Dynamic Benchmarking of Simulation Codes

Abstract

Computer simulation of nuclear power plant response can be a full-scope control room simulator, an engineering simulator to represent the general behavior of the plant under normal and abnormal conditions, or the modeling of the plant response to conditions that would eventually lead to core damage. In any of these, the underlying foundation for their use in analyzing situations, training of vendor/utility personnel, etc. is how well they represent what has been known from industrial experience, large integral experiments and separate effects tests. Typically, simulation codes are benchmarked with some of these; the level of agreement necessary being dependent upon the ultimate use of the simulation tool. However, these analytical models are computer codes, and as a result, the capabilities are continually enhanced, errors are corrected, new situations are imposed on the code that are outside of the original design basis, etc. Consequently, there is a continual need to assure that the benchmarks with important transients are preserved as the computer code evolves. Retention of this benchmarking capability is essential to develop trust in the computer code.

Given the evolving world of computer codes, how is this retention of benchmarking capabilities accomplished? For the MAAP4 codes this capability is accomplished through a dynamic benchmarking feature embedded in the source code. In particular, a set of dynamic benchmarks are included in the source code and these are exercised every time the archive codes are upgraded and distributed to the MAAP users. Three different types of dynamic benchmarks are used:

• a plant transients,


• a large integral experiments,


• and separate effects tests.

Each of these is performed in a different manner. The first is accomplished by developing a parameter file for the plant modeled and an input deck to describe the sequence; i.e. the entire MAAP4 code is exercised. The pertinent plant data is included in the source code and the computer output includes a plot of the MAAP calculation and the plant data.

For the large integral experiments, a major part, but not all of the MAAP code is needed. These use an experiment specific benchmark routine that includes all of the information and boundary conditions for performing the calculation, as well as the information of which parts of MAAP are unnecessary and can be bypassed.

Lastly, the separate effects tests only require a few MAAP routines. These are exercised through their own specific benchmark routine that includes the experiment specific information - and boundary conditions. This benchmark routine calls the appropriate MAAP routines from the source code, performs the calculations, including integration where necessary and provide the comparison between the MAAP calculation and the experimental observations.