A Fault Inference Method under Uncertainty: Case Study on Crankshafts in Fracturing Pumps

Abstract Crankshaft is a pivotal mechanical unit in the power-end system of a fracturing pump and its fault inference could facilitate optimal condition-based maintenance. Fracturing pumps are equipped with advanced instrumentation systems able to acquire vibration information for crankshaft fault analysis, but there exist complex uncertain dependences between faults and symptoms as well as incomplete symptom information, further increasing the difficulty of fault inference by operators. To achieve effective fault inference in the case of uncertain or incomplete diagnosis evidences, a Bayesian network-based fault inference method for crankshafts is proposed in this article. The approach can be utilized to implement cause inference and diagnosis inference by incorporating cause nodes, fault nodes and symptom nodes into a Bayesian network (BN) model. The application of the presented approach in fault inference of crankshafts indicates its strong inference capability under uncertainty. The results from the presented BN model may offer a useful aid to repairers in their maintenance decision-making processes.