Research on Vibration Prediction Method of Aircraft External Stores Based on Deep Belief Network

Abstract

This paper proposes a method for predicting the vibration environment of aircraft external stores based on deep belief network, which based on the relationship between the vibration environment, pressure altitude and Mach number of the aircraft external stores in flight. Firstly, GJB 150.16A empirical formulas were analyzed and then generating simulation data, which a total of 7470 sets. Secondly, linear regression and deep belief network models were set up respectively, and deep belief network number of hidden layer node was optimized, then the expected errors of two methods were compared and analyzed, including root mean square error (RMSE), mean absolute error (MAE) and mean relative error (MRE). Finally, DBN method was applied to a case for prediction the vibration of a certain aircraft store, intending to verify the feasibility and accuracy of DBN method. The results show that the DBN model can fully characterize the non-linear relationship between the pressure altitude, Mach number and RMS value of the aircraft external stores in the flight environment, and the overall expected effect is better than linear regression. The expected mean relative error is less than ± 3 dB. This method has provided a new way to predict the vibration environment of aircraft external stores, and is of great significance for PHM in the whole life of aviation weapon equipment.