IIoT and Industry 4.0 continue to dominate the industry media with continued emphasis on how these technology initiatives will revolutionize industry and increase productivity and profitability. The narrative has since evolved, from a sensor to cloud model, to one that includes edge computing as an essential component for real-time analytics and as an intermediate data collection/storage point.
As a firm believer in this evolution, it is always good when you have the chance to hear first-hand what is really going on inside industrial companies. The Internet of Manufacturing conference (IoM) in Chicago proved to be a good venue to engage with a combination of IIoT-related vendors and end users. As a focused event, it attracted people with a genuine interest in understanding how they might move forward.
We can divide the end users into two distinct groups; those presenting who are already moving down the IIoT road, and end users who are in various stages of evaluation and implementation. From a presenter’s perspective, the range of projects covered most applications that are discussed in the media: predictive maintenance, digital twinning, machine learning applications and enterprise-wide integration – leveraging data from disparate parts of an organization to optimize operations. One common thread here is that successful implementations hinge on the convergence of OT and IT, and without these two organizations working together successfully it is going to be very hard to successfully implement IIoT projects. That’s because data is sourced in the plant (and for some applications elsewhere in the enterprise) and sent to repositories inside and outside of the OT environment, where various forms of analytics are performed to achieve the desired objective. Most of the applications showcased at the event focused on some form of increased asset performance, often around predictive maintenance, or targeted product quality/consistency improvements.
New applications and transmission of data outside the plant generally raise common concerns for automation engineers; disruption of existing control processes and cyber-security. The ability to seamlessly add new applications, using technologies such as virtualization and the implementation of layers of cyber-security are generally the realm of the IT professional. However, as edge computing makes its way deeper into the plant environment, and further from IT and even from OT, it is increasingly important that computing platforms inherently support virtualization technologies, as well and being self-protecting and easy to maintain. The one thing we can say with certainty is that any form of analytics performs badly when data is lost!
The other interesting aspect that came through in several presentations is that while IIoT is often closely associated with machine learning and artificial intelligence technologies, companies are finding that a lot can be done with existing data without the need to immediately invest in those areas. Depending on the age of existing control systems, assets may need to be augmented with additional sensors and careful consideration needs to take to ensure that the (additional) data destined for analytics can be extracted and stored locally. I had one conversation at the event with a company that was struggling with this exact issue as their modern PLC was out of processing power to extract and process data along with running the control loops. With some infrastructure modifications, an edge computing system is ideal for this company’s requirements. Back to real-life examples, one presenter pointed out that in some situations, benefits can be gained from simply efficiently interconnecting processes of systems by replacing manual data collection and recording (handwritten notes and spreadsheet), with electronic systems that eliminate errors and provide almost instantaneous feedback with relatively simple graphical correlation tools. This is one area where reliable edge computing can play an important role.
What were attendees interested in and what were other exhibitors saying?
Most of the attendees I spoke with were in some combination of evaluation or planning stage. In some cases, they had a clear idea of what they wanted to achieve, in others they were trying to determine where to start. A few had implemented pilot projects and were grappling with full-scale deployments. The key concerns voiced by several people I spoke with were the latency of a plant floor to cloud implementation and the costs associated with cloud-based implementations at scale. While many people conceptually grasped the benefits that using the cloud could bring to an IIoT project, real-time feedback for reactive process control, quality adjustments and imminent machine failure prevention were requirements where the latency of plant to cloud, and the associated processing, would make this type of implementation ineffective. Others were concerned about cost escalation. While LTE mobile connection and cloud computing costs have fallen substantially, they are still often usage-based and they both fluctuate and escalate from month to month as an implementation scales, and more data is accumulated over time. Certainly, the cloud can be useful for pilot projects and perhaps the right answer for enterprise wide deployments where data from the plant, sales, supply chain, logistics and other areas can be combined. For purer plant level applications, such as predictive maintenance the clould is not always the right answer.
From attendees’ point of view, edge computing is a new concept that is only now making its way into the IIoT discussion. It provides significant additional options to place compute capabilities for local control, analytics and storage/pre-processing and data filtering for cloud-based IIoT. The Stratus ztC Edge led to several interesting discussions not only about IIoT-like applications but about the requirements for simplified installation, resiliency due to remote locations or locations where IT and even OT are not present, remote management, automated fault diagnosis and serviceability by non-skilled personnel.
To me, one of the more surprising aspects of the Internet of Manufacturing event was the diversity of vendors who were exhibiting. I had expected the show floor to be dominated by analytics vendors offering some form of machine learning and/or artificial intelligence, they only made up about 1/3 or the exhibitors. The remainder were showing a variety of technologies and capabilities including augmented reality, systems level project capabilities, software toolsets, mobile connectivity, CRM(!) and, of course, Stratus with our ztC Edge compute platform. It was interesting talking to the analytics vendors, trying to understand their focus, the applications for their technology, deployment models and customer engagement strategies. Some were best described as “general purpose” with technology that could be applied to many “IoT” applications, others were focused exclusively on the industrial space with solutions such as predictive maintenance and process optimization. Exclusively, they were cloud-based implementations, which are well-suited to the large data storage and compute-intensive machine learning technologies. All these vendors claimed that implementations were easy with the tools they had developed, but it was also clear that commercial-grade implementations required consulting with data scientists, to understand the problems, desired outcomes and to develop the necessary specific machine learning/AI algorithms. This is significant potential cost. For some of these IIoT vendors, edge-based solutions were beginning to become a point in their discussions, and the majority were interested in understanding what Stratus could offer in this area.
Large Fortune 500 companies with deep pockets are investing in IIoT technologies. These projects are yielding results, although implementations are not without their challenges. An increasing number of industrial companies are looking at ways to leverage analytics, outside of the traditional applications that helped them on the production side. Some have a clear idea of what they want to achieve, some are trying to understand what various technologies could do for them. What does seem clear is that some form of IT/OT convergence and collaboration is necessary to achieve success.
While there is a lot of focus on cloud and the use of machine learning/AI, there is an increasing realization that edge computing has an important role to play. This could be to support a single production line, multiple lines, or an entire plant; remote locations such as natural gas compression stations, oil rigs/well pads and water/wastewater facilities and that all IIoT applications do not necessarily mean the exclusive use of machine learning/AI. When it comes to the edge, these remote systems are almost always out of the reach of IT and OT, so simplicity, self-protection, remote management, easy maintenance and self-healing are critical considerations.
Want a closer look at edge computing? Download our white paper and discover the evolution that is taking place in edge computing today, while demonstrating how next-generation solutions can help organizations execute on the new opportunities presented by the transformation of the edge, today and tomorrow.