Dimensionality reduction and pattern recognition of flow regime using acoustic data

Abstract

In this study we investigated the novel application of Principal Component Analysis (PCA) in order to reduce the dimensionality of acoustic data. The acoustic data are recorded by fibre optic distributed acoustic sensors which are attached along a 3500 m pipe with a sampling frequency of 10 kHz and for a duration of 24 hours. Data collected from distributed acoustic sensors are very large and we need to identify the part that contains the most informative signals. The algorithm is applied to water, oil and gas datasets. We aimed to form a smaller dataset which preserves the pattern of the original dataset which is more efficient for further analysis. The result of this study will lead to automation of multiphase flow pattern recognition for oil and gas industry applications.