【摘要】 针对应用于结构健康监测中的径向基函数神经网络(RBFNN)算法和训练中存在的一些问题,提出一种将递阶遗传算法和最小二乘法相结合,用于优化径向基神经网络的结构和参数的新算法,遗传过程中采用自适应的交叉和变异概率有效地加快了遗传收敛速度和避免早熟现象的出现。搭建了复合材料结构健康监测实验系统,并将实验模态分析结果归一化后送入训练好的RBF神经网络进行预测,实现了对复合材料梁的脱层损伤定位和损伤程度的辨识仿真。结果表明基于混合递阶遗传算法的径向基神经网络(HHGA─RBF)收敛速度快,鲁棒性好,精度高。更多还原

【Abstract】 Aiming at some problems of the training algorithm of radial base function neural network used for structural health monitoring,a new hybrid hierarchy genetic algorithm was introduced by combining hierarchy genetic algorithm and least-square method.The hybrid algorithm was able to determine the structure and parameters of the RBF neural network.Adaptive crossover and mutation probability could accelerate the genetic speed and avoid the occurrence of prematurity.A structural health monitoring experimental system was developed,and the normalized experimental modal frequencies were fed into the radial base function neural network to predict the delamination location and its extent.The simulation demonstrates that the radial base function neural network based on hybrid hierarchy genetic algorithm is robust,promising and converges very fast.更多还原