Francesco Cadini

Francesco Cadini is Associate Professor of Machine Design at the Department of Mechanical Engineering of Politecnico di Milano, where he is a core member of the SIGMA Lab research group. He holds a Ph.D. in Radiation Science and Engineering from Politecnico di Milano (2006) and an M.Sc. in Aerospace Engineering from UCLA (2003). His research focuses on AI/ML-based methods for the integrity of complex engineering systems — structural health monitoring (SHM), prognostics and health management (PHM), digital twins, hybrid physics/data-driven modelling, explainable AI, and uncertainty quantification — with applications to aerospace and space platforms, UAVs, lithium-ion and structural batteries, and defence systems. He is currently Scientific Coordinator or unit coordinator of several European (EDA, EDF) and national (ASI, MoD) research projects, including BATTAGE, COMMANDS, FMBTech, SYBIL, DIGES and EXTESA.

1. Adaptive prognosis of lithium-ion batteries based on the combination of particle filters and radial basis function neural networks

   C. Sbarufatti, M. Corbetta, M. Giglio, F. Cadini

   Journal of Power Sources 344, 128-140 · 2017

2. State-of-life prognosis and diagnosis of lithium-ion batteries by data-driven particle filters

   F. Cadini, C. Sbarufatti, F. Cancelliere, M. Giglio

   Applied Energy 235, 661-672 · 2019

3. Neutralization of temperature effects in damage diagnosis of MDOF systems by combinations of autoencoders and particle filters

   F. Cadini, L. Lomazzi, M. Ferrater Roca, C. Sbarufatti, M. Giglio

   Mechanical Systems and Signal Processing 162, 108048 · 2022

4. Towards a deep learning-based unified approach for structural damage detection, localisation and quantification

   L. Lomazzi, M. Giglio, F. Cadini

   Engineering Applications of Artificial Intelligence 121, 106003 · 2023

5. Unsupervised data-driven method for damage localization using guided waves

   L. Lomazzi, R. Junges, M. Giglio, F. Cadini

   Mechanical Systems and Signal Processing 208, 111038 · 2024

1. Anomaly characterization for the condition monitoring of rotating shafts exploiting data fusion and explainable convolutional neural networks

   M. Parziale, Y. Yeung, K. Youcef-Toumi, M. Giglio, F. Cadini

   Structural Health Monitoring, 25(3), 1678-1698 · 2025

2. Convolutional autoencoders and CGANs for unsupervised structural damage localization

   R. Junges, Z. Rastin, L. Lomazzi, M. Giglio, F. Cadini

   Mechanical Systems and Signal Processing, 220, 111645 · 2024

3. Enhancing Lamb wave-based damage diagnosis in composite materials using a pseudo-damage boosted convolutional neural network approach

   A. Gonzalez-Jimenez, L. Lomazzi, R. Junges, M. Giglio, A. Manes, F. Cadini

   Structural Health Monitoring, 23(3), 1514-1529 · 2024

4. Explainability of convolutional neural networks for damage diagnosis using transmissibility functions

   M. Parziale, P. Henrique Silva, M. Giglio, F. Cadini

   Structures, 69, 107583 · 2024

5. Physics-Informed Neural Networks for the Condition Monitoring of Rotating Shafts

   M. Parziale, L. Lomazzi, M. Giglio, F. Cadini

   Sensors, 24(1), 207 · 2024

6. Temperature enhanced early detection of internal short circuits in lithium-ion batteries using an extended Kalman filter

   Y. Jia, L. Brancato, M. Giglio, F. Cadini

   Journal of Power Sources, 591, 233874 · 2024

7. On the explainability of convolutional neural networks processing ultrasonic guided waves for damage diagnosis

   L. Lomazzi, S. Fabiano, M. Parziale, M. Giglio, F. Cadini

   Mechanical Systems and Signal Processing, 183, 109642 · 2023

All group publications

• Lecturer

  Design & Manufacturing

  MSc Design & Engineering · 6 CFU · 1st semester · Milano Bovisa

  Course sheet

• Lecturer

  Methods for Engineering Design

  MSc Design & Engineering; MSc Product Service System Design; MSc Integrated Product Design; MSc Digital and Interaction Design · 6 CFU · 2nd semester · Milano Bovisa

  Course sheet

• Lecturer

  Reliable and Resilient Design of Mechanical Systems

  MSc Mechanical Engineering · 5 CFU · 1st semester · Milano Bovisa

  Course sheet

Courses of 2026/2027, from the Politecnico’s course sheets. All the group’s courses

• SHMAIDeep Learning

  Graph neural networks for damage diagnosis

  SHM systems aim to detect and diagnose damage in real time. Conventional methods, such as numerical analysis, can be computationally expensive and less efficient for large-scale or complex structures. Graph Neural Networks (GNNs) offer a promising solution by leveraging sensor networks as graphs, where nodes represent sensors and edges capture physical interactions. The objective of this thesis is to validate and enhance the use of GNNs for diagnosing damage within SHM systems.

