Scholarlink

Fireworks manufacturing is one of the most accident-prone industries, primarily due to unsafe handling of chemicals, improper mixing ratios, and environmental factors such as temperature and humidity. In regions like Sivakasi, Tamil Nadu, frequent fire accidents cause significant loss of life and property. This paper proposes a machine learning–based predictive framework to classify accident risk levels as Safe, Warning, or High-Risk by analysing both chemical and environmental data. The system integrates IoT-enabled sensors, cloud deployment, and a real-time dashboard with alert mechanisms. Various ML algorithms such as Decision Trees, Random Forests, XGBoost, and Artificial Neural Networks were evaluated, with Random Forest showing superior performance in accuracy and recall. Results indicate that the proposed system can significantly reduce human error in safety checks and provide proactive accident prevention measures. Future enhancements include integration with Industry 4.0 smart factory platforms, computer vision for unsafe practice detection, and blockchain-based compliance logging.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Pyroguard ML: Fire Accident Risk Prediction in Cracker Production Using Machine Learning on Chemical and Environmental Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. Maha Sri,
  • V. R. Bevin Sancia,
  • G. Thiraviaselvi

摘要

Fireworks manufacturing is one of the most accident-prone industries, primarily due to unsafe handling of chemicals, improper mixing ratios, and environmental factors such as temperature and humidity. In regions like Sivakasi, Tamil Nadu, frequent fire accidents cause significant loss of life and property. This paper proposes a machine learning–based predictive framework to classify accident risk levels as Safe, Warning, or High-Risk by analysing both chemical and environmental data. The system integrates IoT-enabled sensors, cloud deployment, and a real-time dashboard with alert mechanisms. Various ML algorithms such as Decision Trees, Random Forests, XGBoost, and Artificial Neural Networks were evaluated, with Random Forest showing superior performance in accuracy and recall. Results indicate that the proposed system can significantly reduce human error in safety checks and provide proactive accident prevention measures. Future enhancements include integration with Industry 4.0 smart factory platforms, computer vision for unsafe practice detection, and blockchain-based compliance logging.