Predictive Safety System for Sivakasi Fireworks

Hardware

Inspiration.! The idea was inspired by the safety challenges in Sivakasi’s fireworks manufacturing industry. Many safety-critical conditions can depend on manual checking and separate monitoring. We wanted to explore whether these warning signs could be identified earlier, before they escalate into serious incidents.

What it does.? The system combines multiple safety functions into one connected framework: Smart batch verification using weight-based checking Temperature and humidity monitoring Storage-load monitoring Machine vibration monitoring Multi-sensor data fusion Anomaly detection and adaptive baseline analysis Dynamic safety-risk scoring Real-time safety alerts Configured machine shutdown for critical abnormal conditions Centralized safety dashboard and event history

How we built it.? We developed the prototype around an ESP32-based embedded system and integrated sensors for weight, temperature, humidity, and vibration. The system is organized into five functional parts: Smart Batch Verification Worker & Environment Monitoring Storage & Machine Guardian Dynamic Risk Engine Safety Command Center Sensor readings are processed and combined to identify abnormal conditions and generate a dynamic risk level. The dashboard provides a unified view of safety status, alerts, and event history. For the prototype, we use safe, inert materials and a small motor as a machine simulator rather than real fireworks materials or hazardous processes.

Challenges We Ran Into.? One of our biggest challenges was integrating multiple sensors with limited hardware resources while keeping the prototype affordable. We also faced challenges in: Handling inconsistent sensor readings Designing reliable sensor connections Creating meaningful risk levels with limited real-world data Integrating hardware monitoring with software intelligence Designing preventive actions without creating unsafe automation Building a practical prototype with limited resources

Accomplishments That We're Proud Of.? We are proud that we transformed a real-world industrial safety problem into a working prototype concept combining hardware, software, and AI-based analysis. Our key accomplishments include: Building an ESP32-based monitoring system Implementing smart quantity verification Demonstrating abnormal vibration detection Demonstrating configured machine shutdown Developing a multi-sensor risk assessment concept Creating a centralized safety monitoring approach Designing the system for future factory-level scalability And we have won 3rd prize at National Level technical Project Exhibition (TECHNOMIND-2K26) held at Karpagam Institute of Technology

What We Learned.? This project taught us that industrial safety requires more than monitoring individual sensors. We learned the importance of: Sensor Integration → Data Validation → Anomaly Detection → Risk Assessment → Preventive Action We also learned to distinguish between detecting abnormal conditions and claiming guaranteed accident prediction. With limited real-world data, responsible safety technology should communicate risk and uncertainty rather than make unrealistic guarantees. Most importantly, we learned how to combine embedded systems, IoT, AI, and human decision-making to address a practical safety problem.

What's Next for Predictive Safety System for Sivakasi Fireworks.? Our next goal is to move from a prototype toward a more comprehensive industrial safety platform. Future development includes: More industrial-grade sensors Larger-scale multi-room monitoring Improved anomaly-detection models using real operational data Explainable AI for understanding why risk is increasing Predictive risk trends Secure cloud/edge data architecture Digital safety records and batch traceability Integration with certified industrial safety systems Extensive validation with domain experts and safety authorities