This blog post was co-written with Matthias Bastian.
Manufacturing is a very data-intensive industry. Most of that data has so far been managed on a different floor than processes like assembly. AI can now process enough information fast enough – and convey it practically enough – to be of immediate use to engineers on the line.
Artificial Intelligence is one of the most engaging emerging technologies today. AI has the potential to impact just about every aspect of human society. You’ve probably seen it already in fields like art and entertainment.
In addition to consumer AI, the technology also holds promise for manufacturing. There are many use cases for AI technologies, such as computer vision, that are useful in everyday environments, but also lend themselves very well to industrial use cases. This article looks at a number of specific use cases and applications of AI in manufacturing.
AI has a number of potential use cases in manufacturing, including automating some work processes completely. Right now, most roles that AI takes on involve helping human workers access more information more efficiently.
This use of AI helps experienced workers work better, but it also helps to transition workers into new roles or help new workers learn the ropes faster. Far from replacing human workers and taking human jobs, AI is really empowering human workers and making sure that humans can enter positions that companies are actively trying to fill.
Improving efficiency and productivity has always been a major incentive for collecting and analyzing data. This used to mean taking data from the floor, analyzing it in offices, and communicating findings back to the workers on the floor.
Industrial artificial intelligence - with sufficient human supervision - can carry out this whole process on the floor, providing insights to engineers in real time when and where they need it. This compresses the information flow from a day or more to fractions of a second. Some processes, like ordering more parts and materials before they run out, have already been automated by comparatively basic AI systems..
Artificial intelligence is already being used across the manufacturing industry to do things like recognize potential workplace hazards, automate materials and components ordering, and guide workers through their day. Industrial artificial intelligence is also used to help analyze information and convey it in actionable documents and communications.
These existing use cases aren’t going anywhere. But advances in AI are introducing emerging use cases and making existing ones even more efficient.
Quality control and visual inspection have already seen massive improvements from AI. PTC’s Vuforia Step Check walks supervisors through the process of training an AI on digital and physical models to create a program that helps inspection engineers identify and even troubleshoot potential product issues. Step Check then automates the documentation process, increasing worker efficiency.
AI in manufacturing can also use information from Industrial Internet of Things devices to generate predictive maintenance strategies. These strategies can optimize output by keeping machines in peak working order, not to mention preventing costly downtime by scheduling maintenance instead of waiting for repair. PTC’s Kepware allows human operators to connect smart devices and see their real-time diagnostics at a glance.
In a similar tactic, manufacturers are patching industrial artificial intelligence systems into inventories to automate ordering essential parts and supplies before they run out. In periods when supply chains are already sensitive, AI can prevent delays from the simple mistake of not ordering components in time.
Generative AI is increasingly proving itself capable of creating usable content from prompts, including in the age-old field of Computer Assisted Design. Accepted industry tools like PTC’s Creo are likely to find themselves increasingly augmented by inputs from artificial intelligence specializing in product design.
Further, the prevalence and usefulness of completely automated processes is skyrocketing, as are the number of “cobots” in manufacturing. Cobots are machines that work alongside a human operator performing guided tasks that might be physically impossible or dangerous for a human.
The expansion of robot and cobot use in manufacturing at the same time as explosive growth in the fields of large language models and natural language processing help to drive dreams of fully intelligent and interactive robots that communicate organically with human coworkers. However, this remains a thing of the future for the time being.
Augmented reality is another emerging technology that already has several established use cases in manufacturing. AR models are increasingly replacing physical mockups in early design phases where it saves material cost and iteration time. These models can also be used in remote collaboration programs to save travel cost, as well as for training modules. These models can even be generated from CAD programs that companies already use in the conventional design workflow.
In this article we’ve already mentioned the interplay between industrial artificial intelligence and AR in manufacturing. When artificial intelligence powers augmented reality applications, the benefits of each technology multiply. For example, Magna International subsidiary Nascote Industries leveraged Vuforia Step Check both for new-hire training and to enhance the visual inspection process. The software was even able to identify a “soft connection” that often got past inspection engineers but would eventually become loose.
AI is initially trained from images of a physical product or from existing models. However, powerful AIs can then generate their own images and models simulating different situations and conditions. Called synthetic data generation, this process can help to optimize product design, as well as prepare an industrial artificial intelligence system for situations that haven’t yet been encountered or included in training sets – similar to how a human might imagine what they might do in a future position or situation.
Augmented reality, by presenting spatial information in an intuitive medium, is also a great way at conveying complex information efficiently, making it just about the only way for workers to effectively harness the volume of information that AI makes available. In this way, AR becomes an interface through which humans are able to interact with artificial intelligence, making industrial artificial intelligence a practical work aid.
