Abstract: The real-time monitoring of the health of the ship main engine is studied,and the traditional AMS alarm system monitoring method is made by using multi-dimensional data of continuous time. The real ship data is experimented by four types of anomaly detection algorithms：Mahalanobis distance,Isolation Forest,DBSCAN,and OCSVM,and the actual performance of each algorithm in each scenario is obtained. By comparing with the actual abnormal data points,the four types of algorithms have good performance in the sensitivity,false judgment and missing indicators of anomaly detection.

Key words: health of ship main engine, data anomaly monitoring, Mahalanobis distance, Isolation Forest, DBSCAN, OCSVM