Doctoral Dissertations

Date of Award

5-1995

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Nuclear Engineering

Major Professor

J. R. Parsons

Committee Members

H. L. Dodds, W. S. Johnson, P. N. Stevens, A. E. Ruggles

Abstract

The Westinghouse AP600 reactor design uses a gravity-forced safety injection system with nozzles in the vessel downcomer. In the event of a severe overcooling transient, this system can deliver soluble boron to the core to offset moderator defect reactivity. To evaluate the outcome of a design basis overcooling event, a tool to predict the transport of boron to the core was desired.

A hybrid computational fluid dynamic (CFD)/fluid element (FE) tracking model was developed for this task. In this technique, the reactor loop and safety injection flow was determined by the 1-D system analysis code LOFTRAN. Given these boundary conditions, a coarse mesh reactor vessel steady-state velocity field and k-e turbulence parameter fields were found using the 3-D computational fluid dynamics code FLOTRAN.

These mesh-point values were used to define the flow field characteristics within the geometry of a fluid element tracking model. In this random walk model, the velocity values were used to define convection while turbulent viscosity and kinetic energy values at a point were used to find a local turbulent diffusion time scale and a distribution of turbulent length scales. A distribution of molecular diffusion length scales were also generated based on the molecular diffusivity. Massless fluid element motion was determined by combining the deterministic convective transport with turbulent and molecular diffusion components chosen randomly from the distributions during an interval defined by the time scale. The transient safety injection fluid concentration distribution was determined from the time and position of the elements as they reached the core inlet plane.

A hybrid model of this nature has several advantages. A CFD model is an excellent tool for predicting velocity fields and turbulent viscosity in a complex. 3-D fluid volume, however, codes of this nature are susceptible to numerical diffusion which lead to misprediction of scalar fields such as concentration. Also, mixing in this type of code is generally based on gradient diffusion, so counter-gradient diffusion cannot be predicted. The tracking model developed, on the other hand, is poorly suited to calculation of any fluid property over a volume but estimate scalar properties over a small region of the model without numerical diffusion effects while allowing counter-gradient diffusion.

An experiment was developed to benchmark the numerical model. A scaling analysis of the reactor system showed that buoyancy and turbulent diffusion effects were of equal importance during overcooling transient conditions. A 1:9 grometric scale model was constructed using air and dense gas to simulate the reactor coolant and safety injection fluid. Experiments were performed over a range of velocities, each chosen to give either Richardson and mixing Reynolds number scaling. Concentration of the dense gas measured at the core inlet using a sonic nozzle/hot film anemometer instrument.

The results of these experiments showed that for buoyancy-dominated flows the injection fluid concentration field was very diffusive, with peak concentrations of about 5% occurring in the center of the core inlet plane, dropping off to about 2.5% around the periphery of the core inlet. For turbulent diffusion dominated flows, the peak concentration e to around 8% on the core inlet periphery, near the injection nozzle azimuthal positions. Concentrations in the core center were on the order of 2%.

The numerical model applied analyze the experimental geometry. Overall, the model proved capable of predicting the experimental transient concentration distribution very well. Transient peak concentration, averaged over the experimental measurement scale, close to the experimental values in all cases. Azimuthal distribution of concentration also well predicted, with a central peak position. predicted for buoyancy-dominated cases and peripheral peaking for turbulent diffusion dominated cases. The numerical model was capable of predicting this drastic difference in azimuthal concentration distribution between cases without any variation of model constants, indicating that the model embodies the physics of the problem. There was a tendency to predict regions of high concentration on the model periphery, believed to be anomalous, and to underpredict the lower concentration values.

This validation against the experiment verified that the numerical model provides best-estimate prediction of injection fluid transport in both buoyancy and turbulent diffusion-dominated flow fields. Given the appropriate reactor physical dimensions, fluid properties, and quasi-steady boundary conditions, the model should therefore be capable of predicting boron transport in the reactor over the entire range of reactor conditions. Application of the numerical model for conservative safety analysis purposes would require consideration of the appropriate uncertainties

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS