基于卷积神经网络的固体火箭发动机内弹道参数辨识

In order to judge whether the interior ballistic parameters of solid rocket motor after long-term storage change, the parameters of solid propellant stored for a long time were identified by means of the convolutional neural network-AlexNet. Firstly, the pressure-time curve images of combustion chamber were drawn by using solid rocket motor interior ballistic program, and several images were used as training sample sets; then, the convolutional neural network model was obtained by training the sample set with convolutional neural network; finally, the identified image was put into the model to obtain the internal ballistic burning rate coefficient and pressure index, so as to calculate the identified burning rate under the certain pressure. The experimental results show that the accuracy of training results increases with the increase of the proportion of training sets. The number reduction of the images in the training set may lead to the improvement of the accuracy, but may reduce the universality of the trained neural network. The comparison between the identification results and the experimental results shows that the burning rate error can be controlled within 1%, especially when the number of images in the sample set is certain. Therefore, the model can be used to judge whether the internal ballistic parameters change quickly and accurately.