AE-Sensors and Multimodal Sensor Data Fusion in Liquid Flow metering

Abstract

One of the biggest challenges in Oil and gas industries is finding convenient method for accurately measuring flow rate of multiphase materials flowing through a system. There are different approaches done to handle this situation and each ended up with different results. To continue research & development on this topic, two such experiments sites in this case rigs are present, one is in University of South-Eastern Norway and other one in Equinor.

This thesis objective is to estimate single phase flow velocity using clamp-on accelerometer sensors fitted on outer surface of pipes. Raw accelerometer data along with other sensor data like temperature and differential pressure was collected at both rigs. Since the main focus was on accelerometer data, complete thesis was done using only accelerometer data. The data was analyzed using FFT and PSD plots, filtered and pre-processed. Feature extraction was done.

The top three features were used to develop classification models to identify the type of flow material i.e., Gas, Oil or Water. The test accuracy of classification model is around 98 %. Then prediction model was developed for estimation of flow velocity. Top accelerometer features selected for prediction gave an RMSE of nearly 10.2.