 The case study is about the Nahara Fidal Bridge in Lebanon. So we started one month ago and it was built in 2007. It is a concrete girdle bridge, a 100-picast extension girdle. It is located on an international highway. So it's important not to have a traffic destruction. I need to say that in Lebanon we have no bridge management strategy and no assessment for the bridges until something really goes wrong. And we have no monitoring systems. So for this case study as well. So I was able to obtain the autogad files of the bridge with the structural details. Usually also companies does not share the files. They don't give it to us. So we have tried to look at this bridge. So we went down the feet mission. The bridge is new. And we were able to identify leakage at this joint. And what else? We have an abutment tool, some humanity. And but other than this, a guardrail is partly failing. But other than this, it is in a good state. The problem is that bridges, even though it was built recently, it is being overloaded because the increase of the population in a surprising way. So we were discussing about what to monitor and what are the methods to use based on the equipment that we have. So we thought the equipment that I have is the ground penetrating radar. And for mostly short term ambient noise vibration method, two velocity meters and 12 accelerometers and two data acquisition systems. So we thought about maybe loss of restress, loss of stiffness, or analysis related to the load increase. And so by using ambient vibration method, what else? So we have filled this fact sheet. So we have the knowledge on the decision context. The decision maker is the public authority and parcel of development and reconstruction. The additional stakeholder is the Aldar, the company that designed the bridge. Also the municipality. So there's no budget allocated for maintenance of this bridge. At the time being, it's functional. So here the data we've discussed. We just have the drawing, structural drawings. And the objective is to minimize operational maintenance and inspection costs to maximize the income. But specifically, I think, minimizing disruption and the direct cost link to it because in case something will happen in the future. So I think the media action is just when I do nothing, we'll repair if needed or if it will not be in a good state. So we'll be looking deflection cracks and model properties by ambient noise vibration and the performance, the availability in the other bridge reliability and safety. So until now, this is where we are. And I think that between an analysis comparing if we have nothing at all, we model the bridge with respect to certain parameters. And we try to monitor the bridge on a short term visit. And then we update the information. And then we will do the decision trace of future something. And you are welcome to suggest. Many suggestions were considered last day about. But then after discussion about if we take an element of the span and we compare it to we model this element of the span and be compared to the capacity of the codes. This is one of them. And the other one was related to if we look at the increase of the loading also. And we model the function of a certain deflection, for example. But those are two possible ways. But also we can start just from the start about we do not have any information with a basic and try to monitor with what we have. So if you have any specific suggestion about identifying a certain parameter that is very suitable to those kind of complete bridges. And my tools to do it with it, it will be also welcome. Yeah, any questions? Yes, can I get a bit of detail on the bridge itself? These are pre-tension beams, are they? It looks like a pre-tension beam girder. It's not post-tension. It's not in the arm. I don't know how it looks to you, but those are on the side also. It's post-tensioned. It's post-tensioned additionally. Yeah, pre-propyl cast, post-tensioned girders. OK. And it's showing signs of cracking at this stage. Because presumably it's designed to be fully post-tensioned. So it shouldn't crack. It shouldn't crack. So there shouldn't be, I'm trying to work out, you said you want to monitor loss and stiffness. And loss of pre-stress. And you were going to do that with accelerometers if I wasn't wrong, but you're not going to be able to identify, we've done a lot of work on this, trying to identify a linkage between the loss of pre-stress and the natural frequency of the structure. And there's no clear linkages between the two. I think the most value you're going to get out of here is like the previous presentation, putting a WIM system on the bridge, determine the value of information in the context of having actual loading data available for the structure. Because you're going to have to do all kinds of condition monitoring, independent of a structural health monitoring system here, to ascertain the condition of the structure. So I would say, to me, your first step is a WIM system. And determine what the loads look like. And do a reliability analysis on the basis of the actual loads that you're seeing on the structure. And then, longer term, coupling, some kind of monitoring. Initially, you have to have a good, you need a condition assessment. This is 10 years old, and you're talking about cracking. In a post-torture bridge, I'll be worried. Yeah, but my comment was that even before we need a model that represents our stages of being afraid. Yeah, absolutely. We need to develop a reliability problem. Absolutely. Maybe first, to get one, you just take the design equations