Delivering the Promise of the Internet of Things in Manufacturing

How can the Internet of things be adapted your manufacturing goals? The key is to ask—and have answered—these five questions.

< Back to Insights
Insights  <  Delivering the Promise of the Internet of Things in Manufacturing

How can the Internet of things be adapted your manufacturing goals? The key is to ask—and have answered—these five questions.

Manufacturing businesses have no doubt heard repeatedly about the value of the Internet of Things (IoT)—how it will help reduce costs, provide a shorter time to market, support mass customization and/or improve safety. These are all heady goals and expectations, and the truth is they are harder to achieve without the right perspective on what IoT can truly deliver.

That is why manufacturers need to think about what it is they are trying to accomplish, and then develop a plan to get there; it is within that plan where several different technologies will be used and some of IoT’s promises can be delivered.

As an example, say a manufacturer is looking for a predictive maintenance solution. How can these technologies be adapted to support one? The key is to ask—and have answered—these questions.

  1. What problem am I trying to solve?
  2. What information do I need to know to solve that problem?
  3. How do I capture the data necessary to create that information?
  4. What technology do I need to interpret the data?
  5. How do I turn data into information?

“What problem am I trying to solve?”

One of the most common causes of failed IoT projects is the failure to clearly define the goal/objective being south. Tackling something like “predictive maintenance” is too large a problem, so it needs to be broken down into component parts.

First, are you trying to tackle this for your own internal equipment, or are you trying to create a new feature and service line for products that you sell to your customers? Start with one machine, and then one aspect of it. First you ask what needs replacing—such as oil, a filter, a cutting blade, or something else?—and then work backwards from there.

“What information do I need to know how to solve that problem?”

So, how do you know something needs replacing? There are leading indicators and lagging indicators. A leading indicator would be that you have a way to know, for example, that the filter is too clogged before it has a downstream impact. A lagging indicator would be something you could know after the fact, like the product you are producing is no longer within tolerance. Though in deal circumstances, leading indicators are preferred.

“How do I capture the data necessary to create that information?”

Let’s say the pressure differential has increased, and that suggests that you may need to replace a dirty filter—that’s the information, so what data do you need to derive that? This is where we start getting into IoT platforms. In this case you’ll want a differential pressure sensor that can provide these values. Alternatively, perhaps just a pressure or flow sensor on either side of the filter that provides independent measures. In either case, you’ll want this device to be able to produce measurements.

Explore This

Internet of Things Services

Learn More

The question then becomes, “Where does the data go?” Traditionally, organizations would store the data on-site, but cloud-based storage may be preferred, depending on the volumes of data that you may need/want to process in order to start to interpret that data.

“What technology do I need to interpret the data?”

Assessing whether a part is within tolerance or not doesn’t require machine learning. Boundaries are established and if the measurement is within the boundaries, everything is working. Going back to our predictive maintenance problem, at what point does the filter need to be replaced? At what point does it start to impact the cutting tool’s ability to machine an in-spec part?

This is where machine learning comes into play and where having an abundance of data is extremely valuable. By storing our data in a data-lake type of solution, we can now use machine learning tools to uncover what the actual relationship is between the pressures and the parts involved.

“How do I turn data into information?”

Now that we have data and we’ve created a machine learning model that tells us what the pressure differential is, we can now finally get to the predictive maintenance piece. We can use a streaming analytics service to monitor the pressure differential and send data to our machine learning model in real-time. It evaluates the input against the model and returns a percentage likelihood that the part will be out of tolerance. We can now create an alerting mechanism (email, text, etc.) that lets someone know when this number reaches, say, 50%. At this point, we can now schedule the downtime to replace the filter rather than waiting until we start getting out-of-spec parts. We’ve finally gotten the payoff for this project.

This approach can be used over and over again to monitor other aspects of the manufacturing process and, ideally, come together with other data to provide a more inclusive and robust predictive maintenance model. However, starting small and following these steps will help lead to success. This same approach can be used to solve more than just a predictive maintenance question, and can be used to better allocate resources, handle material or even influence the manufacturing process itself. And it all goes back to the value that IoT systems can bring to a manufacturing business.

Continue the Conversation with Our Team
Get in touch with us.

Contact Us