基于卷积神经网络和鱼鹰算法优化BP神经网络预测活性粉末混凝土耐久性研究

Research on predicting the durability of reactive powder concrete based on CNN and OOA-BP

由于原材料原因,活性粉末混凝土(RPC)的耐久性表现出高度非线性行为,较难预测。本文研究了两种人工神经网络在预测RPC耐久性中的应用。通过卷积神经(CNN)和鱼鹰算法优化-BP神经网络(OOA-BP),以腐蚀龄期和腐蚀溶液浓度为变量,对RPC腐蚀前后的抗压强度进行预测分析,并对未参与训练的数据进行预测验证。将预测结果与试验结果比较,结果表明,两种神经网络对RPC耐久性的预测均有良好的潜力,CNN有更大的灵活性和准确性。

Due to the limitations of raw materials,the durability of reactive powder concrete(RPC)exhibits highly nonlinear behavior,making it difficult to predict.This paper investigates the application of two artificial neural networks in predicting RPC durability.Specifically,the compressive strength of RPC before and after corrosion is predicted and analyzed using Convolutional Neural Network(CNN)and Osprey Optimization Algorithm BP Neural Network(OOA-BP),with corrosion age and corrosion solution concentration as variables.And the prediction validation is performed on the data that did not participate in training.Comparing the predicted results with the experimental results,it is found that both neural networks have good potential for predicting the durability of RPC,and CNN has better flexibility and accuracy.