Reliability-oriented sensor fault-tolerant strategy for hydraulic systems

Sensor failures in hydraulic systems can generate erroneous measurements and false alarms, leading to unnecessary shutdowns and reduced operational reliability. To address this problem, a reliability-oriented sensor fault-tolerant strategy based on multi-source information fusion is proposed. Physical sensor measurements, model-based Kalman-filter predictions, and artificial intelligence-based Long Short-Term Memory predictions are integrated to provide complementary sensing information. The artificial intelligence component provides data-driven redundant information for sensor-state assessment and fault-tolerant monitoring. Cross-source consistency is evaluated through a threshold-based mechanism to identify abnormal sensor states, after which the model- and data-driven estimates are adaptively fused to replace faulty measurements. This enables continuous monitoring and alarm functions to be maintained during sensor failures while reducing unnecessary shutdowns. The proposed strategy is experimentally validated on the hydraulic system of subsea lightweight intervention equipment under six representative sensor fault scenarios. The results demonstrate effective reconstruction of faulty sensor signals, reduced false alarms, and improved monitoring accuracy and operational continuity. The proposed framework provides a practical approach for reliability-oriented sensor fault tolerance in subsea hydraulic systems through the coordinated use of model-based and artificial intelligence-based redundant information.