Towards 3D Interactive Design Exploration via Neural Ne
tworks
Physical simulations are performed at multiple scales covering diverse physical domains. Macro scale continuum simulations at the part and assembly level, usually employing traditional numerical techniques such as 3D Finite Elements or Finite Volumes, are widely used for physics-based product design because of their high predictive value. However, these simulations are often computationally intensive and may take minutes, hours, or even days to execute. Virtual Twins of Physics Behavior, derived from these computationally expensive 3D models, massively reduce execution times to seconds (or interactive times), allowing a high number of parameter evaluations in the design space.
However, these virtual twins are often of lower fidelity and provide limited information through a few scaler KPIs, such as max stress, max pressure, max intrusion, etc. To maintain the richness of 3D simulation results while significantly reducing executing time, a Neural Networks-based approaches can be used with the ultimate aim to enable 3D quasi-interactive design exploration. Multiphysics-multiscale traditional simulations are used as the starting point: FEA analyses of structural statics, dynamics, manufacturing, packaging and safety, as well as CFD analyses are used as Design of Experiments (DOEs) to generate the parametric design data. The data is processed and used to train fast-executing neural networks as 3D Virtual Twins of Physics Behavior. The choice of neural network algorithms and architectures (deep feed-forward networks, recurrent and recursive nets, etc.) depends on the nature of the physics. The trained neural network models can be deployed in a collaborative design environment for interactive design explorations. The proposed approach extends traditional model surrogates to cover both transient physical responses and 3D fields, potentially resulting in an information-rich and much more productive environment for product design.
Structures benchmarks
Fluids benchmarks
and we can start a one-to-one conversation with you.
Meet the experts
Victor OANCEA@VO earned his PhD from Duke University in 1996 in the area of computational mechanics. He then joined what was the Abaqus R&D development in Rhode Island, which today is part of Dassault Systèmes SIMULIA Corp. Victor has worked in a variety of R&D positions through the years and is today the Senior Technology Director and the Chief Scientific Officer for Structural applications. In the last few years at SIMULIA, Victor has led from a simulation technology perspective a variety of multiphysics/multiscale simulation initiatives including machine learning, battery cell engineering, additive manufacturing, micro-mechanics-based multiscale materials, particle methods for extreme deformation, oil and gas multiphysics formulations, realistic human simulation capabilities, co-simulation-based multi-physics modeling. | |
Jing BI@JB is a Technology Senior Manager at Dassault Systèmes SIMULIA focusing on machine learning technologies. She received her MS and PhD degree in Mechanical Engineering from the University of North Carolina at Charlotte in 2010 and 2012. Jing joined Dassault Systèmes in 2012 and since then worked in a variety of technical roles working with key customers and partners in crashworthiness, composites damage modeling, multiscale material modeling, additive manufacturing and machine learning. |
Further resources
Blog posts
- Accelerating Packaging Design with AI and Machine Learning
- Machine Learning in Electromagnetics
- Ask an Engineer: What is Machine Learning and Neural Networks?
- Transforming Biomedical Research and Integrating Advanced Simulation
- Democratizing Engineering Analysis Through a Universal Material Subroutine
- Use Machine Learning to Optimize Weld Integrity
- Machine Learning Accelerates Centrifugal Pump Design
- Machine Learning in Electromagnetics
- Rapid Aerodynamic Development using CFD and Machine Learning
- Accelerating Packaging Design with AI and Machine Learning
Tech talks and videos
- Now You Know: Everything About Machine Learning
- Machine Learning Accelerated Conceptual Crash Simulation
- How to Enable Next-Level Efficiency in structural concept engineering
- AI for EMC Simulation
- Democratization of EM-Simulation by using surrogate models on the 3DEXPERIENCE platform
- DoEs & Surrogate Model Extraction for Antenna Design
- What does machine learning mean for simulation?
Event Presentations:
- Brown University: Physics-Informed Machine Learning for Engineering Applications | 2024 SIMULIA Americas Users Conference
- Stanford University: Automated Model Discovery - A New Paradigm in Virtual Human Twin Simulation | 2023 Virtual Human Twin Experience Symposium
💬 Leave a comment with your thoughts on the potential benefits of this technology for the Simulation community
💡 Have a challenge/project to discuss now? Comment here, and we can start a one-to-one conversation with you.
