Fukushima: Interview with GRS expert Dr. Martin Sonnenkalb

Safety research for nuclear installations is one of GRS's main areas of work. For several years, scientists have been looking in particular at the Fukushima nuclear accident. One of the experts in this field is Dr. Martin Sonnenkalb, Head of the Containment Department. To put it simply, he and his team study the processes that take place during accidents inside the containment of a nuclear power plant. On the occasion of the 7th anniversary of the Fukushima accident, we talk to him i.a. about what photos from the inside of the reactor buildings have to do with his work.


Hello Mr Sonnenkalb. You have just returned from Paris, where you met with representatives of other international specialist organisations on the topic of the nuclear accident at Fukushima. What exactly was that about?

Specifically, this was the final meeting of an OECD/NEA project (Note: This was the "BSAF - Benchmark Study of the Accident at the Fukushima-Daiichi NPP" commissioned by the Nuclear Energy Agency of the Organisation for Economic Cooperation and Development (OECD/NEA)). In this project, the accident sequences in the nuclear power plant units at the Fukushima-Daiichi site in Japan are analysed using various simulation codes. With the help of these analyses, we hope to obtain detailed information, among other things on the destruction of the reactors, the dislocation of material inside the reactors, and the releases of radioactive substances, in order to support the plans for the decommissioning of the units in Japan. A second focus is in how  to gain new insights into accident sequences, especially in boiling water reactors, and thus to reduce possible deficits in the simulation codes that we use. For example, the findings from the on-site inspections carried out by TEPCO are used for this purpose and requirements for sampling and inspections during dismantling are formulated.

Why is it important to know how the accident developed?

There are several reasons. As was the case after the accidents at Three Mile Island (TMI) in 1979 and Chernobyl in 1986, the aim of the post-analysis of accidents is to gain precise knowledge of the initiating events and the individual processes for each type of reactor in order to be able to avoid the possible occurrence of similar damage in similar plants in the future. On the other hand, there is a scientific interest in identifying uncertain or as yet unknown events in the course of accidents more precisely in order to then develop any missing models in the simulation codes. The further one moves away from the original condition of an installation, the more difficult it becomes to model an accident sequence with the aid of experiments. If fission products also come into play, it becomes even more difficult to investigate such processes realistically and experimentally. Therefore, any knowledge that can be gained from accidents such as in TMI, Chernobyl or Fukushima is very valuable. Of course, this does not mean that we as scientists want something like this to happen -  quite the contrary.

Within the framework of the exchange of experts in the OECD/NEA project, you also review your simulation codes. How should we imagine this?

We are participants from different European countries. But there are also representatives of organisations from America, Korea and Japan present. In the OECD/NEA project, four or five different simulation codes are used, all of which have the same objective, namely to calculate an accident sequence in a nuclear power plant. Each of the participants tries to simulate the accident sequence with his tool in the best possible way. For this purpose, available measured values and known events from the sequence of events in the plant are used for comparison in order to show that what was calculated is plausible and that the calculations of the simulation codes essentially reflect the conditions found in the plant. With each new finding obtained from the inspections that have been carried out in the units  at Fukushima-Daiichi, it has so far been possible to further narrow down the possible spectrum of the accident sequence.

This comparison can turn out well or not so well for each of the three destroyed units at Fukushima-Daiichi. And each of the three accident sequences has specific phases or events where the results of the analyses are still widely scattered. On the one hand, this is so because the models are not yet detailed enough. On the other hand, it is because opinions on what actually happened during the accident are not yet consolidated and the measured values from the plant are incomplete due to the almost complete failure of the power supply.

So one could say that the data from the accident help to optimise the calculation models?

That is correct, this is the second objective of the OECD/NEA project. What we want to do is to eliminate modelling deficits in the simulation codes as far as possible and to extend and improve the models. The aim is to use findings from the dismantling of the units or, at the moment, from inspections inside the units. Each new inspection brings at least one new surprise, like when a largely intact fuel assembly top end piece was found in the room under the reactor pressure vessel (note: RPV for short) of Unit 2 in January of this year.

