How AI is Set to Change the Way We Design Mechanically Complex Products

Why do we need AI?

With great complexity comes great responsibility.

This has not been more true for the manufacturing industry than it is today. As physical products become more complex, manufacturing companies increasingly rely on specialist knowledge to meet regulatory and compliance requirements. We often hear something like, “Bill is the mechatronics guy.” Or, “We need to have Jane from supplier quality assess this RFQ.”

But this knowledge is at risk. Expertise is leaving the market faster than it’s being replaced. And this skills gap is only widening. Experts estimate that 33% of engineering roles in the U.S. remain unfilled.

Also Read: Making Microsoft SQL Server HA and DR Completely Bulletproof

Now, add mergers, acquisitions and new global facilities on top of this and these manufacturing companies not only have expertise leaking out of their organizations, they now have siloed expertise. Design teams in Brazil, manufacturing teams in Europe and engineering teams in the U.S. means knowledge lives in specific teams, countries and individuals.

Finally, with the need for increased IP security in a digital age, certain kinds of information just don’t get documented. Specifically, why certain design decisions get made and why. And because this kind of information isn’t documented or that documentation lives in too many places, engineers just make design decisions without them.

The reality is, in today’s fiercely competitive global market, there is no room for overspend, product delays or expensive recalls, which are frequent consequences of these knowledge gaps.

In this article, I’ll outline the key role of AI in preventing repeat errors, reducing risk and using the collective knowledge of an organization in mechanically complex product design. I’ll look at how these tools work in practice, and what the biggest challenges are for AI adoption in the engineering sector.

The role of AI tools in mechanically complex design

 When applied in the right way, AI will be a huge asset for engineering product design teams. The greatest risk to engineering teams is not that AI will replace them, it’s fear of AI. AI will not replace great engineers. In fact, keeping a human in the loop is essential. And government regulations are already recommending this as a best practice. The AI Act in the European Union proposes that “AI systems should be overseen by people, rather than by automation, to prevent harmful outcomes.”.

This is about using AI to do what a human brain cannot do quickly and comprehensively: consolidate and surface a lot of highly relevant data. What great engineers already know is this is the mundane engineering work that keeps them from doing great engineering. What AI should do is perform routine tasks with human oversight and free skilled engineers to focus on more critical problem solving and product innovation.

I see three major applications for AI tools in mechanically complex product design:

1. Leveraging knowledge
2. Automation
3. Generative creation

Leveraging knowledge

AI tools are essential in collating knowledge from across a number of different systems and sources, analyzing it, and then surfacing that information to the right person at the right time.

Without the use of AI, 87% of engineers spend hours, or even days, finding the right information to make a single design decision. And, even then, without a single, digital system in place, over half of the information is likely not to be recorded at all.

Some businesses are building AI solutions, like bots, to leverage knowledge in-house. But, when it comes to product design reviews, these will prove ineffectual if they’re not integrated into the existing process. Something that’s not possible with these kinds of ad hoc solutions. It’s through this integration into day-to-day activity that AI tools actually become useful, surfacing the knowledge where and when it’s needed to inform decisions.

Automation

AI can automate low-level design review work such as tracking, sorting, and prioritizing design comments on ideas. According to recent research, 23% of engineering time is wasted on non-value add work, like those tasks I just listed. By automating tasks like routine verifications, engineers can get valuable time back for more important tasks. This is the clearest application for AI automation.

Take this one step further and AI can also surface automatic insights, like similar parts, similar issues on those parts and then lessons learned during adjacent reviews. This is critical for engineering teams today. This kind of right time, right place AI automation can free up entire team members to do more creative engineering work.

Generative creation

The final tool in the AI box is generative creation, but it’s last in the list for a reason. Engineers who are strong supporters of AI believe that the technology will be able to perform generative computer-aided design (CAD), where you put in a prompt and get a suggested design back.

The more likely reality is that AI will create a canvas for engineers to use as a starting point, based on a command. It will then find and group all knowledge relevant to that command. In the future, it may be possible for AI to generate designs for parts within a product.

AI as a ‘catch-up’ tool for mechanical engineering

 It’s no secret. Mechanical engineering is much further behind than software development. Using AI in the form of smart analysis tools and collective knowledge, the industry can leap-frog this gap.

Clearly, it’s important to be pragmatic. A process should be effective before it’s made more efficient. This means parts of your design review process, for example, should be first doable without AI. But the tools used in this process, like CAD, PLM, sharing and markup tools can only track information. They do not understand what is similar and when and where this information has to be shared. And sometimes, the reality is a human simply cannot consolidate and analyze all this information at the speed and accuracy required.

This is where AI can act as a catch-up tool for mechanical engineering teams.

Also Read: Mastering Data Sovereignty: Balancing Innovation with Legal Responsibility

How AI tools work in practice

The automotive industry is a good example of how AI can be used in practice. Car manufacturers have a number of models that share the same underpinning designs. These models are often created by design teams in different places, on different systems. Getting AI to compile all that data, and surface relevant information at the right time, is very valuable.

For example, if there are faults reported on a model from 2022, AI can be used to ensure this information is surfaced in design reviews for newer models that have the same part. In this way, the fault is resolved at the design stage, ahead of manufacturing.

This example also works when looking at household electronics. If AI collects customer feedback data that refers to suction problems in a consumer-grade vacuum, this information can be shared with design teams who are using the same suction part in another model or newer versions of the same one.

Again, AI will not replace humans in this process. Just as software developers have tools to test their code, engineers will have pre-determined sets of requirements, which AI tools need to follow when performing tasks such as reviewing files. But AI will be able to take on about 80% of this work, freeing up your smartest people to innovate and do the value-add work.

The biggest challenges to adoption

 I see two massive challenges to AI adoption: fear and caution. I can’t speak to those who fear AI, but I can speak to those who are too cautious. This is a mistake.

Instead, manufacturing companies should be experimenting with AI in small, contained use cases.

The key here is to start with one small piece. Determine where AI tools can be most easily adopted within your organization and use the results from this first step to develop a process and drive wider adoption. A lot of the challenge is to do with processes, but effective AI adoption requires humans to interact and use the technology. In this way, a culture-shift is as crucial as the processes.

The second step is ensuring that AI tools are integrated into day-to-day processes. AI should be used to improve existing successful systems rather than be created and left out on its own.

The final step is ensuring accuracy. We know AI is only as good as the data it is trained on and, in this sector, a lot of valuable data is protected IP. This can create a massive gap. On the one hand: AI needs data to get more and more accurate. On the other hand: manufacturing companies are notoriously protective over their data and IP (as they should be). To ensure a balance between data accuracy and data security, it’s important to choose AI vendors who value your IP. SSO, role-based access controls, SOC-2, TISAX and ISO should be minimum requirements.

Next steps

There is no doubt that AI tools are essential for mechanical product design. Organizations must start building data sets for engineering decisions or risk being left behind. Partnership is another key part of successful AI adoption in the sector. It will not work if everyone tries to build their own things at their own speed. We need to collaborate and integrate with other systems and technologies to develop the AI tools the industry needs.

The most important thing for manufacturing companies to do now is act. It’s not an engrained mentality or action but the stakes are that high. AI will completely change the way physical hardware is designed, engineered and built. The question now is: will you act, or will you wait?