有限元融合RNN对基坑开挖的影响性研究

Research on the influence of finite element fusion RNN on foundation pit excavation

对于基坑开挖卸载导致地表沉降等问题,笔者以贵阳大型污水处理厂为工程背景,选用通用有限元软件ABAQUS和岩土工程专业有限元软件MIDAS GTS NX用不同的本构模型对背景工程中深基坑开挖过程进行数值模拟,并比较分析了两种有限元软件对地表沉降和围护桩测斜的模拟结果。再通过循环神经网络模型和有限元模型结合的融合模型对有限元模型因材料参数取值偏差较大和简化实际工况等问题导致的模拟误差进行修正和优化调整,最终的预测误差能较有限元模型减小近50%,且预测结果既能考虑基坑开挖的力学影响,又能更好地符合数据的波动性特征。

For problems such as ground settlement caused by excavation and unloading of foundation pits,the author uses general finite element software ABAQUS and geotechnical engineering professional finite element software MIDAS GTS NX to numerically simulate the same large-scale deep foundation pit excavation project with different constitutive models,and The simulation results of the two kinds of finite element software on the ground settlement and the slope measurement of the retaining pile are compared and analyzed.The fusion model combined with the cyclic neural network model and the finite element model is then used to correct and optimize the simulation errors of the finite element model due to the large deviation of the material parameters and simplify the actual working conditions.The final prediction error can be compared with the finite element model.The model is reduced by nearly 50%,and the prediction results can not only consider the mechanical impact of foundation pit excavation,but also better meet the volatility characteristics of the data.