Contact:

Dr. Hayden Reeve
Lab: (206) 616-9371
hreeve@u.washington.edu

POF Preform Heating

RESEARCH PERSONNEL LAB FACILITY VISIT US
POF - Processing Next >>

Experimental and numerical investigation of the initial preform heating was conducted in order to understand the heat transfer phenomena involved in fiber production. A better understanding of the system will help minimize preform waste and minimize the time required to reach steady-state drawing.

The experimental setup is shown above. A cold one-inch diameter preform was inserted into the furnace and allowed to heat until it reached thermal equilibrium. Thermocouples at locations 1-4 monitored the temperature within the preform as a function of time. Irises at the top and bottom of the furnace prevented the flow of air into or out of the furnace. Thermocouples were located in the irises and flush with the inner surface of the furnace wall.

Numerical Model

The heat transfer from the furnace wall to the air and preform was calculated using the numerical domain shown in red in the figure above. In order to accurately capture the natural convection heat transfer the conjugate problem was solved. Therefore, the full axi-symmetric continuity, momentum, and energy equations were solved for the entire numerical domain. The initial preform temperature was uniform at 20 C and the furnace was preheated. The experimentally measured temperature profile of the furnace wall was used as the thermal boundary condition in the numerical model. The no-slip condition was applied to all bounding surfaces.

The governing equations were solved using the SIMPLER finite volume algorithm. The air was considered incompressible and buoyancy forces were modeled using the Boussinesq approximation. The exchange of radiative heat transfer was solved using the net enclosure method. All surfaces were modeled as gray and diffuse. The air did not participate in the radiative exchange.

The numerical domain was discretized into 434 axial nodes and 32 radial nodes. The fully implicit SIMPLER method used a time-step of 0.5 seconds. The numerical model displayed satisfactory self-consistency when subject to both spatial and temporal refinement. Results obtained with a mesh of twice the nodal density (866x462) gave results within 0.6%. Using a time step of 0.01s gave temperature fields within 1% of those found using a time step of 0.5s.

Next >>