 Yes, thank you for this introduction. My name is Jochen Köhler. I had the pleasure to lead this working group four on the case studies together with Helmut Wenzel. So this working group was working or was starting working more or less in the mid-term of this cost action and the aim and objective of this working group was to Demonstrate the benefits and the feasibility of value of information analysis and structural engineering decision-making Also this working group, I believe was the platform for a very intensive information interchange so many researchers from all over Europe did actually try to implement the principles of Bayesian decision theory and in particular value of information analysis in the Practical projects they do anyway, so I think there was a very steep learning curve among the participants and also in working group four Maybe this was this working group where the participants took the most benefit From the state from the training schools we organized within this cost action Also now and this is still ongoing after this final conference. We try to use this case studies to Demonstrate the feasibility and the potential of these methodologies to relevant stakeholders from industry and I think with some of the case studies we rather reached and level Where this objective can be can be fulfilled So we we called for a case studies and early 2016 that was at the time where working group one two three already established their principles and Material we could take a benefit from and then a large number of teams actually. I don't mention the exact number Because it was a lot of teams responded that was very nice and initiated Working on case studies at least they did come up with a proposal or how a case study could look like and Then we did try to set this analysis into practice and As it was the objective we wanted to demonstrate not only the benefits and And the feasibility of case studies But we also did find some obstacles and some problems when it comes to the practical implementation This was already discussed before the break. So before the break we saw all this beautiful theory this very consistent and and and nice formulas how the principles work When you really go into a field and look at the real structure as you all know You we realized that the one-to-one Implementation of these principles is not so easy So the perspective we we had to take in this case studies was not from the bottom up where we look at the theory and then we We try to to put some Practical values on these different Corners of the theoretical concept and try to pull it through from the bottom up. I Think a very useful perspective for Implementing the case studies was also come to come from top-down to look at the decisions themselves To look how these decisions are made in practice And these decisions that are made in practice. They have principle properties and The first and a very obvious property, but this has to be mentioned because it's important to start with is Does these decisions they are done right so they are made? Irrespectively whether we have a beautiful a theory of value of information or not so decisions for instance in regard to the continuous maintenance and development of the infrastructure they are just made and Now the perspective that is useful when we want to implement The theory in case studies is that we want to make these decisions better by helping with our principles and this framework of Posterior decision analysis and also of a pre posterior decision analysis So it's not to get to full theory and try to implement it into practice I think for case studies it was a very useful perspective to look The decisions are already somehow made and they have to be made continuously in the future and we have we have to look How can we formulate the decisions in a way that they are made better? then not considering these theoretical principles So We have this individual terms teams that have been working on this case studies and the onset of work was producing fact sheets where the case studies had been described and and then further development of this case studies was then supported by by workshops and as I mentioned by these training schools and This worked quite well and I think we can say that we have reached a big number of case studies that have been far developed and reached reached the status where they disseminate their results It's 13 case studies, which is a number. I would not had believed that this would be Reachable when we started with the work. I think this is a very big accomplishment so Not all of these case studies are documented in a fact sheet, but very importantly all of them had been documented at least in one conference contribution some of them have more than one and We reached also already a rather high number of journal publications that document these case studies So I think that is a very a good way To disseminate the results and also to promote these concepts And as I said these case studies they take all a starting point in a practical application More general publications are planned. So we are planning to set up a special issue In engineering structures There are rough estimate is that we might reach 12 journal publications and We also wrote a summary paper That will be presented in Guma Reich March at the yaps the conference Very summarized the insights and lessons learned from the case study portfolio so now I will take some few minutes to Present you some examples of the case study. So these case studies go into different areas In our build environment and they also reach different levels of detail and they come from different groups That had been participating In the in the case study, so it will be five examples But all the examples or all the case studies will be documented in the PowerPoint slides After my last official slide so you can when you download the slides after the conference from the web page You can get a brief overview about all case studies by having