 Tak så meget for den kære og kære introduktion her. Jeg er den første speaker i morgen på en af de deler af projekten, som du kommer til at høre om. Og sammen med Dimitri Vald, jeg har været spørgsmål om systematiske udvikling og dokumentation af the theoretical framework for valg af information and structural health monitoring. Men før jeg vil prøve med det, jeg vil tage et kig på tidligere, fordi jeg er noterisk. Og også vil jeg mænke noget, Sebastian ikke mænke i hans introduktion. Det er typisk Sebastian, for at holde nogle ting, en lille smule i hans kig. Men det koncernes den første fase af dette projekte her, faktisk. Det koncernes en peber, vi var skrevet for Offshore Mechanics and Arctic Engineering. Og jeg kan ikke mænke tidligere helt. Men det måtte have været mere eller mindre valg af information af structural health monitoring. Vi havde faktisk skrevet et kig, og vi gjorde det i en vejrige øjelig, isolerende måde. Vi havde taget om et tema, og vi havde taget om eksempler. Og vi havde taget om, hvordan vi skulle skreve et kig, og nu havde vi tekst. Og vi havde selvfølgelig, at du nødvendte at være koncernes, for at være sikker, at ingen anden har gjort det, du vil præsentere og tænke på at blive i en peber. Så jeg sagde til Sebastian, at vi rigtig måtte se hverandre i kronerne af litteratur. Og efter to dage kom Sebastian og fandt den posse peber. Og det var faktisk meget vildt, fordi vi tænkte igen, at vi havde indvendt den døde plade, men indvendt rigtig, at vi havde reindvendt den, som vanligere. Men det betyder ikke, at et tema har ikke hvittet fra størrelsen. Og selvfølgelig er mange ting begyndt i dette størrelseaktion. Og der er meget mere momentum og interesser i temaet, og det er sprættet ud mellem flere af størrelserne til vores samarbejde. Jeg vil gerne præsentere her i den næste halvne år, også om det her theoretical framework for størrelseaktionen. Mange af diger har været, eller mest af diger, hvis ikke alle. Jeg er ikke sikker på, at jeg har været en del af det. Jeg vil starte, og det er lidt svært, fordi mange af diger har også hørt om det theoretical framework før. Vi havde haft en meget stor number of events i størrelseaktionen, som Sebastian har already mentioned, og vi har talet om de forskellige komponenter af projektet, og det theoretical framework har været i spørgsmål mange gange. Så jeg vil give det lidt af et forskellige perspektiv her i enden. Jeg vil starte med at præsentere om informationen, hvor og hvorfor. Så vil jeg spise lidt om beslutningsanalysisk i det simpelste form. Og så vil jeg tænke mig at kombine de to, nemlig systemet og præsentation, og beslutningsanalysisk. Og så vil jeg klare op med at se, hvordan strukturelsehjulet monitorer færdes til det, og jeg vil tænke mig at prøve at give nogle konklusionser. Ja, hvilke vi tager, hvilke vi har tænkt mig, hvilke vi har tænkt mig, i liv som individer, men også som indienneser kan det være forstået som information. Så vi kan alle dele med det her information. Det er prøvet som omkring, og det er startet og akkumuleret i flere form. Men det eneste, vi kan faktisk dele med, er information. Så denne fotograf, det er faktisk mig, det er faktisk også mig. Det er et bære resolusion, okay? Måske ser jeg bedre der, men det ideet her er, at jeg har været konverteret ind til 0 og 1. Og det er essentiellt muligt for noget. Så no matter hvad du ser, no matter hvad du sænker, no matter hvad du ser, du kan også bare consider det at være en stream af 0 og 1, fordi for alle de matter, for alle hvad du har gjort, det er det. Så når vi er mennesker af systemet, hvad vi er virkelig gør, er, at vi er bare manipuleret den enormt ocean af information. Og selvfølgelig er det ikke så svært at være konversion af virkeligheden, eller hvad vi synes er virkeligheden til nogle sort af data, hvilket vi synes er simpelthen virkeligheden er en art, og det er svært. Og det har at gøre med systemet i præsentationen, og jeg vil komme til det. Men essentiellt information er alle vi kan arbejde med, og det er også hvad vi er gør. Og derfor, lad os sige, den term, valget af information, er en god term. Hvordan gør vi det? Hvis vi er tænkt til at repræsente og synterisere information, har vi et par tyder. Jeg har bare her hattet, hvad jeg tror er de superede tyderne, namens de matematiske fremværelse af probabilitetsstatistik og informationsserie. Du skal jo også høre andre propositionser fra andre mennesker, men de er fortæller, at de ikke bare er konsistent og komplet, men de er også superede for den type af applicaterne, som vi begynder at arbejde med. Hvis det kommer til at repræsente kvindelsen, again, nogle mennesker vil klare, at der kunne være valget for at sige det, men jeg vil bare stå her, at resultatet af Thomas Bayes rule er virkelig den enkelte, hvorfor