Image reconstruction for an Electrical Capacitance Tomography (ECT) system based on a least squares support vector machine and bacterial colony chemotaxis algorithm

Electrical capacitance tomography (ECT) is a computational imaging technology that may be applied to visualize and quantify the cross-sectional permittivity distribution of gas/solid two-phase flow. In consideration of the nonlinearity and ill-posed characteristics of image reconstruction in ECT, this paper presents a new image reconstruction method based on the least squares support vector machine (LSSVM) combined with the bacterial colony chemotaxis (BCC) algorithm to meet the requirements in the observation of transient flow regimes and their evolution from the spout of a dense–dilute burner at the end of a pneumatic conveying pulverized coal system. Firstly, a nonlinear mapping model is established from the measured capacitances to the grayscale values of the image by using the LSSVM, which has good nonlinear learning ability and high convergence rate. Secondly, because it is difficult to select kernel parameters for the LSSVM model, the BCC algorithm, which has global optimization and rapid convergent ability, is applied to construct an objective optimization function for the kernel parameters. Finally, a pneumatic conveying system with a radial biased whirl burner (a typical dense–dilute burner) was built up, and a sequence of images was collected using a 12-electrode ECT system from the spout of the burner to verify the effectiveness of the reconstruction algorithm under several cold conditions. Experimental results indicate that the algorithm can achieve reconstructed images with good quality and identify the subtle change of the transitional flow regimes from the spout of the burner.

Figure optionsDownload as PowerPoint slideHighlights
► Establish LSSVM model to map capacitances to grayscales of image pixels.
► BCC algorithm is applied to optimize LSSVM model.
► Using ECT based on LSSVM-BCC to observe dense–dilute transitional flow.