Below is the transcript of the Q and A session of the Webinar. Each question was submitted to the panel of speakers throughout the course of the webinar and is answered verbatim. Simply click on the question to view the answer.

A few questions that we did not have time to answer during the webcast will be added and answered here shortly. Can't find the answer you're looking for? Feel free to submit your question here and we'll respond to you directly.
 

How was the justification to view the monitoring program validated?

How do you make data information and intelligence into predictive maintenance?

What are the key parameters for remote monitoring of combined-cycle units?

Can you detect feed water valve issues, accomplish generator monitoring and monitor HRSG health?

To what extent is all this monitoring automated? All of it?

How does Laborelec Quantify the Success of their diagnostic center?

Have you seen an increase or a change in the ratio of corrective maintenance from unplanned outages to predictive maintenance? Has that been optimized through your company, and how well have you done there?

In general, how far in advance can you predict certain fault occurrence that may lead to a catastrophic failure?

What is the process used to detect early fault detection?

How was the justification to view the monitoring program validated?

Brian Nessen: "How was the justification to view the monitoring program validated?"

Yves De Mulder: "Okay. Well, that's not an easy one, because it's like asking for what is the effect of training? It goes back to the time that LVMS was installed, it was used in different power plants. Once it was proven that this system could help us to speed up the resolution of problems overall, and in an operations situation, it was decided to put it on all the machines.

So in fact, you need a couple of cases to prove that the technology works and that it has a big interest in them. In fact, it is very easy to convince the management to invest in it."

Brian Nessen:"Tom or Todd, do you have any thoughts with your experiences from any other folks you've worked with?"

Todd Pistorese: "Well, there's certainly some other case studies on the use of a centralized Diagnostic Center. Entergy had an excellent story, a magazine out maybe two years ago, where they had a significant save by centralized monitoring across their fleet. So I think that the case for doing this type of extended performance analysis and monitoring with the expertise in one location, typically to look across the fleet to see if there's any circumstances where you have a common problem across your fleet is a very powerful way to look at your systems."

Brian Nessen: "Tom, any thoughts on that?"

Tom Toronto: "Yes, Brian. We have a methodology at SmartSignal where we can assist customers in analyzing their past maintenance history and look into that. We can do a gap analysis to see where early warning and predictive diagnostics may have helped them avoid some of the unplanned outages that they had and a lot of the collateral damage that may have occurred in some associated equipment. We can help lead them to that.

Usually they have all the information that they need to do a validation analysis, and especially if they've got OSIsoft's PI Historian there, then all that data is there, and the data tells the story so we can mine it out, give them some assistance on creating a value case before they actually put in the system.

In the case of Laborelec, Yves did a great job in convincing the people there, leading them through the thought process that it is fairly simple to do, to realize that with some early warning, you can avoid unplanned outages and do predictive maintenance on your equipment."

 TOP 
 

Brian Nessen: "Yves, the next question, how to make data information and intelligence into predictive maintenance. How do you do that?" 
 

Yves De Mulder: "Well, it's not that easy, either. I think you're choosing out the most difficult to start with. In fact, using a global tool like SmartSignal is one of the solutions. So the advantage is that it can retrieve the data coming from all kinds of assets, so you can retrieve the data of all the assets with one single tool. I don't say it's simple, but with one single tool. And that has an advantage that you only need to work with one tool, and you can do further and deeper analysis using other tools. That's how to make data and information and intelligence into predictive maintenance, I think, is a larger process."

Tom Toronto: "Yes, I'd agree with Yves...That's where the rubber meets the road. You can have all these great analytical tools, but they need to be seamlessly integrated into the work processes that already exist in an organization. Or if they don't exist, they need to be created so that that is great intelligence and diagnostic information that can come from having all of your data integrated, having analytical applications that will percolate up the exception base, notifications to you. But somewhere along the line, the people have to turn that into action. And so they need to be able to follow a really nice, clean work process to get that done."

Brian Nessen: "Great. Todd, anything to add there?"

Todd Pistorese: ...because he's hooked into a centralized center, he has a lot broader experience on all of the problems that are occurring across, so maybe isolating into one, that expertise is isolated into one plant alone, you may access that knowledge to work on a problem at another location. So that's an important point for getting the expertise and the knowledge base up to par."

Brian Nessen: "Yes, a clear value in having that centralized."

 TOP  

What are the key parameters for remote monitoring of combined-cycle units?

Brian Nessen: "What are the key parameters for remote monitoring of combined-cycle units?"

Yves De Mulder: "Well, the main components, of course, like the gas turbine, the heat recovery steam generator, steam turbine, and the transformer. So of course, now that a lot of these combined cycles are affected by cycling, it's very important to have a close view of how is the life cycle assumption of the typical components like parts of the boiler and parts of the steam turbine."

 TOP  

Can you detect feed water valve issues, accomplish generator monitoring and monitor HRSG health?

Tom Toronto: "Some people asked some specific questions, for instance, "Can you detect feed water valve issues?" "How do you accomplish generator monitoring?" "Do you monitor the heat recovery steam generator health?"

 

Yves, I think you are monitoring all of those in your applications, aren't you?"

Yves De Mulder: "Most of the components are monitored, and valves, again, also will be monitored, or by dedicated valve monitoring. Or you can also make them part of the scope of SmartSignal. So it's certainly done already for the valves which go with the bigger components like gas turbines or steam turbines. All those valves are already included in the models of those assets. But those valves can also be monitored individually or with the same early fault detection tool."

"Valves, we kind of focus a lot on the main components like the gas turbine and steam turbine and the boiler, but there are hundreds and hundreds of valves in the combined-cycle plants, and we see that they are producing a lot of the unavailability. So we shouldn't miss them. Valves are very important in being monitored, too, because they produce a lot of problems.'

