 Thank you Okay, just just to start. I would like to present a good group as the company, okay, because maybe most of you are aware about the Airbus are the commercial brand that we used to for for the commercial edition, but We have three different divisions. We have the the commercial one when we develop and manufacture the commercial aircrafts We have the helicopters division when we also develop manufacture helicopters for civil and military aircrafts and military world and We have the defense on the space Division where me and and francesca belongs Where we have different business one of them is related with the military military aircraft design and manufacture but also we have another business like communications and surveillance intelligence and the UAP's so you can imagine the big company we are so just to show a Vision of what we have in defense and space about the future of the of the military aircrafts. Just let me show a video of the what can be the vision of the Of the future on the on this world You have seen we imagine the future as a big number of aircrafts flying connected and uses big data and data as the new oil for for giving value to our customers So we belong to flight test flight test is the organization responsible to perform the flight test and to the developed the development flight test and the certification test Flight test is not a new business is something that we have done from the beginning you can see in the picture that the The the right brothers just did the first flight test Probably with no big data because at the moment didn't exist but with the data that they could they could have in that moment so just very few information probably no No computers needed to perform this test and using the methodology of a try and error, okay but the things has changed and We have evolved aircrafts has evolve a lot They are able to report a lot of information coming from the avionics Bases also for the instrumentation of the sensors and We are able also to follow up the flight test by by telemetry means so let me show you another Sort the video just to explain what is our business just to centralize the the topic in the flight test This video is from the telemetry center in Toulouse. We have in Spain We have two more we have one in Madrid in Hetafe and the other one is located in in Seville I just you can see in this slide what we we want to to say to you is that we collect Big amount data. Okay, since the beginning the data has grown a lot and today We are we are in 2018 close to a store By flight close to 200 gigabytes per flight Also We have also more capability to record data new engines are able to us to record A lot of different parameters with a high frequency So you can imagine that the big data is something that we have in our daily basics And what we have to face today in in in the flight test business is how to adopt the big data technologies in in the flight test business So this is a summary of the data that we manage today in the in the flight test center for for Spain We store close to four petabytes of information This information is raw data. That means that it's not completely readable by the by the user by the analyst The data has to be converted to engineering values. That means that say if we convert We we do this conversion. We call multiply by 10 the the factor of the of the volume of data Another characteristic that we have in the in this business is that we record almost everything in 10 series so 10 series is something essential in our business and We are not instrument sensor at a very high frequency that in some cases can can exceed the 10,000 Earths So you can imagine that not only matching learning techniques are needed But also matching learning with the with the frequency analysis are needed in this business Just to to focus in in some program like the 400m we we made Sometime ago a calculation of the number of points that we we could collect in a in a database And the number is almost not affordable by the standard Database So as I said before time series is essential for us we need to to take these time series and to Offer to the user in the most efficient way So this is the reason because databases for us is one of the of the key components in our architecture and we have defined it as a set of requirements that are needed to cover this This this business the first one is that we need the scabily escala scalability That means that just with a single node probably we will not be able to serve this amount of data We need the interaction with with other Ecosystems like a Azure like a typical spot fire or other tools as we are been using by the by the analysts Also something essential is the disk usage. That means that the I will explain later, but because some constraints we need to store it to store it's the information on on premises Another difference between the IO Eterio IO ET wall is that the normally we perform a flight test We record information in a in a media We process the data and this data is Is just for for reading normally the data coming from a flight is is never never is going to be modified probably in bridge By by more information, but never modified This this means that they will use a paradigms more close to the right ones and read many that the IOT wall can use And the other characteristics even we have a time series We have all other information for example the flight of report Information that is not well structured. So for that's purpose. We are looking for a Databases that can cover the three walls the relational The relational standard one the no SQL and the time series in in one come in one solution and the the the last Sensual requirement is to have performance performance means that today we are based in In a in home solutions We are happy with these solutions and these solutions are very good in performance So if we go to a big database, we have to take into account this this this Okay, so in that slide I saw some databases that we are evaluating Some of them has some of requirements cover some of them not I'm not going to enter in details, but just to To to give to the audience what we are doing in the in the wall of the of the database analysis The other characteristics we are in the defense business defense business has some special constraints I have divided in three. Okay, we have first of all we have a export control Regulations that implies that some equipment some system that we have to develop are Under these regulations that implies that