vi kan synterisere kvindelsen, og vi kan gøre kvindelsen, med at gøre kvindelsen af kvindelsen, der kan være kvindelsen over tid og i fritiden. Det er den kvinde instrument i hvad vi har gjort og hvad vi har gjort i dette kvindelsen. Så nu har vi startet ud med kvindelsen, hvordan at synterisere og arbejde med kvindelsen. Vi har brugt om, hvordan at bruge kvindelsen, i order for at gøre kvindelsen. Nu er det, hvordan vi kan kombine kvindelsen og kvindelsen af kvindelsen til kvindelsen. Så når vi har forskellige kvindelser og alternativ, og alle hvad vi har, er selvfølgelig, at vi har valgt en kvindelser, og en mulig mulighed, der kan blive sammen med i fritiden. Vi har ranket kvindelser til, hvilke tilfængelser vi har, så vi kan maximise forfængelsen af vores kvindelser. Det her maximisering skal være, at vi skal være bagefængelige af kvindelser, hvis vi forstår kvindelser som funktion, hvilke tilfængelser vi forstår, og hvilke tilfængelser vi har forfængelser. Det her var realiseret meget snart med Daniel Bernoulli, men kun en kvindelser af 100 år og lidt mere. Det var faktisk matematisk begyndt med Neumann og Morgenstern. I denne resultat har vi været afværende meget, så det er meget fundamentalt. Men nu er de billede stående, de kompriseres, let siges, de bære kvindelser, i for at arbejde med kvindelser af informatiseringen og strukturellængden. Og det er med en resultat, hvor mange af diger har også været i detaljer, som betyder bagefængelser og kvindelser, som er beskyttet med Reifer og Slyfer i 61 år. Erligere arbejde med kvindelser, der nødvendigt ikke er her, Freudenthal, Kornel, i strukturellængden og strukturellængden, var meget afværende deres tid og erfarenhverv inden for at sætte mere større rationeller for kvindelser, når det kommer til management af kvindelser og kvindelser. Men jeg tror, at resultatet af Reifer og Slyfer er også, at mange af diger har experience med at lægge den nu relativt gælde bagefængelser, med Benjamin og Kornel. Det virkelig kom ind i kontekten af kvindelser, og vi har benefittet det fra det, men det har været langt gældig på skjælp i mange størrelser, og jeg vil sige, at konceptet af kvindelser og spørgsmålet af kvindelser og kvindelser ikke har været tilfængeligt til de sidste 10 eller 15 år. Og dette kvindelser er en god eksempel for at få nogle momentum på utilisering af de principaler. Det grundlægge her til kvindelser er, at alle har forstået at arbejde med kvindelser, og det er også det, vi kalder det første kvindelser. Og som du ved, er det første kvindelser i sådan en kvindelserfængelser og kvindelser, og så har vi et system, som vi har tænkt til at manage, med at kvinde de rigtige kvindelser og et system som er tilfængeligt med kvindelser og her i denne kvindelser, vi repræsenterer systemet med den rigtige kvindelser og kvindelser. Og i enden af kvindelser og kvindelser, opgåder på kombination af kvindelser og udkommelser af den rigtige kvindelser, har vi fået nogle kvindelser, som er repræsentet i denne kvindelser. Og hvad vi faktisk gør, og jeg er ikke sikker, at dette er superværdigt, at vi er indenfor indenfor, og vi tror om universitet, i termes af kvindelser, vi ser. Men hvad vi faktisk gør, når vi er kvindelser i sådan en kvindelser, som det første eller, som du ser også, præsteriet, hvad vi faktisk gør, er, at med at købe en kvindelseralternativ, vi er ikke kvindelser med et kvindelser- eller kvindelser med et kvindelser- eller materielle kvindelser. Vi er kvindelser med informatiseringen om kvindelser af den strukturelle komponent, der vi har tænkt til at sætte, for det er optimalt. Så vi må også forlade os om fysiske karakterister og fokus på, hvad den her karakteristiske ting er, i termes af informatisering. Så, i præsteriet af kvindelser, hvad vi faktisk gør, er, at vi er kvindelser af kvindelser af informatiseringen. Og selvfølgelig, kvindelser af informatisering har kvindelser af kvindelser, så vi må finde en kvindelser i, for at optimale vores kvindelser, i for at rankere kvindelser. Hvis vi får nyt informatisering, selvfølgelig, kan vi alle se, at vi sætter nødvendelser af kvindelser. Jeg vil ikke spille tid med det, men, som jeg fortæller, er det en virkelig kvindelser i, hvad vi gør. Og når vi har opdateret vores kvindelser af forældre steder, af det random system, repræsentet af vektor X, kan vi gøre det kvindelser. I denne situation, vi kalder det kvindelser af kvindelser, fordi vi har opdateret vores kvindelser for systemet, vi er kvindelser af, så det er præsterier i vores kvindelser, som vi nu har tænkt til, ikke bare kvindelser, som vi startede med, men også kvindelser, som vi har kvindelser. Vi har kvindelser i at bruge af kvindelser. Og det vil selvfølgelig, at det vil kvinde kvindelser, i vores kvindelser, som kvindelser nu, ikke bare for