Application of a Fuzzy Expert System to Analyze and Anticipate ESP Failure Modes

Abstract

Abstract Electrical submersible pumps (ESPs) are employed in a growing number of oil and gas developments. A high level of availability of the ESPs is essential to production assurance. The purpose of this research was to develop a basic ESP fuzzy expert system and to demonstrate its application. An ESP Fuzzy Expert system can provide a consistent means to analyze ESP operational data. The analysis can be used to either predict developing failure modes or to retroactively determine the causes of a prior failures. The main steps involved in the development of the ESP fuzzy expert system are as follows: identify input variables, develop membership functions, develop rule sets and develop output functions. Input variables representing slopes were mapped to membership functions denoted by linguistic terms such as "increasing", "constant", and "decreasing". The input variables are also assigned a member value to indicate the Degree of Fulfilment (DOF) of the individual premise. The Degrees of Fulfilment (DOF) of all the premises are then combined to determine the overall DOF of the rules (Adesanwo et al. 2016), with each rule representing a distinct failure mode. To validate the system, operational data from an offshore oil and gas field located in the Gulf of Thailand was collected and analyzed using the prototype ESP Fuzzy Expert System. This dataset consisted of nine ESP systems selected from a pool of more than 100 ESPs installed in the field. The nine ESP systems were composed of four ESP pumps which were running smoothly and five ESP systems that eventually failed. For the failed ESPs the operational data leading up to the failure was analyzed. Trends were analyzed for periods of 7 days, 30 days and 60 days for each ESP system.