# Elevating Wind Turbine Inspections with Automated Drones and AI-based defect detection

- Event: NestGen ’24
- Published: April 4, 2024
- Length: 37:31
- Watch: https://nestgen.org/library/watch/TgOEuNRIFNM/

> The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's
> captions; automatic captions can mishear names and terms.

## About this session

Join SkyVisor's CEO, Paul Fontaine, as he explores the future of wind turbine inspections using autonomous drones and AI. Discover insights into optimizing renewable energy operations with efficient data collection and defect detection. Learn how SkyVisor empowers operators to enhance asset performance while reducing downtime. Gain valuable knowledge on the challenges and opportunities in the renewable energy sector. Dive deep into the world of drone-based industrial inspections and the importance of leveraging cutting-edge technology for maximum efficiency and reliability.

To know more, visit:

=====
Subscribe to FlytBase YouTube -

Follow Us On -

LinkedIn -
Facebook -
Twitter -
Instagram -

## Transcript

**[0:20]** hello hello and welcome to another hugely important nesen 24 session I am Rohan currently working as a customer success associate here at FlytBase and I'm delighted to welcome you all to this session how are you finding the session so far please let us know in the chat section and do share your feedback at the Gen wall of tab it will definitely help us curate the sessions better in the upcoming Edition as a dependency on renewable energy sources are growing the need for efficient inspections of say wind turbines solar farms and other assets become critical for mankind to shed some light on such crucial use case at the session we are going to Deep dive into elevating wind turbine inspections with automatic drones and AI based defect detection to go into greater detail we have a very special guest Among Us so let's welcome Paul fontine founder and CEO of skyer sky viser is a drone-based autonomous

**[1:20]** industrial inspection software Seed Company based out of France that empow wind and solar farm operators to optimize their operations and maintenance by in housing their inspections it features automatic data acquisition and AI based processing to maximize asset performances while reducing down times Paul's background is an electrical engineering and computer science with a particular focus on drone design Cloud development and big data management before funding Sky Weiser Paul had been designing and deploying industry Focus drones for over eight years leading to the creation of skyriser which is currently operating in a total total of 17 countries worldwide so before I let Paul start sharing his insights I would like to share a few housekeeping rules so please type in your questions in the Q&A tab of the session and it will be answered at the end of the presentation the session is getting recorded and will be published soon we also have a sky Weiser booth in the

**[2:23]** Expo section so please make sure you do visit it during the networking and Booth visit slots there's also an online poll during the session so please do participate finally please do remember to capture your favorite moments of the session and share it over LinkedIn or other social media handles by using # neson few lucky people who share might also have a surprise in store last but not the least I would like to extend my gratitude to our title sponsor DJI Enterprise and sunflower Labs which is also our gold sponsor for the invalu able support throughout our journey and with that over to you Paul so thank you for this intro thanks hi everyone uh and pul I'm the funer and CEO of Sky visor uh as one mentioned we are software company focusing on our Renewables uh inspections and today's sessions we're goingon to try to focus onto the wind turbine or inspections and how the data can be collected can be better collected

**[3:25]** using uh autonomous drones and then what are the different challenges uh regarding um G detections data processing and how AI can actually works um so I'll try to make this session as insightful as possible I know you guys might not be always familiar with what are the different U Renewables uh type of inspections and especially for win ones and we'll try to share as much knowledge as we can today with you so please if you have any questions we'll be very happy to answer them straight after the session into the Q&A or uh have a look at our book Booth virtual Booth uh onto the the event all right so just a quick intro regarding the the company so um we do our um software specialized companies and we're specializing only on to three different uh products uh I will say the first one is the wi turbine inspections uh it's going to be the focus of today's session and but we do also have two other uh solutions that are dedicated for solar plants inspections one of them

**[4:26]** uh featuring uh thermal inspections um capabilities and the other one doing construction monitoring for PV plants um what we do as skyer is um software once again we are uh specializing into the collection but also the processing and I'll try to give you a quick Insight of course of what our products looks like but I'll rather as well share some knowledge about the industry itself uh we've got some experience six G gigawatts of wind and solar asset inspected worldwide and the company has been operating for the six last years and we be are pleasure to share all these insights we got from this um years of experience uh basically Solutions as you may already see on the right hand side is features and um divided into two parts one of them being automatic drone flights which are compatibl with DJI drones and DJ docks and the second one is uh data processing data management which is definitely focused for the end users which in our case today are Renewables

