 So we have seen the deterministic approach in the way hazards are treated in the safety analysis. We can have a look now on the complementary tool to the deterministic approach, which is the probabilistic approach. Probabilistic safety assessment enables the estimation of accident scenarios probabilities. It considers families of initiating events and includes the probability of failure of the various safety functions. The probability of failure of safety functions is derived from failure trees of individual components coming from experienced data. Let's look at more details. So the purpose of these trees is to determine the probability of consequences on the fuel or on the core. And as mentioned in the difference in depth, if we start from an initiator in this example, there will be a serious dilution that is a reduction in the concentration of the boron in the reactor core. There will be a first counter-mercer such as, for instance, the automatic stoppage of this dilution. And if this one is not efficient, a second line of defense will occur with the initiation of a verification system that will inject boron directly into the core. And so depending on the probability of failure or success of these two counter-mercers, we can calculate the consequence. So this is the way this event tree is built. So starting from initiator and there are a very large number of initiators considered in a complete PSA. And the probability of occurrence of this initiator is taken from the experience. So if the first counter-mercer is successful, of course there will be no consequence. If it's failed, the second counter-mercer will intervene. If it's a success, there will be no consequences. But if it's failed, there will be some consequences such as core damage or release of activity. Now the question is how to calculate the success or failure of this counter-mercer. This is done through fold trees such as this one where to calculate the probability of success or failure, in this case failure of this counter-mercer. The system is split in its various components and the probability of failure of any of these components coming from the reliability database of the plant is introduced in order and combined in order to calculate this probability of failure of this counter-measure. Of course we are taking into account not only the probability of equipment failures, but also the probability of human error or wrong human action. So through this is a very schematic view of probability safety assessments which can give at the end calculating all the initiators to have the global probability of unwanted consequences and also the possibility to prioritize the various sequence to look at those which are the most probable and of course to take some additional measure to reduce the probability of the most important sequences. So there are several types of probabilistic safety assessment. Three main, the level one, look at the probability of the chord degradation. The level two goes a step further and take into account the containment function because you can have some chord degradation and if the containment function plays its role there will be no release but if there is a bypass or a leak in the containment you can have a radioactive release so level two allows to calculate this probability of radioactive release and then there is a level three which calculates the impact on the public environment. Of course this depends on the direction of the wind during the release and the probability of the wind is not the same and also of course of the distance of the population. Initially PSE was considering only internal events but now more and more probability safety assessment including hazards and mainly internal or external hazards such as flooding or seismics are being developed and now we have more global probabilistic safety assessment covering all the different risks. In developing PSE and as you can see from the various elements there are six main steps. The first is to identify the initiating events and again even if we group then in some family total there are several hundreds of initiating events. The second step is to construct some scenario taking into account some thermodynamic calculation to see the consequence of the initiating events in terms of for instance temperature or defecting cooling. Then we need to identify the mitigating features the countermeasure mentioned in the previous slide the various system that will intervene to counteract the consequence of the initiating event. This would allow the construction of four trees through the composition of the different system and their different components with the reliability data of the pumps, pipings, instrumentation and control system combining all that can construct the four trees to calculate so the probability of failure of the mitigating feature. And then we will have the defining the different hierarchy of sequences to look at those which are the most significant and which are the first to be looked at to define potential additional measure. Probability safety assessment has a lot of interest and provides very good insight into the safety of a nuclear power plant. It's first an important complement to the deterministic approach. As mentioned previously the deterministic approach focuses on some specific type of accidents and families of accidents and the main purpose of them is to define and to design the safety system that should be line of defense in the various scenarios. Second interest is the identification of weak points in the old plant through this hierarchization of sequences to identify the most important ones will allow to take additional measures to reduce this probability. Third interest is the identification of common mode failures. If you have a redundant system but they have some common points such as for instance the air compressed system that will serve several valves so identifying these commonalities which are sometimes hidden in the complex arrangement of all the systems is something very important to ensure that redundant systems are really redundant not only by their designs themselves but also by all the utilities and facilities that needs to be available for the redundant system to function. And if these common mode failures are identified that will underline the need to put some diversity between various systems and for instance if you have a first train of a system which is powered by electrical pumps it's better that the redundant one be powered by a steam-driven pump that will avoid any kind of common mode failure through different way of actioning the pumps. Last but not least is to allow a more balanced design globally of the plant ensuring that the most important sequence have about the same order of probabilities and for instance if a first run of a PSA shows that you have just one sequence that represent more than half of the global risk you will certainly do something to reduce this part.