混凝土强度预测的混沌优化神经网络模型

BP神经网络方法由于综合考虑了高强度、高性能混凝土强度的各种影响因素，能够实现非线性关系，可以用于混凝土强度预测上。为克服传统BP网络收敛速度慢、易出现麻痹现象等不足，本文采用自适应变步长(ABPM)算法来改进的BP神经网络，提出了基于混沌优化的自适应变步长(ABPM)神经网络模型，并将其预测结果和训练效率进行了分析。该方法主要利用混沌运动的遍历性为梯度算法创造一个良好的搜索界面。仿真结果表明，混沌优化的ABPM神经网络用于混凝土强度的预测，方法简单可行，搜索速度快，预测结果可靠、精度高。

THE CHAOTIC OPTIMIZATION NEURAL NETWORK MODEL OF THE CONCRETE STRENGTH

Because main factors of influencing high strength and high performance concrete strength are considered, BP neural network can deal with non-line problem and forecast concrete strength. To overcome shortage of slow constringency speed and anesthesia phenomenon in conventional BP network, this paper, the back-propagation (BP) network is trained using adaptive variable step-size arithmetic. ABPM neural network model with chaotic optimization is put forward. The forecast result and training effect are analyzed and compared. Its excellence is mainly a good search interface in grads arithmetic using searching characteristic of chaotic motion. Calculating result makes know that this method is used to speed up the convergence and improve the performance.