Fault Diagnosis Of Pump Based On Machine Learning

In industries such as coal coking,petrochemicals,water conservancy and power,rotary pump equipment plays a key role,and their normal operation has a significant impact on the safety of production activities.The machine-pump group,affected by man-made,environmental or self-related factors,can cause abnormal operation of the equipment,which will hinder the production,and even bring about damaging effects such as casualties.Therefore,it is of great practical significance to study and establish a fault diagnosis system for pumps to ensure the stable operation of equipment,to reduce accidents and to apply in practical production.In this paper,based on the understanding of the fault diagnosis technology of the pump equipment,the method of extracting the fault signal features of the core component-rolling bearing,is studied in depth firstly.Starting from the analysis of the vibration signal,empirical modal analysis and wavelet packet analysis are introduced into the feature extraction process.Energy is selected as an effective fault feature and normalized.By verifying the performance of the two methods,it shows the advantages of wavelet packet analysis in fault feature extraction.Its fault feature vector contains rich fault information,which can be used as data samples for fault diagnosis.Secondly,the feature identification method of rolling bearing fault signal is analyzed,and the support vector machine(SVM)method is applied to fault feature identification.Because of the low identification rate of traditional optimization methods,the weight coefficient of the algorithm is optimized by improving the particle swarm,and the adaptive mutation strategy of genetic algorithm is used for reference.The results show that the improved particle swarm optimization(PSO)has a fast conver gence speed and is not easy to fall into the local optimal solution.Then the improved particle swarm optimization is used to optimize the penalty factors and kernel function parameters of the support vector machine,to build a data classification model with better performance,and to further build a combined fault diagnosis model based on wavelet packet decomposition,band energy,improved particle swarm optimization and support vector machine.Finally,the effectiveness of the improved particle swarm optimization and fault diagnosis method proposed in this paper is verified.By comparing the GA-SVM and PSO-SVM classification models,the experimental results show that the improved particle swarm optimization has strong optimization capabilities and can effectively improve the accuracy of diagnosis,comparing the BP neural network model again.The IPSO-SVM model can effectively identify different types of faults and their degree of damage in the case of small samples.Therefore,the analysis ideas and methods provided in this paper have certain practical value for fault analysis of rolling bearings and troubleshooting of pumps.