Multi-source Sensing Based Mine Drainage Pump Fault Diagnosis And Health Prediction

The mine drainage pump is the core equipment for underground drainage work in coal mines.Although there are multiple drainage pumps in underground coal mines,they often need to work simultaneously when there is a large volume of water.If the drainage pump equipment fails,great losses will be incurred in the production of the enterprise.To guarantee its normal operation,it requires rapid identification of the cause and location when drainage pumping equipment fails as well as accurate prediction of its health before the failure happens.It is important for the safe and reliable operation of mine drainage pumps.According to the research on the fault diagnosis of drainage pump motors,centrifugal pumps and large equipment,a fault diagnosis and health prediction system for mine drainage pumps based on multi-source perception is proposed.The system is based on the shortcomings of the current design of mine drainage pump fault diagnosis and health prediction.The overall scheme of the system is designed by combining common faults in mine drainage pump electromechanical equipment.The hierarchical architecture of the system has been designed based on the relevant functions in the system.The specific layers are the multi-source information awareness layer,the edge resolution decision layer,the heterogeneous network transmission layer and the platform sharing migration layer.The bearing audio signal of the pump motor is extracted with features in the time and frequency domains.Frame kurtosis is selected as bearing fault diagnosis index.The vibration signal of centrifugal pump is analyzed by variational mode decomposition(VMD).The cliffness indicator is used as a measure to obtain the fault diagnosis modal component of the centrifugal pump.The stator winding temperature is also used as another fault diagnosis diagnostic indicator of the pump motor.A hybrid Convolution-LSTM fault diagnosis model optimized by the Firefly algorithm is designed.It is used to meet the requirements for fault diagnosis and prediction of drainage pumps.Convolutional neural networks(CNN)are used,due to its advantages for the extraction of local features of the data.Long short-term memory networks(LSTM)are used,based on its characteristics for processing temporal data.This network performance test was carried out based on the bearing data of drainage pumps in Xutuan coal mine.The results show that the optimised CNN-LSTM network has excellent fault diagnosis and prediction output with the highest accuracy rate.It can satisfy the requirements of mine drainage pump fault diagnosis and health prediction.Finally,the system was functionally tested and verified.The optimised CNN-LSTM neural network model was tested for several faults in pump motor bearings and centrifugal pumps.In addition,three networks were selected to compare the effects.The result is that the designed hybrid neural network has the highest accuracy of fault diagnosis results,minimal prediction output error.It is a strong real-time performance and has excellent overall performance.In the system communication and cloud display tests,the heterogeneous network worked stably during data transmission and the relevant functions of the host computer were normal.The system is capable of fault diagnosis and health prediction of mine drainage pumps,and is of high theoretical and practical value.Figure [72] Table [17] Reference [80]...