基于BP网络的混凝土碳化研究

运用改进的BP算法,建立了3 5 1(输入层为3个神经元,隐含层为5个神经元,输出层为1个神经元)混凝土碳化深度BP网络计算模型·计算模型以水灰比、单位水泥用量及砂率为输入,以碳化深度为输出,计算结果与试验结果符合较好·同时,运用改进的BP算法,建立了3 5 1(输入层为3个神经元,隐含层为5个神经元,输出层为1个神经元)混凝土碳化深度BP网络预测模型·预测模型以水灰比、单位水泥用量及混凝土暴露时间为输入,预测模型之一以暴露20年的混凝土碳化深度为输出,预测模型之二以暴露30年的混凝土碳化深度为输出,预测结果均较为理想·

Research Based on BP Network for Concrete Carbonation

The computation and forecast models (3-5-1),which include 3 neurons in input layer, 5 neurons in hidden layer and 1 neuron in output layer, are set up for concrete carbonation depth, based on optimized BP algorithm. In the computation model the water cement ratio, unit cement consumption and sand percentage are taken as inputs with concrete carbonation depth as output. But, in the forecast model the water cement ratio, unit cement consumption and tine for concrete exposure are taken as inputs with the carbonation depths of those concrete constructions/members which have been exposed to the weather for 20 years and 30 years as outputs. The computation results conform to the test results very well, and the forecast results are favorable.