Predicting Choke Valve Erosion in HPHT Gas Wells Using Machine Learning

Production choke valves are essential to control pressure and flow rate of the produced hydrocarbon stream from a well. These choke valves are subject to continuous erosion that results from the interaction of particles in the hydrocarbon fluids. Traditionally, detection of choke erosion in high-pressure high-temperature (HPHT) gas wells is done through physical inspection of the production choke to confirm erosion. This process is highly time consuming and expensive due to the need to interrupt production, depressurize the flowline, decoupling the choke body from the flowline, and performing choke body inspection on site or at the workshop, if erosion is confirmed then a choke replacement is scheduled. In this paper, we present a revolutionary approach that utilizes machine learning techniques to predict the remaining useful life (RUL) for chokes in gas wells. Our findings offer a multitude of business values such as maximized equipment efficiency, cost reduction, enhanced safety, and improved production. The solution utilizes an advanced regression algorithm, incorporating real-time well performance variables. These inputs communally contribute to predict RUL for multiple HPHT gas wells. Furthermore, thorough data discovery, transformation, and quality control procedures, our predictive model can predict the days to erosion based on current well status. The work provided strong evidence for the effectiveness of this AI-driven approach, with an accuracy exceeding 95% in predicting the RUL for gas wells. The combination of established digital dashboard powered with this AI-predicted RUL enables engineers to quickly identify and prioritize wells susceptible to erosion. Thus, enhancing surveillance to ensure well integrity & production efficiency. Application of this methodology also allowed for careful allocation of limited resources toward wells exhibiting erosion vulnerability, effectively reducing the cost, risk and downtime associated with such failures. By utilizing this predictive maintenance approach to predict RUL for chokes, operators can mitigate future asset failures, improve safety standards and reduce downtime.