Application Research Of Deep Learning In Fault Diagnosis Of Fire Pump Motor

Fire pumps play an important role in controlling and eliminating fire disasters.As the core component of the fire pump,the asynchronous motor directly determines whether the fire pump can operate normally.In order to ensure the smooth development of fire protection work,fault diagnosis of the fire pump motor is also essential.However,the existing fire inspection measures cannot make a detailed evaluation of their health status.At the same time,the traditional motor fault diagnosis methods have the disadvantages of high professional threshold,complicated diagnosis process and high artificial dependence,and are not suitable for fire pump motor.Troubleshooting.Based on this,this paper refers to the relevant requirements of the national standard GB50974-2014 for the inspection method of fire pump,and studies the application of deep learning in the fault diagnosis of fire pump motor.The paper first introduces the network structure and algorithm principle of deep belief nets(DBN),and uses three sets of simulation signals with different amplitude sideband components to experimentally test the classification ability of DBN,which verifies the powerful characteristics of DBN.Extraction ability.According to the characteristics of power frequency inspection and low frequency inspection,the fire pump motor simulation model under the corresponding inspection mode was built by ANSYS Maxwell and ANSYS Simplorer software,and the common faults of the motor were simulated to obtain the stator current data of each working condition.Subsequently,the simulated data is used for training and testing of DBN,and the influence of network parameters on network performance is analyzed.The network classification error is the main evaluation standard,and the appropriate network parameters are gradually established,so that the classification accuracy of the network for 10 types of working conditions can reach 91.2% on average.Compared with the traditional motor fault diagnosis method,it is proved that DBN has better fault diagnosis ability and is more suitable for fault diagnosis of fire pump motor.At the end of the paper,the fault diagnosis interface of the fire pump motor was designed in detail,and the deep learning algorithm was embedded in it to realize the one-button operation of the fire pump motor fault diagnosis.Aiming at the relatively low diagnostic accuracy,the method of using the largest proportion of the network output sequence as the final diagnosis result is proposed,which greatly improves the diagnostic accuracy and further enhances the practicability of the algorithm.