Failure mode analysis of hydraulic turbine governing systems in small-to-medium hydropower units under high-cycle regulation service

Hydraulic turbine governing systems in small-to-medium Francis plants experience multi-mode failures—sensor, electrical-control, hydraulic-valve, sealing, and emergency-valve faults—whose frequency dependence remains poorly quantified at the system level. Existing studies treat these subsystems component-by-component or under standalone semi-quantitative scoring, none quantifying how the failure spectrum is reshaped as annual regulation frequency rises—a gap of growing consequence as hydropower shifts from base-load supply to flexible regulation. This study analyses 27 documented failure events from four Francis plants (12–21 MW, 91–132 m head) in a southwestern China river basin under wind–photovoltaic–hydro complementary scheduling. An integrated FMEA–Bayesian Network framework is developed in which a three-tier OREDA-anchored RPN -to-CPT protocol translates FMEA scores into traceable BN priors, removing the subjectivity that limits conventional FMEA. Inference over a 15-node network reproduces the plant–mode failure pattern (Spearman r s = 0.80; perfect plant-level ranking). Sobol analysis identifies oil quality (0.42 ± 0.03) and regulation frequency (0.31 ± 0.03) as the dominant causes of hydraulic-valve failures, reflecting an abrasive-wear mechanism. What-if simulation shows hydraulic-system modes amplifying by 2.4–3.1 × under high-cycle conditions versus 1.4 × for electrical-control, rebalancing rather than uniformly intensifying the failure spectrum (FM3 + 8.8 pp, FM2 − 9.6 pp). Plant-specific maintenance strategies project 24.6 % aggregate annual cost reduction (20.3–28.9 %) with 0.3–1.2 year payback. The framework offers a transparent, auditable approach to multi-mode failure analysis of hydraulic equipment under high-cycle service, transferable to other hydropower assets transitioning to flexible operation.