Wednesday, October 16th 2024
(Milan Time) 15:00-18:00
(Beijing Time) 21:00-24:00
Tencent ID : 443-161-146
Session Aims & Scope
Digital Twin has been considered as a newly emerging technology that benefits the development of many research areas and disciplines. Driven by requirements from the Physics-of-Failure and machine-learning based reliability design and analysis methods, high precise predictions of the failures under dynamical uncertain conditions become the key concern in the synergistic design of products between their reliability and functional performance. This gives birth to a novel multi-disciplinary research area “Reliability Digital Twin (RDT)”, which should be able to fully utilize the multidimensional data collected from the products, including product model data, statistical data of fault events, real-time operational status data, historically environment and load data, etc., to provide more accurate simulation and reliability predictions, by using the Digital Twin technologies. And the related new ideas and solutions have been emerged all over the world in the recent years. To this end, this session is arranged for presenting these innovative researches from both theoretical and application perspectives to academic and engineering circles.
Submissions that reflect the conference scope and current state of the field are welcome in areas including but not limited to:
- Methodologies of combing reliability and digital twin
- Development of RDT of products in its design and maintenance stages
- AI for RDT
- Uncertainty analysis in RDT
- Advanced application researches of RDT
- Intelligent predictive maintenance using RDT
Session Chair(s)
Chair
Yi REN
Professor
Beihang University (China)
Co-Chair
Cheng QIAN
Associate Professor
Beihang University (China)
Co-Chair
He LI
Assistant Professor
University of Lisbon (Portugal)
Session Presentation
1.
Xinyan HUANG
Associate Professor
The Hong Kong Polytechnic University (China)
Title: Digital Twins for Smart Firefighting and Evacuation
Abstract
Over the past decade, big data, Artificial Intelligence, and digital twin have enabled new approaches to improve fire safety. The emerging applications of AI and digital twins enable more intelligent fire detection, fire hazard assessment and real-time fire forecast. This talk will introduce the digital twin future framework for combining AI, sensor networks and intelligent robots to help forecast and fight fire, as well as safe evacuation in fire incidents. I will also talk about the guidelines for constructing a reliable fire database, propose new concepts for building Fire Digital Twin, and review deep learning algorithms that enable super real-time fire quantification, forecast of fire development, and robotic firefighting.
2.
Jiajie FAN
Youth Professor
Fudan University (China)
Title: Thermal and Thermal-Mechanical Oriented Reliability Co-design of SiC Power Device Panel-level Packaging
Abstract
Silicon Carbide Metal Oxide Semiconductor Field Effect Transistor (SiC MOSFET) is mainly characterized by a higher electric breakdown field, higher thermal conductivity and lower switching loss enabling high breakdown voltage, high-temperature operation and high switching frequency. However, their performances are considerably limited by the high parasitic inductance and poor heat dissipation capabilities associated with existing wire-bonding packaging methods. A new panel-level SiC MOSFET power module is developed by using the Fan-Out and embedded chip technologies. To achieve the more effective thermal management and higher reliability under thermal cycling, some optimization methods, i.e. Ant colony optimization-Back Propagation neural network (ACO-BPNN), non-dominated sorting genetic algorithm (NSGA-II) and the improved multi-objective particle swarm optimization algorithm (MOPSO) will be introduced for optimizing the new design of SiC modules, and contrast it with the traditional Response Surface Method (RSM).
3.
Shiyu ZHAO
Associate Professor
Westlake University (China)
Title: Multi-Robot Swarming: Cooperation and Competition
Abstract
Multi-robot swarming is a core research area in the field of robotics. It has also been recognized as one of the ten grand challenges in robotics. In this talk, I will introduce the research progress of the Intelligent Unmanned Systems Laboratory over the past few years. Multi-robot swarming tasks can be classified as cooperative or competitive. Regarding cooperative multi-robot swarming, I will talk about our recent research on a classic benchmark problem in multi-robot swarming: shape assembly. Regarding competitive multi-robot swarming, I will talk about our research work in the field of aerial target pursuit, particularly a series of vision-based motion estimation algorithms and systems.
