Research On Fault Condition Recognition Method For Key Parts And Components Of Gear Transmission System

As a widely used power transmission component in mechanical equipment,the safety and reliability of gear transmission system are directly related to the working life and efficiency of the whole equipment.It is well known that the gear transmission system usually runs under the working condition of high speed and heavy load,and once the key parts and components fail,it will often cause the whole transmission system to fail to operate normally.Therefore,the fault diagnosis technology of key parts in gear transmission system is deeply studied,and a practical and reliable fault diagnosis method is put forward,which is of great engineering significance for accurately predicting and mastering the law of failure of parts and components,avoiding the situation of causing heavy economic losses,heavy accidents and casualties,etc.Because the essence of fault diagnosis is the process of pattern recognition and classification,it is an important content of mechanical system fault diagnosis to classify and identify the relevant running state of equipment by using the operation information of these key parts.The key parts of gear transmission system,gear,bearing,are taken as the research object in this dissertation,and the key problems in the field of fault state identification,such as signal noise reduction,fault feature extraction,fault feature selection,fault state identification and so on,are studied.First at all,combined with the engineering application of mechanical system,the fault evolution mechanism,the data acquisition method,the signal processing and feature extraction method,the state recognition and the life prediction technology of the key components of the mechanical system are reviewed,and the advantages and disadvantages of the research are summarized.The research content and route of this paper are drawn up.Next,the time-domain characteristic parameters,the frequency-domain characteristic parameters and the time-domain characteristic parameters which can be characterized by the extraction from the original vibration signal are analyzed,and the more sensitive entropy characteristic parameters are introduced.In this paper,the high-dimensional feature vector of the mixed domain is constructed to comprehensively describe the fault state of the gear transmission system.In addition,the experiment rig for fault diagnosis of gearand bearing is designed,which provides a platform for further research.Because noise is the main obstacle to fault feature information extraction,the vibration signal acquired in practice is often mixed with the noise,and the conventional noise reduction method is difficult to effectively process the noise in the nonlinear and non-stationary signals.Therefore,a novel noise reduction algorithm based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN),Permutation Entropy(PE)and Time-Frequency Peak Filtering(TFPF)is proposed in this paper.According to the method,the permutation entropy is used as the judgment index,and the intrinsic mode function after the CEEMDAN decomposed is filtered by a TFPF method with different window length,and the contradiction between the noise suppression and the signal fidelity of TFPF method is well balanced.The experimental results of the simulation and the measured signals show that the signal-to-noise ratio(SNR)is obviously improved after noise reduction,and the fault features can be successfully extracted from the weak signal.Aiming at the one-sidedness and limitation of single-scale feature in representing fault state,a fault feature extraction algorithm based on nonlinear entropy is proposed.In this paper,a multi-scale feature vector is formed by using the PE value of each intrinsic mode function decomposed by the CEEMDAN,and the state identification is carried out by a Support Vector Machine Optimized by Particle Swarm(PSOSVM),and the necessity of introducing the entropy feature and the comprehensiveness of the multi-scale entropy are further reflected.The effectiveness and robustness of the algorithm in bearing fault detection and classification are verified by experiments,and the comparison results show that the algorithm has better stability and higher classification accuracy.Finally,in view of the correlation and redundancy of the extracted high-dimensional feature set in the mixed domain,which seriously affects the subsequent state recognition performance,a Supervision Kernel Entropy Component Analysis with Whale Optimization Algorithm(WOSKECA)is introduced in this paper,which aims to represent more effective information with fewer feature components.The algorithm uses the main kernel entropy score data to select and reduce the redundant features;uses the supervised learning mode through the sample category information not only preserves theintrinsic entropy structure of the original feature set,but also ensure the higher accuracy of classification and recognition;uses the Whale Optimization Algorithm to find the kernel parameters most suitable for kernel entropy component analysis algorithm,which reduces the professionalism and subjective error of obtaining fault state information.This method not only improves the effect of the fault feature extraction,but also ensures the integrity and cohesion of key information.Finally,the sensitive feature parameters containing the composite feature information are input into the PSOSVM classifier,and the state classification and identification of the fault types and severity of the key parts of the gear transmission system are realized.