these slides So the first was Choosing as the first one because it was really the first case study that was pulled through by Dimitris Diamantidis and Miroslav Sukhova It was about the monitoring of snow on the roof it's Rather obvious example and Interestingly, it's also very actual again because there was a lot of snow again in the in the alps It was in a very extreme extreme snow event and then of course when you have such extreme snow events and you have a structure at risk Then you ask yourself how much should I invest to monitor the snow on the roof? So this was Considered here as you see in the table. There had been three different alternatives for measuring or for indicating The snow load on the roof. The first was the snow depths on the ground on a nearby methodological station. So this is the most accrued information that was considered and you can Be a little bit more specific and you measure the snow depths on the roof or as an option as free you can also Estimate the snow load on the roof Which is then even a lesson uncertain indicator for the for the load that the structure is exposed to and the Costs had been considered and it was found that investing into the Direct measurement is the most efficient one So this has been Documented and and disseminated in two journal public or in two Journal public in or in one journal publication and two conference publications or one is in the conference on ICASP on the international conference on the Application of statistics and probability in South Korea. So that's very nice something like this is disseminated so the second example I have is also a very early example and that's also nice because it's Implemented on a very well-known bridge in in Stockholm steel bridge and there The problem was that there was some fatigue and this fatigue Did cause a crack in a structural part of the of the construction and this crack propagation was monitored And as you can see some Bayesian network was used to represent the system and and some rather Progressive assumptions had been made of course of pulse to in order to pull this example through But I think this case study also very nicely Did demonstrate the principle and the logic of? the Relationship between the information and the content of this information For doing decisions quite well and as the other case studies did also lead to quite a number of publications So here when the review will finally be successful, then we have three journal publications coming from this case study The next example is from a bridge again It is a concrete bridge and here the monitoring was on the Response of the bridge due to traffic load in order to learn about the real traffic load on the bridge as you all know the Traffic loads that we have in our structural codes. They are quite conservative meaning that they include a bias on the safe side and When it comes to a decision of lifetime extension of a bridge It's maybe a good idea to look at the real Load this bridge is exposed to in order to make an informed decision about the further reliable Utilization of the bridge this was done in this example and Positive effect of this weight in motion measurements that had been analyzed here had been found It's also it should be mentioned that this project was a nice collaboration within the cost Framework it was actually a collaboration between at Zagreb between yeah, University of Zagreb and DTU in Denmark Also short-term scientific missions had take place for this So this is a very good example how the different tools we have available in the cost action and it lead to a nice result Also here we have a few publications as a result So here we have an example from a from a different domain here. We look at the wind energy converter that is placed offshore and the problem here was that In the design phase so before The the wind converter is is built We want to design The foundation and we want to design the structure of the wind turbine and we have to take into account Dynamic response of the of the converter and this depends on the stiffness of the of the Structure itself, but also very much on the stiffness of the soil so in order to be in the right version in the right region to have not not so a Critical fatigue behavior We have to know something about the stiffness of the soil and in this example We looked at possible soil tests in order to update our information about the soil stiffness and To justify the costs of such tests Here we had also to make some Critical assumptions to to pull this example through In particular our assumption about the likelihood how the measurements of the soil relate to the true stiffness had been somehow Roughly assumed but we demonstrated The applicability of this concept and we did come to some interesting conclusions yet some Two journal public air to a conference publication so far if we have this special issue We also consider to have a journal publication on that maybe also working a little bit further on this likelihood Modeling representation of the likelihood for the soil measurements And then the the last example which was also a good collaboration Between different partners in the cost network Facilitated by a short-term scientific mission here. We are talking about Framework for the reassessment of timber structures and as we already have As we already presume there's maybe a benefit when we Increase the level of detail of our reassessment of the structure in a in a sequential manner So first start with a rough Assessment and then increasing the level of detail whether Depending on the result of the more rough Assessments so In this case study we or the the colleagues did show that this is Actually a very cost-efficient way to do so this was a very interesting example From the area of reassessment of timber structures Which did also lead to some? Publications as before it was to yaps a symposia That did serve as a disseminator for this case study But also I got positive indications that