多种群遗传算法在PBX本构模型参数识别中的应用

利用多种群并行结构对标准遗传算法SGA进行并行化处理，引入移民算子和精华种群形成多种群遗传算法MPGA，并设计了自适应交叉和变异概率对算法的收敛速度进行改进。结合ABAQUS软件和改进的多种群遗传算法，建立了材料本构模型参数识别方法。采用该方法对PBX炸药黏弹性损伤本构模型参数进行了模拟识别，并同基于标准遗传算法的参数识别方法进行了比较。结果证明，基于改进多种群遗传算法IMPGA的方法对克服算法未成熟收敛有显著的效果，识别结果更稳定。同时该方法的收敛速度更快，寻优能力更强，适合复杂非线性问题的优化，此方法可以被应用到其他材料本构模型的参数识别中。

Application of multiple-population genetic algorithm in parameter identification for PBX constitutive model

In this work, the standard genetic algorithm (SGA) was parallel processed using multiple parallel structures. Based on the structures, a multiple-population genetic algorithm (MPGA) was established by introducing the immigration operator and quintessence population. Self-adaptive operators of crossover probability and mutation probability were designed to improve the convergence speed of the MPGA. Combining ABAQUS with the improved multiple-population genetic optimized algorithm, a parameter identification method of constitutive model was built. Using the proposed method, a simulation example of parameter identification for PBX viscoelastic damage constitutive model was carried out. Comparison was made between methods based on SGA and MPGA. The results show that the MPGA method can effectively overcome the difficulty of the premature convergence and the identification result is more robust. The method is suitable for the optimization of complex nonlinear systems due to its superiority in the convergence speed and searching ability, and it can be applied to the parameter identification of other models.