Research On Monitoring Online And Fault Diagnosis Of Engineer Structure Based On Neural Network

Monitoring online and fault diagnosis are helpful to detect the state of a structure,to find and locate faults and to determine the extent to which the structure is impaired without delay. So they play an important role in timely structure maintenance,in prolonging service life of the structure and in preventing calamitous accident. Artificial neural network is a kind of nonlinear dynamic network built by imitating human brain nerve cell structure based on modern nerve physiology and psychology. The main feature of artificial neural network lies in its adaptability,that is,a strategic area can be formed automatically through its learning mechanism. For this reason,it is widely used in fault diagnosis.This thesis is mainly concerned with how to apply to monitoring online and fault diagnosis with the transducer technology,signal receiving and processing technology,artificial neural network and BP arithmetic,fault diagnosis technology,structure dynamics,MATLAB software,SuperSAP software,DSPS system. When the structure is excited,the monitoring system can automatically pick up signals,gain feature parameters. The well-trained BP artificial neural network,can determine whether the structure has fault when the parameter is input into it. If so,BP artificial neural network can locate the fault and the extent of impairment of the structure. Monitoring online of the cantilevering beam can be realized through accelerating transducer and DSPS system. In this thesis,a program based on neural network arithmetic has been designed with the help of MATLAB toolbox. In the process of writing the thesis,an experiment was done about cantilevering beam monitoring online and fault diagnosis. Useful training samples result from the experiment and the calculation for the emulational faults via SuperSAP software.Several conclusions can be drawn from the thesis. First,monitoringonline can be realized through DSPS system,transducers technology,and signal processing technology. Second,it is feasible to apply BP artificial neural network to fault diagnosis,and this can bring its adaptability to the best of its potential. Third,whether the structure has fault can be determined through well-trained BP artificial neural network by taking structure frequency as the feature parameter. Fourth,the experiment of cantilevering beam and the digital calculation reveal that frequency is easy to get and can reflect the state of the cantilevering beam. Under certain conditions,the feature parameter can be used to determine whether the cantilevering beam has fault,where the fault locates and the extent of impairment.