CAJ | 학술논문

针对误差反向传播（BP）神经网络易陷入局部极小值的问题,将遗传算法（GA）与BP神经网络相结合,先以遗传算法全局最优的特性对初始化的BP网络的权重和阈值进行优化,再将优化的权重和阈值作为初值带入BP网络训练得到最优解.运用此改进的BP神经网络对竹盖山隧道初期支护钢拱架内力进行预测,取得了良好的效果,精度高、收敛快,为指导和控制工程施工提供了有效的依据.
침대오차반향전파（BP）신경망락역함입국부겁소치적문제,장유전산법（GA）여BP신경망락상결합,선이유전산법전국최우적특성대초시화적BP망락적권중화역치진행우화,재장우화적권중화역치작위초치대입BP망락훈련득도최우해.운용차개진적BP신경망락대죽개산수도초기지호강공가내력진행예측,취득료량호적효과,정도고、수렴쾌,위지도화공제공정시공제공료유효적의거.
This paper combined the genetic algorithm(GA)with the Back Propagation(BP)neural network to overcome the problem that BP algorithm is prone to local minimum.First,using the genetic algorithm global optimization characteristic optimized the weights and biases of a initialized BP network,then put the weights and biases as initial values into BP network for further training for searching optimum solution.It works well on the prediction of the stress about steel arch in the tunnel initial support of Zhugai Shan with this optimized BP neural network,which has high convergent speed and good prediction precision,and it is helpful to provide basis for controling the engineering construction.