Intelligent systems for recognizing dangerous areas and preventing accidents on construction sites: neural networks in machine safety management

Abstract

Intelligent systems for recognizing dangerous areas and preventing accidents are becoming a critical element of safety at modern construction sites. This study provides a comprehensive analysis of the use of neural network technologies in hazardous area monitoring and safety management systems for construction machinery. The research methodology includes the analysis of various neural network architectures, pattern recognition mechanisms, and algorithms for processing data from various sources, including cameras, sensors, and IoT devices. As a result, an integrated model for identifying hazardous areas has been developed based on a combination of convolutional neural networks (CNN) and YOLO algorithms, which achieves recognition accuracy of 94.7% when identifying workers without protective equipment and 93.2% when identifying prohibited areas. The experiments conducted demonstrate that the implementation of the proposed system reduces the number of incidents on construction sites by 78% and minimizes the risks of accidents with heavy machinery by 82% compared to traditional control methods. The study shows that the integration of the proposed technologies can significantly improve the effectiveness of security systems at construction sites, providing preventive identification of potential threats and automatic prevention of emergencies. The results obtained have practical value for improving safety management methods in the construction industry and reducing occupational injuries.