Digital Twin Integration for Predictive Maintenance and Performance Optimization of Maritime Academy Training

Abstract

Maritime training vessels represent institutions' largest capital investments averaging $15-45 million per vessel yet operate suboptimally with 58% average utilization rates and 22% unplanned downtime creating capacity constraints limiting student sea-time training hours critical for STCW competency requirements, while maintenance practices following rigid calendar schedules ignore actual usage intensity variations wasting 35-40% of maintenance budgets on premature servicing or allowing equipment degradation through insufficient intervention. This research presents the design and validation of digital twin technology integrating real-time vessel sensor data with physics-based performance models and machine learning predictive analytics enabling condition-based maintenance optimization, voyage planning efficiency, and fleet coordination for training vessel operations at Sekolah Tinggi Ilmu Pelayaran Jakarta. Employing design science research methodology with qualitative stakeholder evaluation, the study engaged vessel operations managers (n=8), marine engineers (n=10), and maritime educators (n=12) through structured interviews examining digital twin accuracy, decision support utility, and pedagogical integration. The hybrid physics-ML digital twin architecture deployed 150+ IoT sensors across six training vessels capturing propulsion system performance, auxiliary machinery conditions, fuel consumption patterns, and operational parameters with real-time synchronization to cloud-hosted virtual models updated every 2-3 seconds. Thematic analysis revealed strong support for digital twin deployment, identifying critical themes of maintenance cost reduction, vessel availability improvement, and educational enhancement. Pilot implementation across 18-month period demonstrated 34% maintenance cost reduction ($918,000 annual savings), 47% increase in vessel availability (from 58% to 86% utilization), 28% fuel consumption reduction through voyage optimization, and 89% accuracy in predicting equipment failures 5-14 days ahead enabling proactive intervention, contributing validated digital twin architectures and empirical evidence supporting cyber-physical system adoption in maritime education addressing asset management efficiency and student training capacity maximization imperatives.