we cannot give information to the specific nations or our bodies The other is the intellectual property again We have to protect the intellectual property of our products And the last one is the privacy that may be common with the with the rest of the business When also we have to take care about the protection of the of the individual data That means that when we go to the to the airbus wall because we have different Different activities and we have gone to different solutions That means for example of the airbus group level the the platform that has been chosen is based on Amazon services But if we go to a to a defense business Because we can we are not allowed to to send the data outside of Premises we have to build Our own solutions that means that the company has analyzed the different alternatives and what we found we have found is that the VK data cloud solutions are not covering our our requirements So today the company at the fence of the space level has decided to build a digital platform based on a through the stack and Premises infrastructure This is the the the current solution we have Okay, just to explain a bit we will record the information for data We're stored in what we call the the primary Storax we are storing in raw data and we do a data processing where we convert this primary data to a Proprietary format that is very efficient If we if we have taking into account that we are managing time series We have another another Protocol that is also developing house that is called data server that is the responsible to serve to the next ledger The data to the analysis tools and the analysis tool has a connector to this service to Read the the data and also to feed a new data to this to the primary data storage On top of that what we have in the right side is the data set goals when we can connect to the To the no SQL databases SQL databases son exportation son a specific exportation for the for the design offices and so on and On the thought of the right what we have is the different tools that we have developed for the for the Analysis no so as you can see most of the tool has been developed by by yourself We are happy with this solution because this solution is very efficient to be very efficient in place that we can analyze in a very fast way the the different flight tests and At the end we can deliver a product Sooner with this solution, but what what we are looking here is Why not to adopt some big data technologies that we consider that are enough mature in this In this architect to to integrate in a in a new more efficient paradigm So we have decided to to develop a Plan for that tries to combine or in service tools with a new plan for as I mentioned because Azure Stack is just the Solution that the big companies can can provide us we have decided to evaluate it as we're stuck so the proof on concept that we are doing today is to get the batch processing the Activities that we are doing in or in service Architecture and to integrate in the Azure Stack Environment this will produce some advantages for example to give an astartian laser to the user The user will not need to know where is executing some some bats. Also, we will offer a more efficient Computational platform so the idea is to offer a Nest layer for the for the user just to avoid the hardest that they have when they have to execute the The the batch processing in some in in some computers What happened? Azure Stack is is a very new platform as As you may know the the Azure Stack is sell as I read platform and what we have found is that so an essential service or at least the service that we consider essentials are not Still migrated to the private platform So for example, I thought about services that is one of the key services that we would like to use It's not as still migrated to the to the Azure Stack. It's just We can we can use but we have to go to the to the cloud solution So this is more or less the what we are what we are trying to do, okay again We want to combine our existing solution with the new paradigm of a big data But again, we are in the face to test to evaluate and to see what can be adopted and what Okay Just to to finish my my speech later Francesca will will so as some use cases that we are developing in flight test I would like to summarize the takeaways. The first one is that flight test is one of the customers were More data we are collecting, okay For us is very very important to to to manage in the efficient way the time series is time series is essential for us Not only time series, but also to manage time series in some domains like the machine learning and also the frequency domain The second is we have some constraints because our business where we are not allowed to use a cloud solution so we we have to develop our own solutions and The big big data solution have to know Doesn't cover very well this this need And the the last thing is we cannot start nothing from in front of scratch We have to look for the integration of our in-service Tools we are happy with them, but we know that if we go to a Big data generic solution probably we will Take advantage of this solution and additional we will increase the efficiency So now I'm going to give the word to to Francesca that they will explain what BMAT is Okay, thank you. I will talk about BMAT BMAT means Big data manoeuvre automatic detection in flight testing manoeuvres are motion of diecraft Done by the pilot to determine the performance of diecraft So a flight is Composite by manoeuvres and The objective of this Project is to classify and identify Automatically all the manoeuvres of the flight of the all the manoeuvres of Airbus defense and space aircraft fleet Manoeuvres are for example, they call landing climbing. Okay So BMAT is a big data oriented project which aims to automatize the most of the flight test analysis process and I just say before the fascist tape is to identify classify all the manoeuvres of our fleet of aircraft So the manoeuvres are determined with timers lies an initial and fun final time and Particular calculation associated to each manoeuvres as to know These timer lices are selected using interactive tools and even in some cases. I'm not in Store it in a database So We are thinking about to sell it these time lices automatically In some manoeuvres like like takeoff landing that can easily be Identify these time lices using rules, but there are other kind of manoeuvres That we need to use pattern for this identification Once we have these timer