A fusion algorithm for mass flow rate measurement based on neural network and electrical capacitance tomography

• Published in: Measurement : journal of the International Measurement Confederation Vol. 231; p. 114573
• Main Authors: Mousazadeh, Hossein, Tarabi, Nazilla, Taghizadeh-Tameh, Jalil
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 31.05.2024
• Subjects: Concentration; Forward problem; Fusion algorithm; Inverse problem; Multi-Layer Perceptron; Reconstruction algorithm; Fusion algorithm; Reconstruction algorithm; Multi-Layer Perceptron; Concentration; Forward problem; Inverse problem
• ISSN: 0263-2241
• DOI: 10.1016/j.measurement.2024.114573

•Five non-iterative image reconstruction techniques are compared experimentally.•Each reconstruction algorithm has its own pros and cons for special test in phantom.•A new NN fusion reconstruction algorithm is designed and developed.•The developed algorithm would decrease the concentration error by 7–10 folds. The process tomography has received a noticeable attention in industry due to non-invasive and non-intrusive characterise. To asses the potential of electrical capacitance tomography for gravity-fallen mass flow rate measurement, a fusion algorithm based on neural network is designed and implemented. The hardware of instrument is composed of a pipe as circular phantom with 200 mm internal diameter and the front-end electronics. The specially programmed software, runs five well-known, non-iterative reconstruction algorithms (LBP, Tikhonov, SVD, ART and SIRT) and fusions the results in a MLP-NN. The sensor experimentally evaluated by wheat grains. Five implemented algorithms, and the fusion NN are compared using RMSE and concentration error parameters. According to evaluated results in whole of tests, the fusion algorithm reduces RMSE and concentration error by approximately 2–3 and 7–10 folds respectively. Since these five algorithms are completely independent, running the program in parallel by Multi-Thread mode, do not adds run-time.