1.Client Background

A leading shipyard and marine engineering company that maintained and serviced large commercial vessels, oil rigs, and dredgers wanted to improve operational uptime. The client managed over 70 active vessels across ports in India and the Middle East and relied heavily on manual inspections to identify potential issues with machinery like diesel engines, pumps, winches, and turbines.

Breakdowns led to costly delays and contract penalties. The firm approached Warke Technologies to help them predict mechanical failures before they occurred using data-driven insights.

2. The Challenge

The marine environment posed numerous challenges: irregular internet access, high data volume from sensor logs, and the diversity of machinery types. Maintenance was largely reactive, and the documentation of failure patterns was incomplete. The goal was to collect machine health data, identify patterns, and trigger early warnings for ship engineers—even when offline.

A custom AI/ML system had to be integrated into a low-connectivity, high-risk industrial setting, with multilingual alerting for deck and engine room personnel.

3. Our Approach

Warke Technologies started by retrofitting 12 pilot vessels with IoT sensors to gather live machine data (temperature, vibration, pressure, RPM, and fluid levels). Over 8 months, we collected and analyzed hundreds of gigabytes of operating conditions to train models on normal vs. failure signatures.

The system was designed to work edge-first (locally on the ship), with smart compression and sync mechanisms to send updates to the central server when docked or connected.

4. Solution Delivered

We implemented a hybrid-edge AI Predictive Maintenance System specifically built for marine engineering applications:

• Sensor-Based Data Collection:
• Edge devices installed on engines, HVAC systems, propulsion units
• Real-time reading of 30+ parameters per machine
• Local data logging with anomaly flagging
• AI Predictive Engine:
• Trained ML models to detect early signs of wear, friction, fluid loss, overheating
• Confidence scoring and anomaly severity grading
• Suggestions for parts replacement and service priority
• Onboard Monitoring Interface:
• Dashboard for engineers with warning levels and part-specific flags
• Offline access with sync-on-dock mode
• SMS-based alerting for critical failures (backup to email)
• Fleet Maintenance Portal:
• View health status of all vessels in real-time
• Generate maintenance schedules, audit logs, compliance certificates
• Cross-vessel comparison reports and failure pattern mapping

5. Technologies Used

• IoT Hardware: Raspberry Pi + Bosch sensors + vibration modules
• ML Stack: Python (scikit-learn, XGBoost), TensorFlow Lite for edge
• Backend: Node.js
• Frontend: Angular
• Edge Database: SQLite
• Sync & Cloud DB: AWS RDS + MQTT sync layer
• Security: Encrypted vessel logs, role-based shipboard access

6. Results & Impact

• Unscheduled downtime reduced by 48%, saving an estimated ₹12.5 crore in penalties and delays in the first year.
• 90% accuracy in early fault detection, with average lead time of 5.3 days before failure.
• Engineers reported faster diagnostics and reduced troubleshooting time by 60%.
• The system was accepted by 3 international classification societies as part of the vessel’s digital audit log.
• The client began offering predictive diagnostics as a value-add to third-party vessels.