Research On Gearbox Fault Diagnosis Technology Based On Machine Learnin

Gearboxes are the core components that connect and transmit power in machinery and industrial equipment.They have a complex structure and are prone to wear and failure,which can affect the safe operation of equipment and cause economic losses.It is very important to effectively monitor the safe operation of gearboxes to prevent failures during operation.In gearboxes,gears and rolling bearings are extremely critical components,which have the highest failure rate due to their work of transmission and support and high force intensity.The thesis conduct an in-depth study and analysis of gear and rolling bearing fault states,extract fault features,optimize diagnostic algorithms,and perform fault diagnosis and identification.The main research is as follows:Aiming at solving the problems of low accuracy of fault diagnosis using time domain features or frequency domain features alone and redundant information caused by multiple feature dimensions,a rolling bearing fault diagnosis method based on discrete wavelet transform and neighbourhood component analysis combined with random forest is proposed to shorten the diagnosis period while improving the accuracy.The new method adopts the discrete wavelet transform method to preprocess the vibration signal of rolling bearings,to extract time-domain,frequency-domain and singular value features.The method organically combines these feature values into a feature matrix.Then the method eliminates redundant information,extracts main feature parameters,and optimizes the feature matrix by introducing neighbourhood component analysis to select the optimal combination of features.Finally,the selected optimal combination is input into the random forest for fault identification.Its characteristic of not easy to overfitting improves the accuracy of fault diagnosis and shortens the operation cycle.The experiments show that the accuracy of the method reaches 99.92%and the operation cycle is nearly doubled shorten compared with the method without optimized feature combination,which provides strong support for fast and accurate fault diagnosis.To solve the problems of gear fault characterization,low accuracy of traditional fault diagnosis methods and poor stability of the limit learning machine,the thesis propose a gear fault diagnosis method based on the mind evolutionary algorithm optimized extreme learning machine,and introduce the Sigmoid entropy into the gear fault diagnosis.The Sigmoid entropy can reduce the interference of uncertainty information on the eigenvalues,and the mind evolutionary algorithm has high robustness and applicability,and the mind evolutionary algorithm can effectively improve the stability and reliability of the model by using the mind evolutionary algorithm for parameter optimization of the weights and biases of the extreme learning machine.The experimental results show that the method proposed in this thesis has a very high accuracy rate in gear fault diagnosis,with an average accuracy rate of more than 99.6%,which effectively improves the stability of gear fault diagnosis and provides powerful data and theoretical support for the health monitoring and maintenance of gearboxes,with a wide range of application prospects and research value.