基于改进EWT-精细复合多尺度散布熵和GG聚类的球磨机负荷识别方法

Load State Identification Method for Ball Mills Based on the Modified EWT,Refined Composite Multiscale Dispersion Entropy and GG Clustering

针对球磨机振动信号具有非线性、非平稳性特点所导致的负荷状态难以识别问题,提出一种基于改进经验小波变换(EWT)、精细复合多尺度散布熵(RCMDE)和GG聚类的球磨机负荷识别方法。首先,在EWT的基础上,引入滑动频率窗的思想,提出自适应频率窗EWT算法,将其用于对球磨机原始振动信号的分解以获得本征模态分量;其次,通过相关系数法选出能表征原始信号状态的敏感模态分量进行信号重构;第三,利用RCMDE对重构信号进行处理提取负荷状态特征;最后,将特征向量作为GG聚类算法的输入,将球磨机负荷状态作为输出,建立球磨机负荷识别模型。通过磨矿实验验证了该方法的有效性,结果表明,该方法的聚类内紧致性较好,识别的评价指标PC值最高可达0.9989,而CE值仅为0.0013,识别效果显著,能够准确识别球磨机的负荷状态。

To overcome the difficulty of accurately judging the load state of a wet ball mill during the grinding process,a method of mill load identification based on the modified empirical wavelet transform(EWT),the refined composite multiscale dispersion entropy(RCMDE)and the GG clustering was proposed.Firstly,the idea of sliding frequency window was introduced on the basis of EWT,and the adaptive frequency window EWT algorithm was proposed,which was used to decompose the vibration signals recorded under different load states to obtained the intrinsic mode components.Secondly,a correlation coefficient threshold was used to select the sensitive mode components that characterize the state of the original signal.Thirdly,the RCMDE was used to process the reconstructed signal and extract the load state feature.Finally,a characteristic mill load vector was constructed from the RCMDE of the cylinder vibration signals recorded under different load conditions and was used as the input to GG classifiers,which then outputs the predicted ball mill load state.In this way,a load state identification model was established.Grinding experiments were presented to verify the effectiveness of the proposed method.The results show that the clustering of the method is good in compactness,the identified evaluation index PC value is up to 0.9989,while the CE value is only 0.0013,and the recognition effect is remarkable.This method can accurately identify the load state of the ball mill.