While IIoT has been around for more than a few years, what’s new is it’s integrating AI, just like everyone else. The difference is that adopting AI on mission-critical plant floors in the process industries poses different concerns and more potential risks than simply adding it to typical, mainstream, IT-based systems.

“There’s lots of talk about how IIoT in different industries is implementing AI, but process applications and facilities must consider safety. They can’t just give full license and autonomy to AI agents,“ says Sunil Doddi, member of the International Society of Automation’s Technical Assembly and program chair of the ISA Philadelphia section. “IIoT networks also impact how plants are designed and operated, so I use ‘industrial data suite’ (IDS) as an umbrella term because it also affects how data is collected, connected and consumed.”

Doddi reports it’s crucial to resolve these issues because designing and deploying suitable IIoT solutions can be hugely beneficial. “IIoT-enabled sensors deployed in heat exchangers and other devices can collect continuous steam-production data, guide users to suitable tools for their process, and transition from scheduled maintenance to IIoT-based predictive maintenance that’s more efficient,” he explains. “IIoT can also get an assist from cloud computing, which can provide improvement efforts like predictive maintenance with an infrastructure they didn’t have before, and give them a better chance of getting approved, established and succeeding.”

Get your subscription to Control's tri-weekly newsletter.

Drawing on his industrial experience, Doddi adds that digital transformation teams are increasingly equipping coworkers with mobile HMIs, combined with virtualized simulation and testing environments. In real‑world examples from across the process industries, for instance, virtual machines have been hosted using platforms such as VMware's Workstation Pro, and alarm notifications from distributed control systems (DCSs), such as Siemens Simatic PCS 7, have been routed through industrial callout systems such as Cattron's Aquavax to staff smartphones when process alarms occurred. This kind of infrastructure can let operators view and check their processes remotely and, in some cases, make carefully controlled adjustments at a distance. Virtualization and testing also leverage digitalized data presented by historians, such as AspenTech's InfoPlus.21 (IP.21), and manufacturing integration platforms, which display operations information on mobile devices.

“When the DCS sees an alarm, it sends a contact to the callout system, which is integrated with the HMIs and smart phones, so it can text operators and other authorized users,” adds Doddi. “In the future, AI will likely become part of this infrastructure. However, you can’t just add large language models (LLM) or other AI functions to process plants. It’s not just that they might send the wrong text. They could potentially damage processes and products, and possibly cause safety issues.”