Mathematical Modelling of the Relationship Between Heat Exchanger Effectiveness and Separator Efficiency to Aid Predictive Maintenance and Control and Prevent Product Loss
Abstract
Measurements of separation efficiency in the upstream oil and gas sector are conducted by a combination of statistical and deterministic pathways. Statistical methods of separation efficiency are typically based on computational models that utilize equations such as the Souders-Brown equation and prediction of separator performance through machine learning (ML) models. However, in many cases, separator performance is heavily influenced by Heat Exchanger (HX) performance. Most models for separator efficiency ignore HX parameters, creating a blind spot that limits potential insights that may be derived through assessments of their values. In this work, a novel approach was implemented that incorporates the HX parameters in a dataset generated from a public database developed by the Massachusetts Institute of Technology (MIT), fluid properties from a Bonga Assay, obtained from ExxonMobil, and Bonga Floating, Production, Storage and Offloading (FPSO) equipment specifications. ML algorithms such as Linear Regression and Random Forest were trained on the dataset and delivered prediction accuracy R2 score of0.99558 for oil-water separation. Furthermore, a mathematical model with a high prediction accuracy was developed, this model used Stokes’ Law as a foundation and implemented residence-time based separator efficiency. Walther's equation was used as an intermediary, establishing a connection between viscosity of separator fluid and heat exchanger outlet temperature, a theoretical mathematical model relating heat exchanger effectiveness to separator efficiency was derived. Upon comparison with the curve fitting methods applied to the HX–Separator relationship, the accuracy of the model was verified. The mathematical model introduces an extra layer of confidence to separator efficiency measurements by factoring heat exchanger performance, further reducing the possibility of product loss due to errors or delay in transmission, as well as producing valuable insights that augment predictive maintenance schedules.