Research On Gear Fault Diagnosis Method Based On Multi-model Fusion

With the development of industry in the direction of intelligence and informatization,industry has also entered the era of big data following the development of the Internet.In this context,the sensors collect a large amount of data for monitoring the operating status of mechanical equipment every moment,making the intelligent fault diagnosis technology based on data driving become the trend of fault diagnosis.In actual industrial production,rotating machinery plays an important role,and gears are one of the most important parts in rotating machinery.It transmits the power between the shaft and the shaft so that the mechanical equipment can operate normally.The running state and service life of gears directly determine the running state and production efficiency of the entire equipment,which is the key to ensure the operation of mechanical equipment.In gear fault diagnosis,the classification capability of the model and the data collected by the sensors are one of the important factors that affect the diagnosis results.At present,the use of a single model for gear fault diagnosis is still a common diagnostic method.How to improve the efficiency and accuracy of gear fault diagnosis through model fusion is the research focus of this article.Therefore,this study did the following:First of all,a multi-model feature level fusion gear fault diagnosis model is proposed.The model uses a convolutional neural network to adaptively extract fault features from each sensor in the feature extraction stage,which reduces the dependence on expert experience in feature extraction to a certain extent and improves the diagnosis efficiency.In the fusion stage,a feature fusion algorithm is proposed to fuse the features of each sensor.Through experimental analysis and comparison,it is found that the convolutional neural network can effectively extract the fault features from the original data,and the proposed feature fusion method can effectively fuse the fault features of each sensor.Then an improved DS fusion algorithm is proposed to deal with the conflicts of different classification models during diagnosis.The improved algorithm uses the distance matrix and fuzzy preference matrix to generate joint weights to jointly modify the basic probability distribution of the original evidence body,and by modifying the 0 factor in the original basic probability distribution to avoid the occurrence of the "one-vote veto" phenomenon,and finally uses the traditional DS evidence theory fusion method specifically merges the revised evidence body with the original evidence.Then the improved DS fusion method is applied to the multi-model decision fusion gear fault diagnosis method.Different classification models are affected by their own model structure and data sensitivity,which will inevitably produce conflicting diagnosis results.The experimental results show that the fusion of multiple model diagnosis results through improved DS fusion method improves the accuracy of diagnosis.Finally,in order to solve the problem that compound faults often occur in complex and changeable actual production and compound faults are difficult to diagnose because of various coupling relationships,a compound fault diagnosis method based on improved DS algorithm multi-model decision fusion is proposed.This method splits the composite fault into single-type faults,which reduces the classification complexity of a single model and improves the overall performance of the entire diagnostic model.Furthermore,the use of improved DS algorithm fusion at the decision level further improves the accuracy of diagnosis.By comparing the accuracy,precision,recall and F1 value,the model has a relatively good diagnosis effect,which can provide a new research route for the diagnosis of compound faults from the data.