authors are also interested in in the journal publication so some lessons learned as we already discussed in the morning and About the big challenges to to apply a value of information analysis from the case studies maybe the most obvious obstacles had been The formulation of the prior uncertainty we have so I think in engineering in Practical engineering we have a culture to not consider uncertainties explicitly, but implicitly by conservative assumptions and I think when it comes to strict value of information analysis Then this culture of being conservative and being biased in a way Is setting a very big challenge to the analysis? So making really a fair assessment of our large uncertainties we have in the absence of any additional measurements is of course Very critical in order to prove that the measurements become cost-effective so this is a Very a critical aspect also as discussed before the formulation of the likelihood. So how do we link? our observations to our Property of interest this is challenging not only From a from a principal point of view that we hardly have so many measurements From the observation and the property of interest Simultaneously that we can somehow empirically assess at this likelihood a big challenge is also associated due to the fact that we Very often measure locally, but are interested in more global responses So that we have some Problems on a spatial scale Also, we have we have issues over time Yeah Then But I think it's it's important to move on and to consider a value of information analysis as a relevant Tool for practical engineering but I Think it's necessary to to consider this always in relation to a probabilistic assessment of a structure, so this is logical is the first step of refinement of our analysis we have to to in order to to make a value of information Analysis feasible so we have to depart from the partial factor design format What is of course our usual way to do analysis? to probabilistic analysis And I think this was very nicely Demonstrated by the picture if you remember shown by a Michael Michael Farber today He showed the picture where we have a lot of experience In the in the x-axis So from a lot of experience and not so much experience and then we have the consequences on data axis You remember this picture and then of course probabilistic analysis is very relevant for areas where we don't have so much experience and The consequences are high But that's of course very interesting because when you consider an existing structure as something unique Then we also have to appreciate that very often. We don't have so much experience about this particular structure So I think for the assessment of existing structures very much in general not so much related or also relate to value of information Analysis but independent of that. I think when we talk about existing structures It's very hard to calibrate a partial factor design format That is general enough to cover all different examples of existing structures in an accurate manner So I think for existing structures Reliability methods risk-based methods are a very good idea to Utilize especially when the stakes are high when the consequences are high and then the step to do a full Bayesian decision analysis including preposterior decision analysis including volume of information. This step is actually not so big So I think in order to tackle all these challenges we get with our aging infrastructure in the future We anyway have to move on to probabilistic methods when it comes to existing structures We have some further activity As mentioned or we will document all the case studies on our web page We already have a very nice template for that produced by Maria Pina So she will talk about that a little bit later every plan a special issue With approximately 12 contribution you will have held her later on he will talk about that and We will have an industry innovation event Or several of them and also held I will talk in his presentation about that So I think for all this further activities the case study the case studies they play an important role and I think To conclude It was very good to have this working package in order to set a condo weight from the practical side To this beautiful theory But we did actually accomplish to demonstrate That this principle and this way of thinking that's coming from the theory is very helpful also for practical applications I want to thank all the working group for participants. So there was a lot of Workmen work hours invested in this working group. So that was very good and I also thank you for your attention and I want to introduce Helmut Wenzel. He will also have a small presentation five slides. He told me where he will Present something from his perspective because he's really a practical Engineer working with this concept with this concept every day. So thank you Helmut And then I suggest discussion round we can do after Helmut's presentation so Actually, nobody has to convince me of the value of information, you know, if you're an insider value of information That's that's clear and if we look at our situation that you know all these Big companies are selling our data and then we know that this value of information might be billions but how about our Our community so some some remarks from practice So some some facts in practical work, so When we do some practical work what we have to do is we have to explain our approaches and results to asset owners and You know, most of the asset owners are either Economists or lawyers. So this is very difficult Another fact is that civil engineering is 99% deterministic maybe maybe more And the civil engineers are very conservative Very difficult to convince even my engineers, you know are always reluctant to use these wonderful tools Another big issue which we face in practice is that When we get data and have data about our projects if they are 50% complete, we are happy. That's that's a lot But how can you do good statistics