论文部分内容阅读
为实现HMX爆轰临界厚度的神经网络预测,采用长通道楔形槽内直接压药的装药方式,测量了HMX在不同密度、粒度与不同粘接剂含量下的爆轰临界厚度。以此3个因素作为输入节点、爆轰临界厚度作为输出节点建立了HMX爆轰临界厚度的BP神经网络预测模型。通过试验数据对神经网络预测模型进行训练使之达到误差要求,采用训练好的预测模型对不同密度、粒度与粘接剂含量的HMX爆轰临界厚度进行预测。预测结果表明,HMX的粒度、压药密度、粘接剂含量对爆轰临界厚度的影响与相关文献报道相同,说明神经网络可以用于HMX爆轰临界厚度的预测。
In order to predict the critical thickness of HMX detonation, the critical detonation thickness of HMX at different densities, particle sizes and different adhesive contents was measured by using the direct channeling method in wedges with long channels. Taking these three factors as input nodes and the critical thickness of detonation as the output node, a BP neural network prediction model of HMX detonation critical thickness is established. The neural network prediction model was trained to meet the error requirement through the experimental data. The predicted HMX detonation critical thickness with different density, particle size and adhesive content was predicted by the trained prediction model. The prediction results show that the influence of particle size, pressure density and adhesive content of HMX on the critical thickness of detonation is the same as the related literature, which shows that the neural network can be used to predict the critical thickness of HMX detonation.