Machine Learning Models to Predict Buckling Strength of Steel Beams According to TCVN 5575:202X

The draft Vietnamese steel structure design standard TCVN 5575:202X is nearly issued with many changes compared to the current version. Some recent research results have been updated in this version, for example, section classification and plastic stress distribution. Even so, determining the critical moment for lateral-torsional buckling according to TCVN 5575:202X is still quite complicated with many calculation steps. This paper proposes a new approach to predict the critical moment of steel beams using machine learning. Firstly, a large amount of data is generated in which the inputs are the effective length as well as the section dimensions, while the output is the corresponding critical moment values that are determined using the procedure described in TCVN 5575:202X. There are two separate datasets, one for training model and the other for testing model. Several machine learning regression algorithms, including support vector machine, random forest, artificial neural network, and adaptive boosting, are employed to build the prediction model. The performances of these models are compared through three metrics: MAE, RMSE, and the coefficient of determination. The obtained results show that the random forest outperforms the three remaining models for this task. A numerical example is then conducted to confirm the applicability of machine learning for predicting the buckling strength of steel beams. The value predicted by the random forest model is < 2% different from the value calculated by theoretical formulas.