The electrical capacitance tomography (ECT) based on the neural network multi-criteria optimization image reconstruction technique (NNMOIRT) developed by this research group is used to study the choking phenomenon for Geldart Group B particles in a circulating fluidized bed (CFB). The Group B particles employed are sand particles with a mean diameter of 240 mum and a particle density of 2200 kg/m(2). The ECT examines the real-time, quasi-3D cross-sectional flow structure of a gas-solid circulating fluidized bed. The ECT with NNMOIRT reveals a double solids-ring flow structure at a low solids circulation rate and the presence of solids blobs at the center of the bed at a high solids circulation rate before choking in a CFB. This flow structure undergoes a distinct variation during the choking transition forming wall slugs/gas intervals when the gas velocity is below the transport velocity, U-tr, or forming open slugs with blobs/particles at the center when the gas velocity is above U-tr. The standard deviation of the solids concentration fluctuation and the cross-sectional time-averaged solids concentration variation with increasing solids circulation rates exhibit a sharp change during the choking transition for a low gas velocity (