Barriers to digital transformation can include fragmented communication, inaccurate data collection, and inefficient use of assets. These challenges can pose significant risks to people, equipment and operational performance.
In this Emerson Automation Experts podcast, I’m joined by Emerson’s Michael Tworzydlo to discuss the role of operational analytics to ensure operational health and optimize plant-wide performance by detecting abnormal behavior of processes and assets, identify root causes of problems, and predict future performance.
Visit the Plantweb Optics Analytics section on Emerson.com for more on how it helps support your digital transformation initiatives by shortening the decision-making process which helps prevent further performance deviation and safety issues while maximizing plant efficiency.
Jim: Hi everyone. I’m Jim Cahill and welcome to another “Emerson Automation Experts” podcast. Successful digital transformation initiatives require many factors for success. These include clear business objectives and senior executive sponsorship and support, effective measurements that have been removed from their silos and brought together, robust analytics to turn this wealth of measurement and other data from many sources into actionable information, and a step-wise plan and execution approach to gain success and momentum along the way. Today, I’m joined by Michael Tworzydlo to discuss the important role of analytics in these digital transformation efforts. Welcome, Michael.
Michael: Thank you, Jim. Excited to be here.
Jim: It’s great to have you here with us and doing it in person, which is always a great thing. Let’s begin by asking you to share your background and path to your current role with our listeners.
Michael: Okay. Absolutely. Well, I grew up in Southeast Texas, which is really a major hub for refining and petrochemicals, and naturally what I grew up all around. And with that, I went to school for chemical engineering at the University of Texas at Austin. Hook ’em, exactly. Worked in the engineering procurement construction industry for about seven years as a process engineer just designing equipment like heat exchangers and boilers and pumps. Did a lot of project engineering work as well and worked on, for example, like the construction of refineries in the United States.
Then I went and joined Emerson. I’ve been in a couple of different roles here, including a strategic planning role for the platform. But I really found out I was extremely interested in software and more in software was also the analytics piece. I remember we did an Emerson Exchange 2019, which was very much around analytics and that’s where I kind of first got exposed and the opportunity opened up here to be product manager for our Plantweb Optics Analytics software. So I jumped on that chance and here I am.
Jim: Well, that’s great. We’ll get into Plantweb Optics Analytics and a lot more in some of these questions coming up. But to step back for a minute, I know a challenge, many manufacturers and producers face is the occasional abnormal event that disrupts operations. How has this challenge been traditionally addressed?
Michael: Well, traditionally, these events would just happen, and maintenance and operations have to be reactive. So you found out something happened, you’d fix it, and you’d try to find out why it happened. And then after that, you’d work to keep it from happening again. So that was really more of a preventative mode, a lot of what you hear about preventative maintenance, doing rounds every so often to make sure your equipment is in shape, which is a step better than reactive. But now we’re really getting into that predictive, prescriptive space more and more. And that’s what analytics is trying to provide.
Jim: Yeah, it sounds like anything you can do to get in front of it will help you operate much more reliably and everything. The term analytics is thrown around quite a bit these days in many different areas. I guess relative to industrial automation, what role do these analytics play in uncovering the source of abnormal events?
Michael: Yeah, so the analytics, definitely a popular word. To me, that means really turning data into information and not just any information, but useful information. So analytics is really the key in telling you if an event is an abnormal event, right? So if something happens, how do you know if it’s abnormal or not? That’s analytics. So to understand that, you need to understand the baseline normal operations in order to realize that something is off, but just realizing that doesn’t mean you can do something about it. You know, quality analytics really drives action. And to do that, you need to uncover the source. And it’s not often easy to just have analytics spit out the source, but often combined with some human intervention and the human knowledge and experience, analytics can help you get.
Jim: Yeah, that makes sense. And I guess another trend we’re seeing around here is that many parts of the world experienced plants staff are retiring, and companies are losing this wealth of knowledge. What role can these operational analytics play in capturing knowledge and delivering it to who and when it’s needed?
Michael: Yeah, absolutely. I mean, tribal knowledge is inherent in the industry. When I was coming up, I worked at a lubricants plant for a couple of years and you got people there have been there 30, 35 years and they know what’s going on and they try to teach the young guys, but as they retire, it’s harder and harder to keep that information. So a lot of that is just oftentimes that tribal knowledge is written down somewhere and procedures, but more often than not, it’s just something that people know, and analytics really helped bring some formality there, for sure.
One thing that we work to do in our software is to capture prescriptive analysis and our root cause analysis. So when we have a root cause analysis tree, for example, we’ll show what is that event and we’ll work to capture it, and this is working with the operations, what is the potential impact of that event, and what is the advice with that event? So that way, when that event happens again, we can display the impact and the corrective actions. So analytics is really there to formalize that tribal knowledge and work with those experts that have been there for so long as well.
Jim: Yeah, that seems like that’s a great way to supplement some of that and capture it and make it available to others. I guess performance deviations can be difficult to identify, which in turn delays corrective actions. What are some ways operational analytics like Plantweb Optics Analytics help identify and manage these performance deviations?
Michael: Exactly. So what I mentioned before was that in order to identify a deviation, you have to really understand what that normal baseline is. And the way that we work on that is just pulling in something like historical data, understanding what normal operation mode looks like, what are these other operating modes that you might have? And that really is the basis for identifying an abnormal performance deviation. And when we find a performance deviation, that’s something that we call an event, and we know that these events are interconnected, we just need to know how. So for example, you could have something like low efficiency on a heat exchanger, and that’s the deviation, but the question is what causes it? And there are a variety of other things that could cause it.
So if we have that insufficient heat transfer, one root cause could be a high pressure drop, but why is your pressure drop high? You know, it could be accelerated fouling. So the thing that the analytics really works to do is to take those deviations and start figuring out what is causing those deviations and then ultimately load in some of that tribal knowledge we just talked about to alleviate or fix some of these problems.
Jim: Yeah. It seems like if you’re doing it the old way manually, you could be spending a lot of time sorting it where technology and these analytics can help bubble it, you know, to the source of the problem, get you working on it.
Michael: Right. Yeah. Rather than chasing down all of the different possible leads, at least having an idea of what is that best way to chase down.
Jim: Now, I guess another chasing down thing is data, you know, traditionally a silo, different systems keeping different data in many different locations making it difficult to assess where things currently stand, especially when you’re trying to solve a problem. How does Plantweb Optics Analytics remove these silos and increase visibility across the enterprise?
Michael: Oh, absolutely. To do any good analytics, I think you need good data and that data is often spread across a plant. And it’s almost a chicken-egg type situation because the power of analytics has really illuminated the need to put this together. Analytics has become more powerful. People want to use it, and then they figure out, oh, we don’t have the right data in the right place to do that. So to get a view of how your plant is performing holistically, you need to have that plant’s data, and that’s often, as you said, spread across a variety of different systems.
So Plantweb Optics Analytics is part of our Optics platform. And we do have a product called Data Lake, and that’s a great tool to bring together this data. So that really has these various connectors to various systems that you might have, brings that into one repository where you can clean it and contextualize it. And then with Optics Analytics, you can run the appropriate analysis and visualize that. And the goal really for our tool is to be the hub across the enterprise for a variety of different use cases. So to be one place for reliability managers, operations, sustainability managers, etc. to really come together and have that one source of truth and to have visibility across the enterprise as well.
Jim: Okay. So bringing it together. Then as we look at Plantweb Analytics, what are some of the components which enable it to help detect these abnormal behaviors in the process or assets, you know, identifying root causes of problems and predicting future performance?
Michael: Excellent. So in our software, we do have two main pieces of software in Optics Analytics. So that is the Project Studio and the Modeling Studio. And Modeling Studio is really all about data, ingesting data, cleaning it, trending it, exploring it, finding out more about it with techniques like correlation or classification and clustering, or our even more advanced techniques like neural networks, and ultimately working to model it. So that’s where we take historical data and really figured out what the historical data means.
And then we put it online to something called Project Studio, which has a couple of different components, and one is rules and one is root cause analysis. So the rules are really building that logic of if this happens, then this, and that really creates the different events. And then the root cause analysis is all about tying those events together in a logical way. So I understand if this event happened, what event could have precipitated it? And our goal was to make this really easy and have it accessible to the OT audience as well. It’s a very low code, no code tool, very graphical drag and drop interface. And I mean, even I figured out how to use it and not much time at all. So that’s sort of the different components that helps operations overall.
Jim: Okay. Yeah. I can see you get the models with it in there and being able to work with it to move it along. I’ve heard something about a rules module that enables complex event detection. Can you describe and share an example of how this works?
Michael: Sure. So rules are pretty much where…it’s really the main engine of the software. And here you’ll have something like you want to detect something like a deviation in a KPI, so it can be like low performance on a boiler. So here is where we’ll basically, working left to right, pull in the sources. So what information do I have around the boiler? And start building that logic of if the airflow is over this for five minutes, then that perhaps is like a true statement and that is a precipitating event. And so we build these rules for pieces of equipment and for processes. And then as we run the software, it’s going to be pulling in the data, running it through these rules. And ultimately if an event is true, then that’s something that can be highlighted to you. So I think that’s really the foundation of the software is in those rules.
Jim: Yeah, that’s interesting. So that sounds like what an experienced operator just would have these rules in their head and know something like that.
Michael: Yeah, some of them have like that, but the beauty of our software is that a lot of these rules are pre-populated in our templates. So we have a library of asset templates from pumps to compressors to heat exchangers, where all you have to do at that time is pick your piece of equipment, you drag that piece of equipment in and the rules just pop up. So all that domain knowledge is built into the software, and then all you have to do is map your tags that you might have around your process to these rules and you’re very quickly getting up and running.
Jim: Wow. That does sound like it gives you a huge head start to, you know, fit it to your particular plant.
Michael: That’s the goal there.
Jim: So where does Plantweb Optics Analytics fit in a typical automation architecture?
Michael: Sure. Well, we typically talk about architecture levels, where level one is where you have your sensors, your valves, etc. Level two is where you’ll have your control systems, so your DeltaV and your historian, and then there’s level three, which is above the control system. And that’s typically where we sit, right? So that one level above the actual vault of your control system, we can also sit on level four and we’re very flexible, be it on-premises in level three, or even going all the way to the cloud being installed there, we do offer cloud hosting.
Jim: Okay. So being where it is, and you had mentioned the Data Lake earlier, so the connections that you have to bring in that data, are they coming from levels above, levels below in what you need?
Michael: Yeah, it depends. So for the ingress and the egress, you know, typically we’re bringing in the data from a historian for most processes, cases, sitting around level two, but then we can send the data to a variety of places, be it something in the cloud, or there is some circumstances where we can feed back to the control system as well, right? Typically that’s not something that customers want to do, but we can write back to the control system to help better control as well in level two.
Jim: Yeah, that makes sense. And it seems like having that cloud capability in there, more and more companies are putting their experts may be remotely monitoring a number of places in there. So it seems like that architecture would support that kind of remote expertise.
Michael: Yeah. And that’s where the template early comes in as well. I had a customer call literally just yesterday where they were interested in looking at boilers, but rather than installing the software for every boiler and every plant they have, just have that template in the cloud and send up that boiler data from their different sites and do the analysis there.
Jim: Yeah. That makes a lot of sense and then they can optimize, optimize, optimize kind of as they go across it. Okay. This has been a great discussion on the analytics and some of the surrounding elements to be able to operate much more safely, reliably, and efficiently, but I always like to put our experts on the spot. And what haven’t I asked you about any of this that I should have asked you?
Michael: Sure. So I’d love to talk about some of the new features that we’re working on. And one thing that we’re always looking to do is improve performance and we’re working on how to enhance how we actually take in that data and run those calculations. And that’s something called horizontal scalability. So often these days, if you want…if you’re a computer and you want it to run better, faster, you get a bigger computer, but in the future and in the cloud, you can use multiple computers at once, right, and use that extra power in and out as you need.
So we’re really redesigning ourselves to be more horizontally scalable. It’s a bit technical, but ultimately, we’ll be able to decompose projects, run them on a number of different agents. And what that means for the customer is doing analytics faster and more effectively. And that will also allow us to run on a number of different machines simultaneously. So rather than having one extra-large machine, use a couple of different machines that you can either turn on or turn off as you need. So that’s something that’s pretty exciting that we’re working on as well.
Jim: Yeah. That seems like that would really help with scalability to be able to expand as you needed or contract as you don’t.
Michael: Exactly, right. It’s like turning on and off rather than having to get, like, one massive computer just…
Jim: To handle those cases.
Michael: Yeah. Borrow them when you need them.
Jim: Oh, that’s great. Well, let’s wrap things up, and where can our listeners go to learn more and how can they connect with us on any specific questions they might have?
Michael: Great. Well, a great place to start is Emerson.com/PlantwebOpticsAnalytics. That will take you to our webpage, which is really the hub of information where you can contact us, find out more about our products, get data sheets, where you can find out your Emerson local office or business partner, and all of that. So that’s a great start. I know that working with you and the team here, we are increasing our social. So we’ve got some videos on YouTube as well if you search Plantweb Optics Analytics. And so there’s just a couple of places. And obviously, depending on where you are in the world, your local Emerson representative or impact partner can help too.
Jim: Yeah. A lot of different ways in, and I know you’re on LinkedIn, so I’ll provide a hyperlink in the transcript, so people don’t have to spell your last name.
Michael: Yeah, exactly. I know it’s a little bit difficult, but a hyperlink will help, and I’m always excited to connect and talk shop with anyone that wants to talk analytics.
Jim: Well, there, you heard it, folks. You can reach out to Michael. If you find the blog post, go ahead and hit that hyperlink and you can connect in with him. Well, Michael, I want to thank you so much for joining us today and sharing some on the world of operational analytics.
Michael: All right, Jim, it’s been a pleasure. Thanks so much.
-End of Transcript-