In this study the application of machine learning (ML) and data analytics for monitoring and classifying of slugging behaviour was investigated. Artificial neural networks (ANN) and principal component analysis (PCA) were used for slugging monitoring and for classification. Both methods were tested on simulated data which was generated by a multiphase flow simulator. The results showed that these methods are accurate in monitoring both flow rates and slugging characteristics (amplitude and frequency). Additional insights via PCA-based visualizations show that slugs could be classified based on their frequencies.
Slugging in wells result in operation issues such as increased average pressure drop, potentially increased sand production, increased loads on topside piping and increased separator control issues which can even lead to separator trips. These large amplitude slugging phenomena are well known from long horizontal flowlines (terrain induced slugging), and from flowline-riser systems (severe slugging). But also in wells these mechanisms can lead to large dynamic disturbances. On top of that, in wells often slugging is induced by area transitions such as at the end-of-tubing and due to system instabilities such as void waves.