Increase in the identification of risk situations in structural tasks in medium-sized companies using artificial vision in Lima (#2134)

Authors

Huaman Saravia, Jheremi

Segovia Torres, Frank

Ulloa Román, