Prediction of Separation Shock Environment Based on Genetic Algorithm and BP Network

Abstract

The effect of the shock environment induced by clamp band satellite-rocket separation is usually analyzed by the shock response spectrum (SRS). This paper presents a method of predicting SRS data in a new shock environment on the basis of great accumulated shock tests. Firstly, the shock response spectrums that correspond to the time domain signals of the existing shock tests are calculated by means of the improved recursive digital filter method. Secondly, genetic algorithm and back propagation (GA-BP) neural network are introduced to obtain the internal mapping relationship between configuration parameters of the separation mechanism and SRS data. Finally, prediction of SRS data in a new shock environment can be performed according to the trained GA-BP network under the precondition of having known the configuration parameters. In the application example, the data other than the training samples are used to verify the trained network, and the results show the errors are within the permissible range, which shows the proposed method is feasible and effective.