A predictive passenger car seats comfort model by using deep learning and CAD-CAE methods

Car seat comfort is a critical aspect of automotive ergonomics, yet current design practices remain largely dependent on subjective evaluations and costly physical prototyping. This study introduces a data-driven framework that combines statistical analysis and deep learning with attention mechanisms to predict driver discomfort from pressure distribution data. Using a dataset collected from 192 participants, we analysed biomechanical parameters—such as contact area and peak pressure—across 11 body-parts and correlated them with subjective ratings on a 1–10 Likert scale. The proposed hybrid neural network achieved an R² of 0.82 in predicting global discomfort, with peak pressure showing the strongest statistical correlation (Spearman’s r = 0.78). Additionally, Finite Element Analysis (FEA) was used to validate design changes suggested by the model, leading to an 18% reduction in peak seat pressure without compromising the structural performances. This approach demonstrates the potential of integrating AI and CAD-CAE to guide ergonomic seat design early in the development process, reducing physical prototypes and improving user comfort.