Testing of the evaluation methodology for Ship's Planned Maintenance System Database

Abstract

Ship's Planned Maintenance Systems are in use for a long period of time, but database quality, confi guration and content are still not standardized. Factors affecting the database quality, confi guration and content are the quality of raw data, experience and knowledge of database creation team, importance given to computerized PMS and care assigned to evaluation of fi nished product. Evaluation of fi nished product, i.e. ship's Planned Maintenance System Database is an action requiring specifi ed knowledge and qualifi cations, as well as experience in performing the task. Even when evaluators fulfi ll all needed requirements, credibility of the obtained results is questionable due to their subjectivity. One of the tools to decrease that problem is the Evaluation Methodology for Ship's Planned Maintenance System Database. A research, described in the paper, has been performed to test that aspect of the Methodology. Three Ships Planned Maintenance System Databases were evaluated by four different evaluators, fi rst using only their experience, then using the Methodology. Results of the research are presented in the paper together with comparison of fi ndings. All presented points that usage of the Methodology is a necessity during evaluations of ship's Planned Maintenance System Databases.

Multiple Instance Learning With Random Forest for Event Logs Analysis and Predictive Maintenance in Ship Electric Propulsion System

“…Ship's planned maintenance systems are based on periodically scheduled replacements and repairs according to requirements established by the manufacturers and classification societies. It has been in use for a long period of time [15]. In preventive maintenance, the crew is responsible on keeping the equipment in good conditions through frequent inspections, conditional analyses, and preventive measures.…”

“…where 𝑦 # is the t2F (19) and ℐ " is the ℎ th highest total Gini feature importance (15). ℎ ≪ 𝑀 to avoid noise features and a high-dimensional problem due to very few positive samples.…”