kvindelser, men også for kvindelser. Denne koncept kan du imod, at det kan være kvindelser, i vores kvindelser, at vi kan også ikke bare kvinde det type af kvindelser, som vi har i vores kvindelser, men hvis vi kvinder på kvindelser, kvindelser og kvindelser, og vi nu gør en interfaceline her, så kunne vi imod, at okay, så skulle vi, i anden, ikke bare til at rankere de mulige kvindelser, som vi har i vores kvindelser, men også kvinde på et particular klasse af kvindelser, nemlig de, der er kvindelser, hvordan vi kan kvinde bedre informationer, potentialigt, i nogle point i fritiden. Hvordan kan det være af benighed? Hvordan kan det fortælle vores ranker af kvindelser? Kan det være muligt for at kvinde et eksperiment for kvindelser af nye informationer, i sådan en måde, at rankerne af kvindelser i dette stedet her, kan være fortællet. Og det er ikke så, at vi kan kvinde på fjernet, og så kvinde ting, der vi ikke kan kvinde i først. Så når vi ser på fjernet, når vi considerer, hvordan vi kan kvinde et eksperiment i fjernet, har vi til at tage baser i prøvst kvindelser, på hvilken type af resultat, måske et eksperiment kan lide til. Så sætter vi op, let's say, en liste af forskellige eksperimenter og forskellige eksperimenter. Vi kan gøre det med at bruge priorindformationer. Så alt til det bedste af vores kvindelser, hvad skulle forskellige eksperimenter være af forskellige eksperimenter, hvilke vi ville considere. Og så, for at gøre forskellige eksperimenter, vi kan bruge en simple prior, en posteriør type af beslutningsanalysisk, med at opdage, at bruge baser, vores probabilistisk modeling af X, skjære for eksperimenter. Selvfølgelig er der mange forskellige eksperimenter, og for denne rejse er det nødvendigt, at vi sætter den forventende valget. Så vi må integrere hele forskellige eksperimenter. Og det betyder, at når vi opdagerer, vi gør en posteriør assignment af probabilisering af de forskellige eksperimenter, de effekter, de eksperimenter har, er, at de vil influence de optimale ranker af beslutningsalternateter, som vi har her i A. Og så tager vi den forventende valget over alle mulige utgående eksperimenter. Så der er ingen ting i det, der er ingen hos hos, men vi er bare systematisk og vi er inden for, hvordan forskellige eksperimenter kan effektere de forventende valget af de forskellige eksperimenter af de forventende valget over de eksperimenter af de forventende valget over de systemet, og hvordan det kan effektere de optimale ranker af beslutningsalternateter i A. Det er et fundamentale princippel i det, vi gør. Og det er, i den måde, meget simpelt. Okay. Valget af informat... Nej, så er det en del af det. Valget af informatiske analyser kommer ud af kompariseringen af hvad der er et ansætte valget af benvendelsen, som er forventet med det beslutningstil, hvor vi er considereret forskellige eksperimenter også, og ikke bare beslutningsanalyser. Så der er forskellige i et ansætte valget af benvendelsen af de to, som du også ser her, er det, vi kalder valget af informatiske. Og selvfølgelig, du finder alle fx refrimer og reformuler af disse forventninger i en anden kontekst fra forventing, beslutning, men det er altså always the same. Det er det, vi er forventet med. Og her, her har vi, i denne forventing, vi rigtig har et koppling fra at vise frihed informatiske, som er den sidste term her, og vise pre-posterie informatiske. Det er med at considerere, let's say, potential improvements of expected value of benefit due to possible experiments. No matter what, we take basis in a system representation, so the vector x, which was in these equations, concerns the, let's say, the probability assignment of possible different states of reality, the real world. We are working the models of the real world and that also includes the probabilistic models and by manipulating these models of the real world by actions, decision alternatives, we are trying to optimize the expected value of benefit. Right? So it's super important that we get good correspondence between what we think is the real world and the models of the real world. And we have different frameworks for doing this. I just want to highlight the framework from the Joint Committee on Structural Safety, which we are using a lot. But it's a challenge and it's a huge challenge. And I just would like to take the opportunity here to underline that I fundamentally believe that it is not right to decouple the system, let's say the choices associated with the system representation and the decision optimization in the decision problems in which we are actually using the system representations. So these two classes of choices cannot be decoupled. And this is what engineers have been doing in the past and it's a typical engineering thing. First we do the modeling. We