The biggest buzzword in 2023 with regards to anything technology related is “AI” (Artificial Intelligence). With the rapid advancements of widespread tools like ChatGPT and AutoGPT, it’s becoming hard to remember a world without them. Things are no different in the world of product development and engineering. With this blog post, I will talk about some of the tools within the Altair portfolio that include some of these capabilities.
Before I dive into that, one thing that’s important to clarify is how we are defining the terms AI vs. Machine Learning. By its definition AI is the simulation of human intelligence, typically requiring little to no inputs. Machine Learning (ML) uses historical data to make decisions. ML can be considered a subset of AI in that the end goal is to mimic human behavior, ML just requires inputs. To be transparent, all the tools we will talk about in this post require historical data to get a final product. With that in mind let’s jump into these tools.
Altair physicsAI
Operating within the Altair HyperWorks (2022.3 or later) interface, Altair physicsAI combines the power of deep learning algorithms with traditional physics-based simulations. Users upload datasets, which are collections of simulation that can be used to train the tool to obtain desired results. By training on large datasets, the algorithms learn patterns and relationships that exist within the data, enabling the tool to propose innovative design variations.
Fig 1: Workflow for physicsAI
Fig 2: Result from running physicsAI on imported mesh model.
Altair romAI
HyperWorks also includes romAI, an application used for the creation of dynamic reduced-order models (ROMs) and system identification. A ROM is a mathematical approximation of complex systems that preserves system behavior while reducing computational requirements. ROMs allow engineers to simulate large-scale systems with simplified models.
By using machine learning and the laws of physics, it is possible to create high-speed and high-precision dynamic surrogate models and 1D models from 3D CAE simulation results and experimental data. Check out our Blog Post here that covers using Altair Compose with romAI for a real-world application of the tool.
Fig 3: Graphic showing how ROMs connect the 1D and 3D Worlds
Altair DesignAI
One of the key strengths of DesignAI is to streamline the design process by taking various inputs from an end user to automate results for multiple design studies. Similar to Altair HyperStudy, DesignAI can be considered a Design of Experiments (DoE) tool. The workflow is relatively simple: Import simulation models needed for machine learning, DesignAI will scour the file for inputs and outputs, DesignAI determines the number of variables based on these inputs and outputs, users can modify the selections and run the studies. After the studies have run, users can get quick visualizations based on those results.
Models can be automatically updated with design changes based on results from DesignAI. These results can be varied part or element properties within the design space. This truly allows faster design convergence and can eliminate extra work by finding inadequate design changes earlier in the cycle. DesignAI is accessible as a standalone, cloud-hosted app through the Altair one portal.
Fig 4: Example of input variables recognized by DesignAI
Altair shapeAI
The main goal of shapeAI is to recognize shapes, correlations, and patterns within your models. Built within the HyperWorks interface, it gives users the ability to find matching shapes, including mirrored parts, quickly without the need for manual searching. Users can make design changes to a single part, and have those changes synced to matching parts within an assembly. These part changes can also include meshing techniques that can be synced to matching parts. Another time saving use of this technology is the ability to link design variable for gage optimization for instance. If you are working on a large assembly with instances and mirrored parts shapeAI can automatically link the gage design variables for families of parts.
Essentially, shapeAI makes the manual process of finding similar parts in an assembly automated, saving a lot of an engineer’s time and sanity.
Fig 5: Example of shapeAI result
It feels like almost monthly, we are hearing of new ways these tools are being applied to our customers business, and love that these technologies are making their workflows more streamlined and efficient. We will update accordingly as updates to these tools and new tools come out, but it’s exciting to see how Altair is embracing the world of AI to support engineers as they figure out how these tools are going to impact their processes. When combined with the advancements of the data analytics and machine learning portfolio, the goal is to make Altair, and its partners, the best support system for companies looking to innovate and take advantage of these emerging technologies.
As always, if you have any questions or want to dive deeper into any of these tools, don’t hesitate to contact us. You can also subscribe to our YouTube Channel to get the latest on Altair tools!
Submit a comment