Fusing Prognostic Models with Sensor Diagnostics to Predict End of Life Gearbox Projections

Recording Transcription:Natalie Hills: Thank you all for joining us for today’s session, on adding sensor diagnostics into our prognostic model to better predict end of life gearbox projections. My name is Natalie Hills. I am the Marketing Programs Manager here, for those of you who don’t know me. Again, I see a lot of repeat attendees here, so thank you all for joining us and coming back to our series. I am the Marketing Programs Manager and I run the webinar series for us here at Sentient. I’m also the first point of contact. So please feel free to reach out to me at any point in time. My contact information is shown on the screen here. I’m always happy to put you in contact with the correct folks on our team, or get you further resources that you might need on our technology.

Before we get started, I just want to tell you guys a little bit about our webinar series that we’ve been doing over the past year here at Sentient. We run three webinars per quarter, one per month, typically falling on the second Wednesday of the month. We start with a prognostic technology approach, where we take our viewers into a deep dive of our technology to understand how we have developed it over the past years.

Secondly, we take everybody into a deep discussion on life extension applications. Here we show different in applications in the different industries such as aerospace or autocraft or wind as well. And we show how our technology applies to these different industries.

Finally, we like to wrap it all up with a discussion on the Industrial Internet, to show us and our viewers how we are able to connect our technology to the Industrial Internet to monitor and predict the life of their assets today.

For those of you who have missed our previous webinar series, we do host them on our website under the Video Library tab. You’re able to view and download the presentation there. For any of you who would like a certain presentation, please feel free to reach out to me. I’d be happy to send you a direct link to those, at any point in time.

Throughout the presentation, we encourage our viewers to insert questions in the Go-to webinar control panel shown on the right hand side of your screen. At the end of the session, I have allotted 10 minutes to go over these questions, and pass them off to our presenters for a question and answer period. If for any reason I’m not able to get through all the questions, which does typically happen, I’d be happy to reach out to you by the end of the week to get all of those questions answered.

I’d just like to pass it over to our presenters at this point in time. We do have two team members presenting today. First Dr. Raja Pulikollu, who is our Director of Implementations and Chief Materials Scientist, will take you through the beginning of the presentation. Jennifer Haggerty, our Implementations Manager and Research Scientist, will take you through the end of the presentation there. So Dr. Raja Pulikollu, will you please take it over?

Dr. Raja Pulikollu: Yeah. Thanks Natalie.

This slide is the quick summary of the pre-capabilities that we are offering as a service to our customers. Prognostics, life extension and Industrial Internet.

We have been applying prognostic technology on analyzing a system or components which are at the design stage or for the systems that have been already installed in service, to create the life of the individual components and also the full system.

Under life extension, when a particular product is not meeting the design specifications, then we work with the customers in extending the life of the product by performing various optimization and sensory analysis, to meet the design requirements.

The third one, which is Industrial Internet, is linking the prognostic tool with the diagnostic tool on the systems that are in service, and do live monitoring of those individual assets to predict remaining useful life, as it’s being operated. For example, our wind turbines on which we have already successfully implemented this technology. Also we are expanding this technology on monitoring to autocrafts and also into automotive industry.

Next slide please. The technology that we’re going to share with you today is mainly focused on linking prognostics with diagnostic tools. This work has been funded for the past 10 years by various government agencies listed on this slide. The main reason why these funding agencies invested or supported this effort by around $22 million is to commercialize the technology, not just pure R&D, but also to successfully commercialize this technology and solve real business problems.

We spent almost eight years in developing this technology, and we started commercializing this technology in the past two years, and we have successfully solved various business problems for various customers which is listed in the following slide.

These are some of our customers who have been working with us in solving their business problems. We have been adding more customers to this list on weekly, monthly and quarterly basis. When you attend our next webinar, you can expect to see more customers being added to this list. But as you can see, from this list, it’s varies from autocraft industry to automotive industry to wind turbine industry. These are the three major areas on which we are currently working on. But most of our customers are wind-based, and the focus of today’s discussion is applying the prognostics and diagnostic technology for wind turbine assets.

This schematic shows the health state of a system or a turbine once it’s entered the service and how the health state changes over time. So x-axis is the operational hours. You can also look at this x-axis scale as, years since turbine’s installed, and the y-axis is the health state of the system.

