基于参数化有限元与机器学习的钢筋混凝土柱耐火极限快速预测代理模型研究OA

The fire resistance limit of columns is a critical factor affecting the overall fire performance of structural systems.Traditional prediction methods generally rely on empirical formulas or finite element analysis(FEA),making it difficult to rapidly obtain relatively precise prediction results.To address these limitations,this study proposed a method for constructing a rapid prediction surrogate model for the fire resistance limit of reinforced concrete(RC)columns under fire,integrating parametric finite element modeling and machine learning.Through the secondary development of ABAQUS using Python,a parametric automatic modeling program and a feature point post-processing system were constructed,efficiently generating 3 600 thermo-mechanical coupled FEA models of RC columns.The reliability of the program and system was validated against existing experimental results.On this basis,taking machine learning as a surrogate for high-fidelity FEA,a rapid prediction surrogate model based on the gradient boosting decision tree(GBDT)core algorithm was established.Through hyperparameter optimization,the optimal parameter combination was identified,achieving rapid multi-objective evaluations for fire resistance time and failure displacement.The results demonstrate that the surrogate model achieves a high coefficient of determination of 0.93 on the test set,with a mean absolute error below 11 minutes,and a 96%classification accuracy for failure states.It significantly outperforms comparative models based on other algorithms while meeting engineering accuracy requirements,significantly improving the analysis efficiency.