For years we have heard how the next great industrial revolution is either underway or will begin in earnest. Both statements have some truth, but the former is starting to look more and more like a horse in the back.
Industry 4.0 is essentially the interconnection of physical and digital systems achieved through the industrial Internet of Things, which is changing production as we know it. These new modes of testing and production with artificial intelligence are beginning to help engineers monitor physical assets in real time, create autonomous decision-making systems in production lines, and improve value creation processes.
The revolution was so promising that the US Senate voted to approve a bill that was called the most expansive legislation in the field of industrial policy in the history of the country back in June. And while it’s fair to say that most applications of Industry 4.0 initiatives still fall within the scope of the systems of the last industrial revolution, really exciting technologies and applications are beginning to emerge, most of which are based on AI.
The Intelligence Manufacturing Intelligence section has been part of this change for years. The company’s focus on incorporating intelligence into its customers’ design, engineering, manufacturing and quality control solutions has earned them a prominent place as a prominent figure in every Industry 4.0 conversation.
It is therefore appropriate for them to host HxGN LIVE Design & Engineering 2021 virtual conference for three days, starting on October 12. The event will see 220 conversations on topics ranging from acoustics to ICME (Integrated Computing Engineering), multiphysical digital twins, AI and machine learning and how these areas are adapting and leading a new industrial era.
How AI solves a computational scaling problem
It is worth remembering how we got here, as all the conversations we will mention take place in the same contextual industry. FE (finite element) analysis enables engineers to create high-quality, virtual representations of a physical asset and see how it will perform under a number of circumstances.
These virtual simulations are often true to physics, but they also take a lot of time and are expensive. Only the use of these simulations is unlikely to help engineers realize the dream of digital twins, for example, one of the goals characterizing the current industrial movement.
Digital twins are accurate virtual images of products, people, entire processes or even supply chains, offering both a living window on what is happening to them and how they can react to different situations in their lives. By combining real-time data from sensors with the predictive power of the simulation, digital twins help engineers make better-informed decisions at greater speeds, allow for more comprehensive approaches to research and design, and simplify the complex problems of product optimization and innovation.
Awareness of this, however, requires help from AI, as FE simulations have become so complex in recent years that they represent real computational and financial barriers for engineers and designers to use them. AI changes that. The hexagon, for example, has already succeeded with CAE-aware machine learning, using it to create real-time simulations to test automotive hardware, something that is simply impractical with just a physically based simulation.
AI circumvents this impracticality by providing the necessary information to create a ROM (reduced order model), a mathematical approach to identifying the defining features of a traditional model, and preserving them in a simple, computationally efficient way. Machine learning algorithms achieve this by “studying” the dynamics between the input and output values of past FE simulations.
This is an extremely valuable tool in a world where FE simulations of complex systems such as those in CFD (computational fluid dynamics) can take hours or even days to complete. AI can reduce this time to seconds. It also helps lay the foundations for digital twins, whose accuracy in real life will depend on how well researchers can train the algorithms that come before them.
The use of artificial intelligence in production spaces can democratize very sophisticated engineering tools for the practical advantage of non-experts. It also allows for a wider range of collaboration opportunities without risking intellectual property, as ROMs inherently often hide patented information.
HxGN LIVE Design & Engineering 2021: Conversations to watch
Hundreds of conversations are held during the conference, all on fascinating topics, so it’s hard to single out any number of them for an excellent drum. But if you’re curious to see how AI helps shape manufacturing, design, and Industry 4.0 in general, we’ve put together a short list of a few presentations that you might be upset to miss.
The automotive industry is among the most receptive playgrounds. As an electric vehicle (and to a lesser extent, self-governing) the revolution is gaining momentum worldwide, the automotive world is rediscovering itself with AI, even using it to better understand the ergonomics and safety considerations of the seats in both normal driving conditions and accident scenarios.
The construction and execution of simulations for both conditions can take a long time, especially considering that they are often performed hundreds of times while considering different passenger postures to determine injury values.
JSOL Corporation spokesman Masahiro Takeda will give a lecture on a case that attempts to predict such values of injury by linking driving and crash simulation with ROM generated in part by machine learning algorithms. Takeda’s study is a great example of how ROMs can alleviate computational workload, as the average prediction accuracy of their ROMs has reached over 90 percent and cut what would be a 19-hour computer crisis in terms of only a few seconds.
Laurent Di Valentin, CAE Senior Technical Assistant for Stellantis, will present a report machine learning and CAE-replenishment, detailing how the company builds behavioral models using artificial intelligence to extend and replace more traditional physical models for vehicle design. These models increase efficiency from the design stage to validation and are an illustration of how AI can accelerate the entire system.
Another discussion focused on the synthesis of FE simulations and machine learning is the presentation of Gustav Eiffel University researcher Dr. Michel Behr real-time design of 3D printing, Orthopedic insoles. Behr has been working on the biomechanics of impact and injury prediction for the past 15 years, and his conversation will focus on developing and testing new ways to predict how insoles will affect a patient’s gait.
Behr’s work shows how useful it can be to combine the best of model reduction methods and more traditional FE modeling to create better decision-making tools for designers and planners.
One of the most up-to-date presentations we look forward to is Hideki Nakata and Anton Zhuravlev talking about using CFD and AI to better understand and prevent the spread of airborne infections. diseases such as COVD-19 as well as better preparation of cities for disaster response.
Using St. Mary’s Cathedral in Tokyo as a test space, researchers from Ecococa Kyoto’s technology department, in collaboration with the Hexagon Leica Geosystems team, developed a droplet visualization system in which they were able to conceptualize the airflow (and thus the potential range of the virus) in the building.
Leica’s 3D laser scanners and Hexagon’s Cradle CFD software allowed them to construct a model of the building’s interior, visualizing how its architecture and air conditioning systems would distribute airborne particles in the air, which are the most common vector for transporting the virus.
Usually such modeling can take considerable time, but Ecococa managed to use AI to convert the analysis of the measured data into ROM, which allows real-time views in a number of “what-if” airflow scenarios. This has helped researchers develop effective measures to reduce the risk of worshipers from exposure to the virus, allowing the surrounding community to gather for the Christmas holidays.
Interestingly, the techniques used in modeling the cathedral space and developing safety measures are also applied in the analysis of extreme weather events. Organizations like ecoKaku can use the same artificial intelligence tools to provide action plans to city officials and engineers to better prepare for and respond to natural disasters. As climate change promises an increase in extreme weather events in the near future, AI-assisted analyzes could ultimately save lives. The talk promises to be remarkable.
Finally, Leonardo Aircraft representative Stephanie Sorentino will give a presentation enriching the database with materials using AI. Sorentino will discuss how machine learning techniques can be used to predict material behavior — in this case, composite layouts — without actually performing material behavior simulations for everyone, let alone destructive testing of the material and costs; and the time it takes.
These predictive models could potentially be extended to other areas of design, allowing researchers to better decide what tests should be performed to improve the accuracy of the model and how best to combine these. -new tools for artificial intelligence with traditional virtual tests.
Build instead of predict the future of AI
As the world of design and manufacturing moves more and more in the practices of Industry 4.0, Hexagon is holding a timely conference, especially given that these technologies are almost guaranteed to change much of each manufacturing sector soon. .
Just like in any other field to which it applies, AI plays a key role in this transformation. The fourth industrial revolution is still full of opportunities, and predicting how technological tools like this will change the future is a big part of what we’re doing here in interesting engineering. But it is the hundreds of speakers at HxGN LIVE Design & Engineering 2021 who are actually shaping this future. Don’t miss what they have to say on this issue.