Computer simulations: a cornerstone of nuclear power station safety

A thorough understanding of how a nuclear power plant behaves during normal or abnormal conditions is a key component of plant safety assessments and this requires computer simulations. To that aim, the employed computational models must be validated against measured data from the plants. But such data is usually very rare and does in fact simply not exist for most of the accidents considered in safety assessments. Therefore, tests in experimental facilities are instead performed to provide the necessary data for the verification and validation of computer codes. As example, experimental programs aimed at investigating loss-of-coolant-accidents (LOCA) have been and continue to be subject of intensive research. These events are not expected to occur at all during the entire plant life-time but are postulated in order to verify that even during such events, the reactor design would allow to withstand the loss of any of the critical components which are in place to protect the public and environment. Thanks to experimental programs, such postulated accidents can now be modelled much more accurately on computers. Simulating such very rare events as well as other types of phenomena that can occur during nuclear reactor operation, is the speciality of Hakim Ferroukhi and his team. Ferroukhi, the head of the reactor core behaviour group at the Laboratory for Reactor Physics and Systems Behaviour at the PSI, and his team have as primary research activity, the development and validation of advanced computational methodologies for safety assessments of the Swiss nuclear power stations.

The constant development of computer technologies and computational models has led to simulations that nowadays allow for a much greater level of detail. For example, it is now state-of-the art to simulate the reactor core at the level of a single fuel assembly in three dimensions. This makes it possible to determine more precisely the safety margins for each fuel rod against the prescribed limits. Obviously, regulatory authorities also need to constantly follow the progress achieved with modern simulation methods and integrate this knowledge into their supervision activities. In Switzerland, this is required by the Nuclear Energy Act: only by keeping pace with modern developments can the authority make sound decisions on the licensing of new procedures or components (e.g. plant modification, modernization) and ensure thereby the safety of the plants for the long term.

Paradigm shift: from conservative to best-estimate models

As the boundaries of what is possible with simulations continue to be pushed forwards, the demands for accuracy in assessing the safety of nuclear power stations are also increasing. In the past, computational models used for such assessments were deliberately made pessimistic. In other words, the simulations were conducted with physical models and/or analysis assumptions selected such as to “force” the results to be conservative i.e. to produce small margins to the safety limits. However, this approach could eventually lead to over-dimensioning of, for example, the required emergency safety systems designed to deal with a loss-of-coolant-accident.

As a result of very extensive international research, the relevant phenomena can now be better understood and on that basis, the simulation methods can in turn be enhanced such as to allow for a more optimized design of nuclear power plant components and associated safety systems. And consequently, the trend has now shifted towards so-called best-estimate models. This means that one aims at describing and quantifying the physical processes in a reactor more accurately, allowing thereby to relax some of the pessimistic assumptions that were previously made due to lack of knowledge. However, a prerequisite for safety assessments based on best-estimate methods is to gradually complement these with rigorous quantifications of uncertainties in the simulations results.

A second trend is to move towards multi-physics approaches. In the past, it was indeed common to simulate with a given computer program the flow and heat processes (thermal-hydraulics) responsible for cooling the reactor core. But for the behaviour of the neutrons in the reactor (neutronics), only approximate models based on information obtained from another separate simulation tool, would be employed. Similarly, thermo-mechanical interactions between the nuclear fuel and the fuel cladding, due to e.g. thermal expansion of the heated fuel, would be simulated in a stand-alone manner by another separate tool. In other words, each tool was designed for its specialized field of physics, using only approximate models of the physical processes treated by the other tools.

With best-estimate models and multi-physics simulations, the objective is to integrate all these processes in the same simulation so that interactions between various physical mechanisms, which factually take place during operation or events, can be modelled more accurately. Consequently, computer code systems have to be developed that connect the individual simulation tools. However, this is not necessarily an easy task as many of these tools were developed in isolation for decades as they were never intended to be combined later on down the line. Also, the quantification of uncertainties becomes more important and more complex since the connection of several computer programs means that the range of uncertainties will also increase significantly.

Despite all these difficulties, Ferroukhi’s team has in recent years succeeded in developing a novel code system applicable for all types of reactors operated in Switzerland and which combines a state-of-the-art neutronic code with an advanced thermohydraulic simulation program.

Barely detectable instability successfully simulated

The researchers recently reached a major milestone with such type of tools: they managed to simulate the emergence and progression of a so-called regional instability in the reactor core, where the thermal output generated is not distributed evenly across all fuel elements. Furthermore, they could show a very good agreement between the simulation results and the experimental data.

A regional instability consists in the occurrence of power oscillations that might not be easily detected: the thermal power increases in a group of fuel elements and decreases simultaneously in all other fuel elements located in the opposite region of the reactor core. Such stability problems occur only under very specific conditions far from normal reactor operation. But should such oscillations occur and not be detected in time, this could lead to overheating of the fuel rods. Therefore, a correct prediction of this kind of phenomena imposes very high demands on the simulation codes.

The models and methods that have now been developed within the STARS program have not only been shown to reproduce accurately such events. They have also allowed identifying and understanding better some of the basic physical mechanisms that guide the reactor behaviour under these conditions. The researchers believe therefore that such advanced methodologies will gradually enable more accurate assessments for a wider range of applications that are of fundamental importance for the safe and economic operation of nuclear power plants.

Text: Leonid Leiva