InfraGuard

Inspiration

InfraGuard was inspired by the increasing number of real-world tragedies caused by delayed emergency response—collapsed buildings, fires, and structural failure that weren’t detected until it was too late. We asked a simple question:
What if critical infrastructure could speak before it failed?
That idea became our mission: build an intelligent monitoring system that detects risks early, sends alerts instantly, and prevents loss of life.

What it does

InfraGuard is an IoT+AI solution that continuously monitors vital infrastructure—bridges, tunnels, industrial plants, and metro stations—using real-time sensor data. It:

• Tracks stress, vibration, temperature, humidity, and gas levels.
• Uses on-device anomaly detection to assess risk.
• Pushes alerts to operators, maintenance teams, and emergency services.
• Generates visual dashboards and prediction analytics. Ultimately, InfraGuard turns raw sensor signals into actionable safety insights.

How we built it

We built the system in three layers:

1. Sensing & Hardware

We deployed low-power IoT modules using:

• MEMS accelerometers for vibration
• Temperature & humidity sensors
• Gas/air-quality modules (CO₂, methane, smoke)
• Edge microcontroller (ESP32) for local processing
  The design allows modular plug-and-play with existing infrastructure.

2. Connectivity

Data is transmitted via MQTT/REST to a secure broker using:

• Wi-Fi / LoRaWAN depending on distance
• Adaptive duty cycles to preserve battery life

3. Intelligence & Software

• Python backend for ingestion + validation
• Machine-learning model for anomaly scoring
  Sensor patterns are treated as multidimensional time series.
  We approximate stress deviation using: [ R = \sqrt{(x - \mu_x)^2 + (y - \mu_y)^2 + (z - \mu_z)^2} ] If (R > \tau), the system flags potential danger.
• Frontend dashboard in React/Next.js
  Displays health state, trends, alerts, and device status.

We iterated rapidly, testing sensors in simulated load environments and tuning thresholds to reduce false alarms.

Challenges we ran into

• Sensor noise and false positives: Raw signals fluctuate heavily due to traffic, weather, and environmental changes. We had to introduce smoothing, moving averages, and normalization.
• Latency vs. battery life: Live streaming every second drains power fast. We implemented adaptive sampling and edge inference to send data only when necessary.
• Data reliability: Field deployment conditions are harsh—dust, humidity, and temperature spikes. Calibrating sensors in real environments required experimentation.

Accomplishments that we're proud of

• Building a fully working prototype from hardware to cloud.
• Deploying an on-device ML pipeline that works offline.
• Creating a real-time dashboard with meaningful alerts.
• Designing something that could genuinely protect lives—not just a demo.

What we learned

• IoT complexity is not just wiring; it’s field conditions, calibration, and maintenance.
• Anomaly detection for infrastructure requires domain-specific logic, not generic models.
• Real-time systems must balance data granularity, bandwidth, and energy.
• Collaboration between hardware, software, and UX matters as much as the tech itself.

What's next for InfraGuard

• Integrating computer vision for crack/heat mapping.
• Deploying more durable industrial-grade sensors.
• Training models with historical stress data for predictive maintenance.
• Partnering with municipalities and smart-city vendors.
• Scaling from single-node prototypes to full nationwide monitoring networks.

InfraGuard is more than a project—it’s a step toward safer cities.

Built With

• anomaly-detection-models
• aws-iot-core
• aws-lambda
• docker
• dynamodb
• esp32-microcontroller
• firebase-authentication
• flask-backend
• for
• gas-detection-modules
• github-version-control
• grafana-dashboards
• lorawan-connectivity
• mems-accelerometer-sensors
• mqtt-protocol
• next.js
• node.js
• python
• react
• rest-api
• temperature-and-humidity-sensors
• tensorflow-lite
• websockets