Much has been written, and continues to be written, about the virtues of digital transformation, information technology / operational technology (IT/OT) convergence and IIoT. Foundational to why a manufacturer would make investments in these is the acquisition of plant floor data for purposes of converting the data into actionable information. The type of information and the level of analytics largely depends on the maturity level of the factory floor convergence of the automation systems.
The place to start is assessing the foundation of a plant’s convergence model which can be classified into four categories.
- Restrictive – Many OT networks exist with no convergence, no interconnectivity. Harvesting and aggregating data becomes difficult at best.
- Functional – OT infrastructure converges into a wholistic network. Data is restricted to sharing among process cells and work cells on the plant floor.
- Effective – OT and IT network infrastructures converge giving the opportunity to interweave plant floor data with business reporting.
- Innovative – OT, IT, and business applications create a connected factory emergence capable of effective data analytics driving business decisions.
Note the first two classifications above restricts each line of business to focusing on plant floor silos. The third and fourth classifications offer greater opportunities for businesses to focus on mission, vision, and business outcomes.The need for data analytics containing descriptive, diagnostic, predictive, and prescriptive information exists at several levels of a manufacturing company.1
- Analytics germane to operations at the plant floor device level are, “Is a device running OK?” [descriptive], “Why did a device fault occur?” [diagnostic], “I predict a fault will happen soon” [predictive], and “What actions should be taken to avoid the fault in the future?” [prescriptive].
- “System” level analytics are typically relevant to plant floor supervisors and might be “Is Line 1 running OK? [descriptive], “Why is Line 1 quality poor?” [diagnostic], “I predict that Line 1 is moving out of tolerance.” [predictive], and “What actions should the operator take to avoid poor quality?” [prescriptive].
- “Enterprise” level analytics relative to management can be “What facility is performing the best?” [descriptive], “Why is Plant A throughput behind plan?” [diagnostic], “I predict that Plant A will be behind plan soon.” [predictive], and “What action should I take to avoid Plant A falling behind plan?” [prescriptive].
Getting the most value out of data has evolved over time. Traditional plant floor data has relied on live data coming from PLCs, DCS’s, HMI’s, or time series data from historians or event-based data from relational databases, all mostly setup by control system engineers with reporting also created mostly by control system engineers.
Today, data comes from disparate data sources, control systems, data warehouses, and data aggregation from multiple platforms requiring the talents of data integration engineers, data architects and business intelligence (BI) engineers. Future data analytics will most likely consist of the same disparate data sources on the plant floor feeding data orchestration via edge computing; event ingestion, event processing and enrichment, streaming analytics, data hosting and staging; feeding cloud analytics and on-premises self-service analytics, all developed by controls engineer hybrids, IoT specialists, and software developers.
Applying analytic solutions might be best approached using a “crawl, walk, run” strategy.
- Crawl – Trending time series data has been, and continues to be, the most fundamental analytical tool of any control system; use it wisely, don’t over-historize your process. Time-series trending can be extremely insightful when correlated with event context. Overlay alarms and events with trends. This, many times, is native integration within a control system and is 2nd nature to many OT specialists and integrators. Leverage current asset modeling and event platforms to create equipment hierarchy, attributes and events versus just “adding tags” to a system database. Utilize widely-available reporting tools, such as SQL Server Reporting Services (SSRS) to glean meaning insights.
- Walk – Contemporary platforms exist that support OT / IT collaboration by empowering controls engineers to identify the most appropriate data sets and send information directly to the IT layer. Controls engineers can connect, contextualize, and map data in a way to create information models without requiring traditional IT skills.
- Run – Here’s where data analytics addresses all three levels – device, system, and enterprise. The system integrator 1st concentrates on clearly identifying, in each segment, who are the key stakeholders and why is the information they need valuable to achieving their goals. After that the system integrator can best utilize the talents and skills of their control engineer hybrids, IoT specialists, and software developers to collaborate to configure, program, test and deploy necessary software modules and platforms for data creation, data flow, data view, device analytics, edge analytics, augmented modeling, and process/work cell optimization needed to deliver solutions bottom-to-top for a manufacturing organization.
System Integrators knowledgeable in manufacturing and advanced technologies can be an asset to manufacturers navigating the many options to road mapping their use of data analytics. The role of the SI is to understand the client’s business objectives so the SI can provide relevant guidance to the client on the “here and now,” “what’s next,” and “what’s after next.”
Daniel C. Malyszko, Director of Operations and IIoT Consultant at Malisko Engineering, a certified member of the Control System Integrators Association (CSIA). See Malisko Engineering’s profile on the CSIA Industrial Automation Exchange.
1 Footnote: Information contained in this paragraph courtesy of Rockwell Automation.