How Data From Reuse of Electrical-Submersible-Pump Components Can Help in Predicting System Failure

Abstract

Summary A statistical study using extensions to the semiparametric survival analysis (SA) was applied to an individual component of an electrical-submersible-pump (ESP) system. Since its proposal in 1972, the Cox proportional hazards (CPH) semiparametric model (Cox 1972) has become the primary tool for regression analysis of censored data. However, CPH assumes that observations are time-independent and that censoring is noninformative. With extensions to the standard CPH formulation, we demonstrate the analysis procedure by considering a specific component of an ESP system (a motor) such that informative censoring and time-dependent variables may be accommodated. We track this motor from its first deployment, and subsequent redeployments, in different ESP systems to provide an estimate of how its reuse affects ESP run life. Exploiting such model extensions, however, demands a more-involved analysis undertaking than is typical for conventional SA studies. Model limitations and applicability are also discussed.