Research On Gear Transmission System Fault Diagnosis Method And System Based On Energy Signal Analysis

Carrying out research on the mechanical fault diagnosis technology innovation is of greatsignificance for the safe and stable operation of quality upgrading equipment. The faultidentification rules hidden in the energy signal is revealed in this thesis with demonstratingthe correlation between vibration and energy parameter in gear transmission system based onthe energy perspective. a model intelligent fault diagnosis method is established for geartransmission system, etc rotary machine in the process of researching on effective methodin nonlinear energy signal process and feature extracting and fault fuzzy pattern recognition.First of all, the vibration mechanism of gear transmission system is studied.The mappingrelationship between the input instantaneous energy and the deviation of gear generating withthe microscopic analysis of static transmission error variable.At the same time, the experimentresults show there is strong correlation between energy and vibration signal usingcoherence method in frequency domain to analysis their coherence. These studies provide thetheoretical basis for the following work.Secondly, the PSO-ARMA model of wave continuation forecasting for the inhibitionof the end effect in HHT method is researched. The parameters adaptive mutation PSOalgorithm based on Entropy is proposed and studied firstly in model research, and then applyit to parameter optimization of ARMA estimation model,obtain optimum parameters throughglobal search in the solution space based on the initial value from method of momentsestimator. At the same time, the end effect simulation using the model is completed and theresults show that each IMF waveform maintains higher integrity and spectrum energy afterthe EMD decomposing.Then, study on the fault characteristic extraction method and energy signal analysisexperiment using HHT method of the gear transmission system. According to the EMDdecomposition with energy signal of normal and broken teeth and Hilbert spectrum andmarginal spectrum analysis on vibration and energy signal, the superiority that the energy signal analysis results on the characterization of the fault state is verified. At thesame time, the construction method is studied to multi-dimensional fault feature vectorlibrary which containing the first6order IMF normalized skewness, kurtosis, energy, standarddeviation and the approximate entropy and other parameters. These Achievements lay thefoundation for further study on faultfeature recognition method.Next, the thesis presents a novel fault identification method and the KEFKM modelbased on fuzzy clustering strategy with kernel principal component entropy. The model has3major functions: KPCA feature dimension reduction; fuzzy clustering based on kernelprincipal component entropy; fuzzy fault recognition based on fuzzy relative entropy. Torealize this method, firstly, the concept of the main element of information entropy which isused to integrated with FKM algorithm is proposed, and the feature data is clustered after thefirst kernel principal data is clustered to obtain the best classification number and the initialclustering center using KPCA method to reduce the dimension of data in order to reducecomputation, kernel density estimation and maximum entropy principle. The experimentalresults show that this method can significantly improve theclustering results of fault data. Inthe last part of the method, according to the sample to be detected the fault fuzzyrecognition method based on fuzzy relative entropy is proposed. The difference between thegreatest close degree and fuzzy relation entropy method is presented through experimentcontrasting and analyzing. It is also present that the distribution characteristics of whole datacan be measured by fuzzy relative entropy which has obvious effect on the judgmentof similarity between two fuzzy subsets. And to verificated the method, the KEFKM modelwas used to carry out experimental study of gear, the result shows that the model can ensurethere is reasonable sample space from inner class to class distribution to the significant kernelprinciple data derived from training samples by fuzzy clustering. Therefore, the methodshows excellent performance in dealing with the pattern recognition of fault samples.Finally, study on the design of energy signal monitoring and fault diagnosis system withthe analysis on realizing process of the system construct and processing information function.The fault diagnosis system is developed preliminary which realized the energy signal analysis, failure mode fuzzy recognition, etc steps through the embedded Matlab based onvirtual instrument technology. At the same time, the application of wireless sensor networktechnology is studied in gear transmission monitoring using functional nodes. Thus, throughthe optimization of the Zigbee protocol stack and the nodes in ad hoc network, the system canremote collect data for safely and conveniently monitor.