Predicting Ejection Velocity of Ejection Seat via Back Propagation Neural Network

摘要

The ejection velocity of the escape system, which is a primary parameter of the sequencer control subsystem, determines the parachute shooting time. It is found that, in some certain circumstances, a large error of the measured velocity is generated, which significantly influences the performance of the escape system. In this paper, a method that predicts the ejection velocity by a neural network was presented. Based on the mathematical model, a module simulation solver on the MSC.EASY5 fundamental platform was developed and programmed. According to this solver, considerable ejection conditions that were sufficient to contain all the representative situations in the lifesaving envelope of the escape system were calculated. Then, the relationship between the ejection velocity and other parameters could be obtained from the simulation results. Subsequently, a back propagation neural network was established to fulfill the relationship. Further experimental validation indicated that the error could be accepted in engineering application. Consequently, the method that predicts the ejection velocity via a back propagation neural network was proved to be feasible and would be a useful technology for the escape system.