4.
Xingsuo HAI
Research Fellow
Nanyang Technological University (Singapore)
Title: Digital Twin-Enabled Decision-Making for Multi-UAV Autonomous Recovery Oriented to Resilience
Abstract
Real-time decision-making for multiple unmanned aerial vehicles (multi-UAV) mission planning is crucial but challenging due to unexpected disruptions. We propose a resilience-oriented decision-making framework that enables UAVs to generate autonomous recovery strategies in real time. A novel resilience metric is introduced, considering variances in mission completion rate and remaining resource inventory. On this basis, a resilience-oriented joint optimization model is formulated, incorporating maintenance and task reassignment as recovery strategies following accidental failures to UAVs. The optimization problem is formulated as a partially observable Markov decision process (POMDP) and solved using a modified multiagent deep Q network (MADQN) algorithm with enhanced learning efficiency. To leverage real-world experiences, digital twin-enabled technology is adopted for training, where each agent interacts with the DT environment using original and target networks to acquire an optimal assignment strategy. Simulations validate the effectiveness of the proposed methods.
5.
Sifeng BI
Assistant Professor
University of Southampton (UK)
Title: Data-driven Stochastic Model Updating with Conditional Invertible Neural Networks (cINN)
Abstract
This work focuses on a data-driven version of the model updating process. A recently developed conditional invertible neural networks (cINN)-based architecture has been adopted in this work to achieve the multilevel Bayesian updating. Unlike the conventional approaches that employ the artificial neural networks (ANN) solely as a forward surrogate during model updating, the cINN-based model updating is a framework that performs as a bidirectional network where the forward training and inverse calibration are integrated into a uniform structure. The cINN consists of two parts known as the conditional network and the invertible neural network (INN). Both networks are trained jointly in the forward direction and can operate inversely to offer rapid and accurate predictions by given observation data. The application of the cINN provides a more efficient and direct manner to solve model updating problems without calculating the likelihood function, compared to the conventional Bayesian model updating. The cINN is employed to establish the multilevel Bayesian interface, enabling the updating process to be conducted in a stochastic manner. Rather than directly calibrating physical parameters, this approach focuses on the calibration of their statistical moments, e.g. mean and variance, referred to as hyperparameters. The hyperparameters are then utilized to determine the Probability of Damage (PoD), which provides a confidence level about the structural condition, facilitating stochastic damage detection. Two case studies are proposed to demonstrate the multilevel cINN-based stochastic damage detection approach. The first involves a 3-degree-of-freedom (3-DOF) spring-mass simulation model, while the second case study employs an experimental rig testcase with practical measurements, each under various damage scenarios.
6.
Zhongchao SUN
PhD candidate
Aalborg University (Denmark)
Title: Electro-Thermal Digital Twin for Power Modules Temperature Characterization during Power Cycling Tests
Abstract
The thermal behavior of power modules significantly impacts their power cycling lifetime by influencing power loss generation in the electrical domain and subsequently altering the temperature response. This study introduces a coupled electro-thermal digital twin designed to accurately predict the junction temperature during power cycling tests of wide band gap power modules and to provide detailed insights into the package’s internal temperature distribution. The implementation was achieved through the integration of LTspice and COMSOL simulations governed by MATLAB scripts. Moreover, the proposed digital twin is compatible with enhanced electrical schematics by including parasitic and switch loss effects. It also supports the incorporation of innovative packaging patterns for comprehensive thermal analysis and extended thermo-mechanical analysis to identify the degradation mechanisms of critical packaging components.
7.
Mama Diakité
PhD student
the University of Bordeaux (France)
Title: Data synchronization for reliable Digital Twins
Abstract
Data synchronization between a Digital Twin and its corresponding real system is a key feature of the Digital Twin concept, as it is what allows the Digital Twin model to be a faithful representation of the system of concern. However, no academic studies have been devoted to understanding the impact of such a synchronization on the quality and reliability of the Digital Twin’s services. This presentation provides a conceptualization of the problem, and suggest, through a case study, how a sensitivity analysis approach can be formalized to address the issue.
