 Hello everybody, my name is Jorge Mendoza, I'm a PhD student at the Norwegian University of Science and Technology and I will present this case study I'm developing together with my supervisor Joakim Kelle, named Respace Design of an Optionary Terminal Support Structure Using Value of Information. This is the agenda we're going to go through. I will first talk about the context of the case study, the philosophy we are following and the goals we want to achieve. After that we will talk about the methodology and how to implement influence diagram to assess the value of information. In the third point we will talk about the value of structural health monitoring information and also the in a bit redundant about the value of information already at the design phase of a structure and from the point of view of the owner and the concessioner and we will finish the presentation with some open information and discussion. So the aim of this case study is to show how the value of information analysis can be already used to enhance decisions already at the design stage and for that we chose to regard the design of a monopile support structure to support an offshore wind turbine. For those that are not familiar with offshore wind energy I will spend some slides explaining the background of the case study. I guess that's needed for those that are not familiar. So in order to design a monopile we want to have the first natural frequency of the design in between the 1P and 3P excitation regions and I will talk a bit more about that but due to large uncertainty especially in the soil structure interaction this estimation of the first natural frequency can only be done to a certain degree of confidence. On the left figure we can see a typical relation between the rotor speed and the wind speed for each control turbine and on the right figure we see an accountable diagram that relates the frequency of the structure for each rotor speed and also the first natural frequency in this blueish line. So regarding 3P and 1P excitation which are due to the blade passing, one blade passing and for a three blade turbine angry blade passing generates an excitation that we need to avoid from a dynamic point of view in order to avoid the resonance. We are left with three regions the so-called soft soft regions soft stiff and stiff stiff. Designing the soft soft region are not feasible on all the reasons due to excessive flexibility of the compliant tower. The designs in the stiff stiff region are not cost effective due to excessive over design from other limit states point of view. So we are left with the soft stiff region. That's for increasing the rotor diameter becomes narrower and narrower and there is really not large margin in which we can design the structure to be in. So an example Siemens 3.6 megawatt turbine. If we consider the minimum rotor speed and the normal rotor speed we are left with a soft stiff region of 0.21 Hertz and 0.25 Hertz. If we would design a structure to have a natural frequency right in the middle we would be left with less than 10% margin. The American Society of Civil Engineering recommends 10% for onshore-based turbines. Of sure of course we expect to find even higher uncertainties and therefore we need to understand better what are the consequences of this uncertainty. We cannot just be left with a deterministic margin. So this is our probabilistic decision scenario. We have uncertainty in the design performance and we need to manage that uncertainty. The typical ways to overcome uncertainty in a design phase are either by investing in a more robust design that is over engineering the solution or investing in information acquisition that can reduce the uncertainty already design phase. Both strategies can be used to reduce the systemic uncertainties and increase the reliability. In this case I want to do a valid information analysis that give us a trade-off between both strategies and let us assess which one is more cost-effective and yields lower risk. This is the most famous valid information scheme that's in the last workshop we summarized into this table at least the decision scenario was summarized. So in this case the decision maker is the developer of the offshore wind farm. We are in the design phase of the decision point in time is design and we want to design to optimize throughout the lifetime of the structure. In this case I just mentioned the design and operation phase and the objectives are to minimize the cost and risk and to maximize the revenue. An option in there is crucial to optimize or to maximize profit to minimize the levelized cost of energy. The remedial actions we have are to modify design or to modify the operation range and the information requirement strategies that we have at hand. In this case I consider of course to do nothing or to further test the soil or to perform an output on the analysis in a nearby already installed monopiled foundation that we can use to assess the real or to reassess the uncertainty at a near location. Monopiles are designed usually and grouped into clusters with similar soil characteristics and similar dimensions and heights and water depths. So within a cluster we can use information to lower or to extrapolate information. The methodology applied we are using Bayesian networks and here I will describe the main features or the main characteristic of an influence diagram that we can use to map or to represent decision as an area. For those that are not familiar these are the typical components. We have chance nodes to represent our discrete random variables. We have the choice nodes that are our decisions. We have the consequence nodes in which we can assess the utility and to decide for better approach strategies and we have the arcs which are imply direct probabilistic dependency. This is the influence diagram of the case study. So in the color scheme we have in yellow the nodes related to design. So we have a set of designs that we want to assess. Each design has associated stiffness which is in this case a random variable and this accost. On brown we have the soil nodes. We have the initial soil parameters that we can only assess in a very brusque way with low fidelity. And then during the life of the structure there will be phenomena that will modify those parameters. So there will be soil softening stiffness. There will be scour development and as you can see those nodes are related to the stochastic and wind and wave nodes in blue. We have in green the soil testing that we can either perform or not and accost associated with it. And together the design and the soil structure interaction will let us assess in a probabilistic way the first natural frequency. On the left part of the influence diagram we have in blue the nodes related to operation. So we have the for example the cutout wind as a decision variable, decision node and energy production as a consequence node. The first natural frequency and the operation chosen let us assess limit state in this case adverse states of the resonance hazard. And this resonance hazard together with the stochastic environment let us assess the reduction of a deep life. That is what in the end causes the consequences that we want to minimize. So this is an example for example in the natural frequency estimation we will have a probability conditional on the design which is on the estimation of the soil parameters and on the selection of the testing alternative. The discretization of this cumulative distribution will be given the conditional probability table that are going to be used in the influence diagram. And it's very important to notice this conditional dependency because in the end this will give us the value of the structural health monitor. In the end we want to have something as simple as that. We want to have a probability based conditional to design soil and test performed. And we want to estimate the reduction of fatigue life and with that assess the best design strategy. So a tradeoff between design and information acquisition based reduction of uncertainty. So what is the value of the value of information? We can map the complex interdependencies and this is very important. We can see the different variables that affect our decision scenario and it's rates how variations in the nodes will affect our decision. So this is also related to the sensitivity of the input variation and the uncertainty modeling. And also innovation belief network we are able to assess how the system responds to evidence. So that was my presentation. Please look at the fact sheets for more information. I left here the link. It will be updated hopefully frequently. And if you have any questions it's now their moment. Thank you very much. Consequences and the costs. What are the related uncertainties based on your experience in this case? The related uncertainties? When you have consequences. So you have costs related to energy losses. So I don't know you. You are the package of costs which here I cannot I mean what are the related uncertainties if you assess consequences of your case? Yeah it's a difficult one. If you want a good question. Maybe this is an issue for later. It's probably of course meant to be. If we use linear utility relation we are only looking at the expected of the consequences of course. But it's of course going to be a more complex scenario. And in here we have costs associated to the benefit to the revenue in the energy production. We have costs of the test. We have costs of the design and we have consequences of the reduction of fatigue. So it's going to be definitely a major part of the development of this case study. I understand the case study is not 100% related to the interest of RISA. But I hope at least it's a good pedagogical case study for understanding how the value of information analysis can also be used already at the design phase to better rational decision making. Thank you. One challenge is which we are dealing with and which is always an issue when we are trying to model how new information can be related to a benefit at some sort of time scale for the decision maker. That is to define what is actually let's say the package of information which we can collect. So there are so many different things which can be observed all the different indicators. But how do we define in a stringent and consistent manner the package of information which will facilitate that we can make a decision on whether we want one package or two packages or three packages because they cost money. Every package of information costs money. And also the dependency between the information contained in the different packages is something that we of course have to account for because if that information is fully dependent or just strongly dependent we don't have any benefit from buying more than one package. So these are surely some of the issues which you guys have been fighting with in this application also. But thank you very much for your presentation and I think we should give you a hand again.