Step Check can do more than scan for defects. The program incorporates work instructions that guide the inspection engineer through an entire visual inspection workflow, including helping him or her navigate around the object in space and even troubleshoot common problems if possible. Finally, the program automatically generates a report on each inspection, including any issues that might have been found with each unit.
These last processes – AI-powered work instructions and documentation solutions – have futures as their own projects. Inspection engineers aren’t the only kinds of workers currently following physical work instructions and managing their own documentation. Workers in all kinds of roles are currently jostling paperwork instructions and stepping away from their real jobs to file reports.
These hassles are inefficient at the best of times and dangerous at the worst of times as physical documents are just one more thing to occupy a hand and take attention away from the engineer's surroundings.
Further, AR solutions can be engaging in a way that standard processes typically aren’t. While these solutions aren’t deliberately gamified, they can be mentally stimulating and a welcome change-of-pace to what might currently be a mundane task that is repeated throughout a long shift. This can go a long way toward improving job satisfaction.
Allowing industrial artificial intelligence to help in these areas will also help new workers to catch up and to transfer knowledge from more experienced workers and subject matter experts alike. Training the AI from human experts turns the AI into an expert in its own right. Also like a human engineer, the AI continues to learn from each use. That knowledge and experience is then passed on to newer human engineers who see the accumulated knowledge of the AI through intuitive AR displays. With the advancement of natural language models, we may also soon see AIs that onboard and upskill workers more efficiently and safely than conventional training methods.
AI in manufacturing isn’t the end of the story. Applications similar to those that we’ve explored on the assembly room floor can also be implemented after a product ships so that service personnel can maintain products without sending them back to the manufacturer. One day, these tools may be expanded to the average customer.
While the potential for AI in service after sale is huge, there are still some questions that may need to be answered before the practice becomes commonplace. For example, will an AI trained on potentially proprietary product information be a security risk for companies adapting that AI for public use? Will there be a language barrier to overcome when a program meant for engineers begins communicating with non-specialists? Who might be responsible for any consequences of that?
We’re still at a moment defined by companies and individuals alike becoming comfortable with using AI for more and more tasks. Questions like those above shouldn’t stop us from exploring these possibilities; they should guide us as we move forward with new AI implementations.
Some of the concerns mentioned above deal with corporate privacy and security. This is a real concern for companies. In many cases, companies – particularly those with defense contracts or dealing with other sensitive clients and customers – have strict security regulations for services involving the use of a camera. These can often be solved with on-premise solutions that don’t always lend themselves well to AI. However, these concerns are increasingly being solved by private cloud infrastructure or edge computing that maintains information on the device.
One myth of AI adoption has to do with replacing human workers. The fact is that the growing skills gap in manufacturing promises to leave millions of critical jobs unfilled over the next decade. Implementing AI in roles that support human workers gives people the resources necessary to step into roles that are already available.
In many situations, such as those involving cobots, AI changes the jobs that humans perform – rather than performing a dangerous or exhausting physical task, they now need to supervise a machine performing that task. AI in manufacturing doesn’t put humans out of work, though it might put them out of harm’s way.
However, AI is a powerful transitional technology. Making the most of it means building trust in AI systems and ensuring regulatory compliance. Mindful regulation can keep humans and companies safe while using AI and help us to prepare for the ways in which AI may change the workforce. For example, regulation might require that critical decisions be made by humans rather than machines, or that money spent on AI infrastructure comes with a budget for training humans to work with AI-powered devices and systems (upskilling) or to transition into jobs that can’t be filled by emerging technology (reskilling).
We’re in an explosive moment for AI, but AI will only become more practical in the manufacturing industry through the development and adoption of companion technologies like augmented reality and advanced data systems. These technologies allow AI to work through existing infrastructure in ways that are accessible to existing workers, driving workforce efficiency.
The future of industrial artificial intelligence isn’t just the future for the technology, it’s also the future of the people who use it and benefit from it. Companies using AI will see production and environmental costs go down as they save on material, travel, downtime, and rework.
Workers working alongside AI will see job satisfaction increase as AI automates both the most mundane and the most dangerous elements of their jobs. Finally, customers will benefit from more affordable and more reliable products as well as increased ability to maintain and repair those products when necessary.
AI has already been involved in industry in terms of data management and interpretation. We’re just now starting to see the seismic shift that occurs as AI finds its way to the production floor through robots, cobots, generative AI and through emerging technologies like AR that give engineers on the ground the ability to access AI insights in real time.
This is an exciting time for forward-looking companies to start integrating AI into their existing workflows and connecting their existing infrastructure.
Jon Jaehnig is a freelance journalist focusing on emerging technologies, particularly mixed reality and blockchain. He writes for MIXED and ARPost, among other publications.