A New Risk Assessment Method to Prevent Progressive Cavity Pump Failure Based on Bayesian Networks and Noisy-OR Gate

Summary The frequent failure of progressive cavity pump (PCP) remains a critical issue in oilfield operations, with existing methods falling short in providing reliable risk assessment and mitigation strategies. To address this gap, we introduce a novel approach based on the Noisy-OR Bayesian network (BN) to systematically model and evaluate PCP failure risks. A comprehensive case study from an oil field in North China was conducted using Graphical Network Interface (GeNIe) Academic to simulate failure mechanisms. The initial analysis indicated a PCP failure probability of 32%, categorizing the system as high risk. Through generalized failure analysis, the five most influential risk events were identified: excessive sand content (0.498), rotor surface wear and erosion (0.445), stator-rotor friction and heat generation (0.405), rubber friction in the stator (0.401), and excessive displacement (0.398). By integrating the Noisy-OR gate within the Bayesian framework, this research offers a more realistic and computationally efficient representation of multifactorial failure interactions compared with conventional models. A series of targeted mitigation measures were implemented based on sensitivity analysis (SA), leading to a substantial reduction in failure probability from 32% to 13% (low risk). The proposed methodology not only improves upon prior models by incorporating probabilistic interdependencies among failure factors but also aligns closely with field applications.