Research On The Diagnosis Of Rolling Fault Bearing Based On CEEMDAN And Fuzzy Neural Network

With the rapid development of industry in our country,more and more high to the requirement of mechanical equipment,every link of the industrial production system is becoming more and more specifications,this is not only reflected in the equipment precision,quality and cost control in the areas of high demand,also in the whole may be produced in industrial production,the fault forecasting.Along with the needs of industrial development,mechanical equipment has also begun to develop towards large scale,integration,high speed,precision and intelligence.The rolling bearing is electric power,metallurgy,petrochemical,machinery,aerospace and military industry the most widely used by the department of mechanical parts,is also one of the most easily damaged parts in machinery and equipment.In this paper,several typical failure modes and principles of rolling bearing are introduced,and the calculation method of the failure frequency of the fault part under the corresponding failure modes is presented.Then according to the situation of the experimental platform for reasonable selection of the experimental bearings,and related equipment of the test platform to do the commissioning and the related experimental steps were determined,to ensure the accuracy and rationality of the experiment.Secondly discusses the wavelet packet decomposition method and its corresponding improvement method principle,the use of improved wavelet packet method combining threshold noise reduction method,and based on energy entropy and correlation coefficient and kurtosis criterion,puts forward the improved method of rolling bearing fault characteristic information extraction.The method is validated in the experiment,which proves the feasibility and rationality of the feature information extraction.At the same time,this paper discusses the adaptive noise completely set the principle of empirical mode decomposition method and basic steps,then use the method combining wavelet packet to the bearing fault signal feature extraction,using the Hirbert-huang to extract the feature information of envelope spectrum analysis,failure frequency.Later in this chapter,we will apply this method to the selected bearing for verification,and effectively prove the practicability of this method.At last,we introduce the fuzzy neural network(FNN),a hot intelligent algorithm,including its principle and classification.An experimental model was designed by ourselves and its reliability was verified by the neural network method.In the experiments,build the network of two types of diagnosis,one kind is directly using the neural network fault diagnosis of bearing,and those is combined with fuzzy neural network diagnosis and improve the accuracy of the diagnosis of bearing.In the course of the experiment,it is shown that the fuzzy neural network trained on the basis of selecting the appropriate parameter feature input vector can more accurately determine the fault modes of various rolling bearings.