Point A is, when a wind turbine has been installed for the first time, and it entered into the service, started generating the power. From point A to B, that’s when the damage gets accumulated on the individual components in the system, based on its usage. At Point B, you can expect to see cracks getting initiated in the local components. It could be a gear or it could be a bearing in a gearbox.

Once you have a crack or a damage in a particular component, then you can expect to see a drop in the health state of the system, which is shown from point B to C, which we define as mechanical failure. That means the damage is localized in a particular component, and it’s started affecting the health state of the system.

From point C to D, we define that as operational failure. So it’s point of no return. Once the damage has significantly affected the performance of a local component, then the damage will also affect its neighboring components. For example, a cylindrical bearing on a low speed shaft leading to misalignment on a gear mesh, leading to significant increase in stresses, so bringing down the entire gearbox into a failure state. That’s a good example. So from point C to D, you can see a significant drop in the health state, which we define as operational failure and it’s like imminent failure regime.

The innovation that we are bringing to wind turbine industry is prognostics and also better diagnostics. Using prognostic technology, we can predict the tension from point A, B to C and also to D. Using diagnostic technology linking that to prognostics, we can effectively and more accurately capture the trend from C to D. By combining these two technologies, we can extend the life of a specific effort, based on its operating conditions. Today’s seminar is mainly focused on diagnostics, linking that to prognostics. So we’ll be mainly focusing on the operational failure regime. The previous webinars that we have offered are mainly focused on mechanical failure and prognostic technologies. If you’re interested in knowing more about that, please visit our website, or you can contact Natalie to get that information.

By innovation here is using prognostics and diagnostics for life extension. We are able to do that because we start from crack new creation. We know exactly when the crack is going to new create, how it’s going to behave, and we know exactly what the small crack life and the long crack life.

If the system is written this new creation to small crack regime, then we can do life extension on this particular system by working with the customers. But once it crosses the small crack regime and once it goes to long crack, then it’s going to affect the full system. So the life extensions options will be limited by the time you reach the long crack. The innovation here is to use this prognostics technology and diagnostic technology for life extension, and also as much as we can when it gets into the long crack growth regime.

Today’s webinar we have divided this into two sectors. The first segment is based on our model data fusion which is linking prognostics with diagnostics and we will go through the technical approach and some of the advantages and critical points of differentiation. The second part of this webinar will be covered by Jennifer, mainly focusing on applying this technology in solving real business problems on site, on the wind turbine site.

I’m pretty sure that everyone is familiar with this technology – CBM and PHM. Just a quick note: condition-based maintenance is reactive tool and it can be used for corrective maintenance, whereas prognostic tool is a proactive tool that allows customers to perform preventive maintenance. As everyone knows, if it is more preventive maintenance versus corrective maintenance, then it offers a lot of cost savings and increases the return of investment on your assets.

This slide is a quick summary of the conventional approach, the current state of vibration analysis. There’re three major issues in analyzing vibration data. This is based on information that we gather from the industry.

The first major issue is high up-rent cost. CBM system is offered as an additional package. It doesn’t come with the purchase of a new turbine. It’s offered as an additional package. Also there’re a lot of sensors that being installed on these assets, because we don’t know what are the critical locations and hot spots on the turbines. So a lot of hardware is being installed on these turbines, which is very costly. Whereas with prognostics, we can identify the hot spots and critical components. So we can reduce the number of sensors and also reduce the cost on the CBM system.

The second major issue is, a lot of data is being gathered from these sensors on site, but the tough task in the hand of the operators is how to use this data, and how to interpret this data in analyzing the health state of the system.

The third major issue is, a large amount of time in some cases days and weeks, are dedicated to interpreting the data for monitoring these turbines. There’s a lot of false signals involved in this. These are the three major issues that are being faced by the current operators.

The following slides are focused on the critical points of differentiation of our technology versus the conventional approach. I think Natalie, this is a good time to bring out the first poll question.

Natalie Hills: Great. Thank you, Raja. For the first poll question that I wanted to ask the audience here is, how early can you detect a turbine gearbox failure by using CBM systems? And there’s a variety of actions there: less than 1 month, less than 3 months, less than 6 to 12 months, or most of the time everybody thinks it’s too late. I’ll give everybody a few minutes here to answer that question, and then we’ll proceed with a deep dive into the presentation.

(response to poll question)

All right. Great. It looks like the majority of the audience has answered, and our answers do vary across the board there. I don’t know Raja do you want to make a few comments on that? Is that something that you’re used to seeing?

