基于BP神经网络的自密实清水混凝土性能预测研究

Research on performance prediction of self-compacting fair-faced concrete based on BP neural network

针对自密实清水混凝土试验周期长、表观性能影响因素多等问题,应用BP神经网络对其性能预测,有效减少试验量,快速找出外加剂最优掺量。基于BP神经网络卓越的非线性处理功能,将减水剂、消泡剂、引气剂、坍落度作为输入变量,自密实清水混凝土的7 d抗压强度、扩展度、气孔面积、色差作为输出变量,建立含有2层隐含层的BP神经网络模型,利用试验所得12组数据,预测自密实清水混凝土的性能,将预测值与试验值进行比较,确保模型高精确度。结果表明:神经网络模型预测结果良好,强度预测的相对误差最高达到10.8%,其余均在10%以下,其中第11组的混凝土性能最优,预测与实际结果相吻合。

Aiming at the characteristics of self-compacting fair-faced concrete such as long test period and many factors affecting its apparent performance,the application of BP neural network to its performance prediction can effectively reduce the amount of experiments and quickly find the optimal amount of admixtures.Based on the excellent non-linear processing function of the BP neural network,the water reducer,defoamer,air-entraining agent,and slump are used as input variables,and the 7 d compressive strength,expansion,pore area,and color difference of self-compacting fair-faced concrete are used as output variables,establish a BP neural network model with 2 hidden layers,use 12 sets of experimental data to predict the performance of self-compacting fair-faced concrete,and compare the predicted value with the experimental value to ensure the high accuracy of the model.The results show that the prediction result of the neural network model is good,the relative error of the strength prediction is up to 10.8%,and the rest are below 10%.Among them,the 11th group has the best concrete performance,and the prediction is consistent with the actual results.