lices we have in-house show where for these for perform this automatic calculation the name is J processor and This process covers data kitchen automatic maneuver detection Calculates automatics and store all in No a squirrel database and finally to spot the all the information using our web service Why do we use? No a squirrel database is that all relational database because We want to have the same structure of database of table for all the aircraft for all the manoeuvres And this information is not a truth to it. So for example in in I refueling contacts is Necessary to to calculate to to store the the time the duration of the contact or the fuel transfer it but Perhaps this information may not send for example for all the kind of manoeuvres like I will deliver the manoeuvres so this kind of Calculation this kind of information is can be easily stored in a Jason and finally import in a no a squirrel database And Finally we have all this information Information related to the manoeuvres information that to the fly in Together no a squirrel a squirrel database and relational database. We finally exploit all the information in a web service This is the process the world's flow of the big man first We access to the to the data using data servers and automatically we find we Entify the initial a final time of each manoeuvres using rules we have for this kind of manoeuvres in house so we're named J. In the detector and The for the other manoeuvres we have we use wavelets And explain later what waves is and I suppose that you know this and the second step is to Perform all the calculations automatically and store all in a database with all the information of the fly a fly report the objective of the fly video So and finally we can access to all the information in a web service and even Service a video using streaming. Okay, so this is Betty and this is Project that is nothing it we are working on it In line of big data, we have multiple manoeuvres different and fly different prototypes and the solution is Parallel in processing for all the calculus multiple matching working together for the same job and Using a lot of matching we improve the speed of the calculus and Using Q we can to minimize the adult times What the technology We have tested different technology to 25 the this kind of manoeuvres that need a pattern to be Identified so we began to use to test a Short time for a transform, but these only work in frequency domain and For for this came is not good enough. Okay, and we test neural networks, but Our experience was not good and we need a lot of time to train in the net and finally we were tested a Continuous webless transform is similar to the Fourier transform But instead of seeing a cosine we are using web less Waveless model. So what is the way play? way play is mathematical function with some characteristics like zero average is centered in the in the origin and See let's be a series of translation and dilatation of a Waveless name Waveless model the continuous Waveless transform is of a function is the integral of a function by this family of Waveless and Similar to the Fourier transform We obtain Spectrogram in this case we obtain a Scalogram and it's very easy to find a pattern in in the time or We are that I we are looking for and These are we show you some Slides correspond the the steps that we are using to identify the manoeuvres first We in this example. We have a theoretical signal composed by sense And we select this is for us our pattern We find the best Waveless that add up to this pattern And finally using the Scalogram we find the pattern in the time history You can see in the the red points are the center of the manoeuvres and This is where the correlation between the signal and Waveless and the pattern is bigger and the other the scale the ice assist We have the scale That's been where for a sample, you know, I say in the beginning we select the first scene as a pattern and The scale in this case the value is one and in this way You can determine the time of life the initial time of life of this manoeuvre to a scaling for the for the value of the is ace ace so and this This is nice is to apply this technique in a real fly Corresponding to our aircraft seat 295 this manoeuvre is Typical manoeuvre or longer to the longer to the knowledge ability the name is side slip and You can see in this picture diecraft is the time history Correspond to the rather the rather is a surface controller diecraft is that is located in the tail on the tail of diecraft and this this Manoeuvre correspond to the they are moving the rotate the rather Maintain in in the position Like using steps as you see and you can see the picture diecraft is sleeping in this way is this Doing this. Okay It's a typical manoeuvre. So we select one sample in this time history What will be our pattern? second step we Find the wireless model that best add up to this pattern and finally we find all the These manoeuvres into the flight and in this video you can see how the way left words finally is moving around the Scalogram and with when fine the best correlation stop and began to scale the the the way let's model until you have the best the best Implication and in this way we can find automatically the time relays of the manoeuvres and we have a database containing Not all the manoeuvres because we know are working on it, but we have a database with the the characteristic of each Manoeuvres of this weblet that we will use and we Take all the aircraft all the flies and we pass for all the Wavelets and in this way we classify Our manoeuvres and began the calculus of the automatic calculus. Do you have to do with the manoeuvres? Okay? And that's all Thank you So you have any question? If if you don't have we have a last slide It's not mine. It's from the human human resources department that Has told me that here in this room. There are very clever people and very people interested in in the data So just to say that we are hearing we have some open position not in flight test but in general in the company we are in a big company, so There are a lot of departments and a lot of business where we data is is key so you can you can see in the slide that Entering the in the website you can access to these job positions for me are a very interesting company where where you could work and of course V data is one of the of the strategy Topics in our in our business, so I invite you to enter in this in the weapon and to look for for more information Thank you