 Good morning everybody. So we present this study case together with Yann. So our study case deals with a wealth that was built this year. You will see that we have the experience of monitoring of several wealth in France. So just issues about wealth. So 80% of the world overseas trade and 99% in the USA passes through ports. That's a carol in Europe on defense. 3 million people are employed by this sector in Europe. And the state for maintenance is for instance in France that we have 100 km of wealth. That's probably 1,000 km of bills. And among which 64 km are built in reinforced concrete platforms. So what is a platform that's about 300 meters of length structure and about 20 to 30 meters wide and with 1.6 km of beam if you consider a structure like that. In France 10 million euros are given for curative maintenance of this structure. So there are some repos in the USA where you can find other figures. And so that's typical the building of the structure we have insurmounted this year. So there is an existing video on YouTube. So we have a little less views than Riela, but it's promising. And so what we will present today is for sub-study cases. Two topics mainly. The first one is maintenance of course. So I say the benefit of chloride for concrete. And the second is retrofitting. So based on the mechanical understanding of the structure. Why? Because boundary conditions of this type of structure are very difficult to model. So there are a lot of assumptions when you design it. And so the real behavior is very far from the design behavior. So any sensors you put in the structure give information about that. So that could be the sub-study case. But that's too separate problem for us in our presentation. So the cost of... If we deal with the first one, that will be our main purpose today. So the maintenance of structure due to chloride ingress. The cost of NDT, typically a core and you measure the chloride inside, is 1,000 euro per point. The cost of the sensors is 1,000 euro and more or less between 1,000 and 2,000 euros. But you have 3 to 4 points. So the cost analysis is finished now. But it's not so simple. So we have two key questions. The first one is that the uncertainties for NDT and sensors are not the same. And the second one is that the lifetime of the sensor is not the lifetime of the structure. For at least 10 years, not more. For this type of structure, we lose about 30% of the sensors after the first year. So we have to keep that in mind. So that's not magic. That's reality. So that's the plan of the structure. When you can see the tidal holes here in the soil, the river is here and the platform is here. So you can see how it works. So there are tidal holes to support the horizontal loading. And piles to support the vertical loading. So that's a very simple scheme analysis. And you can see that there are a lot of interaction with soil. And of course that's the problem of the boundary condition I have explained before. And so we have five monitoring structures in now. So that's the fifth one. And for this one, we monitored the tidal holes mainly. So we have quite a lot of papers dealing with the understanding of the monitoring of tidal holes. So Jan, if you can speak about what we did this year. So we focus on a particular area of the structure. In fact, two structural elements that we call beam. The work is built with a precasted element with a shape of shell. And between two shells, we have some reinforcement elements. And so the structural elements that we call beam. We have 40 sensors in this area. And just here, we collect all the data. You can see here. And we have started to collect the data when we cast concrete. So we have all the history of the construction. And we also performed some material tests on the same concrete. So six months after the beginning of the construction, we are able to assess the performance of the whole measuring device. So we have some broken sensors now. We have broken some wire during construction. And we have some minor problem with water in electric connection. So for the rest, we are still recording the data. And so we considered the decision context. As I said before, we wanted to discuss with non-authorities and see what could be interesting for them. So the first one, the first question, is the usual quantification of the value based on preventive maintenance due to the monitoring of chlorhagin res. So that's the main focus of our subsidy. The second one, I don't know if we will work on it because I think we need a collaborative work in this action if you want to do that. It has good behavior of the structure. I mean the mechanical behavior. And to see how it could be interesting to have this behavior, this work behavior in case of retrofitting of the structure. So retrofitting is needed because the lifetime of this type of structure is about 100 years. And during 100 years, the use will change. So there will be more oil, more coal, or containers, but it will change. Now it is used for waste turbines, provided by General Electric. But we know that in 10 years, these wind turbines will not be the same. That changes a lot. Or there will be no wind turbine in nonce. So that could arrive too. So the third one is the change of paradigm. So to go from the means-based contract towards the performance-based contract during works about the concrete performance. So for the moment, we put some concrete beside. We make a test. Really the concrete that is inside the structure. And the last one is the detection of initial cracking in view to have a better understanding of the probable development of cracking in the concrete. But we will not deal with that in the cross-section. It's not realistic to work on this. So I will not tell more about the flow chart because that will be presented in the slides after. So first, we will discuss how we want to deal with that, and if it's convenient for you. So the sensor first. So we just speak about the measurement of chloride unrest. So we design a measurement device with a different sensor because we are not sure of the performance and of the life duration of the sensor. So we propose some new sensor, some resistivity sensor on one meter long, some chloride sensor. But we are not sure about the life-different duration of this sensor and some armature potential sensor that you can see here and here. So the instrumentation is redundant and we complete it with some temperature sensor because all the data are also sensible to temperature and we have to correct it. So the resistivity sensor or new sensor provides a map of resistivity under the sensor. So we have one meter of measurement with 200 points and so we are able to make some statistics. So we are not dependent of the material heterogeneity of the concrete. And we have assessed the performance of the sensor by casting different concrete with different content of chloride inside. So we have the probability of detection and the probability of false alarm for different levels of chloride. So what we want to do here is a service limit state so durability due to chloride ingress, induced corrosion. But we will stop at the corrosion initiation. So there is two main reasons for that. It's to simplify the whole analysis because there is quite a lot of parameters that we act. First, the first reason. The second one is that it's not reasonable to change the river on this type of structure because the access is very difficult. So to remove rivers and to install a new river, it's very, very costly. So we tested in another project the efficiency of several river and we consider that we just change the concrete cover in this project so we can change the concrete cover signals and the type of river. So