Research on cross-section gas content evaluation based on ultrasonic tomography system

Cross-section gas content (CGC) evaluation was of significance in the chemical industrial fields. As a non-destructive testing and evaluation method, Industrial Process Tomography technology was one of typical method to apply to gas-liquid two-phase flow in pipelines. It has been proved that UTT (Ultrasonic Transmission-mode Tomography) is a feasible and effective method for cross-section imaging of the pipeline from recent researches. In this paper, two kinds of methods were considered to derive the CGC of the pipe, imaging-based method and non-image-based method. Three imaging algorithms, FLBP (Fan-shaped linear back projection), Tikhonov and Landweber were applied to derive the cross-section imaging results of four different flow patterns at first. A quadrupole evaluation index system was proposed to evaluate the image reconstruction quality. On the whole, the circulation flow (Case2) reconstruction quality is the best, with PE (Position error) and REGC (CGC relative error) as low as 0 and 0.98%, respectively, and ICC (Image correlation coefficient) and PSNR (Peak signal-to-noise ratio) as high as 0.94 and 240.9, respectively. The estimated REGC for Case1, Case3, and Case4 are 15.64%, 0.05% and 0.22%, respectively. GRNN-based approach was proposed as non-image-based method to estimate the CGC under the same conditions. The number of samples for training the GRNN (Generalized Regression Neural Network) is increased from 18 to 138 at intervals of 20. The average REGC tested is 68.33%, 28.17%, 16.84%, 12.51%, 8.57%, 5.26%, 3%. The accuracy of prediction tended to decrease as the number of training samples increases. Simulations and experiments had proved that the method is feasible and effective, and the algorithm has the potential to be applied to the estimation of CGC in industrial processes.