📄 中文摘要
海洋发动机的灾难性故障意味着功能的严重丧失，并会对系统造成不可逆转的破坏。这类故障通常是突发且不可预测的事件，对航行、船员和乘客构成严重威胁。由于其突发性，早期检测成为唯一有效的对策。然而，现有研究主要集中在组件的逐步退化建模上，对突发和异常现象的关注有限。该研究提出了一种新的早期检测灾难性故障的方法。基于来自故障发动机的真实数据，该方法评估实际传感器读数与发动机变量预期值之间偏差的导数。通过随机森林算法获得预测结果。
利用机器学习早期检测海洋柴油发动机的灾难性故障
📄 English Summary
On Using Machine Learning to Early Detect Catastrophic Failures in Marine Diesel Engines
Catastrophic failures of marine engines result in significant loss of functionality and can irreversibly damage systems. These failures are sudden and often unpredictable, posing serious threats to navigation, crew, and passengers. The abrupt nature of these failures makes early detection the only effective countermeasure. However, existing research has primarily focused on modeling the gradual degradation of components, with limited attention to sudden and anomalous phenomena. A novel method for early detection of catastrophic failures is proposed, utilizing real data from a failed engine. The approach evaluates the derivatives of the deviation between actual sensor readings and expected engine variable values. Predictions are generated using a Random Forest algorithm.