(388a) Genetic Algorithm Based Fast X-Ray Ct Technique Applied in Multiphase Flow Measurement

Non-invasive measurement techniques have drawn much attention in the study of hydrodynamics in many chemical engineering processes, which allow researchers to visualize the underlying physics in either opaque (e.g., multiphase flows) or transparent systems (e.g., single phase flows). Whereas, as highly sophisticated techniques, non-invasive measurements meet the difficulty in technical details both in theory and in operating experience such as for the process tomography. In particular, more challenges are superimposed on the current techniques to facilitate the high-resolution measurement in space and in time simultaneously. This presentation is to introduce a newly developed fast X-ray CT (computerized tomography) technique based on the genetic algorithm (GA) for multiphase flow measurement.

It is known that conventional X-ray CT has been widely applied in medical imaging areas, providing high resolution (in space) images of human organs for disease diagnosis. However, these applications always employ very expensive facilities with relatively huge bodies, not convenient for multiphase flow study in laboratory. Meanwhile, a complete data set from large amount of projections is inevitably needed for the image reconstruction, which lowers the time resolution of the X-ray measurement even down to the time-averaged results.

To conduct a fast measurement using X-ray CT, limited projections are expected to speed up the sampling rate and reduce the cost due to the CT hardware and software. While, the problem of tomographic reconstruction from the limited projection data (e.g. limited in number of views or view angles) is an ill-posed problem in mathematics. The conventional techniques such as convolution back-projection (CBP) and filter back-projection (FBP) cannot give satisfactory results under these conditions. In this study, a GA based method is proposed to implement the image restoration, together with the simplification that the most commonly encountered multiphase flows often have two distinct concentration values, i.e. 0 or 1 within a small control volume, such as in a gas-liquid bubble column. The image reconstruction is thus carried out by searching the globally optimal solution to the line integral equations by GA, which also incorporates prior knowledge of the multiphase flow physics. The computer simulation results already demonstrated that the proposed X-ray CT algorithm is rather robust in reconstructing ?bubbles? in liquids with limited angle projections even with noises at a certain level. Some preliminary experimental measurements gave sound validation of the feasibility of the proposed methodology to the transient measurement of multiphase flows.