make a lot of choices. Now we have the model. Now we use the model for decision analysis. But essentially we need to combine the two. Well, we don't really know what the truth is. So in many senses, the real world, we do not fully appreciate how knowledge and information relates to the truth. One could even debate in many cases which knowledge and information is relevant in a given context. There are liars, there are notorious liars and then there are statisticians. So we can do the fine mathematics on basically anything, but the choice of what we are analyzing, the choice of what we are putting into the statistical or probabilistic framework is a choice. And this is where you can start doing the lying at least. Maybe it's not lying, but it's a very subjective choice. In society basically any knowledge and information is on the free market and you also really have to appreciate that. This is so important for what we are doing to appreciate this. And in science and engineering of course knowledge and information could easily be influenced by one fundable, what is expected by stakeholders and what is desired by stakeholders. Along these lines there is also a tendency to mix truths with information and assumptions. So we better maneuver in the right direction and keep this in mind. Information and knowledge really influence all aspects of decision problems. Here in this corner here we have the state of nature. Here is the system. This is what I introduced before and this is what we normally focus on a lot. But we also need to appreciate that we are not the only stakeholders in the process like the risk specialist and maybe also some clients. There are also the societal stakeholders at different levels and the management of the systems which we are dealing with depends on a flow of information. I have tried to sketch possible flows between these different stakeholders and these flows are vulnerable. And you see that there is much more information than just information about the system which deeply affects the management of the systems which we are dealing with. I think it is very useful to realize that information may be associated with different problems and normally this is the ideal case relevant and precise. This is like the deterministic approach which is sometimes reasonable. But most often when we are dealing with information we consider it to be relevant information but it is imprecise and we try to account for the imprecision and we model that probabilistically. Rarely we really take into account that the information could also be utterly irrelevant. So we have observations but it in no way has anything to do with the system which we are trying to identify decisions for. It could also be relevant but simply incorrect. So simply wrong. Or information could be disrupted and delayed. And there are different ways of realizing that if we are trying to optimize actions on a system represented by X in order to maximize expected value of benefit we can support that by collecting information about the states of the system and this is what this decision value of information analysis is all about to find the trade off between collection of information and associated cost and the achieved benefit. But we have to realize that the information could also come from another system. A system which we are not really trying to optimize in any way. And it could also be a system which can be managed. So in some cases information is fed in in the loop by maybe some for instance if we are talking about fake news or malevolence then there could be an active actor with an intent to put wrong information into the system. But no matter what the fact that we may be dealing with information observations which do not relate to the system that we think we are trying to manage is an important point and it's grossly being neglected and we cannot do that. So we really need to appreciate at the different systems are not known. There could be different candidates of systems explaining the same data and the management actions for the different systems would not be the same. So we need to appreciate this and we need to account for all relevant scenarios and include possible adverse consequences which originate from the information flow which I just indicated and focus on how management of information can contribute in optimizing our objectives. When it comes to the management of the system there is no fundamental difference between information which is intentionally wrong and information which is unintentionally wrong so it doesn't matter. It may have a significance in regard to identifying if you want to try to manage where the information comes from and maybe avoid that wrong information is fed into the system then it matters because then it has an