Scholarlink

<p>Controllable pitch propellers (CPPs) are widely used in modern ships. During operation, CPPs are subjected to complex hydrodynamic loads from the propeller, causing the internal oil pipe to experience multiaxial alternating stresses that can easily lead to fatigue failure. An ensemble learning-based fatigue life prediction method under small-sample conditions was proposed to enable rapid and accurate assessment of the fatigue life of CPP oil pipe system. First, transient dynamic simulations are conducted using finite element software to obtain complete stress tensor time histories at each node under nine operating conditions. These stress histories are then input as load spectra into nCode DesignLife software. By applying the critical plane method in combination with the S–N curves of the respective materials, fatigue life calculations are performed to construct a small-sample database incorporating both hydrodynamic loads and fatigue life data. An ensemble prediction model consisting of six support vector regression (SVR) is developed to predict the fatigue life of the oil pipe system. Results show that the ensemble model achieves high prediction accuracy (R<sup>2</sup> &gt; 0.99), significantly outperforming individual models, with a total training and prediction time of only 0.016&#xa0;s—drastically reducing the time required for conventional fatigue assessment. More importantly, once the model is established, it enables rapid fatigue life estimation based solely on hydrodynamic excitation inputs, offering enhanced engineering efficiency and applicability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ensemble learning-based fatigue life prediction method for controllable pitch propeller oil pipe system under small-sample conditions

  • 期刊论文

  • 发表日期：

    2026年7月2日

查看全文
  • Yi Zou,
  • Linhui Zhou,
  • Weiwei Zhang,
  • Ruiyun Shi,
  • Tangqi Lv,
  • Yijian Zhao,
  • Hang Ren,
  • Shaogang Liu,
  • Dan Zhao

摘要

Controllable pitch propellers (CPPs) are widely used in modern ships. During operation, CPPs are subjected to complex hydrodynamic loads from the propeller, causing the internal oil pipe to experience multiaxial alternating stresses that can easily lead to fatigue failure. An ensemble learning-based fatigue life prediction method under small-sample conditions was proposed to enable rapid and accurate assessment of the fatigue life of CPP oil pipe system. First, transient dynamic simulations are conducted using finite element software to obtain complete stress tensor time histories at each node under nine operating conditions. These stress histories are then input as load spectra into nCode DesignLife software. By applying the critical plane method in combination with the S–N curves of the respective materials, fatigue life calculations are performed to construct a small-sample database incorporating both hydrodynamic loads and fatigue life data. An ensemble prediction model consisting of six support vector regression (SVR) is developed to predict the fatigue life of the oil pipe system. Results show that the ensemble model achieves high prediction accuracy (R2 > 0.99), significantly outperforming individual models, with a total training and prediction time of only 0.016 s—drastically reducing the time required for conventional fatigue assessment. More importantly, once the model is established, it enables rapid fatigue life estimation based solely on hydrodynamic excitation inputs, offering enhanced engineering efficiency and applicability.