Dr. Raja Pulikollu: Yes, actually. This correlates well with what we have been hearing from the industry. Highest percentages most of the time it is too late. That’s exactly right. It correlates well with the industry, yes.

Natalie Hills: Great. That’s what I thought. I will now hide the results and move forward.

Dr. Raja Pulikollu: The critical points of differentiation. What is the advantage in linking sensors with prognostics? The first major advantage is better accuracy in predicting the operational failure, specific to each serialized asset.

The second major advantage is using the sensors in a predictive fashion, rather than a diagnostic fashion. That means if you use it in a predictive way, then you can do more preventive maintenance rather than corrective maintenance.

The third major advantage is lower cost. If we know exactly what are the hot spots in a system and what are the critical components, then we can reduce the hardware that getting installed in the turbines and we can also reduce the cost associated with that. Also we can effectively reduce the false alarms.

The following couple of slides are focused on comparing prognostics versus diagnostics. Conventional approach versus Sentient approach based on the work that we have been doing under various government projects. Conventional approach assumes that a system is healthy. There is no fault in the system. And by the time there is a fault in the system, it’s too late as we have seen in the poll, and it becomes more reactive, because there is no physics involved. There’s not much physics involved in the conventional approach. It’s a simplistic approach. It doesn’t account how the system, the gearbox behaves under wind loads scenario or based on one site located in let’s say Texas versus another site located in New York. The effect of a snow storm or the effect of having ice on the blades.

Natalie Hills: Raja, at this point I want to ask one more question to the audience here. I’m going to go ahead and launch that poll. Our question for the audience is how much support do you get from OEMs in using CBM data? On a scale of 1 to 10, 10 meaning excellent and one meaning no support at all. I’ll give everybody a few minutes again to answer that, and then we’ll go ahead and get started with the rest of the presentation.

(response to poll question)

All right. It looks like the majority of the audience has answered in the middle there, meaning they get a little to no support.

Dr. Raja Pulikollu: Thanks Natalie. That’s an interesting poll and outcome too.

This slide is a quick summary of linking sensor data with the prognostic model. Prognostic model takes into consideration the multi-body, dynamic behavior of a gearbox. Diagnostic model takes into consideration the health state of the individual components. So by linking these two, we can accurately predict the stresses and loads on the gearbox based on its operating conditions, the wind loads, and using this information and updating the health state as it’s being operated.

As we mentioned before, one of the services that we offer is Industrial Internet. So this is a part of our Industrial Internet software which we call DigitalClone Live. Where we effectively link the sensor data with the prognostic model and capture the physical behavior of the individual components in predicting the remaining useful life of the system. Also this approach allows us to take into consideration the manufacturing tolerances. Also by linking the onsite sensor data with the prognostics also helps us in reducing the uncertainty involved in predicting the remaining useful life. This is a very innovative approach and that’s the reason why National Science Foundation and other agencies, Department of Energy Health funded this effort, and a lot of operators and OEMs that are taking advantage of this technology.

As we mentioned before, the focus of today’s presentation, webinar, is mainly in capturing the operational failure by linking diagnostics with prognostics. This is how we rank the health state of a turbine, based on its sensor data and prognostic data. We rank a turbine based on four severity levels. Ranking from 1 to 4 – 1 means least severe, 4 means most severe. And we rank them based on the excitation forces, based on the condition indicators that we have been collecting on the turbine, which we then translate that into individual components. Based on this… this is just an example showing the operational remaining useful life based on the sensor flag. If it is less severe, then the life of the system is 12 months, compared to most severe where the life is 1.5 months. We have successfully correlated our predictions with the site data, and that’s the information that Jennifer is going to cover in her slide deck.

Typically we spend on an average 30 minutes to give a quick demo on our product, but due to time limitations, we’re just going to show a couple of screenshots so that you can get a feel for our product.

Industrial Internet which is one of the products that we offer we call as DigitalClone Live, where the customer can log in, and they can get an assessment of all their assets at each of their sites, and they can sort the turbines based on their health state – least severe to most severe or vice versa. They can also focus their attention on the most critical one and work with this tool in terms of extending the life of that particular asset, so that they can increase the return of investment. So there’re a lot of features included in this product. From prognostics to diagnostics and linking this to the return of investment moral, so that it can also be used for life extension of the turbines. Users can access this product either on a cell phone or on an iPad and they get email alerts based on the health state of the system from this product directly, 24 x 7.

