Edge Analytics and the Evolution of the Control System
Thought leaders have stated that Industry 4.0 is about the effective use of digital information to improve all facets of manufacturing, including product quality, keeping assets in top condition, safety and health-related issues, environmental stewardship, optimizing production time, cost, and carbon impact of the product made. To achieve these goals, much of the improvement will be made from impactful use of existing data and complementing the existing data with new information needed for emerging analytics.
As analytics have proliferated and helped companies become more proficient and more agile, the big data analytics are usually performed at a corporate IT level. Some of that, especially when mixed with business data, must remain at that level. However, ОТ analytics is best when moved closer to the edge, where the manufacturing and equipment data is collected. This strategy is attractive for the following reasons:
- 1. The data does not have to go through an extraction and cleansing process, thereby reducing time to insight.
- 2. Fewer people in the company (IT, engineering, production, process engineering) need to be involved. Again, this reduces time and labor costs.
- 3. The strategy drastically reduces the time needed to make adjustments to the manufacturing process.
- 4. The edge analytic results are more easily integrated with EIDI systems, so that the quality of data is enriched by these results.
Peter told Bill that by bringing analytics closer to the edge, the roles of the process engineer and other local engineers at the refinery will change. They will be empowered to create these analytics using their EIDI system data to optimize the process governed by the traditional control system. Bill suggested that Peter look into hiring an engineering student or a recent graduate because engineering students at most universities are becoming much more well-versed in areas such as data science; AI; predictive tools, such as machine learning; and newer programming languages like Python and R. A recent graduate would have proficiency in machine learning and software analytics. This new worker could help develop solutions that would give the refineries greater insight on their equipment time-to-failure. The company could better prepare and budget to replace or avoid large or expensive capital equipment projects with smaller, more nimble projects.
Peter also had a vision of moving the analytics developer closer to the edge, where the developer would more quickly learn the refining process and what issues the refinery units face on a day-to-day basis. Bill digested what Peter related from his conference takeaways and remarked that this type of change will dramatically affect how control systems of the future will be designed and built.
Bill had attended Proclndustries internal planning meetings where the company's top technical people were struggling with how to improve the existing refinery control systems and how to prepare and define what's needed for the next-generation control systems. Bill theorized that because of edge analytics, machine learning, and other breakthroughs in data modeling, the next wave of control systems will be quite different. For years, traditional control systems have been strictly ОТ based. That is, the fundamental components are process control logic, interlocks for safety, and data handling to efficiently make products.
When these systems are linked to modeling or planning software systems, they tend to use first-principles engineering models, ОТ equipment models based on vendor performance information and recommended service intervals, or engineering-based process models. They have not traditionally used modeling or predictive analytics that is, based on the data, for control. Nor have they used many of the IT analytics developed in the twenty-first century. They also don't natively adapt to adding IoT sensors in a multi-vendor environment without some data transmission work involved to integrate these sensor signals.
This is expected to change in the 2020s, however, when new control systems will be fundamentally changed to a more modular design that includes both traditional ОТ tools and IT analytics as well as the capability to ingest incrementally added IoT sensors. The ОТ landscape for control systems and supervisory control and data acquisition (SCADA) will change to incorporate much more IT specific functionality, such as data-dependent analytics, and become more flexible to business and consumer demand for dynamic production requirements.
At some point in the not too distant future, control systems will also be designed to function in lights-out manufacturing facilities without workers. As you would expect, this would start with facilities that do not have a high probability of malfunctions that would cause a fire, explosion, any kind of environmental incident, or harm the neighboring community. Certainly a refinery would be one of the last facilities, if ever, to do this. But it will likely follow the trend toward self-driving vehicles, where autonomous driving is implemented, controlled in stages, and is likely to be more safe than with humans driving.