基于小波包分析和优化KNN的电动开度阀故障检测方法

Fault Detection Method for Electric Opening Valves Based on Wavelet Packet Analysis and Optimized KNN

针对以微控制器MCU为控制核心的电动开度阀控制系统难以集成高效且计算量小的故障检测子系统的问题,基于小波包变换和优化K近邻(K-Nearest Neighbor,KNN)算法提出了一种电动开度阀故障检测方法。对阀门振动信号进行小波包变换,计算小波包节点的能量值与其重构信号的时域特征参数。根据Pearson系数筛选出两种与能量强相关的故障特征参数:峰峰值与均方根,并将两者作为KNN算法的样本评价指标;通过对评价指标进行加权优化了KNN算法的距离计算公式,分别在MATLAB和实验样机中进行故障检测测试,对应最高分类准确率分别为92.5%与86.7%。结果表明:实验测试与仿真分析具有较好的一致性,该故障检测方法的优势在于计算量小、故障识别率较高,并能有效地应用于以MCU为核心的电动开度阀控制系统。

To address the issue of difficult integration of a highly efficient and low-computational-fault detection subsystem into an electric opening valve control system with a microcontroller unit(MCU)as its control core,this paper propose a fault detection method for electric opening valves based on wavelet packet transform and optimized K-Nearest Neighbor(KNN)algorithm.The wavelet packet transform is applied to the valve vibration signal,and the energy values of the wavelet nodes are calculated together with the time domain characteristics of the reconstructed signal.Based on the Pearson coefficients,two fault characteristics parameters with strong energy correlation:peak-to-peak value and root mean square,and both are used as sample evaluation indicators for the KNN algorithm;the distance calculation formula of the KNN algorithm is optimized by weighting the evaluation indicators,and fault detection tests are conducted in MATLAB and the experimental prototype respectively,with corresponding highest classification accuracy rates of 92.5%and 86.7%.The results show that experimental test and simulation analysis have good consistency,and the advantage of this fault detection method is that it has small amount of calculation,high fault identification rate,and can be effectively applied to the electric opening valve control system with MCU as the core.