Need continuing education credits? Join Us For Five Days of Education on the Industry's Leading Topics beginning October 5th!Save Your Seat
DCS, SCADA, Controllers

Use SCADA data for greater profits

On-demand subscription-based predictive process analytics turns SCADA data into greater profitability

By Edwin van Dijk August 25, 2020
Courtesy: Trendminer

Supervisory control and data acquisition (SCADA) systems excel when it comes to collecting and storing data and monitoring automated machine performance. Traditionally they were supported by condition-based maintenance tools and historian technologies. That is, until modern, more economical predictive process analytics and the use of IIoT brought improved efficiencies.

Predictive process analytics, when applied to SCADA data via IIoT, can deliver an untapped gold mine of efficiencies for a plant’s asset performance. For instance, SCADA data can mirror asset health and when predictive algorithms are applied, it can optimize asset health, and produce substantial cost savings.

Due to rise of IIoT, lately there have been various developments in SCADA systems. These include the data acquisition part where data from dispersed SCADA systems are unified in a single system, accessible via web and cloud. Other developments are more related to the supervisory control part of SCADA, using web services and IIoT protocols to collect exposed data and control processes in real time. There has also been progress seen with process controllers enabled to apply complex business logic and predictive maintenance algorithms to operational data and assets.

Just last year McKinsey and Company reported 3 to 5% loss in the industry overall for equipment effectiveness due to unplanned maintenance. For one chemical processing plant, McKinsey and Co. noted that through the use of analytics, the plant could cut at least half the time it took to repair pumps, which amounted to about $120,000 in costs avoided per pump failure.

Predictive maintenance requires the processing of enormous amounts of data and the running of intelligent algorithms, which is costly to apply within local SCADA implementations. It also requires scarce data scientists to be involved to apply those complex algorithms and data models. On the other hand, IIoT platforms can store terabytes of data and can provide global access to data across production plants. These platforms are also capable of combining brownfield sensor data with greenfield sensors and device data, opening new use cases for operational improvement.

The impact of COVID-19

In striving for more stable operating conditions, many process manufacturing companies today have come to understand the value of data. Unlocking the data from the SCADA system is a hurdle difficult to overcome. Due to the COVID-19 pandemic, investments made in IIoT have decreased and market recovery may take some time.

Unfortunately, many manufacturers already face supply chain issues that are further slowing IIoT technology adoption. The National Association of Manufacturers (NAM) reports 80% of manufacturers anticipate their businesses will be negatively impacted by the pandemic. Therefore, in part, manufacturers are looking for small gains by enhancing their existing systems to take advantage of IIoT opportunities.

Figure 1: Computer Integrated Manufacturing Pyramid. Courtesy: Trendminer

Figure 1: Computer Integrated Manufacturing Pyramid. Courtesy: Trendminer

With changed market circumstances and necessary investment reductions, it becomes harder to face the competitive market circumstances. Still, leveraging the data has proven an agile and profitable way to quickly adapt a business to changing circumstance. At each level of the computer integrated manufacturing (CIM) pyramid (see Figure 1), complex algorithms and data models can be applied to get more value out of the data. But this has proven to be inert, time consuming, cost intensive, and requiring scarce resources.

What is needed are new tools and technologies such as self-service industrial analytics, where operational experts (such as process and asset experts), control room personnel, and maintenance engineers can analyze the production data themselves. Self-service industrial analytics tools are based on pattern recognition technologies, combined with machine learning techniques, and leverage data science techniques and algorithms in the background.

This brings the power of data science into the hands of the process experts, without them needing to become data scientists themselves. With self-service industrial analytics, operational experts can search the process data to see what has happened, how often it may have happened, and what potential root causes may have led to it. Modern self-service industrial analytics tools also provide machine learning techniques to recommend to the user potential root causes to explore. These next-generation solutions are developed to take advantage of IIoT opportunities with ease of use, affordability and scalability in mind.

Case in point

Chemicals manufacturer Arlanxeo, Maastricht, Netherlands understood the value of leveraging time-series data. To start, the company worked with different types of analytics models and identified their limitations for scaling up beyond pilot projects. Over time, they developed their deep knowledge of process operations to create “pattern search-based discovery and predictive-style process analytics” for the average user. The unique, multi-dimensional search capabilities of their platform enable users to find precise information quickly and easily, without expensive modeling projects or data scientists.

Using a song recognition app like Shazam or SoundHound as an analogy, the application uses pattern recognition rather than mapping every note in the song to its song database. The pattern recognition software seeks “high energy content,” or the most unique features of a song, then matches that to similar patterns in its database to recognize it. Such pattern recognition works with a high rate of accuracy and speed.

Of course, analytics for manufacturing plants requires more sophisticated algorithms that able to go beyond mere search capabilities. Self-service analytics work by connecting to existing time-series databases and indexes the data rather than copying it in another system of record. The indexed data makes it easy to find, filter, overlay and compare interesting time periods, to search through batches or continuous processes to pinpoint areas for improvement.

Another capability is search for particular operating regimes, process drifts, operator actions, process instabilities or oscillations. By combining these advanced search patterns valuable information that Arlanxeo needs is unlocked. You don’t have to be a data scientist to use it. For example, an operator can compare multiple data layers or time periods to identify what sensors are deviating from the baseline and adjustments can be made to improve production efficiency.

Process data contextualization

Process data contextualization and predictive analytics capabilities add other dimensions and optimize process control systems making them multi-dimensional for better process quality control. By allowing operations personnel to provide annotation, greater insight is gained.

Figure 2: Combining live data with historical context shortens the analysis latency, allowing real-time monitoring and providing an opportunity to take actions even before an event can affect process performance. Courtesy: Trendminer

Figure 2: Combining live data with historical context shortens the analysis latency, allowing real-time monitoring and providing an opportunity to take actions even before an event can affect process performance. Courtesy: Trendminer

Predictive analytics enable early warning detections of any anomalies or undesirable process events by comparing saved historical patterns with live process data. Moreover, the solution calculates all possible trajectories of the process and process variables can be predicted before anything happens. Recent process changes can then be matched against expected process behavior and settings pro-actively adjusted accordingly.

Online subscriptions

In conclusion, now that companies have the option to enhance the value of the investment they have made in high quality SCADA systems, low cost predictive analytics solutions that complement their existing time-series databases unlock more value from the data collected and are able to provide valuable business insights faster and more effectively.

Forward thinking manufacturers are gaining an incredible competitive advantage using this digital business model and leveraging cost-effective plug-and-play analytics that were developed from the ground up to integrate with digital technologies.


Edwin van Dijk
Author Bio: Edwin van Dijk is the vice president of marketing at TrendMiner.