Flow Regime Classification Using Artificial Neural Network Trained on Electrical Capacitance Tomography Sensor Data

Khursiah Zainal-Mokhtar,Hafizah Talib,Junita Mohamad–Saleh ,Najwan Osman–Ali

doi:10.5539/cis.v1n1p25

Abstract

The main goal of the presented work is to analyse the performance of the Multi-Layer Perceptron (MLP) neural network for flow regime classification based on sets of simulated Electrical Capacitance Tomography (ECT) data. Normalised ECT data have been used to separately train several MLPs employing various commonly used back-propagation learning algorithms, namely the Levenberg-Marquardt (LM), Quasi-Newton (QN) and Resilient-Backpropagation (RP), to classify the gas-oil flow regimes. The performances of the MLPs have been analysed based on their correct classification percentage (CCP). The results demonstrate the feasibility of using MLP, and the superiority of LM algorithm for flow regime classification based on ECT data.

Highlights

The results demonstrate the feasibility of using Multi-Layer Perceptron (MLP), and the superiority of LM algorithm for flow regime classification based on Electrical Capacitance Tomography (ECT) data
The results show that the tanh-log combination of transfer functions gives higher classification percentage (CCP) compared to log-log, the difference is rather small
The work proposed a direct method for flow regime classification of gas-oil flows using MLP neural network as the classifier, based on simulated ECT data

Summary

Introduction

Knowledge of flow regimes is essential for investigation involving multi-component flows such as in oil production. The supervised type of ANNs such as the MLP, learn from given sets of input-output examples, establishing a mapping between inputs and outputs by updating their weight connections, to become ‘intelligent’ (Bishop, 1994). They ‘learn’ to predict the correct class for a given sets of inputs based on a suitable learning algorithm. This work investigated the most suitable learning algorithm for MLP in the quest for flow regime classification using ECT data of various flow patterns. The direct approach involves artificial neural networks (ANNs) which have been trained to classify various type of the flow regime based on the sets of ECT data. The results of MLP performances based on various employed learning algorithms are presented and discussed

Development of the MLP Classification System

Results and Discussion

Conclusion