CAJ | 학술논문

针对传统的数学统计模型无法完全描述大坝变形量与多种荷载因素之间非线性映射关系的缺点，引入了一种基于遗传算法的小波神经网络模型，利用该模型对小波神经网络的初始权值、尺度因子进行全局优化搜索，克服了 BP 神经网络初始化的随机性以及网络易陷入局部极小值的不足，将该模型运用于大坝坝顶的径向、切向位移预测，结果表明，遗传算法优化的小波神经网络模型结构稳定性更好，预测精度较 BP 神经网络模型、小波神经网络模型有较大提高。
침대전통적수학통계모형무법완전묘술대패변형량여다충하재인소지간비선성영사관계적결점，인입료일충기우유전산법적소파신경망락모형，이용해모형대소파신경망락적초시권치、척도인자진행전국우화수색，극복료 BP 신경망락초시화적수궤성이급망락역함입국부겁소치적불족，장해모형운용우대패패정적경향、절향위이예측，결과표명，유전산법우화적소파신경망락모형결구은정성경호，예측정도교 BP 신경망락모형、소파신경망락모형유교대제고。
Aiming at the demerits of traditional mathematical statistical model couldn’t completely describe the nonlinear mapping relationship be-tween the dam displacement and various load factors,we introduced wavelet neural network model based on genetic algorithm to have a global opti-mization search for WNN’s initial weights and scale factor,the disadvantages that the net training might be easily fallen into local minimums and the BP neural network’s initialization was random could be avoided. This model was used in forecasting the radial displacement and tangential dis-placement of the dam,the results show that the genetic wavelet neural network’s structure stability is better,compared with BP neural network and wavelet neural network ,its prediction accuracy has improved greatly.