Abstract <p>Square dents are common defects in oil and gas pipelines. In the safety assessment of pipelines containing dents, it is critical to accurately and rapidly predict the strain of dent-affected pipelines. Currently, there is limited research both domestically and internationally on strain prediction models for square dented pipelines, and the related work on predicting the mechanical behavior of dented pipelines remains relatively immature. Building on this foundation, this study proposes a model based on a Backpropagation Neural Network (BPNN) to predict the maximum equivalent plastic strain in the curved dented regions of pipelines. This study employed finite element software to construct a static analysis model for square dented pipelines, and the reliability of the finite element model was validated through relevant experiments. Based on the finite element model, the Pearson correlation coefficient method was used to analyze the interdependencies between maximum equivalent plastic strain and various key parameters. The parameters were ranked according to their correlations to construct a comprehensive training dataset. Using the Backpropagation algorithm and optimizing the number of neurons in the BPNN, a strain prediction model was established utilizing the constructed dataset. The model was used to predict the maximum equivalent plastic strain at the curved dented regions of the pipeline, and its stability in predicting this strain was verified against experimental data and a random dataset. The results show that the predictions exhibit minimal deviation from the experimental data and random dataset, indicating that the model can accurately predict the strain behaviour of square dented pipelines. The predictive model established in this study provides a significant reference value for the assessment of square dented pipelines in practical engineering applications.</p>