Classification of flow regimes using a neural network and a non-invasive ultrasonic sensor in an S-shaped pipeline-riser system

• Artificial neural network-based classifier for objective flow regime identification. Doppler ultrasonic signals online non-invasive sensing. A new Continuous Wave Ultrasound (CWDU), Power Spectral Density (PSD), and Discrete Wavelet Transforms (DWTs)-based feature extraction framework. method classifying regimes is proposed that employs a network with inputs of extracted features from flows using either the Transform (DWT) or (PSD). The are classified into four types: annular, churn, slug, bubbly regimes. used in this work feedforward 20 hidden neurons. comprises output neurons, each which corresponds to target vector's element number. 13 40 PSD DWT respectively. Experimental data were collected an industrial-scale multiphase facility. Using features, misclassified 3 out 31 test datasets classification gave 90.3% accuracy, while only one dataset was yielding accuracy 95.8%, thus showing superiority classification. approach demonstrates applicability industrial applications clamp-on sensor. scheme has significant advantages over other techniques as non-radioactive non-intrusive sensor used. To best our knowledge, first known successful attempt liquid-gas S-shape riser system sensor, PSD-DWTs network.