Once the initial difficulties of adopting the Industrial Internet of Things (IIoT) wears off, and implementing it becomes more familiar, users are incorporating its most useful strategies into subsequent designs and projects. This can all happen sooner as the design cycle tightens, but it requires finding the right combination of networking, digitalization, cloud computing, virtualization, and/or artificial intelligence (AI).

“We see process application designs shifting from isolated control systems toward connected architectures that combine plant-floor control with historians, analytics, virtualization and cloud-enabled applications,” says Heath Stephens, PE, automation solutions director at Hargrove Controls & Automation in Mobile, Ala., a division of Hargrove Engineers & Constructors, and a certified member of the Control System Integrators Association (CSIA). “These technologies are improving visibility, enabling predictive reliability and digital twins, and allowing operations teams to make faster, better decisions using data from platforms such as PI, IP.21, MES and cloud analytics environments.

Stephens adds the primary benefits of linking IIoT architectures include improved operational visibility, better asset reliability, faster troubleshooting, and stronger decision support across engineering, operations and maintenance. Hargrove’s clients have used these technologies to support paper-to-glass workflows, predictive maintenance, quality improvement, and broader optimization of production, maintenance, and business processes.

Work the common network

Just as more useful IIoT strategies emerge and shake out over time, Stephens reports that de facto networking standards such as message queuing telemetry transport (MQTT) are also simplifying  IIoT environments.

“Protocols such as MQTT are helping simplify the IIoT landscape by making it easier to move data efficiently from edge devices and field collectors to centralized platforms. In our experience, standard communication methods reduce custom integration effort, improve scalability, and make it more practical to connect remote assets, historians, dashboards and cloud-based analytics in a secure, supportable way,” explains Stephens. “It’s also important to remember that more traditional communication protocols, such as Ethernet I/P and Profinet, are still being used and still appropriate for devices with larger data sets.”

To determine what combination of IIoT and digitalization is best for their processes, Stephens reports that users should start by identifying specific business problems they’re trying to solve—such as reliability, quality, operator effectiveness, maintenance planning or enterprise visibility—rather than starting with a particular technology.

“Hargrove typically recommends first evaluating existing data sources, system readiness, cybersecurity requirements, and expected return on investment (ROI), and then selecting a phased digitalization approach that fits the organization’s operational maturity and internal support capabilities,” adds Stephens.  “IIoT hardware and software technologies makes it cost-effective to collect information from small data sources.”

Counting on the cloud

Focusing on business problems or goals can also guide users to determining how much computing to do on the edge or the cloud, as well as indicate how much and what kind of data preparation and cleaning is needed before sending data to the cloud, so subsequent analyses will be optimal once it gets there.

“A good rule is to keep time-critical control, safety and high-availability functions at the edge (meaning onsite), while using the cloud for heavier analytics, enterprise reporting, model training and multi-site comparisons,” explains Stephens. “Before sending data to the cloud, it should be contextualized, cleaned and validated, so bad tags, inconsistent naming, missing timestamps and/or poor-quality data don’t undermine the value of downstream analytics. Most of this data preparation and cleaning is rule-based, and AI can perform additional cleaning, but cleaning data is incredibly important for good modeling and analysis.”

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

Stephens adds that Amazon Web Services (AWS), Microsoft Azure and similar cloud-computing services platforms are becoming more common in plant environments, especially where clients want to extend historian data, enable AI or machine learning, support remote monitoring, or scale solutions across multiple sites. Sometimes clients use native AWS and Azure functions, but often those services are underpinning industrial cloud software.

“Hargrove has supported plant applications that leverage Azure along with PI and IP.21 data, and we expect hybrid edge-plus-cloud architectures to continue growing because they balance plant-floor reliability with enterprise-level analytics and flexibility,” says Stephens. “The main drawbacks and risks include cybersecurity exposure, overcomplicated architectures, vendor lock-in, and deploying technology without a clear internal ownership/support model. Clients should also be aware that many of these technologies operate on a subscription basis, so there will be continuing costs that must be budgeted.”

Support, monitor, backup and virtualize

To optimize virtualized, cloud-based, IIoT-enabled and/or mobile environments, Stephens adds that users typically just need the support tools provided by their system vendors. However, he also recommends that every client have a good network monitoring platform, and a good backup and recovery platform. Hargrove often provides support for these solutions after the initial project is complete because not all clients have the internal capacity to support complex systems.

For instance, Hargrove recently succeeding in installing a Proxmox virtual environment for an Ignition web-based HMI software client. Proxmox open-source server platform for managing virtual machines (VM) is  appealing to many clients because it’s open-source and more cost effective than alternative, virtual hosting platforms.

“We’ve also supported digitalization initiatives that included Azure-based applications, process data mining, digital operator rounds, augmented reality tagging, and using PI and IP.21 data in broader analytics workflows,” adds Stephens. “A few key lessons learned are to begin with a high-value use case, make sure the source data is trustworthy, involve operations and maintenance early, and build cybersecurity into the architecture from the start. Hargrove also recommends phased deployment and choosing solutions that plant personnel can realistically support and sustain after implementation.”

Stephens concludes that he doesn’t presently see an endpoint to virtualization, cloud, IIoT and mobility’s evolution.

“The long-term evolution will likely be more integrated, secure and autonomous operations built on hybrid architectures that connect edge control, enterprise systems, mobile tools and cloud analytics into a common decision framework,” adds Stephens. “It will certainly involve more AI with the likely addition of AI-powered humanoid robots, which are just starting to roll out in industrial settings. In Hargrove’s view, the end state is not simply more data in the cloud, but smarter and more contextualized systems that help people run plants more safely, reliably and profitably with better real-time insight and decision support.”