COMSOL, the Stockholm-based company behind the software, is bringing the world’s most essential conference to Burlington, Massachusetts. Join us as we showcase the new breakthroughs in multiphysics modeling. Don’t miss this three-day event where the global COMSOL user community will gather. In tandem, we’ll explore the transformative potential of surrogate modeling systems within the simulation software landscape. This cutting-edge technology makes it possible to evaluate even the most complex models in near real-time, vastly improving productivity and decision-making capabilities for every industry.
From the discipline of engineering, surrogate modeling has been a game-changer in engineering simulation. One approach companies are using to accomplish new levels of efficiency is compressing detailed multiphysics models into smaller, condensed formats. These formats are amenable to rapid evaluation using machine learning techniques. These innovations provide new capabilities for engineers and managers to make data-driven decisions in real-time. Historically, these decisions would sometimes be longer than a coffee break to compute!
Bjorn Sjodin, the senior vice president of product management at COMSOL, emphasized the significance of this technology:
“Surrogate models are an interesting new technology where you take your fully-fledged multiphysics model and compress it down into a compact format that’s quick to evaluate using machine learning.”
Our National Conference will include keynotes, hands-on workshops and demonstrations. In these sessions, you’ll learn how academia, industry, and government are making use of surrogate models. COMSOL’s surrogate system has successfully been adapted into an app directly for Indian farmers. This significant use of AI has had a tremendous effect on our agricultural practices. Strengthening a farmer’s ability to accurately forecast food spoilage in multiple cold storage facilities, this app is literally a lifesaver. As a result, they realize a 20 percent decrease in spoilage rates.
In the automotive realm, European manufacturers are using COMSOL’s surrogate models to more quickly simulate full electric vehicle battery packs. Where traditional simulation techniques might require weeks for most high-fidelity designs, surrogate models bring this time down significantly. Per COMSOL’s own benchmarked analysis, a complete model with unknown inputs can take an upwards of 15 minutes to solve in most cases. Integrating a hybrid approach that combines neural networks with existing mathematical reduced-order modeling (ROM) tools can help accelerate computation time immensely. Indeed, for certain applications, this approach can realize speed-ups of 100,000x faster than traditional models.
Our 2024 industry-wide deep dive into ROMs uncovers some very promising trends. These approaches that happen to use algorithms like neural networks depend on a whole lot more than AI. Researchers explain:
“ROMs are divided into two big families: intrusive methods, in which one manipulates directly the governing equations, and non-intrusive methods, in which only the simulation data are considered.”
Sjodin further elaborated on the role of neural networks in these models:
“Neural networks come into play there, but also other technologies, more traditional reduced order modeling technologies.”
The ability to run complex multi-physics scenarios at high speeds, in turn, leads to real-world advantages when used on the shop floor. Sjodin noted:
“If you want to give this to someone on the factory floor, these surrogate models are really useful because it allows you to evaluate and get results immediately.”
The conference will showcase how these innovations are changing the way industries work. You’ll learn about how they’re giving designers real-time feedback and helping speed up prototyping. Users can even compile surrogatemodeling apps into standalone executables. This interestingly enables distribution worldwide without licensing costs, opening access to the latest, most advanced technology for all.