**[5:27]** operators so uh going to jump right to it this is a vis track record we do have in the 17 markets as ran may have mentioned just before and a couple of logos that you guys may know U just are always operators and energy operators operating either wind turbines or other PV plants so just maybe have a quick look at the Renewables uh Market on the right hand side just here you can see um two insightful Maps one of them featuring the wind turbines uh being installed worldwide and the uity scale photovoltaic plants as well um so as of today 2024 uh we are actually uh having 450,000 turbines in St worldwide that's quite a big amount and quite a big growth over the last years and in term of PV plants we've been reaching the terawatts of uh installations which is just to give you um a quick um quick comparision the size of Tunisia so imagine Tunisia being entirely filled with solar panels um the the

**[6:32]** context as you may all have seen onto the news is very favorable um a lot of turbines are being installled more and more and right now we are um on a key phase in 2024 because 32,000 turbines will be actually um age of 20 years uh and More in 2028 so which means there is a bunch of turbines to inspect there is more and more wear and tear over those turbines and when you have a quick understanding and quick look at um a normal Renewables uh life cycle um you have to understand that the part operation maintenance which is actually when the turbines or the PB plan is actually running it lasts for 25 years in average so there is more and more maintenance more and more inspection to be led over the years and having in mind that all those turbines are quite old there is quite a big challenge regarding the number of inspection that should be carried out and it's exactly where drones uh take uh into place second very interesting factor is the drop of energy guaranted

**[7:34]** price um I don't know if you guys know but uh in order to start this business start the uh renewable business a lot of um governments have been uh actually subvention um and helping those Industries to to start and grow but now that it's becoming more and more mature the price uh the guaranteed price is dropping to make it more um I would say um comparable to coil uh price industry price simply and so this forces basically our clients which are the operators to focus more and more on to the operation men and site so just to give you um a quick overview there's a ton of components in win turbines and PV plants and the one we're gonna focus on today are the blades uh as you may know win turbines are composed of three blades and those blades should be inspected once to twice per year depending on different markets and different of course wear and tear and because a successful drone in a box deployment always comes with a successful as well business plan uh it's

**[8:35]** interesting to know that a blood inspection today is priced around $700 as well so uh quick question for you guys just before kicking on this session it's um what do you think is the daily production loss when a turbine is actually uh done when we stop a turbine based on a 4.2 megawatt turbine so there is the Q&A on the chat um and you may answer either $200 per day a loss of $1,000 per day ,200 per day or $2,000 per day I'm going to give you a couple of seconds to get some uh some answers yeah Paul just a minute the Q&A hasn't started just a second I will turn on the poll so guys you'll be having 30 seconds before you answer the polls so yes

**[9:38]** so yes so the poll has started so you can answer it in by navigating to the poll stab uh when you navigate to the session tab there you can see three toggles between chat polls and Q&A so you need to go to the poll section wherein you can see this question down there so you all have 30 seconds to answer and I will be closing down this poll after afterwards so 0% people say like $200 per day 5.3% said $1,000 per day 52% of them said $1,200 per day and 40% said $2,000 per day so the maximum of the votes have gone for option number three that is op C $1,200 okay let's see let's see together uh what the answer is indeed indeed answer C is the correct answer congratul to the ones with the good one um it's definitely large amount of money that you lose whenever a wind turbine is actually uh being

**[10:39]** stopped and it's um exactly where um the kind of solutions provided by drones is very relevant because as you may have seen number of inspections we should do um we cannot afford uh getting the turbines down turban can be done for inspection purpose but if inspections has been uh poorly carried out there might be as well some corrective menance happening there because of a blade failure that leads to actually uh a very important defects afterwards so when when it comes to Blade uh blad defects uh I wanted to share with you guys um a couple of insights regarding um how blades are made actually and what are the main main defects you can find of it on it so blade is um really made of composite material most common one is being fiberglass reinforced epoxy and as you may have seen on the right hand side there is a bunch of different uh measures and metrics that can be found out of blades here you have a a cut section of of a blade uh you see it looks very similar to actually a a plain

