Exploring the Predictive Performance of Simple Regression Models and ANN in 2D Truss Analysis

This research investigates the performance of various regression models in predicting critical structural parameters within a plane truss model. The study encompasses linear, second- and third-degree polynomial, and artificial neural network (ANN) regression models, which are evaluated for their accuracy in estimating the maximum displacement, maximum (tensile) stress, and minimum (compressive) stress of the truss under specific loading conditions. The findings unequivocally establish the superiority of the ANN model, showcasing its ability to capture complex nonlinear relationships within the data. Moreover, the research explores the influence of model complexity, demonstrating that the transition from simpler to more intricate models enhances predictive performance. The implications of this study extend to diverse engineering applications, offering insights into the selection of appropriate regression models for structural analysis and design. Beyond improved predictive accuracy, the ANN’s predictions provide potential for reducing computational demands, making them valuable tools in structural optimization and similar contexts. However, the study underscores the importance of cautious interpretation, as certain scenarios may yield outlier predictions. Overall, this research contributes to the understanding of regression modeling in engineering and provides a foundation for informed decision-making in structural analysis and design.