Altair Engineering Inc.
12/12/2024 | News release | Distributed by Public on 12/12/2024 06:18
Tips for Choosing the Right AI Use Cases in Engineering
Ever feel like you're missing out on something important? For many engineers, that "something" might just be their first artificial intelligence (AI) use case.
This may come as a surprise since there's no shortage of interest or investment in AI in engineering - research shows that 86% of engineers value AI as an important emerging technology. This may come as a surprise since there's no shortage of interest or investment in AI in engineering - research shows that 86% of engineers value AI as an important emerging technology. However, it's apparent that for engineers looking to start their AI journey, identifying a viable AI use case is a common hurdle. Altair's Frictionless AI Global Survey revealed that 35% of respondents either don't know where to start with AI, or find the path to business value unclear.
Many professionals at Altair's recent "AI for Engineers" workshop expressed similar frustrations. Part of the problem is the hype around AI. Understandably, people on the front line of engineering sometimes find it hard to square bold claims about AI's transformative capabilities with the reality of their day-to-day work.
So how should engineers face this challenge? Inspired by the professionals we met at AI for Engineers, we're sharing eight straightforward tips for choosing the right AI use cases.
Eight Tips for Choosing the Right AI Use Cases
Tip #1: Focus on Optimizing, not Revolutionizing
One of the panelists at AI for Engineers was the chief engineer for structures and simulations for a multinational heavy equipment manufacturer. He stressed the importance of not trying to overdo it. Think in terms of optimizing design attributes to meet customer needs, rather than in terms of revolutionizing products or transforming organizations.
Significantly, AI solutions can help engineers move on from manual design-and-evaluate cycles that are time consuming, expertise intensive, and subject to bias. Our panelist also demonstrated how dramatic the improvements can be. For example, his team is benefiting from reduced-order modeling (ROM), using high-fidelity simulations to create training data for a neural network (essentially a pattern recognition algorithm) that in turn creates an AI-enabled dynamic model. Finite element (FE) models can now be solved in 16 seconds, rather than nine hours.
Tip #2: Identify the Outcome You Want to Predict
AI is a vague term. Right now, when we talk about AI in engineering, we usually mean machine learning. Generative AI (genAI) and other flavors of AI are beginning to make their mark, but the differences are best covered in another blog.
Machine learning uses algorithms to create a model based on sample data, known as training data, to make decision or predictions. Machine learning systems learn from that data without needing to be programmed to do so.
That means machine learning projects all start with a simple but fundamental question: What outcome do you want to predict? Many AI projects fail simply because the objectives were never clearly defined.
Tip #3: Ask Yourself if You Have the Data
After identifying a potential outcome, the next question is straightforward: Do we have the necessary data?
Machine learning needs lots of complete and accurate data. For some organizations, getting that data is a challenge. Many enterprises also struggle with large volumes of messy, unstructured data, which is difficult to prepare for machine learning applications. These problems make it difficult to do any meaningful machine learning work.
Fortunately, accessible and efficient tools can automate the data preparation and cleansing process. AI can also be used to impute missing values in data and even generate additional data points. Altair recently worked with a company in the materials sector to automate data acquisition from various sources and then consolidate and prepare that data for machine learning models. These models generate high-quality synthetic data to fill gaps, reducing the need for physical testing and accelerating early-stage decision-making.
Tip #4: Think Domain Expertise, not Data Science
The shortage of data science expertise was another common refrain heard throughout AI for Engineers. It's a problem that affects all industries. According to the U.S. Bureau of Labor Statistics, demand for data science is set to increase by a staggering 36% in the next decade compared to the 4% average for all positions, making it one of the country's fastest-growing roles.
But don't panic if the data science department isn't returning your calls - or if the data science department doesn't even exist yet. Successful AI applications in engineering are driven by people with a granular understanding of data, processes, and desired outcomes. What's more, democratization is transforming AI. Low- and no-code AI tools are empowering a new generation of citizen data scientists. Increasingly, AI is also being incorporated into existing design workflows. For example, engineers can now easily access time-saving AI-powered tools that automate shape matching and make physics predictions directly from CAD files.
Tip #5: Start Small
The barriers to adopting AI aren't just technical. Cultural resistance is also common. Attendees at AI for Engineers agreed that starting with a small initiative and demonstrating a quick return on investment is a proven strategy for securing wider organizational buy-in.
Don't overlook potential AI use cases just because they aren't flashy enough. AI excels at "boring" stuff like automating time-consuming, repetitive design tasks. As such, it frees engineers to use more of their time flexing their unique skills, experience, and creativity.
Tip #6: Be Inspired by Others
Still wondering what outcome prediction and process optimization look like in practice? Check out some recent AI use cases here, including how one manufacturer is predicting quality control issues earlier in the steel production process.
All these projects highlight common benefits of AI, including reducing physical testing, accelerating time to market, and supporting earlier, more informed decision-making.
Tip #7: Don't Let Perfect be the Enemy of Good
Waiting for the "perfect" use case, dataset, or timing often leads to missed opportunities and regret. Instead of striving for an ideal scenario that may never come, focus on using what you have. Engineers are natural problem solvers, and the iterative nature of problem-solving aligns perfectly with this approach. Start with a viable use case, work with the data you have, and refine as you go.
Progress often comes from learning through doing. Early iterations might not be flawless, but they provide invaluable insights and momentum. By starting small and iterating, you can identify challenges, adapt solutions, and drive tangible results faster than if you'd waited for perfection. Remember, what's good enough today is often the foundation for what's great tomorrow.
Tip #8: Reach Out for Help
Altair is rooted in engineering. We pioneered the simulation-driven design approach, and we're committed to making it just as easy for users to harness AI-powered engineering. We've partnered with key experts such as Factspan to enhance support for our clients, ensuring they're well-positioned to identify and implement AI use cases.
In collaboration with Factspan, we're pleased to offer organizations a complimentary, design thinking-led Data and AI Maturity Assessment and Workshop tailored to users' chosen business or data function. The goal is simple: benchmark your readiness for AI adoption and lay the foundation for long-term success.
AI solutions are delivering game-changing improvements in engineering's speed, efficiency, and quality. Ultimately, AI may do all the revolutionary things futurologists predict. But for now, engineers should use it as another tool in the box. The first use case is almost always the hardest. The good news is that engineers have the ideal skill set to embrace AI, and solutions are within reach that help make a giant leap feel more like a small step.
To learn more about Altair's data and AI capabilities, visit https://altair.com/data-analytics.