with 50% of the data? And the quality of data is not always What really helps helps us There is not enough Determination towards making good data The consequence is first of all, we have to find a way to communicate the value of information to our owners and This is very difficult And we have to simplify the approaches You know what has been presented here is wonderful. It is Super fantastic on the scientific level When we come out to practice We will have to go down to a level first of all where we understood and secondly that The owners are able to know to buy it because it costs money so and Another thing is we have to deal with this unknown unknowns There are a lot of things which we don't know. I give you a short example I cannot tell you which one and where it is. I shouldn't show it, but in offshore In the offshore business There are very complicated structures. This is one of the construction of one of the platforms in the North Sea and The model that was telling us the expected lifetime of this structure Contains about 200 parameters so I'm sure if we know 200 parameters. I'm sure that we Don't know another 20 of them which are which are relevant. So how to how to do how to deal with that? This shows the result of Well people were unhappy with the performance of the platform This shows the results of the diving campaign that there was a big discrepancy between design and Execution and what you see here is that the piles are not Centralized as they should be and they were not fixed on the on the bottom just No details, but if you look at our monitoring system here, I think what is the uncertainty? you know the This jacket is swaying with about a period of 10 seconds and the diver has to have four measurements around the circle of the piles and always with this swing thing Who tells us which kind of Uncertainty we should put into our calculation So what came out of this? You know the the fact was there were four Parties engaged in assessing. What does this mean to the structure? What is the expected lifetime? The original designer said 350 years is the lifetime But considering that there were no mistakes made the construction company which made some over design just to help them during a Reaction calculated 806 years of lifetime of this structure Then came a very famous consultant from Denmark that told us that the lifetime is 0.2 years Which is a little bit ridiculous because this came two years after the structure was found to be not very consistent And the reasonable consultant from Austria said we have 52 years so how How to deal with this I came Finally in and to make a kind of assessment of all this information. I Collected more than the other people had Then we applied some very simple things. This is a Monte Carlo simulation on This is a sliding surface. We see here. I don't see sliding surface and There's a water level and this could push one of the piles for example and It contains very simple parameters not not too many and Here we have our uncertainties So I very much like the the work of the Joint Committee of Structural Safety there in their documents You find a lot of interesting values for that But when I put this to my guys, they say well, this is very difficult to difficult to understand. So if you could make let's say a list of default values for Uncertainties to be used steel concrete all these kind of things this would be would be helpful we started with well simulating with one standard deviation and Finally using two models came to a very simple representation. This is a fake factor of safety This is understood even so it is It is not what it is But this is what you can sell And also in nice and important would be this is the coloring because you can show them If they tell them the factor of safety is one Okay, but if they say you're more to the green than to the red the economists are happy So what we have to do is simplify it and But not give us let's say two simple things where we predict then with Let's say accuracy of plus minus 30% in the cost action 14 oh six Which is in also in relation. I think most of you know that we have collaborated. There were models elaborated to How to represent this and this is the spider, you know in 14 or six we use a five leg spider Okay, doesn't doesn't make much difference, but when we when we use this we get a clear Location of the phenomena for example safety you we use the euro codes 1990 series for safety and durability For service ability we use the sector specific rules and this is new here Economy environment and safety. It's not new of course, but it helps me To convince the economists and the lawyers that they should do this and if we Put the parameters which we now specify in this is a 21 292 Tomorrow, I will have the final meeting on on this and this will be published soon Then we can bring our things also successfully through the process of Accepting it. This is and how it how it works and we have done such a spider diagram and finally we can say The smaller the area below here The bigger the risk And as a last slide, I just wanted to show that one of the this is from the iris project Quite some time ago, but these results are already standardized. This is a new eurocode 16991 standardized last year in May and it contains our Aging model there's a mathematical formulation of aging to use this makes now sense because if we make an assessment right now People ask us so what after ten years this model can say the code according to Most probable degradation in ten years your rating will be such and such So that's it. Thank you very much for listening. So this is Some time for questions. Are there any questions from the audience? Yeah Okay, thank you very much was very interesting to see how it goes from the theory to practice especially Moving to standardization You can do you have some? Results which changed due to the fact of working at four different Levels of detail in the timber structure example So was the first level of