Axial heat transfer correlations in a circulating fluidized bed riser

Abstract

Experiments were conducted to determine the total (gas convective plus particle convective component) tube to suspension surface-average heat transfer coefficients for several vertical tubes placed in the core of a 5.5 m tall, 100 mm × 100 mm square Circulating Fluidized Bed (CFB) riser. The tubes carrying hot water were made of copper having 9.6 mm outside diameter and 0.6 mm thickness. The influence of axial location of the tube on heat transfer was studied using a 0.6 m high vertical tube placed along the riser axis at distances of 0.97, 1.62, 3.0, 4.0 and 5.3 m above the distributor plate. The effect of tube length on heat transfer was investigated using 0.6, 1.2, 2.0 and 2.5 m vertical tubes placed axially at four positions in the CFB riser, the tube mid-points being 5.2, 4.9, 4.5 and 4.25 m above the distributor plate respectively. The tubes are thus occupied about 10–45% of the riser height from the top. Sand was used as the bed material with mean diameter of 143 μm, 256 μm and 363 μm with respective densities of 2631, 2564 and 2740 kg/m3. The fluidizing air velocity and solid recycle flux were varied in the range of 4–8 m/s and 21–72 kg/m2 s respectively resulting in the average cross-sectional suspension density of 10–42 kg/m3, which is the range usually obtained in commercial boilers. It was found that the tube length and location strongly influenced the heat transfer. Heat transfer coefficients decreased with increasing tube length and with increasing axial locations from the distributor plate. Two empirical correlations for average heat transfer coefficient were developed in terms of relevant design and operating parameters including new-non-dimensional parameters for tube length (Ht/dp) and axial location (Hd/Hb). A semi-empirical heat transfer model for core tube to gas–solid suspension flow in a CFB riser in terms of Nusselt and Reynolds numbers with effective suspension properties was also presented. The semi-empirical predicts the present experimental data within ±20% and when other published data was included, the deviation was ±30%. While the modeling approach seems satisfactory, it needs refinement to improve its accuracy and consolidate all data in a more comprehensive way.