Organizing their data across the value chain is a big challenge for businesses in manufacturing. We had a conversation with Jeroen Caré, Market Lead Manufacturing Netherlands, and Suman Kumar Sengupta, Client Partner Manufacturing Netherlands at Cognizant, focusing on the intricacies of OT and IT integration and what it means for increasing business efficiency by not only predicting but also prescribing operational maintenance.
While digital companies lead the charge in embracing digitized processes, the majority of legacy companies still grapple with bridging the gap between IT and OT. It all starts with data hygiene, otherwise, the integration of OT and IT will fail. Integration opens the door for predictive, preventive, and ultimately prescriptive maintenance. But it also clears the way for digital twin technology, made possible by powerful solutions built on hyperscale platforms like AWS.
Where is the manufacturing industry as a whole when it comes to the convergence of OT and IT?
Jeroen Caré: “As in every sector, you will have legacy companies and digital companies. The same goes for the manufacturing sector. Digital companies have created manufacturing processes that are digitized by design. However, about 90 percent are what I would call legacy companies, meaning that they are built on traditional production and manufacturing principles and, therefore, by nature, are not digitized.”
Suman Kumar Sengupta: “However, it is not the level of digitization that marks some enterprises as ahead and others behind. Even companies with a comparable technical environment can differ significantly from each other. It is the misalignment between departments that is the culprit. On the one hand, OT is traditionally managed by the engineering division, which usually reports to the production engineering chief, production chief of manufacturing, or COO. On the other hand, IT is typically managed by the technology office, the CIO, or the CTO. We often don’t see converging ideas or synergy between the engineering and IT divisions. In many cases, this is the main constraint, rather than the technology.”
JC: “The challenge is to capture the data coming in from the OT to the IT. An overall look at the market indicates that about 90 percent of companies have yet to make this connection. In digital companies, OT and IT are integrated by nature. Around 90 percent of all companies however are struggling a great deal in making the switch to digital because they were never designed for this.”
SKS: “Also from a security perspective the IT/OT connection is important. This convergence facilitates remote control and management of operations and monitors certain critical, sometimes hazardous environments where, from a safety perspective, you don’t want your labor force to work.”
“IT and OT convergence also enhance traceability, which is mandatory in upcoming regulations for ESG reporting. From January 2025 onwards, the EU will become much more stringent about where you source your raw material, for example. IT and OT convergence is not just about creating value for manufacturing use cases, customer quality, uptime, and preventive maintenance, but also about helping with regulatory compliance on the traceability of raw material data and production data.”
Predictive maintenance has been on the radar for a long time. Could you elaborate on the adoption in the market?
JC: “Well, we see that where it has been applied, it has a significant impact on costs, as well as the mode of operations, the availability of processes, and machinery. So, where it has been applied correctly, the business case is really strong.”
“If you are doing predictive maintenance on a set of machines or production lines because you have been able to create a mature model on that particular entity from a data perspective – meaning that you are collecting the right data through the right sensors in real-time, then predictive maintenance modeling can work wonders on your shop floor. You don’t have to go back to the whole legacy discussion again.”
What would be the next step in predictive technology?
SKS: “In manufacturing, there is a lot of process integration and data coming from plants and machines, but also data from ERP applications such as inventory and planning data. Take a company that has hundreds and hundreds of plants and day-in, day-out, decade-long operations. The same is going on at hundreds of other plants. How do you collect and present the historical data for the next preventive maintenance?”
“With our APEx solution [see below], we integrate AI and machine learning to get all those historical patterns and enhance the predictability – and preventability – of particular incidents by looking at the data in real-time.
“We see more and more maturity in this area. With hyperscalers like AWS, the data and decision support are becoming real-time, and the ability to consume so much data in real-time or semi-real-time has improved a lot over the last couple of years.”
Now the question is, after predicting a failure, can you also prevent it?
JC: “Well, the next step after predictive maintenance, is what we call prescriptive maintenance. This is where the system determines what exactly you need to do to prevent the predicted incident. It goes beyond merely planning a maintenance session.”
SKS: “And in that final stage of maturity you may not need traditionally planned maintenance anymore – based on the demand and supply or the inventory, you will be able to keep producing and do the maintenance later. So this prescriptive maintenance is the future and it requires all the due diligence that we talked about before.”
Could you discuss the significance of AI in the manufacturing industry, aside from its role in maintenance?
JC: “In my point of view, AI and generative AI are very much connected to the digital twin principle. In the past, it was always difficult to create a digital twin, mainly because there were no good software solutions to build it. But with AI, the technology has gained a certain level of maturity now. Having that digital reflection of the reality that, with lots of data flowing from a particular production line, a particular machine, and a particular process, will help build the baseline data you need to build predictive models or AI models. Typically, the difference between a predictive model and an AI model is that the AI model itself creates a point of view, based upon a problem statement that you had, which we train the model for.”
“There’s a somewhat softer element to this. I did my first AI analytics project in 1996, with neural networks and a kind of ’embedded statistics.’ At the time, it was very difficult to convince people that the outcome of the model had value. People asked, ‘Can you explain how these kinds of values come out of this?’ And of course, you couldn’t, because the nature of the model is that the model continuously rebuilds itself. In today’s world, people accept the outcome of a model that by itself is difficult to understand. People are willing to accept that the outcome is indeed valuable beyond our comprehension of how data and patterns are being developed. I think that is an important element of the acceleration in this domain.”
SKS: “As we stand on the brink of a new digital era in manufacturing, the imperative is clear: embrace the convergence of OT and IT or risk falling behind. By fostering collaboration between engineering and IT divisions, investing in predictive maintenance solutions, and harnessing the power of AI, manufacturers can position themselves for success in an increasingly digital world.”
Smart factories with APEx
Cognizant Asset Performance Excellence (APEx) is a predictive maintenance smart manufacturing solution, built on AWS using multiple technologies like advanced computer vision, AI, and IoT. It can connect to SCADA MES systems and plant-level instrumentation to bring in the right level of decision support alerts. More…