impact on your system identification but otherwise decision analysis is completely unemotional so intent does not really factor in. I have earlier advocated for the necessity to deal with system representations where we so if we are operating at the level of graph representations of systems then imagine that there could be a variety of different graphs which we could choose from and they all in a plausible way can explain the information we observe but the graphs might not be identical and we we definitely need to account for the possibility of these different systems when we are optimizing decisions there are different ways of modeling these systems top-down models is one type of models which we have not looked much at in this course section and also the techniques which are associated with top-down systems modeling are also very strong techniques directly in support of value of information analysis this is one direction we have not really taken too strongly in this course section we have been working consistently almost along the lines of the engineering bottom-up modeling when we account for the possibility of different systems we can actually do it and we mark Maze and myself we did some work on this we can go already where we are looking at parallel possible universes represented by these different possible systems and we take into account let's say the disbenefit associated with optimizing decisions under a wrong assumption of the prevailing system by adding the disbenefit into the utility associated with exactly that and this can also be extended to the pre-posterior decision analysis so we have the tools we have the formulations we have some examples we have been doing it let's say in scientific papers but these things still much more need to find their way into practice I think I will try to be a little fast about the fundamental logic of structural health monitoring we know that so well the idea is well appreciated what are we doing there has been at least in some discussions there seems to be a differentiation on what are we dealing with here are we talking about inspection planning or are we talking about monitoring but from a fundamental perspective it's all the same there's no difference but there are some possibilities of taking benefit of let's say different techniques at the same time when we are trying to collect and develop knowledge about the performance of structures and there let's say mixes of what is normally understood to be continuous in time monitoring with point in time inspections could be a very strong approach and what I am addressing here really is concerned about the possibilities and what is normally accounted for that monitoring results may be significantly biased so the bias modeling is let's say a weak point in most of the works I have seen so far and this is definitely one of the things we need to look at more carefully in the next steps points to keep in mind collected as I said information is not correct or biased the actions will not be optimal and could even lead to let's say losses when assessing the benefit or value of different monitoring schemes and corresponding optimal strategies for adaptive actions the only basis which we have for the modeling of the not yet collected information is of course only our prior information and what we have also seen a lot is that when people are showing what is the value of monitoring they kind of take basis in a plausible outcome of a monitoring strategy so assume some results and then look at the results it could be like results corresponding to an expected value and then looking at what would the benefit be associated with that then we cannot do that we will need to take the expected value over all possible outcomes of monitoring results why is it so really important what we are doing in this course section and how does it fit into the overall let's say standardized codes for the management of the safety in the bill environment here in this figure please look the access level of knowledge goes from this point in this direction so the more here we have a lot of knowledge and experience up here we have no experience and little knowledge and here along the y-axis the consequences of failure are increasing and you could say that in this domain here this is a good domain for code this is a good domain for codification and regulation this is where we can feel safe this is where we have a lot of experience but as we move outwards here in this direction we need more and more and can justify more and more time spent in order to make sure that the available information is accounted for consistently we are moving away from where we can feel very safe with what we know into a domain where we really need to do the more exact accounting calculations and