"The next [question] also about generator monitoring?"

Brian Nessen: "Yes"

Yves De Mulder: "Well, there was a slide on the generator monitoring, on monitoring in general. So on generator monitoring, generators, so partial discharges into the stator bar isolation can be monitored, and we do this on nine units today. These are today mainly air-cooled generators for the hydro generators, the air gap and the stator bar vibrations can be monitored. And for the turbo generators, the untwining vibrations can be monitored, as can the rotor flux. But these are a couple of generator monitorings which we are using currently today. Is that an answer to your question?"

Brian Nessen: "Yes, it sounds like you've got a good suite of diagnostic information there to monitor."

 TOP  

Tom Toronto: "Because some of that data and some of that analytical information that comes in from separate systems, people may wonder how that all gets into the PI data infrastructure. Yves, maybe you can give us some insight into to what extent is all this monitoring automated? Which part? You had that one slide of all the different systems. Some of those, or hopefully all of them, are automated and automatically collected and historized, or maybe you have some remote handheld information that's typed in by hand, but still historized. What type of data automation do you have at this point?"

Yves De Mulder: "Well, into the PI systems, the main process data is captured on a permanent basis and is permanently available for the Diagnostic Center if you want to analyze something. The other systems also are working on a permanent basis, so the vibration monitoring, generator monitoring, the gas turbine condition system, the wind turbine, so everything is working on a permanent basis, stores its data. For some, we back up the data remotely to be sure that the data is always available. That means for the data which doesn't go to PI, we store it from the Diagnostic Center. And so it's mainly all automated."

 TOP  

Tom Toronto: "How does Laborelec quantify the success of their Diagnostic Center?" You gave us a great example in a case study of an avoided cost for that combustor bulging problem. Is there a systematic way that you try to quantify your success and justify the Diagnostic Center?"

 

Yves De Mulder: "Well, I have to admit that today we are not doing this on a permanent basis. We do this when we have to renew some of our purchase orders, because then we need the signature of one of the managers. But today, we are more putting our effort in implementing the solutions in a wider and wider fleet, and the people of the power plants are convinced. And of course, when you show the case data of which I showed one, I think this speaks for itself."

Tom Toronto: "I would agree. That's...very similar to how you justify or validate creating a Diagnostic Center in the first place; but just monitoring your success day-to-day, taking the notifications, and then discussing with the plant what the resolution was for the diagnostic notifications and discussions you had with them, and then you can calculate what type of generation loss or maintenance cost avoidance you were able to have."

 TOP  

Have you seen an increase or a change in the ratio of corrective maintenance from unplanned outages to predictive maintenance? Has that been optimized through your company, and how well have you done there?

Tom Toronto: "When you're feeding the plants these various notifications of problems that could be coming up, and they go out and do their in-depth diagnostics and then actually solve these problems, you indicated one where the temperatures were only up from the model predictions by six degrees or so, and that was able to help you avoid a bearing problem. Have you seen an increase or a change in the ratio of corrective maintenance from unplanned outages to predictive maintenance? Has that been optimized through your company, and how well have you done there?"

 

Yves De Mulder: "Well, I can't put a figure on this, but from the feedback we are getting from the power plants, I can say that what we deliver to them has been integrated into their process. And they really take it into account. I have no figures to illustrate it, but it tells me it has been integrated into the operation and maintenance process."

Tom Toronto: "Well, that speaks well to their trust in the accuracy of your diagnostics that you're giving them. That's great."

 TOP  

In general, how far in advance can you predict certain fault occurrence that may lead to a catastrophic failure?

Tom Toronto: "In general, ...how far in advance can you predict certain fault occurrence that may lead to a catastrophic failure? That one, I'm sure you'll agree, you'll have to answer in more of a general term, as each fault can be different and have a different time signature, can't it?"

 

Yves De Mulder: "So yes. In fact, it depends on the nature and the speed of the degradation, but in the case of the degraded bearing, in fact, we already saw the difference two months before the plant could program a planned stop. So it can be months in advance. It can be, of course, days, weeks, or months, depending on the amplitude of the degradation and the speed at which it's evolved. But some can be detected months ahead."

Tom Toronto: "Yes. Generally, the system is based on the sensor data, so as soon as the sensors indicate that there's a deviation in the equipment behavior from the model's predicted behavior, then of course there's a differential there that you can start analyzing. And that, of course, is going to just depend on the failure mode that you're currently looking at...if you have a crack developing in a rotor shaft, that's going to happen quite quickly, whereas a bearing starting to rub a little bit, you'll get some very early warning, maybe up to six months in advance on that one, I'd imagine."

Yves De Mulder: " Yes, of course...It has to be, yes, in fact, you have to be ready to detect something like the cracked shaft, which Entergy, I believe, detected. I think I wouldn't like to miss that one."

 TOP  

What is the process used to detect early fault detection?

Tom Toronto: "What is the process used to detect early fault detection? The technology that you're using, of course, is you've got your PI data infrastructure collecting information from all your different analytical tools—vibration analysis and so forth—and then you've got SmartSignal that is getting the data from the PI Historian and analyzing that with respect to model predictions and then giving you notifications. Maybe this question refers to what is your actual process for somebody sitting in the Diagnostic Center? What's their first indication that there's a problem, and then how do they step through their process for notifying the plant?"

 

Yves De Mulder: "Well, I showed in the workflow of using SmartSignal how it works. So in the daily WatchList analysis, those things are detected and investigated. So we can use PI to view more data and to make a detailed analysis. And, if necessary, we can go to other condition monitoring like vibration or combustion dynamics when it is necessary to see vectors of frequency spectrums. But for the units which are monitored with SmartSignal permanently, it's through the WatchList that we get the first indication."

 TOP