【作者】 吴昌宁； 程易； 丁宇龙； 魏飞； 金涌；

【Author】 WU Chang-ning, CHENG Yi, DING Yu-long, WEI Fei, JIN Yong (Department of Chemical Engineering, Beijing Key Laboratory of Green Chemical Reaction Engineering and Technology, Tsinghua University, Beijing 100084, China)

【机构】 清华大学化学工程系北京市绿色反应工程与工艺重点实验室；

【摘要】 针对多相流的快速检测要求,本文提出了有别于传统X射线层析成像技术(XCT)的新方法:考虑多相流相分布的离散特征,基于围绕被测流场有限角度的X射线投影数据,建立改进的遗传算法(GA)求解不完整投影数据集的二维截面图像重构问题．对气液鼓泡两相体系,局部相分布满足二值分布,该物理特征的引入大幅度降低了投影角度数的要求,从而提高了截面检测的速率,并可降低硬件成本．考虑重构问题的不适定性,对标准遗传算法的变异算子进行了有效改进．通过数值仿真和物理实验验证了新方法(GA-XCT)的可行性．在3-24个有限角度投影下,GA-XCT 表现出明显优异于传统CT图像重构算法(即滤波反投影算法)的二值图像重构能力,且具有良好的抗噪声能力．更多还原

【Abstract】 A new approach based on an improved genetic algorithm (GA) was proposed to implement the image reconstruction for the application of fast measurement of multiphase flow dynamics using X-ray computed tomography (XCT). Instead of directly using a traditional XCT, we pursued to develop a different discrete tomography (DT) method, aiming to achieve a high resolution in time during the measurements with only limited projection data. The proposed method assumed that the interested multiphase flow can be simplified as having distinct dense and dilute phases so that the local phase concentration can be binary-coded, e.g., 0 or 1 in a gas bubbling system. The mathematical problem under these circumstances is strongly ill-posed, and thus tackled with an optimization approach, i.e., a GA incorporated with the underlying physics as some constraints. The numerical simulations mimicking the physical measurements demonstrated the feasibility of the new approach, namely GA-XCT, especially with high robustness to the noise. Experiments were performed to simulate a transient measurement on the gas bubbles in water, with a portable X-ray tube and a 2-dimensional plane detector as the hardware and a static object rotating in between. The results further provided the validation of the GA-XCT being superior to the conventional algorithm, e.g., filtered back-projection (FBP) technique, in dealing with the tomography of multiphase system with binary local density field.更多还原