Detecting any anomalies in spacecraft operation using AI

Spacecraft, including satellites, are exceptionally complex and expensive machines with thousands of telemetry channels, recording a number of aspects from temperature, radiation, power and instrumentation to computational activities. In addition, advances in engineering and technology are leading to the generation of an increasing amount of such telemetry data. As these data contain valuable information about the operational status of the spacecraft, monitoring of these channels is a constant necessity as a failure to detect and respond to potential hazards could result in the full or partial loss of the spacecraft. Therefore, anomaly detection is a critical tool to alert operations engineers of unexpected behaviour. Artificial intelligence (AI) techniques are becoming increasingly used to detect such anomalies in time-series data. They are capable of capturing interactions and dependencies inherent in the telemetry channels, providing important insights into complex system behaviours. The state of the art in anomaly detection currently consists of deep learning Transformer models, which also lie at the heart of generative AI tools such as ChatGPT. On the other hand, digital twins are virtual models designed to accurately reflect a physical object and can be used to run simulations, study performance issues and generate possible improvements, which can generate valuable insights to be applied back to the original physical object. The concept of a ‘digital twin’ was born at NASA in the 1960s as a “living model” of the mission. As part of the Anomaly detection for Spacecraft TelemetRy dAta using Artificial Intelligence (ASTRA-AI) project, a multi-disciplinary research team from the University of Malta has developed a Transformer-based anomaly detection method to capture various types of irregular telemetry signals, with a precision of 99 per cent on spacecraft telemetry datasets from NASA ( satellite and the Mars Science Laboratory) and the European Space Agency (ESA satellite). One of the outcomes of the project would be to demonstrate that such AI models can also run onboard spacecraft with the limited computational resources available. In addition, the team has successfully developed a digital twin of a typical Earth-orbiting satellite electric power system. This digital twin will then be used for prognostic and diagnostic purposes, for instance to simulate various types of anomalies and eventually evaluate the performance of the AI model in detecting them. Xjenza Malta • Firefly Aerospace’s became the – after Intuitive Machines’ spacecraft in 2024 – to achieve a soft-landing on the moon on March 2, landing in . One of the main goals of the lander’s mission is to gather data in preparation for the arrival of astronauts from the programme in the years to come, with payload experiments expected to run for the duration of one lunar day – approximately 14 Earth days – before sunset over forces the solar-powered lander to shut down. • The spacecraft was launched on September 5, 1977, with the aim to study the outer solar system and the sun’s heliosphere. It made several outer solar system flybys, including the gas giants Jupiter and Saturn and Saturn’s largest moon, Titan. It is now approaching one light day away from Earth – meaning that it currently takes signals travelling at light speed from Earth to the spacecraft (or vice-versa) just under 23 hours and four minutes to reach us. At its current rate, will take a whopping 17,720 years to reach one light year away from Earth! • Several Soviet landings between 1975 and 1984 captured details of the Venusian surface, with one of the more successful landings pertaining to the mission in 1982. successfully operated on the Venusian surface for an impressive 127 minutes at a scorching temperature of 457oC and under a crushing atmospheric pressure of approximately 90 times that at sea level on Earth, capturing the first-ever colour image of the surface of Venus. You can unsubscribe at any time by clicking the link in the footer of our emails. We use as our marketing platform. By subscribing, you acknowledge that your information will be transferred to Mailchimp for processing.