In January 2024, we launched nvision, an open, solver-agnostic, and data-based modeling tool by Noesis Solutions, to enable simulation engineers, designers, and product owners to reduce design time by leveraging existing simulation data. nvision performs real-time "what-if" analyses, speeding up product development while saving both time and resources. In this AI-driven era of product development, surrogate modeling powered by nvision can significantly reduce the time required to evaluate the performance of new designs and predict their behavior in near real-time.

However, as AI technologies such as machine learning, neural networks, and advanced algorithms become more accessible, one critical question remains: Are we, as engineers, ready to fully exploit AI in our design and analysis processes?

The Promise and Challenges of AI Models in Engineering

AI models offer immense potential for engineers, but they also present unique challenges that require a fundamental shift in thinking. The adoption of AI compels a convergence of various disciplines—mathematics, physics, data science, and engineering—all of which have distinct perspectives on data. For mathematicians, truth lies in numbers. Physicists focus on ensuring those numbers align with the laws of nature. Data scientists are driven by patterns, sometimes irrespective of physical constraints. Meanwhile, engineers demand results that can be reliably replicated in the real world.

This is perhaps why many engineers are initially skeptical of data-driven AI models. Yet, here’s the irony: the very data that powers these AI models originates from the engineers themselves. The key challenge is ensuring that the data we input into these AI models is both clean and meaningful.

AI Models and the Data Challenge: Garbage or Gold?

A common adage in AI is, “Garbage in, garbage out.” AI models are only as good as the data they receive. If the input data is flawed, the models are likely to produce flawed results. However, this also underscores an exciting possibility: if AI is fed high-quality, structured data, the results can be transformative— “Gold in, gold out”. This is where engineers come into play. The real challenge is ensuring that the data used in AI models is of high enough quality to unlock the technology’s full potential.

As engineers, we face the daunting task of generating vast amounts of this ‘gold’ data. One solution might be to dig through legacy data—archives filled with valuable insights from past projects. But this isn’t always as simple as it sounds. Often, legacy data is a disorganized mix of outdated simulation files and quick fixes: creating a messy collection that may seem overwhelming at first glance. However, there’s always value to be extracted, even in this chaos. No knowledge is ever truly lost. With the right approach, even seemingly messy legacy data can be organized, cleaned, and used to train AI models. This process, though resource-intensive, holds tremendous potential.

Balancing Legacy and New Data for AI Models

Another option is, using AI only on brand new projects where we will produce simulation data in a very well-structured manner? So, which path do we follow first? Investing time on cleaning up legacy data, or considering AI as a solution only for new data coming? I would say you don’t have to choose! It is in mathematics where there are sudden jumps, singularities, infinitesimal steps... Not in engineering! A sudden jump in theory is just a very high gradient in engineering, and you can choose the partial gradient you like. The automotive industry has a relatively high gradient in switching to electric drivetrain, but a smoother gradient in adapting the chassis and body designs. Maybe we need 5 or 10 more years until we see dramatic variations in the car bodies driving on the streets but, adapting the drive train to decarbonization is much more urgent. So, the same engineers can easily extract data from legacy simulation files and homogenize them for AI applications, while finding their well streamed way of generating new simulation data.

The question is not whether AI will play a role in the future of engineering—because it already does. The question is how well-prepared we are to harness its potential. By implementing data-cleaning processes, engineers can feed AI models with both legacy and new data to extract meaningful insights. This process of "autumn cleaning" legacy data not only repurposes it for AI but also encourages better organization for future use. The AI models on legacy data will serve as the library of your “know-how” in a nutshell, streamlining product development for common design challenges. And integrating a data-driven AI approach can enable organizations and engineers to enhance efficiency and innovation.

For more information, review the latest announcement, webinars, case studies and stay up to date on the latest products and case studies at Noesis Solutions. Watch a free demonstration of the product here and to discover how nvision can accelerate innovation at your organization, request a free non-obligatory consultation here.