**[11:45]** Wing uh and it's basically doing the same um the same job except that instead of of course lifting a plate a plane it's actually rotating so there's a couple of of technical uh keywords to understand very important importantly one being of course the size the different size you can find onto the onto a blade so on this on this section you can find what we call the Leading Edge which is the edge that faces the wind first in in a certain way and then on the other side you do have the training Edge a composite blade is usually composed of two different section um and one of them and both of them being glued uh onto those different edges as you can see here in the dark areas so this is definitely um a key point to actually um understand and to inspect because if the we alter is too important into this section then it can lead to um a shell separations basically and the blade can open itself

**[12:46]** as you may see right here you can see perfectly two different shells being open into what we call the leading en version so uh basically a blade uh has four mon four actually uh very common defects uh one of them being the erosion of the Leading Edge so once again is this area that gets eroded one of them is the erosion of the Leading Edge tape which is actually a protection you can put onto the Leading Edge uh to reduce this we and tear of of theing edge but that actually get used over time as well so you need to change it and change it before it actually gets transform into Leading Edge erosion issue and even worse into an opening uh Shelf issue third uh defects that are quite common on blades you have a bunch of different addons what we call which are additional components that you can put onto blades uh to make sure the air flow is actually as smooth as possible and those missing addons can actually

**[13:46]** uh be damaged as well and create noise or worse uh create vibrations for the blades on the blade and then uh last but not least one of the very common defects is what we call the lining strikes defects so so as you may know wind turbines are very tall structures plates can measure up to 131 mters when it comes to Offshore turbines and of course when a lightning strikes occurs it's very likely going to go through the blade so of course blades are equipped with a specific system that can fail sometimes to time and the strike can actually uh strike onto the composite part of the blade so those are the very uh common defects of course there is very common inp methods turbines as you may have seen they have been over on the market for the last 20 years and um inspections are already being carried out um one of the very Comm common inspection method is human contact inspections which are basically Rob access uh technicians inspectors getting down onto the blades to making sure they are

**[14:48]** they're doing a visual inspections of blades second type of inspections you can find in the market those are camera based inspections so basically um technicians from the ground with very large object jective very high quality camera trying to collect um very good data which is quite hard because you do have to change the rotal angle quite often and then the last but not least ones is the Drone B inspection ones that's one we're going to focus on today of course uh with this event that is very interesting just a quick overview I think it's it's quite impressive one of our U customers actually that does as well R uh inspections and you can see how dangerous it can be uh when you actually get down onto a blade basically to carry out these inspections uh you need as you may see um the rotor to be locked into a very specific position with one blade facing down and then the inspector can go up into the top of the tower uh with the help of an elevator most of the times hopefully for

**[15:51]** him and then he gets down on ropes and then has to switch in between every single side of the blade to make sure he has a very good coverage uh and a good inspection so as you may see it's quite dangerous uh quite long as well to carry out with you have to imagine that such inspections are actually taking about four hours uh to inspect the three Blades of one turbine when drone inspections are way faster so uh Jones because definitely topic of uh of today's today's meeting today's conference um you guys are very familiar with Jones I'm sure you know how beneficial it can be and how efficient it can be on site well here I wanted to maybe emphasize on three different points uh one is of course the speed of uh drone inspections uh compared to the other traditional methods uh how cost effective it can be uh when you understand um the downtime and the power loss uh that leads to a turbine

**[16:52]** downtime then you understand that having a fast inspection is very important and it's a key point to face um the coming market and then the last but not least uh benefits is definitely the quality of data you can achieve with such drones uh there's been large uh improvements into drone quality and drone camera quality over the last years but those drones I mean we are still very much into the manual uh inspections and manual inspections uh is facing actually a lot of limits when it comes to inspecting a large three-dimensional structure like uh wind turbines one of them is actually having a very uh close and precise flight uh usually um when you want to get gather very high definition pictures you need to be very close to blades and when we're talking about very close we're talking about four meters away from a blade and I'm sure no drone Pilots wants to bring a two um $20,000 drones four meters away from a blade um mostly when it comes to getting underneath blades

