Design of a Fuzzy Logic Model for Real-Time Predictive Maintenance of a Tugboat Engine: A Case Study of MV Kiboko

Abstract

Reliable and efficient tugboat engines are essential to safe berthing, towing, and manoeuvring operations in busy port environments. Conventional maintenance strategies, whether reactive or time-based, struggle to capture the complex, dynamic, and uncertain behavior of marine engine systems, often resulting in unplanned downtime and elevated maintenance costs. This paper presents the design and simulation of a fuzzy logic-based predictive maintenance model developed for the main engine of the MV Kiboko tugboat operating at Dar es Salaam Port. Six real-time input parameters, namely engine temperature, lubricating oil pressure, vibration level, exhaust gas emissions, fuel consumption rate, and load condition, were converted into linguistic variables through fuzzification and processed using a Mamdani-type fuzzy inference system implemented in MATLAB and Simulink. Expert-defined IF-THEN rules, formulated in consultation with experienced marine engineers and technicians who operate and maintain the vessel, classify engine condition into three maintenance alert levels, Low, Medium, and High, through centroid defuzzification. The model was evaluated using simulated operational scenarios and validated against historical maintenance records and expert judgement. Results show an average prediction accuracy of 88.5%, a false-positive rate of 7.3%, a false-negative rate of 4.2%, and an average fault-warning lead time of 3.5 operating hours, while expert reviewers rated the system 4.6 out of 5 for overall usefulness and reliability. These findings demonstrate that fuzzy logic can effectively manage the uncertainty inherent in marine engine condition monitoring, providing a practical, interpretable, and cost-effective decision-support tool for predictive maintenance planning in tugboat operations and, more broadly, in maritime engineering practice.