AIFDIR

The AIFDIR project aims to design, develop and validate SW components that are able to significantly improve the effectiveness and efficiency of the management functions of FDIR (Fault Detection, Isolation and Recovery) systems, through the use of artificial intelligence techniques based on Machine Learning (ML) and their integration with symbolic models of the systems and of the possible faults. The project has a duration of 24 months. The main objectives of the project are to improve the development process of FDIR systems, improve the accuracy of rule-based FDIR components using artificial intelligence techniques and increase the diagnosis and prognosis capability of FDIR systems to identify causes of failures and possible propagations. The ultimate goal of the project is to improve the autonomy and reliability of aerospace systems used in future missions, which require a high level of autonomy. The results of the project are applicable to generic aerospace systems, such as satellites or space exploration vehicles.

The main activities of the AIFDIR project include: - The use of a Model-Based System Engineering (MBSE) methodology to specify and design FDIR systems with both symbolic rule-based and AI-based FDIR components. - The definition of a reference architecture for FDIR with generic interfaces to integrate symbolic models and ML-learned models. - The identification of suitable ML algorithms for modeling FDIR components for anomaly detection and failure classification, and their data-driven training. - The use of a specific process (VIVAS methodology) for the validation and verification of FDIR systems with MBSE techniques, which consists in abstracting the learned models and testing the modeled systems through simulation on automatically generated scenarios, evaluating their coverage based on the requirements. - The validation of the approach in the laboratory (TRL4) to demonstrate its efficiency and effectiveness on several potential use cases in different application domains.

The AIFDIR methodology is built and demonstrated using the TASTE tool, developed by ESA, which implements the MBSE approach for SW system design. TASTE has been recently extended in the COMPASTA project to allow the modeling of HW components and their possible failures, and the verification and validation using formal techniques based on model checking. In AIFDIR, TASTE is further extended to allow the ML-based modeling of FDIR components, and the verification and validation of the overall system using simulation-based techniques. Further future activities include the consolidation of the tool developed in the project to reach higher TRL levels, and the demonstration of AIFDIR on more complex case studies in the space domain.

In the space sector, the methodology and technologies developed in the project are applicable to the design of any type of aerospace system and can also be used in complex operational missions, such as satellite constellations for Earth observation or commercial communication, and also for space exploration vehicles.

In general, the AIFDIR approach is potentially applicable to the design of complex systems in domains other than aerospace, such as the aeronautical, railway, automotive, manufacturing, robotics and energy production sectors.

General Info

Start Date: 6 Jun 2024
End Date: 5 Jun 2026
Duration: 24 months
Funding: ASI

Partners

• FBK (contractor)

Contacts

Marco Bozzano <>
Stefano Tonetta <>
Marco Cristoforetti <>