IPFI

IPFI aims to advance the use of intelligent methods and tools for the identification of parameters and faults in aerospace slender-type structures — such as launchers, satellite solar panels and flexible aircraft wings.

These methods improve the robustness, performance and eco-footprint of the systems. The main challenges arise from the onboard aim, which places technological constraints on the computational and reliability aspects of the methods, and from the system-identification aspect, which raises mathematical questions such as the proper definition of the inputs, the persistence of excitation of the signals (closely related to the convergence and robustness of the algorithms), the noise/frequency-filtering levels, and the definition of the identification model structure, among others.

Why AI, and why now

The recent re-emergence of AI technologies has generated massive amounts of data and a paradigm shift for society and the digital economy, made possible by scaling computer and software systems into internet-accessible storage farms and now-familiar generative models such as ChatGPT, Copilot or DeepSeek. This has pushed other domains, including aerospace, to explore the use of AI at an unprecedented scale and scope — with acronyms such as AI, ML, DL and RL becoming household names.

Our approach

IPFI focuses on Machine Learning (ML) methods — specifically supervisory approaches based on sparse regression and Neural Network-based methods such as Deep and Reinforced Learning (DL/RL). Deploying these AI/ML solutions in aerospace systems poses three key challenges: