Name: From time-based signals to understanding: Investigating industrial process behavior
Uploaded: 2026-11-12T17:16:05.633Z
Duration: 52 min 48 s
Description: From time-based signals to understanding: Investigating industrial process behavior

Transcript for "From time-based signals to understanding: Investigating industrial process behavior": Hello, everyone, and welcome to today's webinar, part of the Spotfire product webinar series from time based signals to understanding, investigating industrial process behavior. We're thrilled to have you with us today. I'm JP Richard Charman, principal product marketing manager here at Spotfire, and I'll be your host for today's session. Now before we dive in, I'd like to cover a few quick housekeeping items so you get the most out of, today's session. Now the webinar will last for approximately forty five minutes with a ten to fifteen minute, question and answer segment at the end of the presentation. Now if you have any questions during the presentation, please use the q and a panel located on the right side of your screen, and we'll address as many as many questions as we can during the q and a segment at the end. We've also made a few assets linked to today's webinar available for you, which is... Which can be accessed through the doc section of a webinar platform. So please feel free to access these, and these are right next to, the q and a tab on the right hand side of the platform. After today's session, a recording of today's webinar will be made available on demand, and we'll email you the link shortly after the event. With that, we will dive right in. I'm excited to introduce our presenters alongside me today. So Dan Ropp, our senior principal product manager here at Spotfire, and Jose Silvester Acuna, our data scientist here at Spotfire. Now I'll start by setting the scene, the challenges teams face when they investigate industrial process behavior. Dan will run through a quick overview of Spotfire Industry Pro before we dive into a demo that Jose has prepared for today. Now industrial operations are driven by processes that evolve over time, and every one of those processes tells a story. The challenge is turning time based signals into a clear understanding of what is happening and why fast enough to act on it. A well declines, a compressor degrades, a process tool drifts between maintenance cycles. Every one of those changes has to be understood across thousands of signals before anyone can make a decision they can stand behind. And that's the starting point for everything that you'll see today. So we'll solve by focusing on two industries where we see this the most, oil and gas and high-tech manufacturing. In oil and gas, millions of tags stream into historians while the events and maintenance records that explain them live in other systems. In high-tech manufacturing, a single wafer can carry more than a 100,000 dyes, each measured many ways, and engineers need to connect those measurements to equipment history and defects. Different worlds, same problem. The answer is never in one place. Now the same pattern holds in other process industries like refining, chemical production, and bioprocessing too. So what makes industrial processes, process behavior hard to investigate? We've noticed three main challenges, and each one is a reason why watching data is not the same as understanding it or understanding a process. On each of the next slides, you'll also see the key Spotfire features that help address them. Challenge one, slow drift hides inside the limits. The same process rarely behaves the same way twice, and a slow drift stays inside its alarm limits until performance is already affected. On the left, you'll see an illustration, so not real plant data. Here, the signal is flat, starts to creep upward, and never crosses the dashed alarm limit. So no alarm ever fires. What you want is the limit and the reference in the same picture as a signal. In Spotfire, reference elements let you place control limits, targets, and maintenance or shift intervals directly on the visualization. Match data tables on column names applies limits and rules held in one table to the measurement in another. Transformations, including missing data handling, resampling, and smoothing, prepare noisy signals before you judge them. And process capability analysis with quality control charts checks whether the process is staying in control. The second challenge we've we've observed is that the signal that matters is buried. A single process can generate thousands of signals, and the one that explains an upset is rarely the one in the alarm list. Each square that you see in this illustration here is a signal, and only the pink one explains the upset. As mentioned, this is an illustration, not real data. But here, behavior also depends on signals moving together with lags and context. So looking at one trend at a time hides the relationships. What you want is to reach that signal quickly. Time series analysis including, includes auto and cross correlation, which finds lag relationships between signals, and dynamic time warping, which align series to reveal similarities. Dimension reduction helps you find the best predictors using, feature selection and extract features across thousands of variables with principal components analysis and clustering. Range based marking lets you drag across a time window to drive filtering and drill down, all for root cause investigation. And Spotfire inside agents invoked from the inside panel, identify trends, patterns, and anomalies, and recommend next steps with you, the domain expert, in control throughout. Now the third challenge really ties the other two together. Insights are fragmented across tools and teams. Specialized tools are essential, but each one holds part of the story in its own format. On the left, oil and gas signals in historians and control systems, while maintenance events, well logs, geospatial data, and economics sit somewhere else. On the right, high-tech manufacturing, design, process, metrology, defect, electrical test, operations, and WIP, and yield and quality, each with its own tools. When something changes, the explanation lives in another system. So engineers spend their time finding, moving, and reconciling data instead of understanding it. Specialized tools are working against cross discipline decision making. What you want is really one place where signals meet context. So the specialized tools keep doing what they do best. And just to be clear, Spotfire doesn't replace them. The Aviva PI connector gives native access to time series and OT data alongside OPC UA and log ASCII standard LAS files. Join tables and columns brings data from different systems together, and manage data table relations keeps it linked. And lastly, brush linking means a selection in one view drives the rest. So just to make that a bit more concrete and starting with oil and gas, specialized tools are essential here in geology, petrophysics, seismic, reservoir, drilling and operations, production, and well and equipment surveillance. Every discipline depends on them, but each discipline works with its own data in silos created by those tools and industry data formats. Collaboration gets messy and slow when every part of a project needs its own tool. And upstream teams face tight deadlines and unforeseen issues while trying to keep a competitive edge. Now the same story plays out in semiconductor manufacturing. Specialized tools are essential here too. KLA and applied materials for defect inspection, PDF solutions, and yield works for yield and quality onto an Hitachi for metrology, Adventast, and Terradyne for electrical tests, and many more. But each one generates data in its own format with its own interface and its own export process. When the yield excursion hits and it always hits at the worst moment, engineers pull data from half a dozen systems, reconcile formats, build their own views, and try to correlate signals that were never designed to be seen together. So as I mentioned before, specialized tools are working against cross discipline decision making. Now siloed analysis, manual handoffs, slower decisions, that all has a cost. It costs time, multiple hours per incident across multiple teams. It creates confusion with manual handoffs and fragmented workflows and frustration from working in isolation, and it leads to errors like miss signals, mistakes, and delayed decisions. And, ultimately, it means lost opportunities, millions in potential value left on the table. None of this is a failure of the specialized tools or of the people using them. It's what happens when insights stay fragmented. So the question is, what if the signals and events and the context could come together in one investigation with the domain expert in control? Now that's what Dan and Jose will show you next. And with that, I'll hand it over to Dan for an overview of, Spotfire Industry Pro. Alright. Thanks a lot, JP. Okay. So I thought I'd take a few minutes here just to sort of set the stage from a product perspective. The problem that we're building for, where we're taking the product, and what Spotfire does with process signals today. And then Jose will make a concrete with a demo. So if you run a plant, align, fab, you already have things like like trends, control systems, alarms, SPC even. These are things that are good at telling you something changed and when, and that's monitoring. And that's largely been a solved problem for many years. The hard part starts right after that with three questions that every process engineer would ask. First of all, is this normal? A trend shows you the value. To judge that, you need limits, you need target, reference period, and the statistics behind that. You need it in the same picture as the signal and not in a spreadsheet that's sitting next to it. So why did it change? The explanation is rarely sitting in the signal itself. A compressor train starts drawing more energy for the same throughput. A batch time drifts over a few weeks. The reasons could be in the operating mode, the equipment, the maintenance event, the batch, a batch that ran at the same time, and that information can live in a totally different system as well too. So now what else moved with it? So you have thousands of tags patterns. It builds up... Things build up gradually. You have multivariate relationships versus individual trends. That's too much for a trend screen and too too subtle for an alarm. And so that's where Spotfire comes in. So the signal really belongs inside that whole investigation and not isolated in a separate historian or a trending tool. So a visualization in Spotfire then becomes the central place for your investigation, for your thinking. So... And when you select a period on a trace, that means every other view follows that selection. And so you're comparing, conditions and outcomes. You're running statistical tests, supplying calculations, and all those results follow into the next few. And so the historian signals, the events and the asset context are all within the same analysis. And that means... So what what is happening, why it's happening, and what to do next become a single investigation instead of three separate tools. And that's where we're taking spot fire. The signal stops being in a separate place and in a separate tool. And that means we have engineers that go from monitoring what changed to understanding it better, controlling it, improving processes within the investigation, signals that carry their context with them, changes explained against the reference run, and stepping from, like, from characterization to testing a change before you actually make it on a live process. Next slide. Okay. So let's talk about a little bit about what spot spot fire does with process signals. And a lot of this has been available, in Spotfire for years up until '16, which was just very recently shipped. So first of all, connect. Spotfire reaches the data where it lives. Native Aviva PI connective... Connectivity retrieves interpolated values on a common time basis. OPC UA for vendor neutral route access to historians, or SCADA or MES systems. And, of course, we have the usual databases, lakes, and files through native connectors, SQL, Snowflake, Databricks, and many, many others. You can also access real time streams when you need live monitoring and alerts. And when the data grows too large to move, our new push compute capability runs the heavy analysis directly within the data platform itself. Next, we have preparing. So built in functions normalize, do imputations, resample, and smooth data. Missing data analysis can find gaps before they bias a comparison. The data canvas inside of Spotfire handles the tall to wide reshaping, the pivoting, joins, and calculations that tag data always needs. Python and r data functions and also the recently added statistical algorithms can cover whatever else is specific to your process. So exploring in time. So every view in Spotfire is linked so that when you filter, select, drill down, the rest always follows. You can select the time range directly on a signal and the whole analysis response to it, which which is really the difference between knowing something happened in March versus having everything focused about what's going on in March, for example. Time axes adapt as you zoom to... From years to milliseconds. You can trellis by asset, chamber, or batch, and page through those groups as you wish. And finally, comparing and explain. Reference layers, put limits, bands, and c... Even c p... CPCPK zones directly onto a chart. Distributions are compared side by side using box plots, violins, ridge lines, and funnel plots. And when two populations look different, you can run statistical tests directly from within the visual, and Spotfire AI can even pick out the right test and explain it. So SPC charts correlation, curve fits, and forecast are also available as easy to use actions. And and insight agents can run, your own methods on on your data. So underneath all of this sits, Spotfire AI. So as Spotfire lets you describe what you need, and you get an interactive analysis back instead of a static picture. And that means it's something that you can further inspect and refine and learn more from. Skills that you can create describe a validated trusted technique. So that way, you know, the way the way your best engineer runs an investigation, you can make that reusable by others in the organization. Insight agents carry out specialized analytic work, choosing the right statistical test, preparing data, finding commonalities on your data. So the point here is is that using proven, reusable methodology in conjunction with AI really assist the investigation, and it keeps the engineer in control of the of the whole thing. And then once an investigation was repeating, Spotfire, of course, can automate this and share it. We have scheduled jobs and notifications with governed interactive analysis in the browser for for every engineer. And everything we're talking about here is available, today. Next slide. And so this is the place... Spotfire industry pro, this is the place where we have our most advanced analytic techniques, visualizations, and industry focused capabilities, actions, and so forth. All of that is what is is what Spotfire industry pro is all about. And that... And that's... We're we're very excited to carry this this forward. Next slide. Okay. So that's kind of the stage, and Jose is gonna now make it more con... Concrete with, the product as things are available today. So there's a few things to watch out for during the demo. Looking for several signals being investigated together in context rather than individually, how patterns and drift, that develop gradually can hide in high volume data, relationships across many variables, how these things move together and not just how each one trends. Moving beyond monitoring. So from from identifying something that might have changed to why it changed without actually leaving the analysis and doing all of that, at industrial volumes of time series data in in minutes rather than days. And so with that, I will hand it over to Jose for the demo. Thank you, Dan. Welcome, everyone. Happy to be presenting for you guys today, for everyone here today. So the example that I have, prepared for today is, an example for the oil and gas industry, specifically for detecting long term drift or long term, wear and tear on, rotating equipment. But for... I I was looking at some of the audience members here today, and, looks like we have some pharma and semiconductors people. So if if you can just think about this example applied to any long term, like, mechanical wear drift or process baseline drift or thermal drifts, that... That'll be the the same kind of application that that we'll be discussing here today. So what I'm showing right now is an application built fully inside of Spotfire. So I have this, Spotfire analysis open. The data has already been preloaded, but like Dan and JP mentioned, there's the ability to use PI data. We have our our PI connector, in here. This dataset was already, prepared using a PI, a PI server and a PI as a framework and a little bit of, asset management or historical, asset management data. And going into, so the... This analysis is looking at two two pumps, that share, a pipeline, and there is a pump state. So here in the in the center area, we have our time series data being shown. Right now, we're looking at, the December 21 to the December 31, so about, ten days worth of data. And we have, a tag for our pump state with some history here. It looks like the the pump tripped... Oh, let me expand this a little bit. It looks like the pump tripped, for two two point eight hours, on this day, and if you hover over it, it'll give you that information. We're visualizing the suction pressure, the discharge pressure, and flow rate for the pump. We have a... So everything here is native, Spotfire. The only thing that's not native is, we... I have created a custom action mod... Visual mod using our visual mods framework to have, like, an easy UI for my time series library that I'm going to be using, and manipulating some of the time series information here. But, I guess, to get straight into it, we have that, state that state tag for our pumping active or trip. And first, we wanna visualize the other, data that we have here. So here I can see I have four out of my seven process variables for this tag shown, suction pressure, discharge pressure, flow rate, and pump state. But I also want to see my bearing temperature, so I'm just going to add that. So I click the add button and you can see bearing temperature gets appended to that trellis, that ballpark trellis in the time series visual. And I also want to visualize my vibration data. Now you have vibration, varying temperature as well. We want to start preparing our dataset to use some of Spotfire's reference elements so that we can visually see these overlaid on top of our time series data. The first thing I'm going to do is, coming over here to the library, I'm going to create an event from a state change. I'm just gonna select my pump state tag, and I'm going to look for the state's, tripped, which you can see the the value for that is right here. If I add this step to my calculation sequence, you will see a a new overlay of a of a red reference element band, on top of the time series data. And a couple of things happened here. There's a new panel up here for events. So this this pump state tag is just, something coming from our historian or alarm management system. Now we have an actual event, being quantified here and and visualized, and then that also creates the ability for us to to look at our events, stacked on top of each other. So, in this particular example, this pump has, it looks like, one, two, three, four events across this one year of data. So I'm I'm gonna, zoom out a bit and try to find those. So I'm using this bottom control here, but same thing with the with the window. You can just, expand your, your viewing, the the data that you're viewing with with calendars or with this, slide bar down here. And now we can see that there are indeed one, two, three, four events. And for those four events overlaying the suction pressure, discharge pressure, and flow rate, let's us let's us see how the process variables are, are proceeding across that event. Now that we have this one pump, I'm going to zoom back in here. We created one reference element, but this pump only has one type of category for this pump state. So using this asset selector, I'm going to change my reference to pump b. Pump b, has more history of, events. It also has this. This one has a stopped event here. I'm going to add that as another type of event in my analysis. So I'm going to select state change again. And, again, I'm going to reference my pump state tag, and I'm I'm going to look for any stop times in this dataset. And we should we should get more, events for this pump now. So you see how it's missing from my event one, which was trip. Now I'll get a new event up here to highlight this band, the sequence of time. There it goes. So now I have my my pumps tripping as my red bands. I'm going to zoom in. Actually, I'm gonna zoom into this area just so I can see this a little bit easier. Oh, there it goes. So I I zoomed in here. Now you can see that I have two events, the red ones, the the yellow ones, and I can swap between the events that I that I have shown here. But the next thing I want to do is find any time that the pump is cavitating, and all of this is in preparation for, for me to create a long term drift detection, analysis on this pump. It looks like this pump has... This pump experienced a long term, here it is, in long term vibration, or bearing failure. So your bearings is being worn down right here, so your vibration is increasing over time, and it looks like it happened across, like, three months. So this is a this is a long term drift example. And what I'm doing right now is I'm preparing my dataset, my analysis to, efficiently block out times where I don't want my, SPC analysis to account for bad data or events that will skew the SPC analysis. Right now I'm creating these events. Eventually, I will use these events to be removed from the analysis. The next one is, I want to create an event from a condition, and here, the condition will be, a combination of suction pressure and, vibration. So... Oops. I might have to let's go up in from condition. And then here, I should be able to just input this. I'm gonna I'm gonna go to my other screen because I think it's, zooming, too far out. Yeah. Give me one second. I'm gonna zoom out to my other screen. Oh, that's that's because I was using the wrong one. Okay. It's actually, I want to create a formula first. That's what it was. So let me just look for my formula. There it is. So my formula is going to be whenever my suction pressure is less than 45 PSI and my vibration, is greater than 0.2 inches per second. This is going to be the the condition that I'm... The formula for the condition, and I'm I'm going to reference my suction pressure and my vibration. So I need to add that into the configuration for my inputs. Now after I add that, I can create... So this is just creating a a new, condition, but the condition doesn't create any events. It just creates... Oops. It might have might have broken it. Let's go back. Give me one second. Okay. I think that... I think, there it goes. Okay. So so I created the, that condition formula that has that that formula of suction pressure, vibrations. And, and then after that, I mean, what it create an event from this from this formula. So going into, going into my events, I can do event from a condition, and I wanna choose my condition formula. And now I should be able to find, any time of cavitation for this pump, And it'll be overlaid as as event number three. So it should be a couple of cavitation episodes for this pump, and the... These are events that you want to, take out of our, analysis as well. So now we can go to, start creating our SPC analysis. And, we're we're going to use the exponentially weighted moving average because, if we use the x bar here, it might... We might get a lot of events using exponentially weighted, moving averages. We can, incorporate that historical, historical points into that. I'm actually going to set this to, start on January 1. I think there's... There was, yeah. So you you can see that long term, this pump was operating fine on the vibration and then it kinda started, curling up here so that that bearing started experiencing wear. So what I'm going to do is I'm going to zoom into the time period where the pump was operating fine, and then I'm going to say for my training window... So for the window of data that I wanna use for this SPC, I'm gonna use, some of my current view, whatever it's it's currently on. So it populates that, And then I'm gonna I'm gonna keep the weighted, the weights for the lambda and the limit, for the standard deviation limits as zero point two and three. This just means that 0.2, incorporates more of the long term, the historic... The... It weighs the historical points higher. So if you change this up to 0.9, then the present terms are weighted higher in the analysis, and it'll it'll skew the analysis. So we we do want long term drift, so that's why it's set to a low value here. And then I want to exclude, this is the reason why we were preparing the analysis earlier is we wanted to exclude the event sets. So we want we want to exclude all these times, that are highlighted right here. Then, and then we add this, and what this does is it creates a couple of a couple of things. It creates the upper control limits, the control limit, and the lower control limits for the vibration, the vibration data. And and then, so so now we get a lot of events. And the reason for that is, so so I'm going to go back to the to the recipe and and edit this because, this this looks like it's, too many. So we go over here to the recipe. You can edit it if you needed to, adjust any items. This is one of the settings that we're we're going to want to change here. I'm going to set this to twelve hours because this is a long term drift. If if it doesn't persist for at least twelve hours, we're... What we're looking for here is, like, months or weeks worth of of data. And, if it doesn't persist for at least twelve hours, then it's probably not a real, long term drift. So now I'm I'm redoing that and, I get a little bit better tuning for for my analysis here. I think I'm gonna do one more thing here because, it's still a little bit out of the... Let's do let's just do it to June. I think that was a little bit nicer of a time frame. And I'm gonna set this to one day. So earlier, we were mentioning, whenever you're doing these time series analysis, you want to have all of your data with your events and your analysis being able to reload on on demand, basically, and re visualize how your analysis is going. So that... That's kind of what I'm working through it right here right now is I'm retuning my analysis to figure out when is... What is a good set of parameters and training data for for this specific application. Everything, every one of them will be different. So, so now we see... I'm gonna add the vibration upper and lower control limits here. And so so we can see them overlaid on the vibration. So now signals, I have 10 shown, so I'm I'm gonna take off the flow rate. I don't want to see that for a bit. That's actually what happened. I applied it on my flow rate. That's what it was. That's something good to pay attention to if you're doing these analysis. Whenever I selected my SPC analysis, I zoomed into my time period, but I didn't actually select my vibration data. I selected my flow rate data. Let me go back a little bit, which explains why all of this was very, erratic. Back a little bit more. I should be able to just edit the recipe at this point. So here on the input, this was a a mistake on my side. I should have selected, vibration for this. And now now, I noticed that because it added the flow rate, upper and lower control limits up here. Let's see if it's gonna like this. There it goes. I did. What I changed was the reference measurement for that SPC analysis, and, I changed it to be the vibration data now. Again, similar to before, if we zoom up... If we zoom out here, we get a couple of erratic events, but the main one is this where the vibration starts curling up. So you can see it does highlight most of that, but we wanna get rid of all these smaller ones. So now I'll change this back to, twelve hours. It should work with their upper lower case. So after I retune that, it should just basically highlight this this area where you see the vibration, having a long term, upward trend. There it goes. That looks so much better. Okay. So what so what do we see now? We see, a set of highlighted, times in our analysis for that long term drift. So now we can easily, so now we were able to quantify that using the SPC method. The last thing I'm going to do here is I'm I'm going to clean up these gaps because this really should just be one long term, drift events. Last thing here is I'm I'm going to do an event merge. So I wanna merge everything from my vibrations, from my vibrations, events that's shorter than one day. If the gap between the events is less than one day, so this should just come out to be one long, period. And then I'll and then I'll zoom in a little bit to these. So now you can see, you have a couple of trips and stops. I'm gonna stop looking at those for a bit just so I can clean up my analysis a little bit. But I'm gonna leave the SPC analysis, on there. And I'm also going to remove the the flow rates, the motor currents. I don't need that one either. So two things that happen, is your vibration... Your bearing temperature will increase, and your vibration, will also increase over time with this bearing with this bearing wear. There it goes. And earlier, whenever we added the SPC analysis onto this vibration data, we also, calculated the upper and the lower control limits. So I'm I'm gonna add those horizontal reference elements onto this analysis now. So it should just be horizontal lines. There it goes. Kinda hard to see right here, but if I zoom in a little bit more, they should be more visible. There it goes. I'm gonna zoom in a little bit more. Let's zoom in right here to this area. I just wanna I just wanna highlight this, the reference on it. There it goes. So this is a a little bit cleaner view. So in the, so whenever the vibration goes above or below the, upper and lower control limits, then you'll start accumulating that that deviation. But since these are short spikes, we're not we're not seeing the events for, for that long term, because we used a 0.2 in the Lambda factor for the SPC analysis. We're not seeing that here. But, once we start accumulating that, you see one right here. I think I used twelve hours, so that that must have been out of the limits for more than twelve hours. And the real one is the the long term one. So the long term one stays above those control limits. You start accumulating that, and then you have this, this one long event, which is, nine hundred nine hundred hours. So can't convert that in my head right now. But but here, the the the point here is, using the SPC, the reference elements for the upper and lower control limits, and the the highlighting, you can have a more holistic view of your analysis. And the last thing that I'm gonna add here is, a forecast, for my for my bearing temperature. There is a, as you can see, like, there's seasonality on your bearing temperature, and that's because this pump is located outside. So it's it's subject to ambient temperature, and ambient temperature has long seasonality to it. So if you just wanted to add a forecast for your bearing temperature, you can do forecast trend season. You can use sixty days worth of history, and this is just an example of of adding, a forecast here. I use seven days and now I use a 90% interval, and I'll do it on my bearing temperature. So that's already selected. So when I add this, it's gonna do two things. It's going to append that future, that fuse... That future time series indexing to my time series array, and it's going to create a new, a new events. So it's creating this future events, and and now what we see here is for for only my bearing temperature data, I can see this this forecast, because the forecast is... Oops. I'm gonna zoom in a little bit here. I'll stop looking at my condition one. My forecast starts from the last valid point and then the forecast is the the red line. So it it caught that seasonality, that that daily seasonality from your ambient temperature. And then that that band, that red band is using the spot bar reference elements to highlight that 90% confidence interval for the forecast for the bearing temperature. And, yeah, so that so that wraps up analyzing, the history and and finding, good, control limits and long term long term bearing wear or mechanical drift on your equipment. The last thing that I'll show here is the recap. Oops. That that is the wrong one. The recap slide. So what we covered was for a crude pump station, we had two pumps. One of them experienced that long term drift on the vibration, so we added seven steps into our calculation recipe. We found the trips, the stops, and the cavitations so that we can exclude that from our SBC analysis. Then we went to on and did that exponentially weighted moving average chart for the vibration, for the vibration sensor data. And then we merged some of the the events so that we can get just one log events, and then we forecasted the the bearing temperature using that seasonality forecast method that we have. Pop pop a, I I didn't show, but pop a didn't have any, long term drift alarms. Then I'll, I'll give it back to you, JP. Thank you very much, and, thank you so much, Dan and Jose, for that insightful presentation today. And thank you to everyone who joined us today. Now before we get on to our q and a segment, so please, if you do have any questions, don't hesitate to put them in the q and a panel located on the right side of the webinar platform. But there's just a few things that we wanted to share today. So as I mentioned at the start of today's webinar, for on demand access, a recording of today's webinar will be made available soon, so do keep an eye on your inbox for the link. Next steps, if you're interested in learning more, please do feel free to visit our website at spotfire.com or contact us directly. There are lots of ways to interact with us, whether it is via our socials, through our community. Additionally, our blog site has lots of great content where we share the latest on visual industrial analytics, dive into Spotfire industry pro in more detail. And last but not least, if there are any enhancements that you'd like to see or have ideas that you'd like to share with us, please don't hesitate to visit our ideas portal. Now with that, I'll give it a few more seconds, with regards to our q and a. Perfect. We've just got a question that's been added. Thank you very much. So the question being, in the past, connecting to OSI PI was possible using a free connector being available. If I interpret your presentation right, it would now require an industry pro license, and enterprise would not be enough. I'll hand that over to yourself, Dan, if you've got anything to add to that. Yeah. What was the connect? Can you remind what the specific OSI PI. source was? Oh, a viva pie. Yeah. The... I I just want to check on exactly how that works now. So I... The... Yeah. I need I need to actually check on that and get back to you. No problem. So we'll be in contact with you directly, Peter, with regards to the art, but thank you very much for your question. I'm just having a quick look. Yep. I can say, though, that the, the connector, which we continue to be available, the one that you're using on the community, if you're using community one. Thank you, Dan. I don't. seem to see any questions, but as we do with all of our webinars, if there are any questions that pop into your head after the fact, please do not hesitate to reach out to us directly, and we will be in contact. And we can address your question, offline line or so of this webinar and have a chat with you directly. Now, I will leave it there as we don't have any further questions that have come in at this moment in time. But just wanted to thank all of you once again for joining us today, and we hope to see you at one of our future webinars. Notably, our next webinar is taking place next week on the November 19. So thank you again, and wishing you all a fantastic day. Thank you.