The Importance of Embedded AI PCs in Manufacturing

When I worked at a smaller manufacturing operation, we migrated our ERP system from a program that resembled DOS to a cloud-based, modern-looking ERP that promised to have all kinds of new capabilities we would enjoy. Upon launching and utilizing it on the factory floor computers, the shipping department found that processing a single shipment in the new system took 20 extra minutes of computer work. The reason was simple-our infrastructure was not where it needed to be to handle this new software.

Hardware Requirements Can't Be Neglected

The unfortunate truth is that this is not a rare occurrence. Software is only as good as the hardware it runs on. There's no question that AI is the hottest topic in business these days, but like any software, it needs to be run on the proper hardware. Since the most familiar form of generative AI are public LLMs that can be run on practically any computer, it comes as a surprise to many that business AI use cases require powerful devices designed specifically for use with AI-powered software. That's where embedded AI PCs come in.

Despite the name, embedded AI PCs do not usually come with AI applications installed. They are simply PCs (usually in a micro form factor) that are tailor-made to run AI-powered software and can quickly respond to input, making them crucial to certain operations. They can also be built to withstand the rugged conditions of any manufacturing environment.

AI has proven itself able to truly excel when given tasks involving pattern recognition, so most applications that yield good results are created with this in mind. Thus, it makes perfect sense that AI-powered camera vision can effectively perform quality control-it's been trained that products must be a certain way, and anything that doesn't conform to that pattern needs to be flagged. Even better, it can see these issues in real time during the production process, so manufacturers can easily determine when and where the defect was caused. If the same issues are popping up repeatedly, AI can also analyze the data and make process improvement recommendations.

However, running this software with a traditional PC could mean it will take several minutes before a result is returned, and in a high throughput facility, this can be especially troublesome. For example, a quality control inspection performed by a human on a circuit board can take several minutes or even hours depending on the complexity of the product. AI programs running on hardware with the right processing power can utilize high-resolution cameras to check a board in under a minute, allowing for much faster processing and often more accurate test results. Such rapid testing means that it can also be done multiple times throughout the facility instead of one formal inspection at the end. The days of wasting precious time and resources on defective products that should have been pulled out five operations prior to final inspection can finally be over.

The Use Cases Are Limitless

Once you have hardware and software that can react quickly, the possibilities are endless-which can make it tough to know where to start. My advice is to tackle the persistent challenges in your business that are shared by most operations: keeping your machines up and running and keeping automation reliable.

Factories have been aiming for true predictive maintenance for decades, but it's been a very difficult challenge for most operations. However, once you're equipped with devices that can quickly respond to machine conditions, the goal of zero breakdowns is within sight. Embedded AI PCs have the processing power needed to look at all your real-time factory data and immediately adjust machine operations to prevent errors and breakdowns. They can also automatically notify maintenance personnel of changing machine conditions so that repairs can be scheduled when the machine is not in use.

Considering that 76% of manufacturers are using automation in their factories, the value it can bring is undeniable. However, even simple automated operations can have errors and crashes because the machine is blindly following a program and will carry out those instructions regardless of the circumstances even when a human would know to stop. However, embedded AI PCs running your machines can adjust the program based on circumstances just the way a person would.

I once worked in a facility that involved a lot of liquid handling. Ultimately, all you needed to do was draw liquid from a source and dispense it in another location from a syringe-like container. This is extremely easy for anyone to do, but automated machines had issues due to variation in container size, container location, and several other factors. I spent days upon days fine tuning programs to handle all of the exceptions, but an embedded AI PC running the right software could have simply ran the basic program and seen "this is a different container, stop short of the container bottom and dispense liquid" or "the container is off center, adjust robot arm half an inch right," and we could have run our operation with practically any variation. This same type of functionality can be applied to almost any operation that involves automation. All of this is only possible due to the near instantaneous reaction time offered by embedded AI PCs.

Don't Go It Alone

If you've ever built your own computer, you know it can be frustrating and time-consuming making sure you have the right parts to run the software you need at a reasonable price, which is partially why prebuilt PCs have become so popular.

There's no reason to add to the difficulty of implementing AI in your factory-that's what Connection is here for. If you're considering harnessing the power of AI on the manufacturing floor, engage our Manufacturing Practice today. We will make sure you have the right PCs for your specific use case so that everything will run as intended, and you can start reaping the benefits of this exciting new technology.