utilization of what can be observed and it's out here it's out here that monitoring becomes really important and of course also into the full domain of probabilistic methods so I've written here modeling analysis detail quality control during design construction inspection monitoring during service life here you also see I would like to highlight here I listed up a number of characteristics non-linear structural response accidents, natural hazards and human error structural system as opposed to component failure modes new materials, new designs, new uses if you look at the last three ones and you just briefly consider the needs for sustainable development what we are going to see in the next years is a bunch of new materials, new designs and new uses in structural engineering and let's say one of the benefits of structural health monitoring is that it can release some of the redundant passive safety in structures and the build environment in exchange for some active systems so to footprint from the active systems monitoring are much much smaller than those of conservative engineering stronger cross sections, more material more high quality material etc so structural health monitoring in this picture will be much much more important yeah well this figure here you definitely have seen before as I mentioned already structural health monitoring can save human lives, reduce CO2 emissions and increase competitiveness I think this is a way we still this is a way we need to look at it in the future so it's not reducing cost and saving lives no it's saving lives, reducing CO2 emissions and then comes cost considerations we all know these potential applications maybe I'll just summarize so service life management of structures prototype development, code making, code calibration early warning systems and of course dedicated leave for optimization of service life maintenance I will drop these and go directly now to the conclusions so knowledge and information it forms a basis for decision making in a very fundamental way Bayesian probability theory is an adequate framework for representing knowledge and also knowledge development true collection of new information structural health monitoring aims to do exactly that namely to develop knowledge in support of the management of structures and the built environment and the value of information analysis from the Bayesian decision analysis the field of Bayesian decision analysis it really facilitates in a very consistent and elegant way assessing and optimizing the benefit associated with structural health monitoring information and knowledge modeling which has not been focused too much on but I think it will be in the future especially in the future called technology 4.0 or 5.0 information and knowledge modeling are really essential parts of structural health monitoring and do not forget but try to account for it in your understanding in your modeling and in your decision analysis that we are dealing with a range of possible competing systems we don't know exactly which one we are dealing with but collection of information can help us to identify the most likely candidates and to deal consistently with these possibilities Thank you for your attention Thank you very much for the description of your presentation of the theoretical framework We have time for questions and also microphones for questions I have one and there are a few other ones Thank you Michael for the presentation I have a question that is not related with the framework but since you are an expert on this field and you are the best person for this type of question I have been dealing with a lot of work on it but facing some work that I am reading on machine learning and artificial intelligence My question to you as an expert on this How you see this possible crossing of these two fields You talk about there on the bottom up or up bottom approach which are about the model that you are on the system Do you envisage that these two fields might cross this framework and the artificial intelligence I know mainly machine learning because I am reading a lot of things and I know that there are a lot of project proposals about this machine learning approach for data analysis What is your opinion? For sure, it is happening while we speak We are working with the coupling of top down modeling and bottom up modeling and as we are more and more in the position really to harvest like terabytes of potentially relevant information data mining is of course super important and this is where we are important as engineers we need to use sound reasoning also to synthesize the information coming out of data mining out of the big data in this combination and when you go in and analyze the typical algorithms which are used in data mining they