<p>This study explores the potential of artificial intelligence, particularly machine learning techniques, for predicting the critical loads of shell structures. Experimental data on the effects of various types of damage on the critical load of orthotropic cylindrical shells have been used as a training dataset. Eleven parameters have been considered as input variables influencing the critical load, including geometric characteristics, loading type, and damage configuration. The input data have been labeled and randomly divided into three training sets of varying sizes. The models trained using the Random Forest algorithm on these datasets have demonstrated high prediction accuracy and clear correlation between training set size and model accuracy. The results have confirmed the effectiveness and potential of machine learning methods for multifactorial analysis and stability prediction of shell structures, particularly in scenarios involving multiple influencing parameters.</p>