摘要
以BP神经网络在解决非线性问题方面的优越性为指导思想,建立了再生混凝土28 d抗压强度预测模型。利用具有全局搜索优势的遗传算法对BP神经网络进行优化,提高BP神经网络的预测精度。经过查阅相关文献,筛选出较为合适的L-M算法作为训练函数,把收集到的再生混凝土配合比数据代入网络进行训练与预测。结果表明,GA-BP神经网络预测精度较BP神经网络有了进一步提高。
Based on the superiority of BP neural network in solving nonlinear problems,establishing the forecast model about the 28 d compressive strength of recycled concrete.Genetic algorithm with global search advantage is used to optimize BP neural network to improve the forecasting precision of BP neural network.The collected recycled concrete mix proportion data was sent to the net for train and forecast,using the algorithm called Levengerg-Marqardt which is selected after refering to the reference.The results suggest that the forecasting precision of GA-BP neural network is preferable.
作者
罗巍
王浩然
甘凯元
谭静雯
温宇嘉
Luo Wei;Wang Haoran;Gan Kaiyuan;Tan Jingwen;Wen Yujia(College of Civil Engineering,Guangxi University,Nanning 530004,China)
出处
《山西建筑》
2021年第20期128-130,150,共4页
Shanxi Architecture
基金
广西壮族自治区大学生创新训练计划项目(202010593031)。
