Author(s): Tao Yu; Xiang Li; Longhui Xiao; Fang Li; Sen Han; Shifeng Ding; Li Zhou
Linked Author(s): Xiang Li
Keywords: Polar ship; Podded propulsion system; Health index; Reliability assessment
Abstract: Polar podded propulsion systems are subjected to severe ice-induced loads during dense ice navigation, making degradation monitoring and reliability assessment essential for safe and reliable operation. However, the coupled effects of thermal accumulation and transient ice-impact loads lead to highly nonlinear degradation evolution, posing significant challenges for quantitative reliability evaluation. To address this issue, a physics-guided reliability assessment framework is proposed and validated using full-scale condition-monitoring data collected from the polar icebreaker Xue Long 2 during the 15th Chinese National Arctic Expedition. A multisource health index is developed from bearing-temperature information and propulsion-load fluctuation characteristics, from which equivalent degradation time and Weibull-based relative reliability are established. A physics-guided neural network (PGNN) is further employed for sequential degradation prediction. The results show that the degradation rate under Condition 2 is approximately 2.67 times higher than that under Condition 1, while the relative reliability decreases to 0.3679 compared with 0.7650 under Condition 1. The comparison between the two operating conditions suggests that actual ice-induced load responses play a more dominant role in propulsion-system degradation than environmental ice parameters alone. Furthermore, the proposed PGNN achieves R² values of 0.9996 and 0.9998 for relative reliability and degradation-rate prediction, respectively. The proposed framework provides an effective tool for degradation monitoring and reliability assessment of polar podded propulsion systems under dense ice conditions.
Year: 2026