Next slide please. This is an example of our user interface, showing how we generate the vibration reports and how we present the data to the customers so that they can interpret that easily. Leaving all the physics behind the screens, presenting the data in a form that can be easily used by the customers on site. In this case, we’re looking at a turbine which is ranked as most critical. And also you can see we have listed the most critical components of this gearbox, and also the condition indicators on the individual components. Also there is a lot of information if the customers [inaudible 00:25:22] on below, showing why this turbine has been ranked red, and what’s the current health state compared to historic health state.

I think this is a good transition in showing few examples on how we use this technology in solving some of the real business problems for our various wind customers. Jennifer will go through that information in the following slides. So Jennifer, please take it over.

Natalie Hills: Raja actually, before we pass it on to Jennifer, I do want to ask the audience one last poll question here. Our last question for the audience is, understanding the costs and benefits, how likely are you to add a diagnostics system to any new turbine asset, very likely or not likely? I’ll give you guys a few minutes again, and then we’ll proceed and pass it over to Jennifer.

(response to poll question)

All right. At this point in time we’re going to close the poll. It looks like a higher majority of the audience are saying, “Very likely.” Thank you guys for your answers there and your input. It’s very appreciated.

Jennifer, at this point in time I’d like you to take it over.

Jennifer Haggerty: Thank you Natalie.

Here we will walk through some real data examples, to applications in the wind industry as to how we’re utilizing the prognostic technology along with diagnostic information from the sensors in this framework of model data fusion as we call it, to assist users to make decisions based on gearbox operational failure – projected timing of that. It also enables decisions regarding when and if an up-tower replacement should be made, and in more thoroughly understanding the effects of operate and derate of your wind turbines.

Here, this is an example where… and I believe everybody’s familiar with speaking about in terms of condition indicators, but here a little bit different from the conditional-based maintenance. Our condition indicators here that come out of this fusion of our diagnostic information from the sensors and our prognostic models, we’re able to look at excitation forces within a gearbox for example, between components and really just reduce the size of data that we’re sifting through and automate this process of identifying jumps in our data which correlate very, very well to failures.

Here this is an example where on the x-axis we’re really looking over a range of time. In this case, the condition indicator that we are looking at is sort of average value of the magnitude of excitation forces. We’re going to see that it’ll correlate with a spalled intermediate pinion tooth. Earlier on, in August of 2014, again we’re trending this live in an automated fashion based on the combination of the prognostics and diagnostics but we can see that our data for our condition indicator falls within a certain range from about 1 to 3, [inaudible 00:29:24] the 4 newtons. Again, data driven based on real data coming from the system. Within our physics-based technology, it’s quite straightforward to… for example when we got to September, we were able to automate the identification of a large jump in excitation forces, which correlated very well with it’s spalled intermediate pinion tooth. So we were able to give more information ahead of time for planning for imminent failure of this gearbox.

You can see after that point in time, the excitation forces were very high, where we built an alert into the software and the gearbox did fail some time in December. So that’s one view of the data. I’ll keep moving forward through a couple of different ways in which we utilize the data behind the scenes.

Okay. Here’s another example. Again, this is a real life example taken from true gearboxes. You see two plots. We’re looking at two different condition indicators, meaning excitation forces at different locations within the gearbox. You can see in an easy way, that it’s easy to sort of visualize the difference between a healthy and a damaged gearbox. Now, there’s a lot of physics and data that’s taken into account and very, very robust algorithms that have been validated and around for a long time, but here on the x-axis we actually collect data and we’re able to look at a healthy gearbox and a damaged gearbox, and their condition indicators at different operational speeds.

On the left and the right you can see that the data for a healthy gearbox over a range of operational shaft speeds is grouped together in one region with lower condition indicator values, lower excitation forces – on the left of the location of the meshing gears. We can also see a damaged gearbox. These are validated regions and values for our condition indicators that we can use, and that we have used successfully in picking out damage that’s going to lead to imminent failure, and also pinpointing the specific component that the user would want to take a look at.

On the right, it is the same sort of plot at the location of the intermediate pinion tooth. To close this slide, it does its job in accurately estimating the health state of the faulted gearbox and a healthy gearbox.

