A new method for the construction of cross-sectional velocity field of array electromagnetic flow sensor based on Tikhonov regularization- CNN

Abstract

Electromagnetic flow sensor is a reliable flow measuring instrument that can usually be used as a standard meter, The number of electrode pairs determines the resolution of multi electrode electromagnetic flow sensors in measuring cross-sectional velocity fields, and the multi electrode electromagnetic flow sensors resolution of measuring the velocity field of cross-section is relatively low. This paper proposes a region weight function theory to address the issue of low resolution in the velocity field of multi electromagnetic flow sensor measurement cross-sections. By combining the Tikhonov regularization algorithm with the Convolutional neural network (TR-CNN), a nonlinear model based on TR-CNN is constructed. This model forms a mapping relationship between the potential difference on the electrode and the cross-sectional velocity field inside the pipeline. This work utilizes the TR-CNN algorithm to improve the resolution of pipeline cross-sectional velocity field measurement. The potential difference is used to predict the velocity at different positions inside the pipeline and compared with the experimental data of a pitot tube. The results show that the array electromagnetic flow sensor can effectively predict the eccentricity trend of the fluid in the pipeline using the TR-CNN model, with a predicted cross-sectional mean squared error(MSE) of 0.015, mean absolute error(MAE) of 0.95, and root mean squared error(RMSE) of 0.123. This also demonstrates the feasibility of the TR-CNN algorithm in predicting the velocity field of the pipeline cross-section.