The future of engineering: Blending AI speed with simulation accuracy
The synergy between AI and simulation was a hot topic at last year’s Witness User Conference, and it’s also a hot topic in academia. Michigan Technological University (MTU) offers a well-known manufacturing and mechanical engineering course that uses Witness simulation modelling software in the curriculum, and course coordinator Dr. Vinh Nguyen is also director of MTU’s Center for Artificial Intelligence.
Dr. Nguyen chatted to us about how he’s seeing AI and simulation evolve, and the implications for training tomorrow’s engineers.
(This is part 2 in our series with Michigan Technological University. Read part one here, where Dr. Nguyen explains how they use simulation in their curriculum to boost student engagement and provide a stronger link to real-world manufacturing challenges.)
Where do you see opportunities for AI and simulation to work together?
Dr. Nguyen: There are clear use cases for AI to support with faster model creation, iteration and interpretation. For example, there are growing opportunities to use GenAI chatbots with natural language interfaces to help define and refine models, and to analyse model outputs. By building surrogate models and guiding iterative simulations, AI could help engineers tune parameters more efficiently and identify the best-performing process much faster.
In return, simulation generates contextual training data and can guide machine learning.
What AI-related limitations are you seeing?
Dr. Nguyen: The major limitations relate to data. We’ve had students take a screenshot of Witness, upload it to ChatGPT and ask it to rearrange the layout to achieve a specific throughput. The results were terrible because an image doesn’t contain the data structure the LLM needs. The ChatGPT outputs improved dramatically when students represented data in a tabular format.
The Center for Artificial Intelligence has encountered similar issues. Teams are wrestling with the challenge of how AI can explain simulations or act as a natural language interface for managers to ask questions conversationally. In theory, ChatGPT could integrate CAD data into Witness and calculate cycle times automatically. However, for this to work, every tool would need to be aligned and interoperable from a data perspective. I’ve been part of working groups trying to standardise this, but it’s hard. The challenges are logistical rather than scientific: they aren’t to do with AI itself, but rather the proprietary ecosystems around it.
In a manufacturing context, are you seeing a preference for AI vs simulation?
Dr. Nguyen: In a manufacturing environment, precision is everything. Recently, I took computer science students to a manufacturing engineering conference, and they were surprised to see “old-looking” robotic arms compared with the more advanced robots they’re used to seeing. I had to remind them that production processes operating at Six Sigma levels deliver 99.997% accuracy, and that need for precision affects the ability to try experimental tech.
It's the same with AI right now. Simulation leaves AI in the dust from a precision perspective. Even when it performs well, AI deals in approximations rather than certainties. This is problematic in production, where even small deviations can have huge impacts.
In a manufacturing environment, precision is everything… Simulation leaves AI in the dust from a precision perspective.
Is there still a role for humans as engineering evolves?
Dr. Nguyen: Without question. AI will change manufacturing, but humans in the loop will remain essential. You need human judgement to quality-assure inputs and outputs and apply them to decision-making.
While AI can accelerate optimisation or suggest scenarios to test, it can’t replace the systems-level reasoning and constraint awareness that engineering demands. The future belongs to engineers with critical thinking and a solid foundational understanding – who can use technologies like simulation and AI as tools to enhance rather than replace.
----
Dr. Nguyen started his appointment as an Assistant Professor at Michigan Technological University in 2022, where his research focuses on advanced manufacturing through Industry 4.0, human-robot-machine interaction, and physics-based/data-driven modeling. Dr. Nguyen has developed solutions for a variety of production processes including machining, additive manufacturing, metal forming, and robotic assembly to promote smart and sustainable manufacturing. Dr. Nguyen also conducts research in general industrial automation including autonomous vehicles and industrial robotics. Prior to his appointment at Michigan Technological University, Vinh was a National Research Council Postdoctoral Fellow at the National Institute of Standards and Technology from 2020 to 2022. He received his PhD in Mechanical Engineering and his MS in Mechanical Engineering and Electrical & Computer Engineering at the Georgia Institute of Technology in addition to receiving his BS in Mechanical Engineering and Electrical Engineering from Rensselaer Polytechnic Institute.