One last note. Again as Raja pointed out here, in an automated fashion, we are able to in a real time way, constantly update our physics-based models to match the current health state of each individual asset. That’s particularly important in understanding and predicting accurately operational failure. All of this information is useful and can be looked at in both a short term and in a longer term.

Thank you, Natalie.

Okay. Here’s an example of two turbines, and one way in which we use the data stemming from our model data fusion, the combo of the diagnostics and prognostics information. Here turbine one, this was a real example. If we go back to our color scheme across the fleet, if we can take a look and any turbine that comes up is red, has been identified based on its excitation forces and other things as at risk of imminent failure. So it’s going to fail in one year or less.

Also you can see I’ve listed the critical components here: the low speed pinions, cylindrical bearing. On these two assets we could pinpoint the problem ahead of time and inform the customer to make a decision regarding replacement of the intermediate bearings.

We can see over time – and these are data driven changes in the current health state of an asset – before replacement of the intermediate shaft bearings, both assets were at risk of imminent failure. After replacement, we can see that the first asset at first shifted to a yellow state, which on our color scheme is a significant overall improvement in gearbox operation. Again this is for turbines that are between point C and D on the graph that Raja showed you. So there is no stopping the failure, but again we’re extending the life here. And we’ve significantly improved the operation of these assets and are giving it much more time to survive at this point, and operate.

It makes sense if you look at from June to July for turbine 1. There will be continuing damage progression, but we’ve squeezed a lot more life out of that asset allowing the user to make that decision to replace certain components. Again, we’re able to pinpoint which components would be beneficial to extending the life; the same with turbine 2. We’ve looked at the critical components, flagged this asset and the customer performed a replacement of the bearings. This greatly improved in operation of the gearbox.

Bottom line here, the combination of our prognostics plus our diagnostic data from our sensors allowed a better understanding of the effects of operate and derate. The user was able to make decisions about megawatts to operate at, and the effects of up-tower replacement.

Okay. This is one more example. The last example of another way of sort of visualizing the data that we glean from this model data fusion, of particular interest for those turbines with a year or less left. Here again, we’re thinking in terms of condition indicators and from model data fusion at this point these are examples where we’re looking at excitation forces, and we have plots for each component. Here we’re showing an example for four components – bull gear, intermediate pinion, downwind main shaft bearing and upwind intermediate shaft bearing.

In April, based on our combination of prognostics/ diagnostics, we flagged all of these components as in imminent danger of failing, which also put the full gearbox at high risk of just full operational failure. These were the results in April. In the next slide you will see the result after up-tower replacement. Again, just another way of looking at the effects.

Here these are the same four components. Again, the x-axis just to clarify that was showing a variation over different operating shaft speeds and again just looking at excitation forces. But when we are in a yellow category or an orange category, we have significantly improved the operation of that gearbox, especially in this case, the bull gear, the upwind and downwind bearings. And it’s pretty typical to have trouble with intermediate pinion, most intermediate pinion.

This gearbox went form a high risk critical turbine, very close to complete imminent failure and has now had a greatly improved operation and extension of life after replacement. This is evident in our data.

Just a quick summary of this combination of the prognostic information and these sensor information. If you already have a condition-based system, prognostics will provide assessments, 1 to 10 years going forward it gives you that look into the future. Also with this technology we can get a global view of an entire fleet. It’s not like traditional condition-based maintenance where one asset comes up as good, they’re good or bad. We get a full look at the fleet and can make decisions along the way to really optimize return on investment and solve your business problems.

If you do not have an existing condition-based maintenance system, we can still use our model data fusion technique, again combining our physics-based models along with, it could be oil or vibration sensor, and we can generate highly accurate solutions in this window of time between remaining useful life of a year or less. Give you time in preventing a full operational failure and extending that amount of time.

Last thing I would want to highlight really is that by adding these prognostics and getting this look into the future across the whole fleet, we’re also able to, because of the use of the physics-based model and bringing that in, we can really pinpoint at which components are in danger, assess the current health state of each individual asset, better serialize a model of an asset to match the current health state of that asset and that improves the operational failure projections and predications. We can react sooner instead of just waiting for imminent failure and flagging an asset as good or bad.

With that said, I guess we will open up to questions.

Natalie Hills: Thank you. At this point in time I want to open the floor to questions. If anybody has any further questions that they weren’t able to input during the presentation, please do that now. I did get a lot throughout the session.

