Manufacturing Process Data Analytics

Leverage IIoT Analytics to Drive Operational Performance

Companies throughout process industries are big users of traditional modeling, simulation, and data analysis technologies ― and increasingly of advanced Industrial Internet of Things (IIoT) analytics that make use of manufacturing data directly from plant operations and machinery. However, the way they use this analytics technology differs from other industrial sectors in some fundamental ways. At LNS Research, we’ve measured the use of analytics in process industries over several years, and earlier results showed that process companies were focused almost exclusively on a single-use case: predictive maintenance for critical risk assets.

However, our most recent studies show a growing trend toward a wider range of use cases, extending far beyond uptime and asset performance to process improvements, quality improvements, and better forecasting, to name just a few. This push beyond predictive maintenance has led companies in process industries to provide access to IIoT analytics directly to plant managers (62%) and plant level personnel (41%) at rates between 50% and 100% higher than other industries. This aligns with our findings on companies that are further along in the Industrial Transformation journey. That is, more mature companies empower plant personnel with analytics that deliver real-time insights, and with authority to make decisions based on the output of those analytics ― not surprisingly, these companies have achieved more significant improvements over time because of this empowerment.

Along with these findings, our research has also shown that a specific operational initiative can greatly increase the positive impact of IIoT analytics programs across a range of these emerging use cases: a rigorous continuous improvement program supported by digital tools. The discipline imposed by continuous improvement aligns well with the new use cases for IIoT analytics within process industries.

Defining IIoT Analytics for Process Industries

The term “analytics” is widely used ― and misused ― by most industries. For this report we use the specific term “IIoT analytics.” By this we mean advanced analytical software and related applications to glean insights from data created within plants, both in real-time (using streaming data for predictive maintenance in a chemical processing plant, for example) and from historical data (looking for correlations between ingredient temperature and quality as in food processing). Because of the unique formats and characteristics of data created within these environments, and because of the unique challenges that factory workers are trying to solve, not all analytics solutions are capable of delivering valuable insights from this wildly different but highly-impactful data. In particular, the unique requirements of time-series data formats, and the speed of data streaming from these systems are often beyond traditional statistical methods and databases.

Several years ago, LNS Research defined four stages of analytics adoption, usage, and maturity:

1. DESCRIPTIVE | Analytics show what has happened (and what is happening at a given moment in time)
2. DIAGNOSTIC | Analytics to show root causes and enable users to associate cause and effect
3. PREDICTIVE | Analytics show what could happen and when it’s likely to occur in the coming days or weeks
4. PRESCRIPTIVE | Analytics extend beyond a prediction to show what action to take to prevent an unwanted outcome

Key IIoT Capabilities for Modern Platforms

Most of the top use cases reported in the process industries are at the predictive level or beyond, and are delivering compelling value and ROI. LNS Research defines IIoT analytics as having distinct features and capabilities: Capable of analyzing plant-centric or manufacturing data in a real-time streaming fashion, and analyze historical structured, semi-structured, and unstructured data repositories

• Can run as a standalone application or on an IIoT platform; may also be part of a layered analytics model that includes some combination of:
  • Business analytics, which may include applications for financial, sales and marketing, customer support, and other disciplines
  • Layered analytics including role-specific applications that are not dependent on data science teams, and instead can be implemented and utilized directly by end-users, e.g., process engineers or quality professionals
  • Edge analytics focused on real-time asset monitoring as close to the asset as possible
  • Industrial analytics, providing real-time asset monitoring in a data center or the Cloud

• Built to run in Big Data environments with a common industrial data model
• Has IIoT connectivity capabilities, e.g. OPC-UA, MQTT, etc.
• Can incorporate non-industrial data in analysis (e.g., financial database, maintenance management systems, Laboratory Information Management System (LIMS), out-of-expectation (OOE) results, batch systems, and unstructured data, among others)
• Has embedded or extensible artificial intelligence/machine learning (AI/ML) capabilities tailored for an industrial environment
• Has embedded or extensible simulation capabilities tailored for the industrial environment
• Built to optimize storage, compute, and user experience across Edge, on-premise, and Cloud
• Has a modern, flexible, and no-code/low-code application development environment

The definition of a layered approach to analytics allows for a wide variety of solutions, ranging from simple machine-focused real-time analytics for predictive maintenance (where process industries started) to fullblown IT/operational technology (OT) data science projects with data science consultants leading the way. More importantly, the layered approach empowers individual departments or teams with deep subject matter expertise to utilize analytics directly, without depending upon a data science team. It’s clear that the democratization of analytics is speeding the rate of industrial transformation.

Trending Now: Self-Service and User-Friendly Analytics

There appears to be a trend towards industry- or use-specific solutions that are more lightweight from an IT or user perspective, enabling even resource-constrained IT or OT organizations to deploy rapidly and achieve substantial ROI quickly. This shift away from dependence on a centralized data science team may also be a result of the challenges companies are encountering with data; the distributed nature and disparate formats of plant data are challenging for IT teams not accustom to working with the unique data formats, volumes, and speeds inherent with manufacturing operations and processes. Incidentally, data differences is certainly one of the driving factors for blending the IT and OT organizations. To shortcut data challenges, companies in process industries have deployed IIoT analytics applications designed for process engineers in specific industries. Solutions that are industry-specific or use-case specific are faster to set up, i.e., self-service, and they can handle industry-specific data types and formats “out of the box.” These advantages enable companies to avoid the delays and costs of enterprise data science projects, achieve ROI more rapidly, and accelerate time-to-value and time-to-impact of these projects.

