Integration of CFD and machine learning for vehicle cabin thermal management using innovative materials

• A detailed CFD model integrated silica aerogel to enhance cabin thermal insulation performance. • Transient thermal simulations replicated real solar loading and ambient conditions. • HVAC compressor energy demand was assessed across insulation configurations. • Machine learning models accurately forecasted cabin temperature dynamics. • Aerogel insulation led to a 57% reduction in HVAC energy consumption under peak load. Forecasting thermal comfort within the car interior is crucial for optimising the energy consumption of the heating, ventilation, and air conditioning (HVAC) system. In this work, various innovative insulation materials and combined data-driven machine learning (ML) methods, along with detailed computational fluid dynamics (CFD) models, were considered to enhance HVAC performance and reduce its energy consumption. Based on climate measurements, a CFD model of the car interior was methodically utilised to create training data under different boundary conditions. The energy consumption of each insulation material during air-conditioning operation was calculated. Three machine learning algorithms were used to forecast the driver's cabin interior temperature using CFD simulation data: random forests (RF), artificial neural networks (ANNs), and linear regression with stochastic gradient descent ((LR-SGD). The CFD results highlighted the effectiveness of thermal insulation in reducing the energy use of the compressor. Compared to the most used material in thermal car insulation (polyurethane), Aerogel Silica can improve energy performance. It reduces energy requirements by over 50% while preserving the same level of thermal comfort. Additionally, the proposed LR-SGD-based surrogate model achieved an R² of 97.91%. and a root mean square error (RMSE) of 2.24 × 10⁻³ ± 2.89 × 10⁻³ K, enabling rapid cabin temperature prediction with negligible loss of physical accuracy compared to transient CFD simulations. By reducing computational time from several hours per CFD run to near-instantaneous prediction, the proposed physics-informed framework offers a scalable solution for multi-scenario thermal optimisation of electric vehicles.