12.147. Severe accident modeling has received increased attention by the U. S. Nuclear Regulatory Commission since the Three Mile Island and Chernobyl accidents. Such modeling has helped to provide guidance for reactor design and to supplement risk analysis studies. Generally, individ­ual codes are used to describe the various steps in an accident sequence and coupled together. Since this is a rapidly changing field, and many of the codes in use differ significantly in their modeling approach, our purpose is merely to outline the functional requirements and cite some examples.

12.148. A major objective of the modeling is to describe the response of the containment building to a core melt accident in a light-water reactor. We are reminded that this building is the final barrier to fission product release to the environment. Therefore, it is necessary to describe, normally by coupled computer modeling, the stepwise progression of accident events

that lead to the buildup of containment pressure. An associated objective is to provide a basis for risk assessment studies (§12.208). Detailed me­chanistic code modules, developed in close connection with experimental programs, are used for benchmarking simplified codes used for source term quantification and risk assessment.

12.149. Typical code systems provide for input from a fission product inventory code such as ORIGIN to an in-vessel thermal-hydraulics model of the accident which may be initiated by such conditions as a station blackout sequence in which all power is lost. Following failure of the vessel, containment effects become important. A lumped parameter approach is followed both for the in-vessel and containment analyses. The respective volumes are divided into interconnected compartments or cells and mass, momentum, and energy conservations calculations are performed for each cell (§12.137).

12.150. Several different code module systems are available to model severe accident phenomena, with the pressure load on the containment generally as the first objective. The modeling approaches used vary in several ways, but a description of code details is beyond our scope. There­fore, we will merely identify several of the code packages used and indicate the “flavor” of the approaches followed. The MAAP code system was developed as part of an Industry Degraded Code Rulemaking Program (IDCOR) for predicting severe accident source terms [16]. The CONTAIN and MELCOR codes were developed for the NRC [17]. To determine containment pressurization, such in-vessel phenomena as blowdown and boil-off of the primary and secondary coolant must first be modeled. Next, the effects of zirconium oxidation, core heatup and meltdown, and debris relocation are described. Since the pressure load on the containment de­pends on both gases generated and energy produced, steam and hydrogen produced during debris-water interactions and molten core-concrete in­teractions are modeled. Energy production and removal as well as the impact of engineered safety features must be considered.

12.151. The modeling of fission product release to the environment from a severe accident is an essential feature of the evaluation of reactor risk (§12.231). In general, the same code systems described above for contain­ment loading evaluation are used to describe fission product behavior. For example, chemistry considerations are integrated into the MAAP code with the concentrations of 34 species followed. It is emphasized that over the years, numerous other codes have been developed to describe fission product transport up to containment failure. However, for emergency re­sponse planning, meteorological considerations must also be modeled. Codes such as CRAC2 [32] develop off-site dose probabilities by sampling local meteorological data for a Gaussian dispersion model (§12.160).

12.152. As a result of the large number of interrelated processes that require description, severe accident modeling tends to be inherently com­plex. Furthermore, it is necessary to use many simplifying assumptions in the representation of individual processes which may differ among code packages. The methodology used may also differ. Therefore, the overall results yielded by one code package may differ somewhat from the results obtained from another. Thus, users should study the details of a given modeling procedure and appreciate the confidence levels developed.