Design And Implementation Of Equipment Monitoring System Based On Sound Recognition Technology

Large-scale equipment such as transformers' s sound have certain identification and stability characteristics when they are running,but they are easily interfered by various environmental sounds.The traditional method of detecting whether the transformer is operating normally is to monitor the running sound of the transformer through a technician who has many years of experience in equipment maintenance.However,the traditional detection method has many drawbacks.The voice recognition system has the advantages of high recognition efficiency,low cost,and labor saving.Therefore,it is considered to use voice recognition technology to monitor the running state of large equipment.In this paper,the requirements of equipment abnormal sound monitoring system are analyzed in detail.Using sound signal processing,sound feature extraction,sound data training,pattern recognition and other technologies,a feasible system construction method is proposed and the prototype system is designed.The system is mainly divided into four core functional modules,including sound data communication module,sound data model matching module,sound data comprehensive analysis module and system management module.The sound data communication module is responsible for acquiring the sound data through the asynchronous message communication.The sound data model matching module is responsible for pre-processing the input sound data,extracting the sound features and reducing it's dimension,finally,the sound feature data is matched and classified with the pre-trained voice recognition model.The sound data comprehensive analysis function module monitors,analyzes and abnormally alarms the operating state of the device according to the sound data model matching result.The system management function module is responsible for assisting the overall management of the system.The equipment monitoring system based on voice recognition technology built in this paper can not only identify the equipment fault well,but also give the fault cause that can be referenced according to the fault sound,even more,it can retrain the model when the new sound type appears to reduce the false alarm rate and the false negative rate,Effectively improved the shortcomings of traditional detection methods.