pop out one model explaining different variables how they are related and they optimize using max posterior concept or something of the like and then you get your model but of course when this optimal model is selected by the algorithms quite often a larger number of potential models are thrown away they all have let's say smaller likelihoods than the one identified but they are not implausible then they can easily explain the data as well in many, many cases and in many cases also we see that the decision ranking would be different if you would have assumed one of these alternative models so yes, we need to dig in much more into the potential merits of data mining in our systems modeling and I'd say especially when it comes to structural health monitoring it's like the area where we really, really have the possibility to use all sorts of monitoring Thank you very much, another question Yes, follow Please go to the microphone This is the short way Okay Can you hear me? Yes Okay, I have a question It's more related to possible applications because I have in mind last year everybody knows in Italy one bridge collapse around mid-August and after that there was a long discussion about the use of structural harmony might have avoid the collapse of the bridge or at least not the collapse but maybe to do some maintenance activities or to avoid the debts and so on Later on there have been some studies and they figured out that eventually some monitoring on the cables one of my colleague from Polytechnic already did that analysis and interpreted data but it was very far away from saying that the bridge would have collapsed just identify some change of model frequencies and so on So now I'm thinking how do you eventually could use the techniques for decision making in large scale applications like for example bridges and because data nowadays more and more bridges are monitoring there are lots of data but then the difficulty is how to interpret this data and eventually push into the decision especially to the owner of the bridges they might not want to do that because of cost In fact I was looking last week a curve of the cost of maintenance in that bridge before the company was in charge of the bridge was public, there was investment after it became private the cost of maintenance dropped completely and so I don't know how this maybe helped in that direction It's more a comment than a question but if you would say something that would be great Well thank you for that comment question I definitely don't have the right answer if you are asking specifically about this bridge in januar if structural health monitoring would have made the difference maybe it would have been able to detect something early enough to stop the traffic I have no clue but the important I think even if we would have had let's imagine that structural health monitoring would have been like a requirement it would have been in everybody's minds all entities, the public, decision makers, owners, operators they would all know that yes we are doing structural health monitoring and we would be doing it then what is a very very important part of structural health monitoring or let's say inspection planning or any activity related to achieving information about what is going on in the structure you are trying to understand your system and in that process which is a super important process you will realize if there are potential degradation mechanisms going on in your structure you put everything on the table all experts are involved you put all degradation mechanisms on the table you try to map those which could be relevant for the structure and then you also identify so using this monitoring technique using this inspection technique are we able actually to see anything about what is going on until the thing actually breaks down and this process you could call it a risk screening process will identify if in the first place structural health monitoring can do anything for you in a given context and if it cannot then it is also really really important information because then you simply if you cannot inspect let's say an important critical condition in your structure it simply means that either you take the structure out of use or you trust that it's okay I don't know by experience by statistical information or you find other ways simply to make 100% sure that there are 100% but at least to a reasonable degree of certainty that the dangerous degradation mechanisms are not taking place so I think yes in either case a culture where health monitoring plays a big role would have helped also for the bridge in January because it would have directed some important attention to the potential problems with the bridge much much more clearly yes of course okay thank you very much