-
-
Complete hand-built SAIM demonstration setup combining the physical building model, live dashboard, and mobile alerts.
-
End-to-end earthquake early warning demonstration: physical structure, real-time monitoring dashboard, and mobile notification output.
-
Internal hardware layout of the SAIM prototype showing distributed vibration sensors mounted inside the building model.
-
Real-time monitoring interface showing live vibration graphs, event logs, and system health indicators.
-
SeismoArch IoT dashboard displaying a simulated earthquake event and system-wide alert state during testing.
-
Mobile alert interface for upper-floor occupants, delivering tailored safety guidance based on floor-level risk during seismic events.
-
Mobile alert interface for lower-floor occupants, providing real-time earthquake warnings and location-specific safety instructions.
-
Core SAIM electronics: microcontroller, multiple vibration sensors, breadboards, and wiring prepared for distributed sensing.
-
Manual fabrication process used to create structural components, foamboards and hand-cutting tools used during model construction.
-
Base plates and vertical alignment elements hand-prepared to maintain uniform spacing and structural symmetry across floors.
-
Hand-cut and assembled structural floor panels used to construct the scaled multi-storey building model for vibration testing.
-
Individually fabricated wall and floor components arranged by floor level before manual assembly of the building model.
-
Manually assembled scaled building towers constructed layer by layer to replicate real building mass and height distribution.
Inspiration
Earthquakes cause injuries not only because of structural damage, but because people receive no clear warning or explanation before destructive shaking begins. Most existing warning systems operate at a regional level, which often delays alerts and provides limited context for individual buildings where people actually live, study, and work.
We were inspired by a simple question: what if buildings themselves could detect danger early and explain what is happening in real time?
SeismoArch IoT Model explores how localized sensing combined with AI-based interpretation can improve disaster preparedness, clarity, and response during seismic events.
What it does
SeismoArch IoT Model is a building-level earthquake early-warning and monitoring system enhanced with Gemini 3–powered interpretation.
The system continuously monitors structural vibrations using distributed sensors. By detecting early seismic signals (P-waves) directly at the building, SeismoArch can trigger alerts before the arrival of destructive shaking (S-waves).
When an event occurs, Gemini 3 analyzes structured seismic data and system context to generate clear explanations, risk assessments, and recommended actions, transforming raw sensor readings into actionable intelligence.
Key capabilities include:
- Real-time vibration, tilt, and motion monitoring
- Early-warning alerts based on seismic signal patterns
- Gemini 3–generated explanations and event summaries
- Live dashboard for visualization, alerts, and system health
- Simulation mode for testing and demonstration
The focus is on early awareness, fast response, and clear human understanding during emergencies.
How we built it
SeismoArch is built as an end-to-end IoT system combining hardware sensing, real-time data processing, cloud connectivity, and AI-driven interpretation.
Vibration sensors placed at the base of the structure capture early seismic motion from any direction. Sensor data is processed in real time and streamed to a monitoring dashboard via cloud services.
When abnormal seismic activity is detected, key features such as vibration intensity, timing, and alert classification are passed to the Gemini 3 API. Gemini applies its reasoning capabilities to:
- Interpret the nature of the seismic event
- Assess potential severity and confidence
- Generate human-readable explanations and guidance
The dashboard displays both raw sensor data and Gemini-generated insights, allowing users to understand not just that an alert occurred, but why it occurred and what it means.
The dashboard provides:
- Live charts for sensor data
- Event logs with warning and critical classifications
- System health and performance monitoring
- Interactive structural visualization
The system is designed to be lightweight, scalable, and deployable in homes, apartments, schools, and hospitals.
Challenges we ran into
- Differentiating meaningful seismic signals from environmental noise
- Ensuring real-time responsiveness with minimal latency
- Integrating hardware data with cloud services and AI interpretation
- Designing explanations that remain clear under emergency conditions
These challenges required careful tuning, testing, and system-level reasoning.
Accomplishments that we're proud of
- Built a working, real-time IoT system, not just a concept
- Successfully demonstrated early-warning logic based on seismic principles
- Integrated Gemini 3 as an intelligence layer for event interpretation
- Designed a dashboard that supports live data, simulation, and AI insights
- Created a scalable architecture suitable for real-world deployment
What we learned
- Early-warning systems require both fast sensing and clear interpretation
- AI is most valuable when it explains complex physical signals to humans
- Hardware-driven systems expose real constraints that improve design
- Even small gains in response time and clarity can significantly improve safety
What's next for SeismoArch IoT Model
Future work includes:
- Deeper Gemini-powered reasoning for adaptive risk assessment
- Expanding sensor networks for large buildings and campuses
- Field testing with real-world seismic activity
- Integration with smart city and emergency response infrastructure
SeismoArch aims to contribute to safer, more resilient infrastructure by combining physical sensing with intelligent, low-latency AI interpretation.
Live Dashboard (Recommended: Full Screen)
This link opens the real-time SeismoArch monitoring dashboard.
The dashboard loads instantly. The structural model may take 10–30 seconds to load depending on internet speed.
Since the dashboard is not connected to the physical prototype in this deployment, system behavior can be demonstrated using the Simulate button in the top-right corner.
Dashboard Capabilities
Structural Visualization
- Interactive structural model for situational awareness
- Controls: right-click drag to rotate, Ctrl/Shift + right-click to move
- Scene management allows saving, reordering, and timing different viewpoints
Tower Control
- Select individual towers to view live sensor charts
- Selecting multiple towers displays the average of selected readings
- No selection defaults to all-tower aggregation
Live Monitoring
- Real-time vibration, tilt, and piezoelectric data
- Charts are minimized by default and can be expanded for detailed analysis
- Time window, playback speed, and signal type are adjustable
Event & Alert System
- Event log records all sensor data
- Events categorized as normal, warning, or critical
- Logs can be cleared or exported as CSV
System Health & Performance
- Sensor health, data quality, and response time monitoring
- Performance metrics including FPS, latency, CPU, memory, and GPU usage
- Status indicators show data stream, sensor, and render activity
System Status Bar
- Displays current system time and runtime state
Note: Some features require a live physical sensor connection and may be limited when using simulation mode.
Log in or sign up for Devpost to join the conversation.