偏高岭土高性能混凝土抗压强度预测研究

以水胶比、用水量、砂率、水泥强度、水泥用量以及偏高岭土、矿粉和粉煤灰用量为输入向量,采用SPSS回归方程分析和基于Levenberg-Marquart算法的BP神经元网络预测模型对偏高岭土高性能混凝土的抗压强度进行了预测研究,并与试验值进行了对比.结果表明:与SPSS回归方程分析预测结果相比,BP神经元网络预测值与实测值线性拟合度高,拟合值为0.997,两者之比的平均值和标准差分别为0.999和0.010,网络预测最大相对误差不超过2.1%,模型预测精度高,结果可靠,为偏高岭土高性能混凝土的抗压强度预测提供了指导依据.

Prediction of Metakaolin High-performance Concrete Compressive Strength

With the water-binder ratio , water content , sand rate , strength of cement , cement content and the content of metakaolin , slag and fly ash as input variables , SPSS regression equation analysis and BP neural network model based on Levenberg-Marquart algorithm were developed to predict the compressive strength of metakaolin high-performance concrete and compared with the experimental values .The results show that:compared with the SPSS regression equation analysis forecast results , predictive values of the BP neural network and the measured values have a high fitting degree , the fitting result is 0.997.Ratio between the BP predictive values and the trial values is 0.999, and standard deviation is 0.010.The maximum relative error of the predictive values less than 2.1%.The BP neural network model has high prediction accuracy , reliable prediction results , and provides guidance basis for the compressive strength prediction of metakaolin high-performance concrete .