基于复杂网络社团聚类的复合故障特征分离诊断方法

摘要

针对复合故障多种故障特征相互叠加彼此干扰，给全面准确诊断带来困难，提出了基于复杂网络社团聚类的复合故障特征分离诊断方法。该方法首先应用 EMD 将复合故障信号分解为若干个 IMF 分量，由于不同单一故障的特征会在不同频段得以体现，提取每个 IMF 分量的特征量，建立故障数据网络模型；然后将每个 IMF 分量视为网络中的社团，根据复杂网络社团结构的特性，进行同类社团合并，合并所得每个社团与单一故障相对应，最后对合并的信号进行分析，实现复合故障特征分离。以转子不平衡和轴承内圈、轴承内圈和滚动体复合故障特征分离与诊断为例，验证了该方法的可行性。%The diagnosis of composite faults in mechanical systems is a challenge at present.Due to features of composite faults interfering,it is hard to diagnose composite faults fully and accurately.Here,a diagnosis method for composite fault features separation based on complex network organization clustering.Firstly,a faulty signal was decomposed into several intrinsic mode functions (IMFs)with empirical mode decomposition (EMD).As different single-fault characteristic could be reflected in different frequency ranges,the characteristics of each IMF component were extracted and the network model of fault data was built.Each IMF was taken as an organization in the network.According to the characteristics of complex network organizations structure,the organizations in the same type were merged.The merged organization corresponded to a single fault.At last,the separation of composite fault features was realized for the decomposed signal.Taking rotor unbalance and bearing inner race,and bearing inner race and bearing roller composite fault characteristics separation and diagnosis as an example,the feasibility of this method was verified.