M&TT Colloquia

03-10-2024 - Lecture Hall A

PhD presentation

Towards sub-grid scale bubble modeling and Characterizing numerical artifacts in multiphase simulation

Tzu-Yao Huang

Simulating bubbly flows is challenging, particularly at high Reynolds (Re) and Weber (We) numbers, where bubble sizes vary significantly. Direct numerical simulation (DNS) is computationally infeasible for high-momentum cases like those induced by Olympic divers. To address this, we aim to develop a sub-grid scale (SGS) bubble model that ensures coarse grid solutions align with real-world data. Along the way, we identified an intriguing numerical artifact in sharp interface simulations, which we call ‘numerical shape diffusion.’ This presentation will cover our progress toward creating an SGS bubble model and our insights into these numerical artifacts in multiphase simulations.

03-10-2024 - Lecture Hall A

PhD presentation

Towards Adaptive Manufacturing Systems: Planning and scheduling methods for more sustainable and resilient factories

Rafael Leite Patrão

The manufacturing sector faces increasing uncertainty due to systemic changes like the green and digital transitions. These changes offer opportunities for factories to become more sustainable and resilient, but they also create significant challenges for decision-makers at strategic, tactical, and operational levels. To help factories adapt while maintaining production goals, we explored two key technological enablers: reconfigurable machines and machine learning techniques. We defined a manufacturing system that is both highly sustainable and resilient, using such enablers, as an adaptive manufacturing system. To drive existing factories towards greater adaptability over time, while minimizing production disruptions, we develop production planning and scheduling models that consider sustainability and resiliency metrics and integrate both technological enablers.

19-09-2024 - Lecture Hall A

PhD presentation

On physical-constraints in Scientific Machine Learning for fast flow field predictions

Manuel Cabral

Traditional fluid simulations, while highly accurate, are often too slow for many practical applications. Recently, machine learning has emerged as a promising alternative due to its ability to deliver fast predictions. However, ensuring the accuracy and physical consistency of these models requires the incorporation of physical priors. In this talk, we explore methods for embedding physical laws — such as symmetries and governing equations — into machine learning models, with a focus on fluid dynamics. By integrating these priors, we develop models that are not only more data-efficient but also smaller, making them well-suited for real-world applications where data and computational resources are often limited. Furthermore, we show that these physics-based models generalize better to previously unseen flow regimes, offering improved performance over purely data-driven approaches.

19-09-2024 - Lecture Hall A

PhD presentation

Digital Potential in Shipyards: A Path to Transformation

Miguel Calvache

Before diving into academia, I spent several years in the navy and the shipbuilding industry, where I saw firsthand the challenges shipyards face in keeping up with digital transformation. Now, as a PhD researcher, my work explores how digital technologies can reshape shipbuilding processes, addressing declining throughput and competitiveness. Using a mathematical model, I (try to) analyze the financial benefits of various technologies and apply innovation theories to understand why shipyards have been slow to adopt new systems. Join me this week as I share my journey from shipbuilding to digital transformation, and discuss the challenges and opportunities for the industry in the digital age.

12-09-2024 - Lecture Hall A

Mini-Symposium: Technical Talk

Perspective and applications of data-informed CFD

Bernat Font

The widespread adoption of data-driven modeling tools and advances in hardware architectures across the scientific computing community have also shaped the future of CFD solvers. In this context, we will review the opportunities and challenges arising from the use of ML models in scale-resolving turbulent flow simulations. As examples, the cases of data-driven turbulence modelling, and active flow control using reinforcement learning will be discussed. We will also introduce our flagship CFD solver, WaterLily, and discuss future research directions.