基于BP神经网络对上海纤维加筋土体强度仿真预测

Simulation Prediction of Strength of Shanghai Fiber-Reinforced SoilBased on BP Neural Network

摘要: 纤维加筋土是一种新型环保土体加固方法，为预测纤维加筋土的强度，通过BP神经网络对纤维土的强度预测进行研究。通过纳米二氧化硅和玄武岩纤维单掺数据来训练BP神经网络从而预测复掺数据。结果表明：预测的剪应力最大值与实际试验剪应力最大值差异范围在−2.350%到7.874%之间，预测得到的剪应力–剪切位移曲线决定系数基本都在0.9以上，说明了神经网络预测是可靠的，是一种对纤维土强度进行预测的一种可行办法，从而减少试验，降低成本，响应我国的“双碳”目标。

Abstract: Fiber reinforced soil is a new environmentally friendly soil reinforcement method. To predict the strength of fiber reinforced soil, a BP neural network was used to study the strength prediction of fiber reinforced soil. Train a BP neural network to predict complex doping data through single doping data of nano silica and basalt fibers. The results show that the difference between the predicted maximum shear stress and the actual test maximum shear stress is between −2.350% and 7.874%. The determination coefficients of the predicted shear stress shear displacement curve are generally above 0.9, indicating that neural network prediction is reliable and a feasible method for predicting the strength of fibrous soil, thereby reducing experiments, lowering costs, and responding to China’s “Dual Carbon Goals”.