Flow regime identification of particles conveying in pneumatic pipeline using electric charge tomography and neural network techniques

Abstract

A plastic beads (solid particles) flow in a pipeline is a common means of transportation in industries. Monitoring and controlling materials flow through the pipeline is essential to ensure plant efficiency and safety of the system. The pipeline transportation used in this project makes use of electrodynamic sensors which are charge to voltage converters. The process flow data is captured fitting an array of 16 sensors around the circumference of the pipe to capture the inherent charge on the flowing solid materials. A high speed data acquisition card DAS1800HC is used as the interface between the sensors and a personal computer which processes the data. A Radial Basis Function (RBF) neural network based flow regime identifier program is developed in Matlab environment. Baffles of different shapes are inserted to artificially create expected flow regimes and data captured in this way are used in training and evaluating the network’s performance. The results of this work show significant improvments, the dataset which was check as the input gave good results, especially for full flow, three quarter flow and inverse quarter flow are 100%, and 95% has been succeed for each of quarter flow inverse three quarter flow and inverse half flow, and for the others flow regimes (center half and half flow) 90% succeed.