Application Of Improved BP Algorithm And Fuzzy Logic In The Two-Axis Turntable Fault Diagnosis

Turntable is a test and simulation equipment of inertial navigation system, which is a complex set of mechanical and electrical and plays a huge role in the aerospace, aviation and marine areas. Therefore it is very necessary to carry out the research on fault diagnosis technology to make the turntable work stably and reliably for a long time.In this paper, fuzzy neural network, which has a powerful fuzzy information processing and self-learning ability, is utilized to deal with the complex nonlinear mapping relationship between the failure and symptoms of the turntable. A fault diagnosis system for the turntable based on fuzzy neural network was built and software programs to achieve the establishment of diagnosis system were finished.First, the common failures in the turntable is divided into such five categories as mechanical failures, angle measurement system failures, communication failures, controller failures and actuator failures based on a in-depth analysis of the system. According to the relationship between the turntable failures and the fault signal characteristics, a decision-making table was established. In order to eliminate high frequency noise in the signals, this paper used the wavelet analysis theory and designed a second-order low-pass filter. At the same time, the threshold function was improved, and achieved good results.Secondly, the paper systematically introduced the fuzzy neural network model used in diagnosis system and its learning algorithm. Three methods, including the genetic algorithm, the introduction of inertia and the dynamic adjustment of learning rate, were utilized to improve BP algorithm for the shortcomings of BP algorithm. And the improved BP algorithm was applied to fault diagnosis of the turntable. The method of anti-noise training is utilized to train the neural network, and this method can make it overcome the noise under certain amplitude. Compared to the standard network, fuzzy neural network which was improved has good global approximation ability to achieve accurate identification of faults.Finally, the establishment of diagnosis system was achieved in a mixed way including Delphi and Matlab programming technology. The diagnosis system has function modules: neural network training, wavelet denoising, database management and diagnosis. It can real-time monitor the state of equipment.