The first question I’m going to start with here is, how accurate is model data fusion in predicting operational failure life? How close can you get to the failure date? Jennifer, do you want to go ahead and answer that one?

Jennifer Haggerty: Yes. Hi. The answer would be we can get within +/- 45 days, based on our correlations with the field failure data.

Natalie Hills: Great. Thank you, Jennifer. Raja, anything you want to add to that or can I move forward to the next question?

Dr. Raja Pulikollu: Yap. You can, please. Please move forward.

Natalie Hills: Great. Thank you. The second question is, is prognostic and sensor approach specific to one platform such as GE Vestas or Seamen or is it a generic model?

Jennifer Haggerty: I would say it is absolutely generic in that it’s physics-based. So we can apply it to a variety of different platforms. It’s ready to go for a variety of different platforms.

Natalie Hills: Great. Thank you, Jennifer.

The next question I have here from the audience is, can you use existing CBM system along with your prognostics and sensor model?

Jennifer Haggerty: Yes. Yes. By applying the model data fusion technology, we can process and utilize data coming directly from your current condition-based maintenance system, if you wish. Yes, that’s one way in which we can do it.

Natalie Hills: Great. Thank you. The next question I have from the audience is what are the turbines that you modeled and implemented this technology to-date? So if you could just take them through some of the turbines we’ve applied our technology to, Jennifer, I think that would be great.

Jennifer Haggerty: Sure. We have modeled and implemented this technology on the Clipper 2.5 MW, GE 1.5 MW. That work is complete, and we’re working on expanding it to Siemens 1.3 MW and Mitsubishi 2.4 MW.

Natalie Hills: Great. Thank you, Jennifer.

The next question I have from the audience is, what is the implementation project completion time? So maybe just go into detail on how long the process is from start to finish and if it changes depending on what model it is.

Jennifer Haggerty: Okay. For turbines that are already in our database, such as the Clipper 2.5, GE 1.5, takes about 60 days. For newer turbines, about 90 days, maybe more dependent on what data is provided by the customers. I don’t know Raja, did you want to add anything to that?

Dr. Raja Pulikollu: I think you covered pretty much everything.

Jennifer Haggerty: Okay.

Natalie Hills: Wonderful. Looks like we have time for about one more question, and I have here from the audience is what is the uncertainty in the prediction model, if there is any? Raja, do you want to go ahead and take that one?

Dr. Raja Pulikollu: Yes. As I mentioned before, uncertainty methods and prognostic methods are a part of this prognostics and diagnostic model. In terms of the uncertainty involved, that comes from the way a specific turbine is being operated, and also it does not include the load, the power output, the operating temperatures of the oil. Also uncertainty is involved in the manufacturing of the parts. That’s also taken into account. Also uncertainty involved in terms of how the products were finished and installed. The [inaudible 00:46:46] and all those… the assembly process. Uncertainties involved in these various steps are specifically modeled and taken into consideration in this remaining life prediction.

Back to you, Natalie.

Jennifer Haggerty: Oh, yes. I would only just add one more thing. We’re also including models of uncertainties from our actual sensors themselves. We include stochastic information there and can show in a very deep dive of the technology that we can prove the reduction in the amount of uncertainty in our predictions.

Natalie Hills: Great. Thank you, Jennifer. Also to add to that, we did do a webinar quite a few months ago on a deep technological discussion on our six-step process to our technology in which our CTO presented, Nathan Bolander. That is available on our website, so you’re able to view those PowerPoint slides and listen to that presentation as well. If anybody wants to see that, you can reach out to me and I’ll send you the direct link.

At this point in time, I’m going to have to close the question period as we’re running out of time. I do have a lot more questions. So I’ll make sure to follow up with everybody before the end of this week. Again I wanted to reiterate that we do host one webinar per month, three per quarter and they are all hosted on our website under video recordings, and you’re able to listen and download the slides to better follow along. So please feel free to reach out to me, I’d be happy to pass you on any of those resources. We have a variety of whitepapers and case studies as well, and our team is always here to help if you have any further questions.

Our next presentation is on November 12th, and it’s entitled Using DigitalClone Live to Investigate the Impacts of Operating or Derating on the GE 1.5 Gearboxes. I believe Dr. Raja Pulikollu will be again presenting for that session. So please everyone, we welcome you to come back and join us, and if you’re not able to make it at that time, it will be on our website as well.

Thank you all for joining us. Thank you, Jennifer and thank you, Raja.