detail which was approximate did show other results compared to the last Do you have a sensitivity study with that or because I have not studied the whole example there for a masking I find it very interesting especially with respect to assessment of existing structures Yeah, I would like to lead over to the author the main author of this study So basically what we've done it was an actual case study we've done all of these assessment methods and What we analyzed me and Daniel was was successive or Successive assessment or non-non successive assessment worthwhile what shall we do so? when we applied it We noticed that if we go more into detail in the beginning we can save money in in this sense I don't know what was the actual question Sorry lecture question is The level of detail plays an important role, and I know for timber structure It's already somehow standardized, and it is standardized in Germany for the reassessment of preachers We have several level of it is we can go up to reliability based assessment. Therefore. I asked just to promote this I like this example, and therefore I try to take this to to promote it further And I would like to add something here to Helmut Helmut in the joint committee. We have a Report on a probabilistic assessment of existing structures with a very good example how the Central safety factor for the foundation has changed based on new data on wave loading for an offshore platform a new Measurements of the soil so it's similar to what you presented with an updating Just to add on that. Thank you Thank you of all these of all these works. This is these are the wonderful words But in practice, you know the engineers they don't have enough time, you know, they are all pushed You know you have to deliver your research tomorrow. So how can they read all this? documents, you know, I have a big folder of all your Publications, and I like the work very much, but to have a standard catalog of standard deviations for example Yes, yes, thank you very much for these points. But we have educational activities from the JCSS We have documented Not not the JCSS has documented as Studies rare to go for probabilistic assessment and what the what the value actually Cost saving is so I I would also ask the question or I would point to the direction of Education of engineers first and second rigorous management to do the best analysis For the cases we have seen for the cases where it is really sensible I think the responsibility of the engineer to point to That it's necessary to make a more detailed analysis. Yes Problem that this is no problem. No, no question, but you know education is a disaster or it is not non-existent This is one of the major things we find all the engineers. I have This is about 300 have never heard anything about it. Really. So is there something to do? Yeah, you have to give us also some let's say more hands-on tools The JCSS is this is excellent top-level science and you know the small engineers are afraid of it. Okay The first hand was by York, please. I have a short question. I thought What you said is calibration of the unknowns of unknowns. I think this is very important for like real structures so my question is Like if you do the theory it's all based on some model assumptions Say be it the numerical model or be it like the stochastic model From your experience when I look at the example where he said the lifetime was point two years or it's 50 or is it 50 Something so like the results really diverge on based on the assumptions that you're making or based on the data From your point of view, what is the important in practical applications? What is the important? like Based on these assumptions the range of results that you're getting and so is it really important to do a probabilistic analysis and focus on more going into the methods itself or Better understand what the unknowns of unknowns that you were talking about is so kind of understanding What is the the model inaccuracies either the numerical model in terms of say find an element or the probabilistic model Well in this specific case, you know the big difference comes from modeling, of course This is how do you construct your model? What are your your assumptions? This is no question this unknown unknowns is just giving you an Additional Corridor of possible results that you say for example I have a structure of sure where I cannot maintain or where cannot go every time So I have to be more safe so with within the corridor. I have to design it Overdesign it. This is where we where we use that and you know what we done here It is very brutal as a Monte Carlo simulation and we put another factor and the factor has a Variation of 10% and then you get you just widen your corridor of possible results This is what you can do and this is very very well accepted because you can show this How does it work? Straightforward and how do you if you say we don't know everything? Okay? We make a wider corridor of results This is what we do in practice and this is so well That's not much. Let's say all the sophistication that we have heard here Any more question Questions dimitris because Helmut said the education. This is the main problem I pointed that out also in in elizabeth the education of young engineers. I am representative of the alumnus of Civil engineer in my university. I mean in contact of four or five thousand of them But they cannot use the methods which I teach so the the problem It has to start start at the low level as Jochen is doing for example at the bachelor degree So it's not only the master degree. We have selected and if you see the courses of the joint committee They are attended mainly by PhD Students so this is not only the level we aim to reach at least long term Fully agree. You don't have to convince the PhDs. This is no question. You have to start with Basic education Okay, I think for the moment we will conclude it is the discussion. Thank you very much for your presentations and insights