基于 BP 神经网络的水中双爆源爆炸冲击波峰值压力预测模型研究

In order to obtain the calculation model of peak pressure at shock wave coupling center when two explosion sources with equal mass exploded simultaneously in water, Autodyn was used to compute peak pressure data under different charge amounts and detonation distances. On the one hand, calculation formula of peak pressure was obtained by fitting the data in the function form specified by dimensional analysis. On the other hand, logarithmic transformation and normalization were performed on three types of data: total charge, detonation distance, and peak pressure, which were divided into training set and test set. The training set was then fed into the BP neural network for training, yielding a BP neural network prediction model with a relatively simple structure and a lowest mean square error. The findings reveal that the peak pressure predicted by the formula calculation model and the BP neural network model agrees well with the actual value. The average relative error between the calculated formula value and the actual value is 1.08%, while the average relative error between the projected BP neural network value and the actual value is 0.52%. It means that BP neural network can achieve higher accuracy predictions with a smaller data sample size compared with formula calculations.