**[17:55]** as well there is very strong perspective point of view When You Reach 200 meters of height and so it's quite risky it's where automation really um adds a adds up a lot of value the other point is definitely uniformity of pictures I just showed you a quick example down there uh you see the three pictures they look very similar but in fact those are three different area of the blade indeed blades are large white structures and it's very hard to actually localize yourself onto a blade and it's very hard for manual flight manual Pilots to actually make sure there is a uniform data collection so it's something quite important as well to take into consideration if you want to do a fast and effective inspection and last but not least it's the limiting knowledge of about blade defects as you may have seen in the previous slides uh blades is quite uh challenging and um difficult component to inspect you have to understand um all the um techniques that

**[18:57]** is be behind Beyond uh blade constructions including composite material knowledge and it's something that companies like sky is bringing as well on the market so quick question for you guys again um how much time do you think it takes to inspect one full turbine using an automated drone okay not a manual one but an automated drone using a solution like Sky Bas four possible answers uh first one is 80 minutes second 25 minutes third one minutes and last one 37 minutes so guys who are joining in now uh the question are published in the poll section so you can navigate to The Sessions Tab and can go to the poll section wherein you can answer your questions so we'll be having one minute so yeah go for it nine vots have gone for 18 minutes again nine have been given to 25 minutes Four vs for 30 minutes and two vs for 37 minutes so so A and B they both have got the

**[20:00]** same vote percentage okay okay so the correct answer is actually 18 minutes all right um indeed indeed Solutions automated Solutions like the one brought by scav on the market enables very fast inspections and with total uniform um data acquisition which is very very important on the market um when it comes to Modern data collection I just wanted to have a quick word and a quick clarification on it um of of course you guys know about digital twins it's a quite trendy word and um digital twin is actually digital representation of a component or an asset in this case those are win turbines and especially composite blades that are being digitalized uh thanks to the the skyer solution and um when it comes to digital twin something that is key is to get I would say continuous data flow of inspections because a lot of times you do have very high quality inspections but if you do them only every five years or every 10 years it doesn't make much sense when the defects

**[21:04]** can basically appear the day after your inspection so that's why sky visor we really focusing on to deploying drone at scale and make sure this data acquisition is as cheap as possible and it needs to be very well organized and ordered so that's why we've divided into four different steps one of them first one being planning of course of inspection second one being the data acquisition third one being analysis of those data and last but not least and something um John um JN specialist should not forget is that once this data has been acquired and processed it should lead to an industry specific decision making all right and it when it's actually where it comes to maybe leading to a new maintenance strategy or maintenance campaign or simply the the End customer can decide not to inspect and go back to an inspection B later on that can be led maybe six months after to actually closely monitor the defect propagation so quick Overlook of the solution uh and how it's divided it's actually divided two into two parts one of them is really focusing on to the acquisition making sure we

**[22:07]** have uniform data um using DJI compatible drones talks as well and the other part is being a win PL web platform where actually can gather all those data process them and then collaborate and take decision those two softwares are really focused for wind turbines and PD plants operators could over look maybe of the mobile app just to so that you guys maybe project yourself into a a daily usage of such solution uh we're talking about dedicated app that you install directly onto the Drone controller or you can as well install onto DJI dog using FlytBase uh interface as well and then it leads to a custom 3D flight we're talking about four meters uh flight we have one picture that is taken every two meters and the four sides of the blades are covered so it's very very important to get always the same type of data acquisition because we're going to see just after how important it is to monitor uh defect propagation of course

**[23:10]** data quality is key and something that is quite important is once you're on site you can actually do also live defect annotations which means if you see anything on the blade if you see any defects instead of sending back the data waiting for the process and then uh getting back on site to uh conduct the reps you can actually on site already label a couple of pictures and mention that there is a strong interest into checking those pictures quite fast after the inspection because they might lead to an important defect and in the worst case we're GNA stop the turbine straight after the inspection all right um something quite important to the uh data collection is that data is correctly ordered um you know blades have three BL have three turbines have three blades and it's important to make make sure you know on which of those blades the defect is so usually they are named ABC it's important as well to know on which side the defect is so we've been seeing together the leading gauge the trading ugge and the two other sides there are the suction side and the

