Project Proposals

Proposed system designs, architectures, and research initiatives — ready for deployment upon engagement.

ProposalProposal

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

1

Deploy 12 multi-parameter sensor stations across 4 urban zones

2

Edge processing on ESP32 gateways with MQTT data transmission

3

Cloud ingestion pipeline using stream processing and time-series database

4

Correlation engine cross-referencing environmental, structural, and traffic signals

5

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.

ProposalProposal

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

1

Instrument 8 critical motors with vibration and temperature sensors

2

PLC data integration via Modbus TCP protocol adapter

3

Real-time dashboard with equipment health scoring (0–100)

4

Temporal pattern analysis for bearing degradation prediction

5

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.

ProposalProposal

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

1

Design a unified signal ingestion framework supporting heterogeneous data types

2

Implement temporal alignment across asynchronous data streams

3

Develop cross-domain correlation algorithms using sliding time windows

4

Build hypothesis validation pipeline — test proposed correlations against historical data

5

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.