基于机器学习的船舶柴油机异常点检测技术研究  

Research on Abnormal Point Detection Technology of Marine Diesel Engine Based on Machine Learning

作　　者：吴德阳 都劲松[1] Wu De-yang;Du Jing-song(Shanghai Marine Diesel Engine Research Institute,Shanghai 201108)

机构地区：[1]上海船用柴油机研究所,上海201108

出　　处：《内燃机与配件》2022年第8期81-85,共5页Internal Combustion Engine & Parts

摘　　要：本文结合某船历史航行数据提出基于孤立森林和长短时记忆网络(LSTM)算法的柴油机异常点检测方法。运用孤立森林算法对柴油机气缸排气温度数据进行异常点检测;针对缺乏异常数据的船舶柴油机热工压力参数根据其和柴油机转速的强相关性提出基于LSTM算法的异常点检测方法,再通过实船数据验证两种算法异常点检测效果。研究表明基于LSTM和孤立森林算法的柴油机异常点检测算法具备可行性。This paper proposes a diesel engine abnormal point detection method based on isolated forest and long short-term memory network(LSTM)algorithm based on the historical voyage data of a certain ship.The isolated forest algorithm is used to detect abnormal points in the exhaust temperature data of diesel engine cylinders;for the thermal pressure parameters of marine diesel engines that lack abnormal data,an abnormal point detection method based on LSTM algorithm is proposed based on the strong correlation with the diesel engine speed,and then through the actual ship The data verifies the detection effects of the two algorithms for abnormal points.Research shows that the diesel engine abnormal point detection algorithm based on LSTM and isolated forest algorithm is feasible.

关 键 词：长短时记忆网络 孤立森林 异常点检测 

分 类 号：U672.2[交通运输工程—船舶及航道工程]