Research On Rotating Machinery Health Detection Method Based On Deep Learning

The complexity of the environment makes rotating machinery more susceptible to damage,failure,loss of remaining service life,and early obsolescence.Failures in machinery are difficult to observe with the naked eye and the remaining service life is difficult to estimate manually,which can lead to serious production accidents if failures are not diagnosed,remaining service life is not calculated,and reasonable maintenance plans are not made.Therefore,the study of fault diagnosis and remaining service life in health inspection for rotating machines can not only improve the safety of production in workshops,but also provide safety assurance for production plant personnel.In this paper,we take the fault diagnosis and remaining service life of rotating machinery as the main research content,take the whole life cycle vibration signal as the main object,use signal processing,intelligent algorithms and other methods to carry out research,build a set of signal analysis,feature extraction,fault diagnosis and remaining service life prediction of health detection system,and establish a variable mode decomposition(VMD)based on optimized parameters.mode decomposition(VMD)-convolutional neural network(CNN)and principal component analysis(PCA)-parameter optimization(back propagation(BP)residual life prediction model.Based on the above models,a rotating machine health condition detection system is built,and the system is tested and analyzed.The main research contents of the paper are as follows:(1)To address the problem that the core parameters of deep learning are determined randomly or determined by human experience leading to low model capability.We propose the Elite Opposition Adaptive Mutation Sparrow Search Optimization Algorithm(EOAMSSA).By adding arnold cat map,Elite Opposition-Based Learning(EOBL),improved Tent perturbation,and Corsi mutation strategies to the original Sparrow Search Algorithm(SSA),and optimizing the explorer and follower formulas.The superiority and feasibility of this algorithm are verified by using 10 benchmark functions with the remaining 7algorithms for simulation experiments.(2)To address the problem of insufficient intelligence and accuracy of rotating machinery fault diagnosis.The research on the deep learning-based fault diagnosis method of rotating machinery is carried out.This paper establishes a fault diagnosis model combining variational modal decomposition and convolutional neural network by noise reduction processing and feature extraction of the vibration signal of rotating machinery,and also optimizes the penalty factor and number of modes of the variational modal decomposition method and the relevant hyperparameters in the convolutional network by using the EOAMSSA algorithm proposed in this paper,combines the envelope entropy and cliffness index to improve the diagnosis accuracy and verifies the reliability and The reliability and superiority of the model are verified.(3)In order to improve the accuracy of the remaining service life prediction of rotating machinery,the research on the remaining service life prediction of rotating machinery based on neural network is carried out.In this paper,through the whole life cycle data set of rotating machinery,after smoothing the signal,extracting the degradation features of rotating machinery,establishing the remaining service life model by using principal component analysis method and BP neural network,selecting 29 time-domain and frequency-domain features and choosing some features with the largest correlation based on the analysis of correlation between features,optimizing the weight value and kernel parameters of the model by using intelligent algorithm,improving the model The prediction accuracy is improved,and the validity of the model is verified through the evaluation index,and the superiority of the model is verified.(4)To address the problem that the model results are not intuitive enough,a rotating machinery health detection system is built.In this paper,the improved fault diagnosis model and remaining service life prediction model of the elite inverse adaptive sparrow search algorithm are integrated into the system programmed in C# language,and the system is tested and analyzed through the human-computer interaction interface and visualized health status detection.The results surface that the built system can be more intuitive for users to see the health status of the machinery and can meet the usage requirements.