Research On Bearing Condition Monitoring And Fault Diagnosis Of Heat Medium Circulating Pump

The heat medium circulating pump is an important equipment in petrochemical industry,and the running status of the bearing directly affects working performance and safety of the pump.The on-line monitoring of heat medium circulating pump bearing can effectively ensure its safe operation.The classification and identification of faults can provide maintenance basis for the production site to prevent serious accidents.The combination of on-line monitoring and fault diagnosis can improve the economy and safety of production.Traditional bearing status monitoring system of the heat medium circulating pump is restricted by backward communication technology and low degree of intelligence that has some disadvantages such as wiring difficulty and poor real-time performance,and the specific fault types and degree cannot be identified,which cannot meet the requirements of safety,efficiency and intelligence in modern industrial development.In this paper,a wireless communication status monitoring system of heat medium circulating pump bearing based on ZigBee was developed,which can realize the real-time monitoring of the running state of bearings.The vibration signals and temperature signals of heat medium circulating pump were collected,transmitted and processed,and LabVIEW software was used to design the upper computer program of the monitoring system to qualitatively diagnose whether the heat circulating pump bearings had a fault.In order to further realize the classification and recognition of specific bearing fault types and degrees,in this paper,the vibration signals were analyzed and samples of the right length were divided.A method of data augmentation was proposed that augments the time domain data of vibration signals.The concrete algorithms of neural networks were designed,and the model with optimal parameters was established by using MATLAB.The vibration signals produced by rolling bearings with different fault types and degrees were classified and identified,the results of which were compared and analyzed.The results show that the neural networks model with augmented processing of original data has better effects on fault classification and recognition.It can be verified that by using the method of data augmentation proposed in this paper,the gapof recognition rate between the training set and the test set is reduced,and recognition ability of the test set in neural networks is improved.