Research Of Fault Diagnosis Methodes Of Beamline In Ssrf Based On Neural Networks

SSRF(Shanghai Synchrotron Radiation facility) is the third generation synchrotron radiation sources, is currently one of the largest scientific devices in China, and is made up of the full energy injector, electronic storage ring, beam lines and experimental station. It is an extremely complex systems engineering, contains many of the sub-system. The device normal operation must be established on the basis of safety work of each subsystem, which requires the running state of each subsystem for reliable monitoring, in order to keep abreast of and handle the faults; and an effective way in treatment of system fault is to carry out system fault monitoring and early warning, will timely detection of fault or eliminate the investigation, accomplish a nip in the bud. The fault pre-warning is useful with national economic and social benefits, and has built relatively perfect processing system in many areas, which is protecting the normal operation of the industry related to the national economy and the continuous development, and brings gratifying profits. The SSRF can also make a contribution to the development of the national economy, can provide users with more time to carry out experimental research and service to scientific innovation, enhance the enterprise’s ability to create; furthermore reduce the fault time can reduce waste, reduce operating cost, improve the operation efficiency of the SSRF.The research of the equipment operation state analysis and fault early warning method in SSRF aimed at improving the beamline equipment operating parameters meet the growing user data processing needs, and to explore the effective method of large plant operation data analysis and processing, and prevent fault occurrence or ahead of time to deal with the possible failure, to assist in the implementation of fault diagnosis and reduce the fault processing time. The beamline fault pre-warning method research adopts the MATLAB neural network, comparison and analysis of several other data processing methods. And the paper gives out the analysis result of the temperatures record data of the monochrometer at BL17 U, and pointed out this method can achieve the pre-warning of the beamline running data. The operation data of the beamline is complex, the sensor data up to massive, when fault occurs, each subsystem also interfere with each other, so the selection of suitable fault diagnosis methods is very important for fault prediction. Neural network is to imitate the biological nervous system, the neural network has good robust, adaptive and nonlinear mapping ability, has been widely used in dealing with nonlinear problems; the beamline operation data are nonlinear, complex, a number of features, and the expression form and characteristics of the recorded data of each device are not the same, no fixed analysis model for reference in SSRF. The work is going to analyze the beamline operation data by artificial neural network method, observation results of neural network application in SSRF, and for equipment fault diagnosis to find more effective means or method.Through the analysis of the running data of the beamline equipment, and put forward the feasibility of fault early warning analysis by using Matlab software. Via programming, the data was extracted from the Archive database under EPICS, and to filter or to clean, in order to ensure the authenticity and practicality of data analysis; application of neural network data analysis, selection of nerve network pattern recognition, based on the liquid nitrogen cooling monochromator temperature signal of the two beamline, the analysis shows that Matlab neural network can according to the change of data gives the corresponding signal warnings; and will get knowledge of the data set available through neural network, neural network training and adjustment, achieve the purpose of identification and early warning. The paper is also carried out on the off-line and on-line test to check the correctness of the fault diagnosis. The result proves that the artificial neural network can be applied to the early warning research of the beamline running data in SSRF, and can successfully carry out the warning prompt, and can classify the warning.The innovation of this paper is put forward a neural network data processing method of analysis of the operation data of the beamline equipment, services for the SSRF; analysis with Matlab neural network method for the beamline equipment pre-warning, reduce because of equipment failure and stop for a time, to further improve the efficiency of the beamline safe running; introduction of the Matlab programming, using Matlab’s powerful data processing function, effectively provided high beamline operation data analysis and processing capacity.