Acoustic signal processing with robust machine learning algorithm for improved monitoring of particulate solid materials in a gas flowline

Abstract

The flow of particulate solid materials in a gas flowline can significantly erode mechanical equipment. Hence, real-time quantitative monitoring is a timely need for the oil and gas industry to achieve real-time control and production optimisation. Although a considerable amount of research has been conducted employing acoustic signals for qualitative monitoring, there is still an unmet demand for a simple and robust real-time quantitative monitoring system. Acoustic signal processing with machine learning is a simple and robust method that has the potential to meet this demand but has not been previously exploited for real-time quantitative monitoring of particulate solid materials in a gas flowline. This paper proposes a novel instrumentation system for on-line measurement of solid flow rate, solid concentration, line pressure drop and gas velocity in a gas-solid multiphase flow using acoustic sensing technology coupled with signal processing techniques and machine learning algorithm. The acoustic sensor is used to capture the acoustic wave emitted from the impingements of the solid particles on the bend component of the flowline. Signal processing techniques are used to extract relevant features about the impingements. An integrated, conventional Artificial Neural Network (ANN) is used to capture the distribution of the acoustic feature vectors in order to establish the relationship between the measurands and the acoustic signal. However, conventional ANNs are mainly concerned with capturing systematic patterns in a distribution of measurements fixed in time and in this case the dynamics of the generated acoustic signal varies with time. A modification, called Time-Delay Neural Network (TDNN) is used to capture such dynamics. The proposed system compares the performance of the classical ANN and the TDNN models. Results obtained demonstrate that with the classical ANN, the normalised root mean square error (NRMSE) is 0.66, 0.29, 0.26 and 0.46 for the solid flow rate, solid concentration, line pressure drop and gas velocity respectively. With the TDNN model, the NRMSE is 0.18, 0.17, 0.20 and 0.16 for the solid flow rate, solid concentration, line pressure drop and gas velocity respectively. In comparison with the ANN model, the TDNN model has better performance as the NRMSE values are lower for all the models for the measurands. Overall, this study lays the basis for employing signal processing techniques and machine learning algorithm in the development of a simple, reliable and low cost real-time quantitative particulate solid flow monitoring system.