Research On Multiple Fault Diagnosis Of The Radio Compass Based On Multistage BP Neural Network

Piezo Neural Network is a multidisciplinary field of high technology. From nowon it has raised a number of physiologists, mathematicians, computer and informationengineering researchers and psychologists’ greatly importance and in-depth research.Most focused on the application of pattern recognition of neural network system, faultdiagnosis, economic forecasting, optimization, control, and so on, these apply to oureconomic life, provides many facilities for scientific research. In-depth study of theneural network is especially important.Generally the difficulty of positioning in electronic equipment fault diagnosissystem characterizes two aspects. It’s hard to ensure the accuracy and efficiency.Besides, it has a variety of forms: single fault, multiple faults, connection failure, etc.The single fault assumption is not always occurs in the traditional troubleshootingactivities. However, because of the phenomenon of multiple faults, we spend moretime in solving the problem of fault diagnosis.Firstly, the paper describes the background and significance in the neuralnetworks, and then it instructions advantages and disadvantages of existing BPalgorithms from the structure and characteristics. Through the MATLAB software, wecan simulation and analysis the existing BP algorithms. Finally the paper puts forwarda new algorithm based on the existing algorithm. According to the simulation results,the paper draws a conclusion that the improved neural network algorithm is moreeffective from the aspects of training speed and mean square error in the circuit faultdiagnosis.This paper analyzes and summarizes the related research results based on neuralnetwork fault diagnosis; Then it introduces a tear by network method is the premiseof circuit decomposition methods and network design method based on BP neuralnetwork; And according to the simulation examples I constructed the correspondingneural network topology structure, MATLAB is used to train the network and applyfor the fault detection.The method of BP neural network theory and network tearing principle isapplied to the WL-7radio compass circuit in this thesis which presents a multiple fault diagnosis method based on multilevel BP neural network. Complex avionics aredecomposed into several sub networks that is Built up the fault sets. Finally, it getsthe verification results through the simulation software in the paper.