Neuro-fuzzy methods for slug detection and control in multi- phase flow based on differential pressure and electrical capacitance tomometry (ECTm)

Abstract

Avoidance of slug flow is of paramount importance in processes involving multi-phase flow. As one of many safety measures, modelling based prediction of flow regime and identifying inflow conditions favourable for slug flow is done in the chemical, process industries, and refineries and in the exploration for and production of oil and gas. These efforts are all meant to develop methods to deter any catastrophic run-away phenomena in these diverse processes. Multitude of parameters associated with the multi-phase flow can affect the results from such simulations. However, these algorithms are still not robust enough to tackle real time control of these processes. The end-user needs a timely indication of some critical parameters so as to control or shut down a process to avoid process calamities. This paper focuses on data fusion of sensor data from an array of capacitance transducers arranged on the surface of the pipe with multi-phase flow along with differential pressure (DP) sensors. The capacitance-based measurements are from a electrical capacitance tomometric module. By studying the time series of the capacitance values logged in continuously from this array of capacitive sensors and DP, slugs can be identified and their parameters quantified. Neural network using self-organising maps (SOM) is used to classify the slugs in a rapid manner giving a good overview of the slugs and their parameters. Important parameters such as slug size, frequency and velocity can be estimated using neuro-fuzzy techniques thus facilitating a model free approach (MFA) for the control of the complex process of multiphase flow.