**[24:12]** pressure side after that once if you want to send any maintenance team on on onto the blades they need to know how far it is from what we call the blade root distance so from the blade base and how far is the defect located which is something we measure with skyer and then to actually conduct a good defect propagation monitoring it's key to know the defect size in centimeters most of the cases all right so here's a couple of shots of a very nice defects taken by our solution um on the left hand side you can of course recognize an onshore turbine um this there we about 200 meters height andone is four meters away from from the blade and there you can see some Leading Edge erosions already you can see the layers underneath what we call the gel coat all right and you can see some fibers which is a bit more brownish on the right hand side there we see um an offsh wi turbine and then you can recognize uh leading Ed erosion there might be a first layer Pro

**[25:15]** protection like Leading Edge tape that you can see is actually getting out um once you've collected all those data um the question is how you to process them because you have now cheap inspections uh maybe it's going to lead to more frequent inspection as well thanks to docks thanks to inh housing as well and then you end up with more than 360 pictures per blade inspection so it can represent quite a lot of data to process quite some high definition data as well just to give you an idea but the GISD so brand sampling density is actually a 5 mimet per pixel okay so those are very very large data and pictures and so the challenge is first how to organize them as we may have seen nothing is more similar to a bled to a bled picture than another bled picture maybe a couple of meters away so that's important to really wisely organize them um wisely look at them as well to make sure you can find um the correct defect and you need a bit of knowledge for that so it takes time

**[26:18]** and expertise and then once again we we talked about how important it was to see how fast a defect will propagate and there you see good example of um six different inspections uh on the same and thanks to skyer Solutions all those inspections have been done the same way the exact same way even though it was different Pilots different hardware and then you can wisely monitor this kind of crack okay which is a spoiler induced crack so where where comes of course the second piece uh very important data processing platform um on our side and SC is providing a side software as a service solution which is available worldwide of course that does input a lot of different um data so of course we've been talking about blade so far you can also input some tower inspection some internal inspections as well so here you have some screenshot of the platform um you can see some inspections taken from the inside part of the blade um some pictures taken from the outside as well and when it comes to doing a

**[27:21]** menance on side is quite important to get a tablet or phone access once you're on those dropes uh you don't want to carry a computer so that's why we do have as well uh a mobile phone versions of that uh what we bring the industry on this side is definitely some best practices uh in term of data organization in term of defect labeling uniformity and then you have very strong uh EMP powerful PDF export uh options onto the platform and you can of course as well check the propagation of those defects something important as well on this platform and something quite trendy of course uh over the last year it's the so-called artificial intelligence defect detections I just wanted to take the occasion of this uh of this show to uh maybe precise a little bit about AI the AI word um maybe some of you know but it's important to understand that machine learning is actually part of the AI uh universe and then the next step is deep learning which is uh neural networks training and it's what we are

**[28:24]** using here at skyer which is a very specific part uh of artificial intelligence um models we actually customizing them to client requirements depending on client's data quality clients uh defect labels and there you have a very high level um defect pipeline when we come we start with the raw blade picture and there's quite lot of challenges around this picture is high definition it has uh very different backgrounds which can be the sky which can be land which can be fog sometimes so there's the first Big Challenge which is to to isolate the component that we want to inspect which is the bed in this case and then how make sure to make sure that we find correct defects onto these blades so this is a pipeline is of course uh within our Solutions we're going to disclose much more informations uh as you may understand but it's um definitely something important it helps usually our customers to reduce by 60% the time they spend onto blade

**[29:24]** analysis I'm reach I'm I'm I'm arriving almost to the end just here a couple of shots of automatic defect detections of very very small and uh small defects you can't even see here on the on the picture and uh yes I I think what's what's best than actually getting a a code from one of our clients um Jose uh which is one of our one of our Pilots overseas um has been um actually uh testifying about our Solutions and uh that we actually meet requirements and also we do have a maintenance

**[30:26]** manager so that's definitely um of course Maxim's uh testifying of the use of of the company of the software and just I see the time is running we are a little bit out of timing just to have a very fast conclusion Al together um key takeaways regarding U blade um inspections is definitely the usage of drone is growing you may have no into the different locations your guys working from um there's some key challenges regarding data collection as well as data processing and it's I believe it's a great opportunity for all the Drone actors to actually focus on this uh onto this Market

