The structural integrity and fatigue life of engine connecting rods are critical to ensuring reliability and performance in internal combustion (IC) engines. Traditional Finite Element Analysis (FEA) methods for stress and life prediction are computationally expensive, requiring extensive simulation time for varying loading conditions. This study proposes an Advanced AI-driven approach using Graph Neural Networks (GNNs) which is subset of Geometric deep learning (GDL) to predict stress distribution and fatigue life of a connecting rod based on historical simulation data.
The methodology involves training on past high-fidelity FEA results, enabling the model to learn spatial stress patterns and fatigue behavior across different design variations and loading conditions. Unlike traditional models, GNNs effectively captures the geometric and topological dependencies inherent in the connecting rod structure, providing robust predictions with minimal computational overhead.
Experimental validation is performed by comparing AI-predicted stress and life results with full-scale FEA simulations. The proposed approach achieves high prediction accuracy (>90%), significantly reducing computational time by up to 90% while maintaining engineering precision. This framework enables rapid design iteration, optimization, and near real-time stress evaluation, making it a valuable tool for automotive and aerospace industries seeking efficiency in structural component analysis.