基于BP神经网络的固化红土抗压强度预测

Compressive strength prediction of solidified laterite based on BP neural network

为分析不同掺量的偏高岭土与石灰共同掺入玄武岩残积红土中对土体的改良效果,本试验选取偏高岭土的掺量分别为0%、2%、4%、6%和8%,石灰的掺量分别为0%、2.5%、5.0%、7.5%和10.0%,同时掺入玄武岩残积红土中,制作25组不同固化红土,对其进行28 d无侧限抗压强度正交试验,并用MATLAB软件建立神经网络预测模型,预测固化红土养护28 d的抗压强度。研究结果表明:本模型预测误差最大为4.56%,拟合度为0.997,且本方法比常规回归分析法更简单、更准确,可预测不同固结材料和掺量的固化红土抗压强度,提高试验效率。

In order to explore the improvement effect of different dosages of metakaolin and lime mixed with residual soil of basaltic weathered soil,this experiment selected kaolin dosages of 0%,2%,4%,6%,and 8%,and lime dosages of 0%,2.5%,5.0%,7.5%,and 10.0%,simultaneously mixed with basalt residual laterite to prepare 25 groups of differently solidified red soils.An orthogonal test on the unconfined compressive strength of the soils was conducted for 28 days,and a neural network prediction model was established by MATLAB to predict the 28-day compressive strength of solidified laterite.The results show that the maximum prediction error of the model is 4.56%,with a coefficient of determination of 0.997.Furthermore,this method offers simplicity,higher efficiency,and greater accuracy over conventional regression analysis techniques,which enables the prediction of the compressive strength of solidified laterite with different consolidation materials and dosages,thereby improving experimental efficiency.