SURROGATE MODELING FOR RAPID THERMAL ANALYSIS OF ISS PAYLOAD

Efficient thermal analysis is vital for spacecraft reliability, yet traditional simulations like Thermal Desktop are computationally intensive. This study presents a machine learning-driven surrogate model that rapidly predicts node temperatures in the ISS Japanese Experiment Module using parameters such as yaw, pitch, roll, incident heating, and beta angle. Trained on precomputed data, the model delivers predictions in 1 second-200,000× faster than conventional methods-while maintaining a mean absolute error of 1.181°F and R² = 0.9976. Automated scripts enhanced case setup, data extraction, and visualization, reducing manual processing times by up to 93%. This efficiency allows engineers to focus on design optimization and anomaly detection. The surrogate model supports the development of real-time digital twins for spacecraft thermal management and has applications in lunar habitats and planetary rovers. Future work includes expanding datasets, integrating uncertainty quantification, and using real-time sensor data for adaptive predictions. This approach bridges physics-based simulations with real-time analytics, offering a scalable, accurate, and high-speed solution for spacecraft thermal analysis.