大型机械设备低频故障信号检测仿真

Low-Frequency Fault Signal Detection Simulation for Large Mechanical Equipment

研究大型机械设备故障检测精度优化问题。大型机械设备在横向振动的情况下,故障信号受到设备产生的横向振动的干扰,呈现信号低频化特征。传统的均值聚类方法在检测复杂大型设备故障时,由于信号低频化,要先进行故障收敛聚类,在收敛聚类过程中K均值算法没有考虑到低频信号相位近邻点的收敛特性,导致算法收敛太慢,检测时间过长,精度不高。提出了一种相位重组近邻点收敛的聚类算法,并有效应用到大组件机械设备故障检测中。聚类算法在传统的遗传K均值算法基础上,引入特征数据点之近邻点排除少数局部最优特征数据的干扰,通过近邻点收敛性的促进,提高了收敛速度,同时避免过早收敛到局部最优解中。仿真实验表明,改进的聚类算法收敛速度提高15.7%,故障检测有效识别率提高11.8%。

In this paper, the optimization problem of large machinery and equipment precision fault detection was studied. In the case of lateral vibration of large-scale machinery and equipment, fault signals are interference generated by equipment lateral vibration, which shows the low frequency of the signal characteristics. This paper presented a clustering algorithm based on the phase neighbor point restructuring convergence. Based on traditional genetic K- means algorithm, the neighbor＇ s point of feature data points was introduced to the clustering algorithm, and a small number of local optima and interference characteristics of the data were eliminated. By promoting the convergence of neighboring points, the convergence rate was improved, avoiding the premature convergence to local optimal solution. Simulation results show that the convergence rate of the improved clustering algorithm can be improved by 15.7%, and effective recognition rate of fault detection by 11.8%.