Q&A

Turn Data Into Doing

Making information actionable requires filtering out the noise. Often, there is too much unstructured data to sift through. There has been an overemphasis on the data itself. Dan Malyszko, Vice President at Malisko discusses what to do with that data and how to do it.

Contemporary control systems contain a wealth of data. How is that data typically captured? What/where are the data sources?

Dan: When it comes to the plant floor, data typically is captured in time-series historians or relational databases for event-based data. Time-series trending of sensor process variables is the most fundamental form of data analysis and typically relies on the process knowledge of the user to glean meaningful insights. Time-series data coupled with event data can be very powerful, but often these data sets are siloed, and still rely on the user to have knowledge of how and where the data is captured.

How does Malisko approach converting captured data into actionable information?

Dan: To make information actionable, we must first focus on filtering out the noise. Often, there is too much unstructured data to sift through. We solve this by focusing on data organization and asset modeling to develop operational twins, which are optimized for analysis. Templatizing the data structure of equipment or processes makes for a much more streamlined integration path for analytics and alerting.

What types of “actionable information” can come from the captured data?

Dan: There are various personas within an operational setting who consume data. For example, an operator acts on near real-time process data in making decisions while an operations manager focuses on bigger picture summary data across several lines or even across several plants. While much of the data in both cases is related, the actions could be more immediate and prescribed versus an action taken to affect a longer-term business outcome. It is important to understand data consumer personas to deliver the right kind of data to the right person at the right time.

What are the end results of the actionable information and how/where is it (or can it be) used?

Dan: For an operator, one of the most fundamental sources of actionable information is a human-machine interface (HMI) alarming system they can trust. Many legacy HMIs we come across lack basic alarm hygiene where alarm banners are filled with unacknowledged nuisance alarms. In these cases, operators often disregard the alarm banner altogether. To expect any kind of meaningful action, we must be disciplined on the front end of a project by focusing on failure modes and effects analysis. A better understanding of what could cause a process or piece of equipment to fail helps us develop more intuitive alarms, but also primes the discussion for predictive analytics.

We are seeing more adoption of executive-level dashboards where production data merges with business system data. This is where the Industrial Internet of Things (IIoT) comes into play as we must gather, cleanse and visualize data from disparate sources within the enterprise. A good example is correlating overall equipment effectiveness (OEE) to financial performance for a given line or facility. Bringing together these types of data sources takes a lot of careful planning regarding how to summarize data at a line level, rolling up to the plant level and ultimately to the enterprise level. An effective key performance indicator (KPI) rollup data structure enables better decision making at the top levels within the organization.