Design of an Experiment to Measure Heat Transfer to a Massive Object Engulfed in a Full Scale Regulatory Fire

This paper describes the design of a full-scale experiment to measure the heat transfer from a large pool fire to an engulfed 1.2 m diameter, 2.54 cm wall thickness hollow cylindrical calorimeter. This work is focused at determining whether temperature measurements made on the interior calorimeter surface can be used with a one-dimensional inverse conduction solver to accurately determine the heat flux from the fire. The Cask Analysis Fire Environment (CAFE) computational fluid dynamics code is coupled with a commercial finite element program to predict the heat flux to the calorimeter and the calorimeter interior temperatures. Time dependent temperature results for the inner surface are used as input to the Sandia OneDimensional Direct and Indirect Thermal (SODDIT) solver. This computer program predicts time-dependent surface heat flux profiles that are capable of causing a given temperature variation within one-dimensional solid conduction problems. However, conduction within the calorimeter is multidimensional since the heat flux varies around its circumference. This work shows that circumferential heat flux variation is sufficiently mild to allow accurate use of the inverse solver. A simple correction technique is developed that allows the use of the inverse solver even near the Curie point of the calorimeter material. INTRODUCTION Large packages that transport significant quantities of Type B radioactive materials must be qualified to withstand 30 minutes in a fully engulfing pool fire without significant release of contents. These regulations are described in, for example, Title 10, Part 71 of the Code of Federal Regulations, known as 10CFR71 (U.S. Nuclear Regulatory Commission, 1992). Analysis of such fires is complicated by several factors. First, the range of length scales encountered in the physical phenomena is large. For example, turbulence and combustion consist of physical phenomena with scales that range from submillimeters to several meters. Secondly, air and thus oxygen are introduced into large pool fires through a complex turbulent mixing process that controls both the location and the intensity of the combustion. Because the fuel-air mixing is limited, combustion temperatures are much lower than the stoichiometric limit. The central region of a pool fire is starved of oxygen, and a vapor dome exists immediately above the pool where evaporated fuel does not have sufficient oxygen to burn. Finally, the presence of soot particles strongly influences the flames. Soot is formed through the inefficient combustion process that is typical of large pool fires. These particles radiate thermal energy with the characteristic orange-yellow glow observed in fires. Recent studies indicate that the absorption length for thermal radiation in sooty flames is short, on the order of a few centimeters (Gritzo et al., 1998). Evidence of all these effects has be experimentally observed (Koski et al., 1996). THE CAFE COMPUTER MODEL Computer simulations of large pool fires using generalpurpose computational fluid dynamics codes require massive amounts of run time on specialized computing platforms. These general-purpose codes are not well suited for cask design studies or transportation risk analyses since these studies require multiple runs under varying conditions. The Cask Analysis Fire Environment (CAFE) computer code is currently under development at Sandia National Laboratories to meet the need of design and risk studies. CAFE models the physical phenomena that dominate heat transfer from large pool fires to massive engulfed objects. These models greatly reduce the computer turnaround time. The CAFE code can be coupled with finite element analysis (FEA) computer codes that model the interior components of specific package cask designs. The calculation procedure is as follows. The cask temperature is set to some initial distribution, T1. CAFE calculates the net heat flux from the fire to the engulfed cask. This flux is a function of location and time and it is responsive to the cask surface temperature. The FEA code uses the heat flux to calculate new cask temperatures. The new surface temperatures are fed back to CAFE, which then calculates new heat flux distributions. CAFE and the FEA code run alternately for the duration of the simulation. CAFE uses computational fluid dynamics (CFD) with a one-equation turbulence model to predict fuel and oxygen transport and mixing as well as convection heat transfer to the engulfed object. It uses Arrhenius kinetics to predict reaction rates and Rosseland conduction to model diffuse radiation heat transfer. CAFE also incorporates four approximations that are specifically designed to maximize its accuracy for this particular problem while minimizing computer time. First, it performs coupled two-dimensional simulations of the fire rather than a full three-dimensional simulation. (It actually performs quasi-three-dimensional simulations by using a variable mesh depth in the direction normal to the plane of the simulations. This variable depth allows the code to produce accelerated velocities near the center of the fire that are caused by radially in-flowing air. However, the code cannot model tornado-like vortices). Secondly, it uses a small computational domain with carefully chosen thermal and velocity boundary conditions. Thirdly, CAFE does not run continuously for the duration of the fire. It runs for short periods of time until the fire conditions are in quasi-equilibrium with the cask. It then stops running until the FEA program determines that the cask surface temperature has changed by some user-defined value (for example 5°C). As a result, the CAFE CFD simulator only runs for a fraction of the fire duration. Finally, radiation heat loss from the fire to the surroundings is modeled using an artificially reduced heat of reaction and not directly simulated. A more detailed description of CAFE may be found in (Suo-Anttila et al., 1999).