Chemical Engineering Science 62 (2007) 2068 – 2088 www.elsevier.com/locate/ces Dilute gas–solid two-phase flows in a curved 90 ◦ duct bend: CFD simulation with experimental validation B. Kuan ∗ , W.Yang, M.P. Schwarz Cooperative Research Centre for Clean Power from Lignite Division of Minerals, Commonwealth Science and Industrial Research Organisation, Box 312, Clayton South, Victoria 3169, Australia Received 31 May 2006; received in revised form 20 October 2006; accepted 21 December 2006 Available online 13 January 2007 Abstract Computational fluid dynamics (CFD) simulations of dilute gas–solid flow through a curved 90 ◦ duct bend were performed. Non-uniform sized glass spheres with a mean diameter of 77 m were used as the dispersed phase. The curved bend is square-sectioned (150 mm × 150 mm) and has a turning radius of 1.5D (D = duct hydraulic diameter). Turbulent flow quantities for Re = 100, 000 were calculated based on a differential Reynolds stress model. The solids mass loading considered is 0.00206 and hence justifies the application of one-way coupling between gas and particles. A Lagrangian particle-tracking algorithm which takes into account the effect of shear-slip lift (SSL) force on particles and particle-wall interactions (PWIs) has been utilised to predict velocities of the dispersed phase. The predictions were compared against the experimental data measured using Laser–Doppler Anemometry (LDA). The study found that the predicted gas flow field has a strong influence over the predicted particle velocities. PWI model considerably affects the prediction of particle velocity and distribution of particles at the inner duct wall within the bend. Inclusion of the SSL force also helps the distribution of the particle tracks towards the duct centre in the vertical duct downstream of the bend. Within the bend, particle velocities near the inner wall have been grossly over-predicted in the simulation, especially at mid-bend. The present study thus highlights the importance of the predicted gas flow field, SSL force and particle-wall collisions to Lagrangian particle tracking.  2007 Elsevier Ltd. All rights reserved. Keywords: Multiphase flow; Mathematical modelling; Simulation; Penumatic conveying; LDA; Turbulence 1. Introduction Elbows and bends are commonly used in pneumatic convey- ing systems to change flow direction so as to transport the sus- pended material to the desired delivery point within a limited space. In the case of coal-fired power plants that operate on a continuous supply of pulverised coal to furnaces, mal- distribution of pulverised fuel often occurs as coal particles are pneumatically transported from the mill through ducts consisting of numerous bends and straight sections. Apart from the duct geometry, the coal pulverisation process is also a strong contributor to mal-distribution of the pulverised fuel. Field measurements at a lignite-fired power plant carried out to support this study found that, depending on factors such ∗ Corresponding author. Tel.: +61 3 95458687; fax: +61 3 95628919. E-mail address: benny.kuan@csiro.au (B. Kuan). 0009-2509/$ - see front matter  2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2006.12.054 as feed size and mill speed, the fuel pulverisation process pro- duces coal particles ranging between 10 and 1000 m in size. Fig. 1 shows a typical coal particle size distribution measured at the mill outlet of a lignite-fired power station (McIntosh and Borthwick, 1984). The extreme non-uniformity in particle size combined with a centrifugal effect arising from the duct bend are believed to lead to the formation of a stratified gas–solid flow, known as a particle rope, downstream of the elbow even at a low solids mass loading, L< 0.1. This invariably creates difficulties for the plant operators to monitor and control the pulverised fuel supply to individual burners, and hence to main- tain an optimal combustion condition inside the furnace. There are a large number of documented studies, both nu- merical and experimental, on particle roping in dust conveying systems with solids mass loading L> 0.3, but most of them focus on particles with a size distribution that is either heavily skewed towards the lower end of the size range ( Yilmaz and