Research On Intelligent Fault Diagnosis Of Heavy Petroleum Gear Transmission System

Heavy petroleum gear transmission system is the key transmission system in oil drilling equipment such as high-power fracturing unit,cementing unit and pumping unit.With the development of modern science and technology,oil equipment is also developing towards the direction of high automation,complexity and integration.When the gearbox fails,the whole oil equipment will not work normally,which will affect the operation of the whole oil production.Therefore,it is of great significance to study the technology of fault diagnosis and predictive maintenance of gearbox in petroleum equipment.The characteristics of heavy duty oil gear transmission system are the complexity of operating conditions,such as high speed and heavy load in the oil drilling process.At present,the fault diagnosis of the gear is mainly focused on the condition of fixed working conditions,while the variable working conditions cause some difficulties in the fault diagnosis of oil gear.At the same time,the fault diagnosis technology has developed from the stage of post-fault diagnosis to the stage of fault prediction and health management.When the oil equipment is fault-free,it is necessary to predict its remaining service life.Therefore,this paper studies the intelligent fault diagnosis of heavy duty petroleum gear transmission system from the above two aspects.First,this paper studies the failure mechanism of the gear,starting from the vibration model of gear,gear mesh are analyzed as excitation source,and then according to the characteristics of the heavy duty gear oil complex working conditions,in variable condition and working condition of fixed gear fault dynamics simulation under two different conditions,the analysis of gear in different fault and variable condition of vibration signal spectrum,from the Angle of the simulation shows that the complex working conditions is the difficulty of heavy oil gear fault diagnosis.Secondly,in view of the oil gear box under complex conditions under the complex characteristics of load,studied the feature extraction method based on Hilbert envelope spectrum analysis,the time-domain characteristics of original signal are extracted on the basis of,in order to get more useful fault characteristic information,affect the sideband modulation in different fault information is extracted,using the Hilbert envelope spectrum method to filter out high frequency,the influence of side band extracted from low frequency signal modulation frequency information,and the features of the original vibration signal in the time domain index set,as the data samples for the following machine learning algorithms.Further,in view of the heavy oil in the gear fault diagnosis of variable working condition and fault types cannot be diagnosed problems,based on the SVM are studied supervised and SOM unsupervised type gear fault intelligent diagnosis model,using intelligent fault diagnosis model with different conditions,different fault types and three different sampling position fault diagnosis cases of data analysis;In the SVM diagnosis model,the performance of the model is evaluated from three aspects: data normalization processing,number of training and test samples and number of feature selection.The results of the two algorithms are compared and analyzed,and the advantages,disadvantages and applicability of the two algorithms are summarized.Finally,in view of the remaining service life prediction of heavy oil gear transmission system,in order to gear box key parts of rolling bearing as the research object,research the degradation model based on the index of the residual service life of rolling bearings,because the problem of life prediction of state index selection,using the spectral kurtosis technology of rolling bearing vibration signal feature extraction,and the characteristic indexes are monotonicity analysis,using PCA method to extract the characteristics of data dimension reduction and feature fusion,make certain that the state of life prediction index,bearing the whole life cycle of data using accelerated experiment,verify the validity of the index regression models.The result shows that:(1)through the operation mode of simulation analysis found that with the change of the rotating speed and load,especially high overloading,no obvious faults in the spectrum characteristics of frequency spectrum analysis method cannot satisfy the complex conditions of heavy oil gear fault diagnosis,because of the actual operation condition of heavy oil gear operation mode is more complex than simulation,based on the analysis of the traditional Fourier transform methods have limitations in the heavy oil gear fault diagnosis.(2)the accuracy of supervised diagnosis based on SVM was 85.7143%,and the accuracy of unsupervised diagnosis based on SOM was 75%.The analysis results show the effectiveness of the model.In the SVM algorithm,data normalization processing and increasing training data samples can improve the accuracy of fault classification.The svm-based supervised gear fault model is more suitable for fault diagnosis of heavy duty petroleum gear under complex working conditions than the SOM unsupervised model.(3)when only the time-domain features of the original vibration are used,the accuracy of fault classification is 67.8571%,and the accuracy of classification is 78.5714% when the same time-domain features are extracted only after Hilbert envelope spectrum processing.The accuracy of the combined feature sets is 85.7143%,which proves the effectiveness of the feature extraction method based on Hilbert envelope spectrum.(4)The prediction results show that the significant change of slope can be detected at the time of 37 min,indicating that the exponential degradation model can be used to predict the remaining service life of the bearing.According to the remaining service life curve,the bearing can be replaced or repaired in advance,so as to achieve the purpose of predictive maintenance of the heavy petroleum gearbox.