Session: ASME Undergraduate Student Design Expo

Paper Number: 173504

Surrogate Modeling for Rapid Thermal Analysis of Iss Payload 

Efficient spacecraft thermal analysis is critical for optimizing mission performance and reliability. Traditional high-fidelity simulations using Thermal Desktop are computationally expensive, often requiring significant time to solve complex thermal models. This limits the ability to conduct large-scale parametric studies and real-time thermal assessments. To address this challenge, a machine learning-driven surrogate model was developed to rapidly predict node temperatures in the Japanese Experiment Module (JEM) aboard the International Space Station (ISS) based on key thermal parameters.

This study introduces an integrated workflow that combines automated data processing with machine learning regression models, significantly reducing computational costs while maintaining high accuracy. The model is trained on precomputed simulation data, capturing the relationships between spacecraft thermal behavior and external factors such as yaw, pitch, roll, incident heating, and ISS beta angle. With this approach, temperature predictions can now be obtained in 0.01 seconds, achieving a 200,000× speedup over conventional simulation methods, while maintaining a mean absolute error (MAE) of 1.5539 degrees F and an R2 coefficient of 0.9972.

The automation scripts significantly improved the efficiency of case setup, data collection, and visualization. The batch modification tool reduced manual input time for boundary condition adjustments by 83%, from 60 minutes to just 10 minutes. The data extraction script streamlined nodal temperature analysis, cutting processing time by nearly 89%, from 45 minutes to 5 minutes. Additionally, the visualization tool enabled rapid interpretation of temperature distributions, reducing analysis time by 93%, from 30 minutes to just 2 minutes. These improvements not only reduce human effort but also enhance accuracy and consistency in thermal analysis, allowing engineers to focus on model refinement and decision-making.

Beyond improving simulation speed, this work contributes to the development of a digital twin framework for real-time spacecraft thermal assessment. By replacing computationally expensive simulations with an efficient machine learning model, engineers can rapidly evaluate multiple design scenarios, optimize thermal management strategies, and detect anomalies in space environments. The surrogate modeling approach has broader applications in other space structures, such as lunar habitats and planetary rovers, where rapid thermal analysis is critical for long-duration missions.

Future work will focus on expanding the dataset to improve model robustness and integrating uncertainty quantification techniques to ensure reliability under a wider range of conditions. Additionally, incorporating real-time onboard sensor data will allow for adaptive prediction models, further enhancing spacecraft thermal management and operational decision-making.

By bridging the gap between physics-based simulations and real-time predictive analytics, this study demonstrates the potential of machine learning-driven surrogate modeling as a scalable and computationally efficient alternative for spacecraft thermal analysis. The results confirm that this approach provides a practical, accurate, and high-speed solution for space mission design and operation, significantly advancing the efficiency of thermal engineering workflows.

Presenting Author: Jayson Johnson Afterlab

Presenting Author Biography: Jayson Johnson is a senior at Howard University and a Karsh STEM Scholar pursuing a bachelor’s degree in Mechanical Engineering. He is the co-founder of Tree Tech, a startup focused on facilitating improved information access in his college communities through AI. This venture aligns with his commitment to helping underserved populations through data science and engineering. Jayson accumulated extensive experience through various projects at the Johns Hopkins Applied Physics Lab, Howard, Johns Hopkins University and MIT, involving heat transfer for hypersonic vehicles, International Space Station (ISS) payloads, self-cleaning solar panels, as well as solid state synthesis for superconductors. His career aspirations involve obtaining a PhD in Aerospace Engineering to continue helping underserved communities. Outside of school, Jayson enjoys physical activities such as gymnastics, going to the gym, and playing soccer.

Authors:

Jayson Johnson Afterlab