  Details and apply
• AIDeep LearningNumerical Modelling

  Graph neural network-based surrogate modeling for dynamic structural simulation

  Simulating the dynamic behavior of structures using traditional methods such as the Finite Element Method (FEM) can be computationally intensive, especially for large-scale or complex systems. Graph Neural Networks (GNNs) present a powerful alternative by representing physical systems as graphs, enabling efficient learning of local patterns with strong generalization capabilities. The objective of this thesis are:

  Details and apply
• SHMSensingAIMachine Learning

  Dataset and algorithms for damage diagnosis in real-world scenarios

  Structural Health Monitoring (SHM) systems aim to detect and diagnose damage in real time. Traditional acquisition systems for SHM are highly accurate but exhibit significant limitations in terms of weight, size, cost, portability, and power consumption. Therefore, the development of systems capable of overcoming these constraints is increasingly crucial to enhance their applicability. Moreover, in the context of SHM, the use of Machine Learning (ML) plays a central role, making it essential to integrate ML even into computationally constrained systems through appropriate optimizations and customizations. In this context, our research project is built up as follows:

  Details and apply
• SHMElastic WavesAIDeep Learning

  Damage Localization using Autoencoders for Varying Temperatures and Frequencies

  Damage localization through the analysis of ultrasonic guided waves can be influenced by an array of cofounding factors e.g., temperature and humidity. Besides, the choice of excitation frequency affects not only the sensitivity of the waves but also the size of damage that can be detected. In this context, properly trained convolutional autoencoders that leverage the temperature and frequency information as points in their latent space might be able to overcome such drawbacks and provide a more robust damage localization framework.

  Details and apply
• PHMAIBatteriesUAV

  Development of artificial intelligence-based HUMS for UAVs HPS to support optimization of operations and maintenance

  In recent years, there has been growing interest in using electric propulsion for aeronautical applications due to its benefits of reduced emissions, vibrations and noise, as well as improved efficiency and performance. Specifically, HPS which integrate batteries, solar panels and fuel cells, have been proposed as a promising solutions for powering High Altitude and Long Endurance (HALE) drones. These drones can operate in extreme environmental conditions for long duration, making the development of advanced HUMS crucial for ensuring their safety and efficiency.

  Details and apply
• SHMPHMBatteries

  Offering an achievable and advanced health monitoring strategy for structural batteries

  With the higher demands of performances in the automotive and aeronautical fields, more battery cells are required for the higher power and energy of the pack. The trade-off between the power and weight is an open research problem. Structural battery (SB) packs potentially offer a compromise whose multifunctional materials serve both for energy storage and load bearing. However, for SB on aeronautic applications, the maintenance and replacement are complex, which might rely more on an efficient PHM system.

  Details and apply
• ExperimentalBatteries

  Towards a Structural Battery Coupon: Development and Experimental Validation under Electrical, Mechanical, and Thermal Loads

  With the higher demands of performances in the automotive and aeronautical fields, more battery cells are required for the higher power and energy of the pack. The trade-off between the power and weight is an open research problem. Structural battery (SB) packs potentially offer a compromise whose multifunctional materials serve both for energy storage and load bearing.

  Details and apply
• PHMAIMachine LearningBatteries

  Lithium-ion batteries PHM by exploiting electrochemical models and machine learning

  To ensure safe and efficient operation, the power batteries of EVs should be controlled on a suitable operative environment, which includes temperature and pressure as well. Therefore, an accurate battery management system (BMS) which mainly rely on the estimation of battery health states including SOC, SOH, RUL and ISC early occurrence is highly required which means, an efficient prognostics and health management (PHM) system.

  Details and apply
• AIMachine LearningBatteries

  Electrochemical-Thermal Modeling and Learning-Based Control of Li-ion Battery Packs

  Battery packs operate under coupled electrical and thermal constraints; conservative control ensures safety but sacrifices usable power/efficiency and accelerates degradation under suboptimal thermal gradients. Physics-based models enable interpretability and extrapolation, while RL can learn control policies for complex multi-objective operation-provided the sim-to-real gap is addressed.

  Details and apply
• Digital TwinsPHMBatteries

  Early-Stage Internal Short Circuit Detection and Proactive Thermal Runaway Prevention in Li-ion Battery Packs via High-Fidelity Digital Twins

  Battery packs face a critical trade-off between performance and safety, specifically regarding latent Internal Short Circuits (ISC) that trigger thermal runaway. This research develops a physics-informed Digital Twin to identify micro-ISCs during the "slow drop" phase, preventing progression to the "unstoppable" terminal stage. By integrating model-based algorithm with surrogate/lumped-modeling, we enable proactive, detection algorithms that bridges the sim-to-real gap and ensures safety-critical operation without sacrificing efficiency.

  Details and apply
• PHMAIMachine LearningSpace

  Time series anomaly detection (TSAD) algorithms for satellite telemetry data

  The complexity and multidimensionality of telemetry data, encompassing both sensor measurements and commands, present significant challenges in analysis, underscoring the continued necessity for experts to check system integrity. To tackle these challenges, the implementation of intelligent prognostics and health management (PHM) algorithms for telemetry data processing, incorporating both machine learning and standard statistical methods, advances a predictive approach aimed at evaluating the health states of space systems.

  Details and apply

All thesis topics