Evaluation of Electrical Resistance Tomography imaging algorithms to monitor settling slurry pipe flow. (August 2019)

• Record Type: Journal Article
• Authors: Kotzé, R.; Adler, A.; Sutherland, A.; Deba, C.N.
• Is Part Of: Flow measurement and instrumentation. Volume 68(2019)
• Publication Date: 2019-08
• Subjects: Fluid dynamic measurements -- Periodicals; Flow meters -- Periodicals; Fluides, Dynamique des -- Mesure -- Périodiques; Débitmètres -- Périodiques; 681.2805
• DOI: 10.1016/j.flowmeasinst.2019.101572
• Languages: English
• ISSNs: 0955-5986

Abstract

Abstract: Electrical Resistance Tomography (ERT) is used in process tomography to measure multi-phase fluid flow in pipes. ERT has the advantage of a non-invasive interface, but has challenges such as noise, a relatively low spatial resolution and the ill-posedness of the inverse problem. Many different image reconstruction algorithms have been developed in the medical imaging community, which offer promise to help improve ERT performance. However, no evaluation or methodology for comparison of different algorithms for industrial applications is available. To provide such an evaluation, we tested six ERT reconstruction algorithms for the identification in static and dynamic flow situations. Metrics were developed to evaluate the algorithms in terms of image quality and accuracy, different objects/shapes and coarse solids bed levels were tested statically in a spool piece, and bed levels dynamically in a pipe loop. An algorithm comparison methodology was developed and used to evaluate the different images based on the results obtained. Overall, results show significant variability between reconstruction algorithms, with some giving poor results at the pipe boundary and others poor results at the centre of the image. We identified two high performing algorithms and show that averages of individual algorithm images can achieve improved performance. Highlights: Six image reconstruction algorithms were evaluated and compared using EIDORS. Two static and one dynamic (with fluid flow) experimental setups were used. A methodology was developed for comparing different ERT algorithms. A new hybrid image reconstruction algorithm was developed for pipe flow. The performances of the different algorithms were not similar and some algorithms are more suitable for specific ERT applications.