that have been used for the design of the bridge and translate them to a probabilistic form. You can make a simple reliability analysis, which is then the reliability of a design outcome from the standard. Actually, you're going to have the same approach, right? And then you can think about it. You can look at this flip-flip function. You can say, OK, how can I now include features I'm worried about in this reliability problem? So you have to extend the model a little bit. The model has to be more complex in order to have the links and the interfaces to include some information about your doubts. So you have to have some model representation of loss of retention. You have some model representation of the possible processes and the causes that might be associated with. And then you set some prior information on that reflect your doubts. That's a heavy part, because that's really guessing. But you have to somehow bring your uncertainty into this problem. And then based on that, you can think about what would measurements bring. The typical situation an engineer does not do, the typical situation that's exactly followed MN's step. You have cracks in this bridge after 10 years. You have to do something. You have to look what it is, because it's not acceptable. But what we need for the value of information analysis, what we first have to demonstrate in the calculation, that it's not acceptable. So we are very unsure about the situation that we get in these indications, usually. We have to represent this. And then we can compare this to our refined things after we have the measurements. And the value of the reliability analysis that you're talking about doing initially is if you do, like we saw earlier on, if you extract the importance factors from the analysis, it will tell you which are the variables that you need to focus on. So where do you need to consider gaining information on the different parameters? Do you give me a good treatment or something? OK. That's absolutely clear. We are more very intent to ignore solid parts. I also would like to ask you about this method to see if it would be helpful or not. There's the method used by VCE, called the Grimos, where in order to monitor, to see if they have a loss of pre-stress, in the mid-span at 1 over 4, they do just one measurement. Then they compare it to drain cuts, and they see how it is changing, and they can scale where they are now in the state. Would it be also useful or not, just from your experience? So looking at vibration, and trying to link. Yes, ambient noise vibration. I tore my hair out with this for about four years, trying to link loss and pre-stress to a variation in the natural frequency of the structure. And one author will tell you one thing, another author will tell you a different thing. And when you do your own experiments, you'll find completely inconclusive results. So, Excalibur, I wonder if they're not going to tell you anything about loss and pre-stress. I think the drain cuts show just variations of natural frequencies, but they do these drain cuts to check if there are cracks and losses of stiffness, not loss of pre-stress. Loss of pre-stress, maybe you can find something with a static test, measuring this prism. Yeah, but I don't know, but they did write a lot of it. I'm trying the cards forward to detect losses of stiffness, not of pre-stress. And the article is about loss of pre-stress, so I don't know. Anyway, I completely agree with that. Yeah, but maybe this is very variable. This is a really good case study, because it's getting it at the ground level now, and following the steps that you're talking about, to demonstrate the value of information in the context of structural stress. So I think it's a very interesting question. Because the situation is typically in the sense that we come to a bridge and we develop some doubts, right? That when you go to the road authorities, they will be very happy about such an example, because that's what they always have. Always have, absolutely. What, how do you, and I think what is under is the male. Normally, we jump into these measurements, measurements that would be overlooked. Most of the time, in order to facilitate it for more formal approaches, we have to represent our doubts in terms of the limits there in the appropriate uncertainty representation. That's of course very difficult, because it requires assumptions. We just have to assume what would we think. And there are some techniques, actually. We did not teach about that in this training school. But we, of course, talk a lot about this Bayesian representation of uncertainty, but that requires sometimes that we translate our doubts, not our data, but our doubts into a model, into a quantitative model. And that's, of course, difficult. You find techniques how to do that. You find a certain way to ask yourself in order to find, for instance, all the standard ideation of my doubt. So that's very interesting topic. So it could be also something for a future training school. Not as the entire topic, but as one of these lectures. Yeah, we actually spoke with, I guess, a bit expert elicitation and the cooperation of it. Exactly. But if you have a good client here, if they're all already here on board, you have an excellent opportunity to demonstrate the value of information in the context of the structure. In the same techniques, Maria can also be used in consequence estimation, just to guess. But, of course, it should be controlled and more less rational, I guess. And a techniques way. Yes, any more questions? Thank you very much for your time. Thank you.