Physics-Informed Evolutionary Artificial Intelligence for Intelligent Multi-Objective Collaborative Optimization of Vehicle Door Closing Performance

Abstract

To resolve the complex multi-objective conflict among high sealing airtightness, door closing energy, and ear pressure comfort in commercial vehicle cabs, this paper proposes a physics-informed evolutionary artificial intelligence (AI) framework for forward collaborative optimization. Firstly, a variable-parameter flow-field and energy coupling platform with continuous, independent adjustability (cabin volume, pressure relief area, sealing reaction force, and hinge inclination angles) was developed to realize the physical decoupling of multiple coupled fields. Based on an L27  orthogonal design, a physics-informed evolutionary symbolic regression algorithm driven by genetic programming was introduced to automatically discover explicit, interpretable governing formulations, establishing a high-precision nonlinear prediction model (R2>0.92). Furthermore, to overcome acoustic boundary differences and unstructured mechanical dissipation between the test bench and the actual vehicle, a physics-guided cross-domain transfer correction strategy was proposed, which successfully suppressed the real-vehicle prediction error to within 5%. Engineering verification on a production commercial vehicle demonstrates that through this intelligent collaborative optimization, the minimum door closing energy was remarkably reduced from 39.61 J to 12.36 J (a decrease of 68.8%), effectively mitigating the engineering contradiction of “difficult door closing” and “ear tenderness” while maintaining superior vehicle sealing performance. This study provides an efficient, interpretable, and prototype-free intelligent methodology for forward vehicle body development.