Expert system design for fault detection and solutions in ship machinery systems

This study focuses on diagnosing failures in a ship's engine room and developing an expert system to address them. A comprehensive database was built by examining 20 systems, 137 subsystems, and 546 distinct failures. Interactions among failures were analyzed using natural-language-processing techniques and a Bayesian network model; the effects of each failure on others were identified, and their probabilities were calculated via a recursive probability method. The resulting expert system, ShipMach. Expert, understands user-described failure symptoms in natural language, matches them with relevant failure records, and—through its custom interface—provides clear, detailed solution recommendations. A color-coding scheme visually highlights severity levels so that critical issues can be identified immediately. Testing showed that ShipMach. Expert achieves high accuracy in both failure diagnosis and solution suggestions. Its responses align with causes and effects reported in the literature, demonstrating that it can serve as a reliable advisory and decision-support tool in marine engineering. Overall, the study delivers a successful application of advanced technologies for efficient failure management and optimized maintenance processes in ship engine rooms.