基于BP网络的粉煤灰混凝土钢筋握裹力计算

Calculation of Bonding Strength of Fly-Ash Concrete Based on BP Network

基于改进的BP算法，建立了2个粉煤灰混凝土钢筋握裹力BP网络计算模型，模型1为2-6-1型，即该模型输入层为2个神经元，隐含层为6个神经元，输出层为1个神经元，模型1的输入为水胶比及粉煤灰掺量，输出为混凝土钢筋握裹力；模型2为2—8—2型，即该模型输入层为2个神经元，隐含层为8个神经元，输出层为2个神经元，模型2的输入为水胶比及粉煤灰掺量，输出为混凝土强度及钢筋握裹力．模型1混凝土钢筋握裹力计算相对误差为0．01908％～3．12892％；摸型2混凝土强度及钢筋握裹力计算相对误差分别为2．24855％～6．80800％和0．11274％～9．77329％．计算结果较为理想．

Based on the optimized BP algorithm, two computation models are developed for the bonding strength of fly-ash concrete which wraps reinforcements. The model 1 is of 2-6-1 type, i. e., 2,6 and 1 neurons are in input, hidden and output layers, respectively, where the input is defined including the water-binder ratio and fly-ash content and the output as the bonding strength of the concrete. The model 2 is of 2-8-2 type, i.e., 2, 8 and 2 neurons are in input, hidden and output layers, respectively, where the input is the same to that in model 1 but the output includes the compressive strength in addition to the bonding strength of the concrete. The relative error of the bonding strength calculated through model 1 is 0.019 08 %- 3. 128 92 %, while the relative errors of the compressive strength and bonding strength calculated through model 2 are 2.248 55% -6.808 00% and 0. 112 74% -9.773 29%, respectively. Such results are quite .satisfactory.