基于粒子群优化的共振稀疏分解在轴承故障诊断中的应用

Application of Particle Swarm Optimization Based Resonance Sparse Decomposition on Bearing Fault Diagnosis

共振稀疏分解是振动信号中脉冲成分提取的方法。与基于频率的信号处理方法不同,该方法同时参考频率和带宽两个因素,从而在分离信号不同成分的过程中能够很好处理信号不同成分的重叠问题。然而共振稀疏分解的分解效果受到品质因子Q、权重系数A以及拉格朗日乘子u的主观选择影响,针对此问题,将粒子群优化算法(Particle Swarm Optimization,PSO)应用到参数的选取中,通过粒子群优化算法的全局优化特点对实验参数进行自适应选取,进而实现振动信号的有效分解。将基于粒子群优化算法的共振稀疏分解应用到轴承故障信号的诊断中,证实了该方法的有效性。

A method for extracting pulse signal was proposed here. Unlike traditional signal decomposition based on frequency, resonance sparse decomposition does well in the separation of signal overlapping according to frequency band and bandwidth. But the performance to extract pulse signal with resonance sparse decomposition is affected by the parameters of quality Q-Factors, weight coefficient A and Lagrange multiplier u. Aiming at the problem of parameters selected subjective. Particle Swarm Optimization is applied here for parameters optimization with the characteristics of global optimization to achieve the decomposition of vibration signal. Results to the decomposition of experiment and simulation signals shows the effectiveness of the method proposed.