基于模型融合的风电机组齿轮箱故障诊断  

FAULT DIAGNOSIS OF WIND TURBINE GEARBOX BASED ON MODEL FUSION

作　　者：孙文卿 邓艾东 邓敏强 刘洋 程强 Sun Wenqing;Deng Aidong;Deng Minqiang;Liu Yang;Cheng Qiang(National Engineering Research Center of Turbo-Generator Vibration,School of Energy and Environment,Southeast University Nanjing 210096,China)

机构地区：[1]东南大学能源与环境学院火电机组振动国家工程研究中心,南京210096

出　　处：《太阳能学报》2022年第1期64-72,共9页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(51875100)。

摘　　要：针对单视角特征及单一模型对齿轮箱复合故障的诊断准确率较低的问题,提出基于比例冲突分配规则的模型融合故障诊断方法。首先,对齿轮箱振动信号进行特征提取,针对复杂复合故障,从时域及时频域角度构造特征;然后,将多视角特征送入多个子模型中进行初步诊断,得到互补性强的诊断结果;最后,模型输出的分类概率由第6类比例冲突分配规则(PCR6)进行融合。实验证明该融合模型所得出的结果稳定性较高,并在一定条件下可提高故障诊断的准确率。Aiming at the problem that the single-perspective features and single model have low accuracy in the diagnosis of gearbox composite faults,a model fusion fault diagnosis method based on the proportional conflict redistribution rule is proposed.First,feature extraction is performed on the gearbox vibration signals,and features are constructed from the perspective of time domain and frequency domain for complex compound faults.Then,multi-view features are sent to multiple sub-models for preliminary diagnosis,and complementary diagnostic results are obtained.Finally,the classification probabilities of the model outputs are fused through the sixth type of proportional conflict redistribution rules.Experiment results illustrate that the fusion model has high stability and can improve the accuracy of fault diagnosis under certain conditions.

关 键 词：风电机组 故障诊断 信息融合 齿轮箱 PCR6 

分 类 号：TG156[金属学及工艺—热处理]