Research On Intelligent Fault Prediction Of Chemical Rotating Equipment Under The Background Of Big Data

With the rapid development of modern science and technology,under the background of the current industrial big data,the intelligent manufacturing strategy represented by "Made in China 2025" proposed by China has been implemented in various countries,among which the rotary machinery equipment with gear box as the core is widely used in aerospace,high-speed rail,materials and chemical industry and other fields.In the field of materials and chemical industry,the operating environment of gear box has the characteristics of strong load,high temperature and high pressure,high speed,periodic impact and so on.Its internal gear parts are bound to produce a variety of faults which will affect the overall normal operation of the equipment.The particularity of the chemical production medium makes the failure of the equipment,which will cause huge economic losses to the enterprise and even threaten the safety of personnel.It is necessary to study the development condition monitoring and fault prediction of gearbox.In this paper,the gear,a key component in the gearbox,is taken as the research object.According to the development concept of industrial big data,the intelligent fault prediction of equipment is studied by using the data-driven method.The main contents are as follows:(1)Improved quality of gearbox big data.The failure mechanism of gear box was analyzed,and the failure trend of gear,a key part,was analyzed and studied from the lubricating oil and vibration signal in use of the equipment.However,both the oil sample and vibration signal would form noise due to the sampling environment and actual operation,which would cover up the effective information in the original data and affect the accuracy of the subsequent analysis results.To solve this problem,the basic principle of wavelet transform and wavelet packet transform is analyzed in this paper,and the noise reduction effect of the two is compared by simulation,and then the wavelet packet transform is used to pre-process the sample information for noise reduction.On this basis,the gear fault feature set is extracted to reduce the computational complexity,which provides a reliable data basis for the subsequent fault prediction research.(2)Fault characteristics were extracted by dimensionality reduction and decomposition.Aiming at the problem that the complex correlation between oil data and the nonlinearity between oil samples can reduce the efficiency of wear fault prediction,this paper uses KPCA method to analyze oil samples to achieve fault feature extraction;In addition,the vibration signal is analyzed using the CEEMDAN decomposition in the time-frequency domain analysis method.By calculating the correlation coefficient between each IMF component obtained from the decomposition and the original signal,the components that contain most of the information in the signal are selected,and the sample set of the vibration signal is obtained based on the sample entropy corresponding to the components,achieving the extraction of fault features.(3)Model establishment and application.According to the knowledge of intelligent diagnosis of big data,this paper uses neural network in the field of artificial intelligence to establish a model for predicting gear parts faults.By analyzing the basic principles of BPNN,RBFNN and GRNN commonly used,it is determined to use GRNN as the basis of prediction model by comparing its theoretical advantages.Aiming at the problem that the value of smoothness factor in the network will affect the reliability of the model,the sparrow search algorithm in the intelligent optimization algorithm is used to search for the optimal value of the parameter,so as to establish the SSA-GRNN model.The model was applied to analyze the actual oil samples of a shipyard,and the RMSE value of the wear prediction result was 3.98%.Meanwhile,oil samples and vibration signals were collected based on the GDS-type test rig for experimental analysis.The results show that compared with the traditional GRNN,BPNN and RBFNN models,the RMSE values of the model based on oil samples are reduced by 23.08%,43.18% and 44.13%,respectively,and the average accuracy of the prediction results based on vibration signals reaches 98.40%.It shows that the prediction model of SSA-GRNN established in this paper is effective and accurate.