改进BP神经网络的再生粗骨料混凝土强度预测*

抗压强度作为评价再生混凝土主要性能指标之一,从四篇文献中搜集了43个抗压强度样本。基于改进的BP神经网络理论,建立了以再生粗骨料取代率、水灰比、龄期为输入,抗压强度为输出的神经网络模型,其结构形式为3-7-9-1。对该网络进行训练和学习,对并测试样本进行检测,结果表明,训练效率不仅大大提高,且而误差较小,并结合回归拟合系数,进一步表明该模型能够有效的的预测再生混凝土抗压强度值,能满足工程需要。

Strength prediction of recycled coarse aggregate concrete with improved BP neural network

As one of the main performance indicators for compressive strength of recycled concrete, 43 samples were collected from four literatures. Based on the improved BP neural network theory, a neural network model was established with the replacement of recycled coarse aggregate, water-cement ratio and ages as the input parameters and the compressive strength as the output result. Its structure is 3-7-9-1. After the training and learning of the network and testing of test samples, the results show that the training efficiency is not only greatly improved, but also the error is very small, and the regression fitting coefficient further shows the model can effectively predict the compressive strength of recycled concrete. Moreover, the model meet also the engineering needs.