基于BP神经网络的混凝土抗压强度预测模型的建立

本文采用误差反向传播神经网络(BP)建立了混凝土抗压强度的预测模型,在不同的误差目标值及隐含层节点数条件下,探讨模型的预测精度。实验结果表明,当误差目标值设定为0.01、隐含层节点数为3时,模型的预测精度最高,平均误差百分数为6.6%。当误差目标值设定较大时,样本的预测值与实测值会发生明显的偏差,预测效果不佳。

In this paper,the error back propagation neural network(BP)was used to predict the compressive strength of concrete.Under different conditions of error target value and number of hidden nodes,the prediction accuracy of the model was discussed.The experimental results showed that when the error target value was 0.01 and the number of hidden layer nodes was 3,the prediction accuracy of the model was the highest,and the average error percentage was 6.6%.When the error target value was set large,the difference between predicted value and the measured value would be huge,and the prediction effect was not good.