基于BP神经网络遗传算法的药型罩优化

线性聚能装药爆破效果的影响因素有很多,且各因素的影响多是非线性的,而且非常复杂,其中药型罩结构的优化设计一直是重点。为了探索有效的药型罩优化方法,本研究对楔形罩运用正交试验法设计方案,利用ANSYS/LS-DYNA进行数值模拟获得结果,再以结构参数和最大射流速度分别作为BP神经网络的输入和输出进行训练,并将预测值作为适应度,结合遗传算法对药型罩进行最优结构药型罩参数和最优最大射流速度搜索。结果表明,该方法能够结合正交试验法和BP神经网络遗传算法的优点,快速精确地进行药型罩结构优化。

Charge optimization based on BP neural networks and genetic algorithm

Many factors can affect the performance of LSC, and most of these effects are nonlinear and complicated. The optimization design of charge structures is noticed widely. For efficient optimization methods, the wedge charge respectively was considered as research objects. First, the orthogonal experimental method was used to design different programs, and the ANSYS/LS-DYNA was used to obtain simulation results. Then, the structural parameters and the jet velocity maximum were set as the input and output of BP neural networks for training, and the prediction result was set as the fitness. Finally, the genetic algorithm was applied to search best structural parameters and jet velocity maximum of the wedge charge respectively. The study results indicated that this method can combine advantages of the orthogonal experimental method, BP neural networks and genetic algorithm for efficient optimization of charge structures.