usually practice consider some administrative testing so course whereas we have demonstrated with Manuela Salta from NEG that there are large uncertainties that it was published very recently. This would be testing so I mean autopsy that could help to make the link between the chloride and detection twist and so corrosion initiation because there is not a direct link and so for a session, so the action are here for this and the action for a session could be the distributed sensors that we promote, the standard problem and the corrosion initiation, so the potential measurement and you can add of course, you can consider these two actions by adding these two SSM systems. We have, we can add something else is to make accelerated tests in lab and to compute the factor of acceleration of the testing lab so we do that with Bayesian network and we have already tested with non-destructive testing but not with SSM but it will be made. So the cost function is the usual cost function so that the cost is linked with the risk of corrosion initiation and considering the inspection, the structural monitoring and the repair. Okay, repair is that. And the variable as the chloride content at the level of the rebar and the critical pressure for corrosion. Okay, so we will have a component approach so we use only one component but we want probably to, we don't know if we will do that in the cost section but what could be interesting is we have a structural approach for that we need spatial variability so spatial variability can differ from the structure to another one so what we do now is to compare the data we have in several structures in Nantes the data that Bernthal in intrandheim and the data that Alan Han has in Trinity so and from that we will see if we can find bond to the interval or variation of the length of correlation or if it's too random. Okay, for the moment I think we will be able soon to say something about that but I think it's probably not realistic to have the whole analysis including spatial variability. So probably we have not more time so I just go very quickly not just this aspect about the retrofitting but on the last question that we had with Vim yesterday how we will model the things so we don't have any information about the reality because it's just to monitor the structure and so we have to model a virtual reality that could happen in the future so for that we have to use models and so the first way could be to use simplified models so there are about between 7 and 8 I don't remember usual models in the literature and we have benchmarked this model in a project in France and there are two main drawbacks the first one is there is correlation between parameters not because that's physically correlated but it comes from the correlation itself from the calibration itself if you use a fixed function you correlated the random parameters due to the fixed function not due to the correlation of the parameters and the second one if you use that in a probabilistic framework the capacity of these models to predict is very bad due to the form of the model and the role of the parameters in the model so we have shown that we have a report about that if you are interested that was made in a PhD in a national project in France and so what we will use is for the simulations is a finite element and finite difference model that we have in lab and so that's good for simulation to simulate the reality but the problem is very difficult to update because there are a lot of factors inside temperature outside all our stochastic processes so what we will do is to use gamma processes for the updating so that could be by-gen network or gamma processes so this process we have already presented that in the cost action I think it was two years ago in Copenhagen so we were quite a bit with the prediction and the use of these gamma processes so that's why we will use it with the uncertainty of measurements of course and the behavior the change of behavior of the sensor depending of the client content so that's what we plan to do and that's why the the discussion is open I think now thank you very much this is extremely complex it's a lot of work a lot of efforts a lot of models are already there yeah I hope it was not too complex can we go to the slide where we have actions there was this one there's actions described can you go through this again yeah the actions that will be used for the optimization of the inspection and repair without structural estimate are the semi-destructive testing the destructive testing and the repair actions the repair actions we have three actions and these repair actions the cost of the repair is not the same and the efficiency is not the same so we have the information for this project the actions if you have SHM they look a little different than the ones if we add the SHM so that's the decision on the measurement devices the A SHM is the decision on what to measure and A is the action what we can do with the system so that was different that's the first decision what do we measure yeah but I'm after this but okay so that's so if I'm after the point that's in the value of we call it the value of information analysis we quantify the value of the information and then maybe this is the case and I don't really see it but only if the actions and the prior decision analysis and the pre-posterior decision analysis are the same then we are quantifying the value of information if the actions are different it could be because you have more options if you have more precise data then we don't quantify necessarily only the value of information but also the value of information and actions yeah because here we mixed in this presentation we mixed the actions due to the collection of information so the difference can be considered no accession with accession so the repairer actions are the same yeah okay so it's two actions right first action what do we measure what brings us new information but I have a question about the RPO we want to start the RPO so we have an RPO model about this yeah that's a simulation and then the uncertainty is increasing over time in your time horizon so it's also you find also different values of information of different times that's also interesting to look at so you can more find optimal timing of your yeah that's that was made for NDT already so the optimum time was computed in this mixture so we have to to compare that the cost is cost with the cost with accession very good thank you your probabilities of detection are for different levels of proportion yeah and these come from other data previous data if you didn't have this data sorry if you didn't have this data all this would not be possible yes but it would be it would be probably no no because yes I agree I was just trying to shift from corrosion to another type of damage if you have a damage on which you don't have this probability of detection would it be possible to apply an analysis like this accordingly yes so the probability of detection is very very important for concrete that's a very important input for instance we just discover that for resistivity so on the point of bridge that we probably know the deco-free project that we have with with Bruno Bada so we discover that the coefficient of variation of the measure of resistivity in a bridge concrete is not constant, is not linear and is exponential depending on the resistivity so that means that the uncertainty of measurement will be complex to model and so that if you don't have that you will assume that the coefficient of variation is constant or the standard deviation is constant dependent of the resistivity but that's not the case and that changes completely the result okay