Technology Research On The Condition Monitoring And Fault Diagnosis Of "Electronic Long Light"

"Torch lignt" is a device dealing with harmful emissions from oil,metallurgy,chemical and other fields in a way of burning natural gas.As the traditional long-lamp work consumes a lot of valuable gas resources and anti-environmental interference is poor,now "electronic long light" has been designed to solve the problems."Electronic long light" is achieved through the principle of high-voltage discharge ignition function,and it has the advantages of saving resources and strong anti-interference ability.However,because the "electronic long light" is a high-voltage discharge equipment,which makes troubleshooting difficult and has the security risks of electric shock.So,an intelligent device needs to be designed to monitor and diagnose the working conditions of "electronic long light".For the realization of "electronic long light" state monitoring and fault diagnosis function,the paper elaborates the research form two aspects of a real-time acquisition and recognition device design and recognition algorithm of "electronic long light"working conditions,which reads as follows:Firstly,for the design of a real-time acquisition and recognition device,the architecture of FPGA + DSP is used and extends the hardware interface of AD data acquisition,SD card storage,serial communication and so on.According to hardware design,the software driver needs to be designed and includes four parts.First,the control timing of AD chip is designed,and a FIFO cache storaging acquised signal is designed.Second,the communication DSP reading the signal in FIFO cache is designed.Third,it is designed that DSP transplants FATFS.Fourth,it is designed that DSP and FPGA configure registers of serial communication chip,and communication funtions of RS232 and RS485 are designed.Finally,the device achieves the functions of signal acquisition,conditional storage of SD card and serial communication.Secondly,for the recognition algorithm of "electronic long light" working conditons,feature extraction is researched deeply.The features of working condition for "electronic long light" are resecrched in different aspects from time domain statistics,spectrum envelope and combination of signal processing and cepstrum to extract the characteristics of high recognition degree.Besides,the theory and design method of BP neural network are introduced.Finally,BP neural network will treat extracted characteristic as identification parameter to be build and identify working condition of "electronic long light".According to experimental verification,diagnosis accuracy of neural network can reach more than 90% for "electronic long light".