**[31:29]** um and of course we're here for you to to help you on getting this industry best practices uh to work for both Winter by inspections as well as thermal inspections but also uh PV plant construction monitoring using different type of hardware and there is bright future for all those drone inspections uh I would say uh mostly onto large Farms where dring the Box technology can be definitely useful and um mostly for offshore inspections as well you do have usually 50 60 turbines on the same location and then you don't want to bring a technician or Pilot offshore can be some very changing conditions so there's a a huge amount of work to do on this this direction and I'm sure you uh uh you guys can actually help onto deploying such solution thank you very much for for listening and hope to uh yes talk to you soon uh and don't ex to visit our booth and now I think ran we do have a Q&A session righto yes yes absolutely thank you so much Paul for sharing your insights it's

**[32:32]** quite like interesting to see how the technology has progressed so far and with better Reliable Hardware as well as optimized software coming in like just the future is not very far I'm sure it's definitely it will definitely benefit the audience looking forward for an autonomous inspection software for wi turbine inspection and now we would like to open the house for a round of Q&A so I guess we have receive few Q&A questions and I will quickly check that if not uh people who are joined in uh you can like put out your questions on the Q&A Tab and yeah so Paul uh we have like four to five different questions so I will just read out those for you so one was that how do you tackle the issue of insurance coverage for such missions given the assets being inspected are very expensive so it would be great if you can throw some light on that just it's a it's it's a big challenge for any inspection technology um drones contact inspections like like R access um today uh Insurance are used to join

**[33:37]** inspections and they are as well very aware of the limits as well of such inspections being drawns obing rope access because here we are talking about the same category of inspections which are visual inspections visual remote inspections and uh so it's part of the inspection procedure it can be complemented with all the type of inspections as we mentioned in side bed inspection that's why it's to us to actually um embed all those data into this very same platform um so this is how the the issue is tacken into the industry um insr are present on one side but they are not as strong as in other Industries neither and operators are very aware of this absolutely thank you so much for answering that we have one more question that stat say I'm using some other software for flight operations since I'm accustomed to its UI however I would like to use skyer software for post processing so like is it possible to export the mission plan in Sky visor to the software that I'm already using yes definitely so software is um is a generic platform um to gather all type

**[34:41]** of data so we've been talking a lot about drone data but we can actually input internal data Rover data R Access Data into it so it can be other drone providers data as well we do have a we do support a couple of providers already into the import system um something that is worth mentioning is that we've been developing a dedicated mobile application because data quality is key into this into this procedure and so it's important that your data is actually meeting our requirements uh regarding metadata Pro provision as well as GSD requirements so it's definitely something we can talk about and see a bit the the data set you guys have so if you just want to contact us uh right after the session we be pleasure discussing about that definitely so one more question was that what is the exact percentage of the average return on investment that you have observed across this industry Yeah so basically the wind turbine and and rotor blades uh defects are actually leading in

**[35:42]** average to a 4% um annual power loss and it's something that has been demonstrated by a couple of studies um and so here uh when it comes to u to Sky visis and drone frequent inspections uh more generically um you can't realize uh how much operators and turbines owners don't know about blades for them it's a three big black boxes uh rotating and producing energy and they are not aware of all the defects so one very important part and challenge that this kind of inspections is um meeting is actually to bring back the knowledge to The Operators that they can challenge uh people we call the manufacturers manufacturers are the win turbines original equipment manufacturers uh basically and they have a total control over those assets so it's important for them for operators which are the end clients to understand what is the real health status of those turbines afterwards to challenge those contracts and then they can see up to 4% of um benefits okay thank you so much Paul for answering

**[36:44]** all the questions and would you like to share any more clo like would you like to share any closing thoughts no well very um well thank you for of course for the invitation uh looking forward on uh on deploying drone at scale using autonomous drone boxes and we be pleasure yes you guys having us at at our virtual boost F now um to actually have a deeper look at your use case and see how this kind of Technology can help you deploy drone at scale into your respective businesses either wind or either photo voltes thank you so much for that so thank you everyone for joining in from different parts of the globe and taking out your precious time and hope you are having a great time at as let us know how you find out the session in the comments