Project Proposals
Proposed system designs, architectures, and research initiatives — ready for deployment upon engagement.
Urban Infrastructure Monitoring — Colombo
12
Sensor Nodes
5
Data Domains
99.2%
Uptime
48 hrs
Detection Lead
Problem
Colombo's urban infrastructure lacked continuous environmental and structural monitoring. Decision-makers relied on periodic manual inspections and citizen complaints — both reactive and incomplete. No system existed to correlate air quality, noise, vibration, and traffic data across zones.
Data Sources
- PM2.5 / PM10 particulate sensors
- Noise level monitors (dB)
- Structural vibration accelerometers
- Traffic flow counters
- Meteorological stations (temperature, humidity, wind)
Proposed Implementation
Deploy 12 multi-parameter sensor stations across 4 urban zones
Edge processing on ESP32 gateways with MQTT data transmission
Cloud ingestion pipeline using stream processing and time-series database
Correlation engine cross-referencing environmental, structural, and traffic signals
Real-time dashboard with zone-level health scoring and alerting
Projected Outcomes
The proposed system targets continuous 24/7 monitoring with sub-second data ingestion latency. Cross-domain correlation is designed to identify pollution source attribution. Structural vibration anomaly detection is projected to provide 48-hour early warning over scheduled inspection.
Industrial Process Automation — Manufacturing Line
8
Data Streams
72 hrs
Prediction Window
67%
Detection Rate
24/7
Monitoring
Problem
A mid-scale manufacturing facility experienced recurring unplanned downtime on a critical conveyor-fed assembly line. Equipment failures were detected only upon breakdown. Maintenance was entirely reactive, resulting in 6–8 hours of unplanned downtime per month with cascading production delays.
Data Sources
- Motor vibration sensors (3-axis accelerometers)
- Temperature probes on bearing housings
- PLC operational data (speed, torque, cycle count)
- Power consumption meters per motor
Proposed Implementation
Instrument 8 critical motors with vibration and temperature sensors
PLC data integration via Modbus TCP protocol adapter
Real-time dashboard with equipment health scoring (0–100)
Temporal pattern analysis for bearing degradation prediction
Alert system with tiered severity (watch, warning, critical)
Projected Outcomes
The proposed system is designed to detect bearing degradation signatures up to 72 hours before failure. The health scoring model will provide operators with intuitive equipment status. Power consumption correlation is expected to reveal motors operating above baseline due to misalignment.
Cross-Domain Signal Correlation Engine
4
Signal Types
40%
FP Reduction
2 sec
Temp. Sensitivity
Historical
Validation
Problem
Traditional monitoring systems operate in silos — environmental data, mechanical telemetry, and network health are analyzed independently. Correlations between domains (e.g., temperature spikes coinciding with equipment vibration changes) are missed, leading to incomplete situational awareness.
Data Sources
- Environmental sensors (air quality, temperature, humidity)
- Mechanical telemetry (vibration, pressure, RPM)
- Network health metrics (packet loss, latency, throughput)
- Event logs from control systems
Proposed Implementation
Design a unified signal ingestion framework supporting heterogeneous data types
Implement temporal alignment across asynchronous data streams
Develop cross-domain correlation algorithms using sliding time windows
Build hypothesis validation pipeline — test proposed correlations against historical data
Confidence scoring for each detected correlation
Projected Outcomes
The proposed research aims to validate that cross-domain correlation significantly improves anomaly detection accuracy. Multi-domain analysis is projected to reduce false positive rates by up to 40% compared to single-domain monitoring. Temporal alignment is identified as critical to correlation detection performance.