Elastic Buckling Load Prediction of Tapered Steel Columns Via Artificial Neural Networks

Resumo

This study presents a novel approach for predicting the critical buckling load of slender, cylindrical,
tapered steel towers commonly used in wind turbines and telecommunications equipment. These towers are prone
to instability issues caused by buckling loads, which necessitates accurate evaluation. To overcome the limitations
of existing instability load formulations and regulatory codes, we developed an artificial neural network (ANN)
model. The ANN model utilizes a comprehensive database of 1,440 finite element models to accurately predict
the critical buckling loads. An MLPRegressor model instantiated with the 'adam' solver and the 'tanh' activation
function in the hidden layers demonstrated a significant alignment with the data, as the model accounted for
approximately 97% of the variance in the dependent variable. Furthermore, the outcomes obtained from the ANN
model closely aligned with the original values, surpassing the predictive precision of the classical shell and beam
formulations, and offering insights into the complexities associated with transformed data.