ODYSSEE A-Eye will use artificial intelligence to create enhanced DOEs.
Hexagon’s Computer-Aided Engineering Gets an AI Boost
Like many simulation software packages, Hexagon seeks to make the process of engineering development better. Better might mean that the process takes less time, costs less money or delivers more quality parts to the customer. The idea is that more work and optimization being done by engineers upstream in the process will save time, effort and money at the end of the process. Finding a mismatch in material or processing on a digital model is easier to fix than making the change to 12 physical prototypes on a workbench, and definitely preferable to fixing thousands of units out in the field.
Hexagon’s Manufacturing Intelligence Division focuses its efforts on design and engineering, production, and metrology. The production arm of the division works to maintain design intent for components, managing and optimizing toolpaths for many different materials and processes. The metrology division works to give manufacturers information beyond whether or not a part meets dimensional requirements. Design and engineering functions of the software work to the various specifications for a product and create the best possible design.
For example, Hexagon recently announced ODYSSEE A-Eye, an artificial intelligence (AI) tool that will work with existing computer-aided engineering (CAE) tools (Structural, Thermal, CFD, Acoustics, MSC Nastran, Marc, Adams, Cradle CFD and Actran) to run studies without requiring expertise in CAE preparation or simulation.
Pattern Recognition for Machine Learning Analysis
Hexagon says that ODYSSEE will apply pattern recognition to images, pictures and videos so it can bring that information into machine learning algorithms for CAE applications. Existing CAE data can be compared with the pattern recognition images and create like-for-like predictions. The company said that applying machine learning to complex engineering problems usually requires expert knowledge and massive amounts of training data and processes. However, the ODYSSEE A-Eye platform will give users digital twin capabilities and use CAD models and image files to help solve problems.
Smaller companies might not have access to a team of experts and the computational resources needed to implement AI, but they can access the Hexagon software and take advantage of this new system. The ODYSSEE A-Eye platform is integrated with the current Hexagon CAE software and will let users take advantage of the image recognition tools at launch. A-Eye combines several different forms of data analysis, including:
- Pattern recognition
- Image processing
- Data fusion
- Data mining
- Process delivery
- Machine learning
- AI
- Automation
- Optimization robustness
The hope for this machine learning tool is that instead of performing several simulations that are done to aid an engineering decision, users can take a few strategically chosen points and run fewer studies. Less computing power overall will be used on this specific problem. Then, the AI components of the tool will let users get smarter over time and save even more effort.
Predicting Noise Transfer Functions
Two case studies are available on the Hexagon website, detailing specific uses of the new software tool. Satven used ODYSSEE A-Eye to reduce the time to produce trimmed body noise and vehicle harshness (NVH) simulations. Satven wanted to use AI and machine learning services to give customers a competitive advantage with respect to the NVH problem. Trimmed body noise transfer and vehicle transfer functions were optimized by studying the effect of BIW panel thickness and Young’s “E” modulus changes in the system.
The National Highway Traffic Safety Association’s data for a Honda Accord was used as the baseline for the study. The team hoped to study, predict and optimize the effects of component thickness and material sensitivity on the noise transfer functions. The novelty of the study was that the team didn’t use traditional CAE simulations and instead used ODYSSEE A-Eye. Key components were chosen for the study and run using the Parser Tool and Nastran as a solver.
Comparing the results between ODYSSEE A-EYE, CAE and Nastran showed a 91 percent correlation. The traditional finite element methods required around three hours before completion. ODYSSEE A-Eye took only a few seconds. The case study noted that identifying the parameter sensitivity in the parts took a few seconds per study, so this delta in solver times would be even more pronounced on a full series of simulations.
Optimizing Crash Parameters
The second case study again featured Satven as the customer, and the company’s efforts to predict the effects of different materials and thicknesses on crash parameters. The objective of this study was the comparison of traditional test methods and ODYSSEE A-Eye. This study looked at energy absorption forces and deformations of a vehicle crush can using different combinations of materials and thicknesses.
The current process for crash test simulations takes a full day using super computers. If the results show that the crush can does not meet design requirements, engineers then recommend changes and run tests on another 24-hour cycle. Satven assumed that this would be a computation-heavy process that would be improved using ODYSSEE A-Eye. Twenty different CAE studies were used as training data for the A-Eye software, allowing the tool to predict force and deformation values.
The AI tool showed a much lower time requirement for each run, but the engineers also decided to run multiple designed experiments on the system because each iteration could be run so quickly. Satven noted that animations were also predicted accurately with the tool, for all node deformations.
The Democratization of CAE?
The implications for the use of image-based machine learning could be huge. Currently, smaller companies have only a few options to run big studies, hiring contractors or bringing the work in-house. Contractors are expensive and might require workers to verify the results after studies are run. Adding another company into the design process stream also adds time and complexity. Bringing the work in-house requires the company to have an expert-level simulation employee and computing power strong enough to run the high-level simulation software.
Hexagon’s vision is that companies without a big simulation budget, and without heavy computing firepower, can work with the ODYSEE A-Eye software to get comparable results. It doesn’t seem likely that CAE simulations, as we know them, would go away or even have their use diminished in the field. However, this new tool might allow new players into the product development field. This might be good for competition and innovation in the long run.
“AI is an increasingly valuable tool within design and engineering, helping push virtual engineering to the next level,” said Roger Assaker, president of Hexagon’s Design and Engineering Software Business Unit. “It has the potential to shorten the time taken to complete labor-intensive design tasks that may have previously taken days or weeks down to minutes or hours without losing simulation fidelity. Furthermore, the user-friendly design of ODYSSEE A-Eye makes it simple to integrate into modern engineering practices, democratizing a highly advanced process for use by non-experts and producing the results in a very accessible format.”
Looking with a narrow scope, this new software might help manufacturers develop consumer electronics faster or create more optimized cars, buses or planes. Big picture, these types of machine learning simulations might help to solve our grand challengesof engineering. If methods can be discovered to use image-based machine learning to help speed up the overhaul of our infrastructure or bring solar energy to wide adoption faster, then it’s definitely worth the effort.