Successful Stacking: Operational Excellence, Analytics

In process industries, a small subset of companies powers the operational excellence initiative with advanced analytics. In a recent LNS Research survey, only 18% of process companies reported using advanced analytics for operational excellence. Recall that at the plant level, operational excellence is focused on two key levers: take full advantage of the plant’s production capacity potential and maximize throughput and yield ― safely and responsibly, with minimum energy use, and in the greenest manner possible. These two levers are the major economic drivers of plant performance. In the first case, the focus of digital tools so far has been on condition-based monitoring leading to predictive maintenance on devices, e.g., pumps, compressors, and other equipment. However, the missing link has been tools designed to analyze the production process in conjunction with the performance of devices and equipment. To date, spreadsheets have fulfilled this role, aside from fullblown process simulators, which plant engineering personnel often do not have time to use or maintain for analysis. Further, it’s difficult to make the leap to digital twins of assets and processes without first understanding how analytics tools can deliver insights to maintenance and operations. We can learn two relevant lessons from continuous improvement programs, which are a subset of operational excellence. First, Six Sigma and Lean tools are not widely adopted because they are designed to be used by only a few specially trained personnel, i.e., black belts. Thus, when there is a process problem or opportunity, the plant brings in the Six Sigma / Lean subject matter experts. Second, there are many continuous improvement tools for business process optimization while those for production process optimization are rather few in comparison. In analytics, this is analogous to requiring a data scientist for every analytics initiative plus having to teach that individual enough about processes and physics to make sense of the data being analyzed.

Fortunately, industrial software companies have solved this conundrum by creating IIoT analytics solutions specifically designed to meet the unique needs of industrial engineers. When we talk about the democratization of data and analytics, we mean that the process industry needs easy-to-use tools that can be used by many so they can contribute to the continuous improvement program and goals.

Some Companies Falling Behind ― But Why?

Despite the business advantages for industrial organizations using IIoT analytics, more than one-third of companies in process industries have not yet adopted or implemented IIoT analytics. Reasons vary widely, but the most commonly stated argument is that existing internal systems already achieve the same purpose (35%). Certainly we see some business applications that include rudimentary analytics capabilities; however, these are limited in scope. For process performance analysis, companies need to make use of data across silos to capture insights about operational performance. They can do that with user- or role-specific dashboards with data from trend and contextual analysis. We have also seen companies defer analytics projects because they believe they lack internal expertise or, of course, budget. Companies cite many other reasons as well ― they say digital tools are too difficult to use, engineers don’t have time to learn programming, they can’t find or attract enough data scientists to support initiatives. What these firms may not know is that there’s a strong correlation between companies that have fallen behind in deploying other industrial technologies like ERP or MES/MOM, and those that report no plans for IIoT analytics or related technologies.

LNS Research has shown that a company can overcome this technology gap by pursuing a digital operational excellence program, then evolve from metrics to analytics. This strategy has proven successful for many across a wide variety of industries. Additionally, some firms are using IIoT analytics to leapfrog competitors, in an attempt to quickly move from technological and operational laggard to leader. Outcomes of this approach remain to be seen. The process industry should note that companies that delay analytics implementations due to budget or resource constraints face a real risk of falling behind competitors in these key technologies ― and the advantages they deliver. Moreover, the business advantages they offer do compound over time, meaning that late adopters will continue to fall further behind.

Recommendations for Process Industries

Yes, companies in process industries are in some ways ahead of counterparts in other manufacturing areas. While predictive maintenance for critical assets has proven fruitful, there is much to be gained by considering a wider range of use cases and addressing whatever challenges (real or perceived) prevent the organization from adopting IIoT analytics. Guide your organization in the steps needed to adopt IIoT analytics for business advantages.

EXAMINE ALL POSSIBLE USE CASES FOR IIOT ANALYTICS, and prioritize according to value and effort required; what are the specific business benefits your company wants? Can you make projections on savings, efficiencies, improvements, or cost reduction? There is no shortage of use cases such as “improve reliability,” “reduce energy consumption,” “increase product yield,” or “improve product quality.” Defining the desired outcomes is an important step but many companies incorrectly relegate it to a secondary priority. LNS Research recommends this as the first step for every company looking at the best uses for IIoT analytics. DETERMINE THE DATA NEEDED TO SUPPORT EACH USE CASE. Knowing which data and what analysis is needed (and where) will determine the overall architecture and which tools to use, from Edge to on-premise to Cloud. Historically this task has soley been the responsibility of IT and data engineers. However, with today’s advanced IIoT analytics solutions, process engineers with in-depth knowledge of the processes and operations must play a role in choosing which tools they need to do their job.

LOOK FOR IIOT ANALYTICS TOOLS WITH BROAD APPLICATION that can be used by business analysts and process engineers, and don’t need data scientists at every step. There are several self-service analytics tools designed for engineers and other operations professionals, just as there are BI tools for business analysts.

MAKE USE OF NEW ANALYTICS APPLICATIONS that serve operational excellence initiatives. They can do so much more, faster, more easily, and with more and better insights than manual or out-of-date tools. To maximize and sustain long-term value, companies in process industries need to build a foundation of operational excellence processes and programs supported by advanced analytics; they also must suitably empower workers and professionals at all levels with the right tools. Democratization of new analytics tools enables scale across the organization and drives substantial value.