Picture of a fuel assembly top found in the area underneath the reactor pressure vessel of unit 2 of the Fukushima plant, using a remotely controlled device.This part of a fuel assembly that you have just mentioned was clearly visible in an image that TEPCO published of the inspection in Block 2. To what extent do such data help your work?

They influence our work to the extent that they help to exclude certain accident scenarios but also to consolidate certain assumptions.  The discovery of the fuel element top end piece in Unit 2, for example, was an event that we had not previously had on the radar from our analyses. Until then, it had been considered rather unlikely that such relatively well-preserved structures could be found outside the RPV in the containment, because this requires a relatively large opening in the RPV through which that part can pass. And the same applies in principle to Unit 3, where the findings from the evaluation of the inspection in the summer of 2017 are actually a bit weirder. Below the RPV, two clearly assignable parts of the control rod shielding pipes with control rod drive shafts with a diameter of about 20 centimetres each were found. These are normally arranged in the lower part of the RPV and must also have got into the containment through an opening after RPV failure.

What are the current questions?

We have several open questions. A central issue is the degree of core destruction. It will remain unanswered until you can take a look at the top of the RPV for the first time. However, the basic views on this no longer differ widely: The reactor cores in Unit 1 and Unit 3 have been massively destroyed, with Unit 1 probably to the most severe extent. The least damage has occurred at Unit 2 - whereby "least" is relative. On the question of release of core melt from the reactor pressure vessel into the containment, it is also clear today that this has certainly taken place to a large extent in Unit 1, even though there are still no pictures. In Unit 3, photos taken last summer from inside the containment confirmed that a very massive destruction and melt relocation into the containment has also taken place there. In Unit 2, as I said, it is not yet certain whether any material has been transferred.

What happens when you "feed" the simulation codes with new numbers, data, facts?

When we get new information, unless this is about any “events” as such – like e.g. "Venting of the containment took place at 7.43 a.m." - the person working with the code cannot "feed" his computer program directly with these numbers but records the measured distributions or values as comparison points in order to compare his calculation results with them. And the "input of numbers" always takes place at the very beginning, when the geometry of the plant (editor’s note: diameter of the reactor, number of fuel assemblies, and pressure and temperature data, etc.) is mapped in the simulation code.

The time range across which we want to analyse the accident sequences is currently three weeks - starting on March 11, 2011. The actual calculation with the simulation code may then take from a few days to several weeks. This also depends on how detailed a model is built, how powerful the code is, and which processes are calculated. The calculation of the core destruction phase, for instance, always takes a little longer.

And what happens to the simulation code once it is complete?

A simulation code is never fully complete. As a rule, these computational programs are always subject to further development, numerically but also in terms of content. And as I said before, during the course of serious accidents with core destruction, there is a whole series of processes and phenomena in the late phase that are  not completely clear; new findings from dismantling or other experiments enable us to close gaps and improve models. As a rule, there is always a product that is ready for use available, which is then merely further refined and improved.

When was the last time you were in Fukushima?

I went to Fukushima twice in 2015. Once with an OECD expert group dealing with open questions relating to research on serious accidents. More specifically, on what information should be gained from the dismantling of installations that can help us close our knowledge gaps? We had the opportunity to enter the turbine building in Unit 1, always complying with the enormously strict radiation protection regulations on the plant site, of course. In will never forget the impressions I got from looking at the severe degree of destruction, the water markers on the electrical switching cabinets at a level of two metres, the pushed-in entrance gate, and of course the many broken containers which were destroyed and shifted by the sea water. Even if you have often seen it in photos, you become aware once again of the forces that worked there during the tsunami.

During another visit to within the framework of the  OECD/NEA BSAF project, we were able to inspect Unit 5, which was not destroyed. We wanted to know what the situation was like inside the containment and, for example, what the reactor looked like from below. We then spent two hours under full protection in the containment so that we had an idea of the space of the rooms for when we later create our computer model. This visit was particularly important for the analytical work.

Thank you very much for the interview!


Sven Dokter, GRS