The presence of stratified liquid-gas interfaces in vertical flows poses difficulties to most classes of solution methods for two-phase flows of practical interest in the field of reactor safety and thermal-hydraulics. These difficulties can plague the reactor simulations unless handled with proper care. To illustrate these difficulties, the US NRC Consolidated Thermal-hydraulics Code (TRAC-M) was exercised with selected numerical bench-mark problems. These numerical benchmarks demonstrate that the use of an average void fraction for computational volumes simulating vertical flows is inadequate when these volumes consist of stratified liquid-gas interfaces. In these computational volumes, there are really two regions separated by the liquid-gas interface and each region has a distinct flow topology. An accurate description of these divided computational volumes require that separate void fractions be assigned to each region. This strategy requires that the liquid-gas interfaces be tracked in order to determine their location, the volumes of regions separated by the interface, and the void fractions in these regions. The idea of tracking stratified liquid-gas interfaces is not new. There are examples of tracking methods that were developed for reactor safety codes and applied to reactor simulations in the past with some limited success. The users of these safety codes were warned against potential flow oscillations, conflicting water levels, and pressure disturbances which could be caused by the tracking methods themselves. An example of these methods is the level tracking method of TRAC-M. A review of this method is given here to explore the reasons behind its failures. The review shows that modifications to the field equations are mostly responsible for these failures. Following the review, a systematic approach to incorporate interface tracking methods is outlined. This approach is applicable to most classes of solution methods. For demonstration, the approach to incorporate the tracking method into the field equations of TRAC-M is described in steps. The success of this approach is demonstrated by exercising TRAC-M with the same benchmark problems that were previously used to illustrate the difficulties the code suffered from in the presence of interfaces. Besides improvements to the accuracy of the code predictions, one of the benchmark problems, which simulates a strong condensation at the liquid-gas interface, shows that the code’s runtime is improved significantly where the alternative methods like water packing fails.

This content is only available via PDF.
You do not currently have access to this content.