A Study on Structural Health Monitoring of a Large Space Antenna via Distributed Sensors and Deep Learning

Area of Science:

• Spacecraft Engineering
• Structural Health Monitoring
• Machine Learning Applications

Background:

• Modern spacecraft feature large, flexible appendages crucial for high-precision observations.
• Increased appendage complexity elevates susceptibility to performance degradation from structural damage.
• Traditional methods struggle to detect localized damage on large satellite structures.

Purpose of the Study:

• To develop a deep neural network for detecting failures in orbiting satellite structures.
• To investigate sensor sensitivity for damage classification on large mesh reflector antennas.
• To address limitations of traditional structural health monitoring for complex appendages.

Main Methods:

• Utilized a fully coupled 3D simulator to model in-orbit attitude behavior of a flexible satellite.
• Employed finite element techniques to model satellite appendages.
• Trained and tested a deep learning framework using time-series data from a distributed accelerometer network.

Main Results:

• The deep learning framework accurately detected structural failures, including complete breaks and intermediate damage.
• Identified critical areas of the structure susceptible to damage.
• Demonstrated the effectiveness of the proposed sensor configuration and deep learning approach.

Conclusions:

• The proposed deep learning framework offers a robust solution for structural health monitoring of large satellite appendages.
• The study provides insights into sensor distribution and geometrical properties for optimal damage detection.
• Enhances satellite reliability and data integrity by enabling early detection of structural issues.