Based On The Time-Frequency Analysis Of The Detection Of Low Speed Reducer

Worm gear and worm reducer with compact structure,smooth transmission and reduction ratio is big and widely used in industrial production.But the worm gear and worm reducer occurs failure,usually,waste a lot of manpower material resources,so the worm gear and worm reducer fault diagnosis has always been the hot spot of the experts and scholars’ study.Main purpose of this article is to extract the fault signal of reducer worm gear fault diagnosis,seek a much simple and effective method to locate the worm gear fault location,and complete the realization of the on-line monitoring system of the worm gear and worm.Therefore,this paper’s main work expanded by followings:First of all,study the mechanical structure of the RV series worm reducer and the vibration principle,the common failure forms of the worm gear,causes of the failure mechanism of worm gear and its fault signal characteristics.Then introduce the vibration reducer used in the experiment platform and its related hardware information,solve the problem of in the process of building the sensor fixed and installation issues.Secondly,this paper introduces the common problems in the HHT algorithm,mainly studies the endpoint effect causes and solutions.Neural network integration algorithm is studied and its application in predicting data,and studies the local characteristic scale boundary continuation method and its optimization,the fusion of this two methods to optimize the HHT algorithm.Compared with traditional optimization method,show the advantage of the proposed method in this paper on the worm gear and worm reducer failure data detection.Then,limit of machine learning algorithm is introduced,due to extreme learning machine has some defects,then introduces the concept of kernel function and the relationship between kernel function and parameters in nuclear extreme learning machine.Considers the optimal kernel function parameter selection,then further introduces the genetic algorithm,and adopt genetic algorithm to optimize the parameters of kernel function in the extreme learning machine selection,then proceeds fault diagnosis to the speed reducer fault diagnosis test data.Compared to other artificial intelligence algorithms’ diagnostic results,the method proposed in this paper has great advantage,in terms of accuracy and operation time has improved.Finally,this paper introduces the programming based on MFC dialog box which can be used for the worm gear and worm reducer field monitoring fault diagnosis system.The system can realize the following functions: acquisition and preservation of the vibration data,real-time display of time domain and frequency domain graph,computing and storage of the threshold value,features value’s calculation and preservation,and the final reducer fault intelligent diagnosis.