can reduce the noise in the signal through sparse decomposition. Then, the frequency domain analysis is made and the fault diagnosis for gears is realized according to the frequency domain analysis results. Comparative analysis of the modulated vibration signal of the simulated gear with the actually collected gearbox vibration signal has shown that this method can extract fault feature frequency from the vibration signal with strong noise quickly and accurately.%齿轮箱传动结构复杂,其出现故障时的振动信号往往含有强噪声.在强噪声背景下微弱信号的特征提取是振动信号处理领域的难题.稀疏分解方法能够自适应地提取强噪声背景下的微弱信号特征,但其在寻找最优匹配原子时计算量特别大.为加快匹配最优原子的速度,提出利用遗传算法优化匹配追踪的信号稀疏分解算法,优化后的算法大大降低了匹配追踪算法中寻找最优原子参数的计算量.齿轮故障振动信号的主要特征是调制现象,通过稀疏分解对含有噪声的信号进行降噪,然后进行频域分析,根据频域分析结果实现齿轮的故障诊断.对仿真的齿轮调制振动信号和实际采集的齿轮箱振动信号分析表明,该方法能够从含有强噪声的振动信号中快速且准确地提取出故障特征频率.
展开▼
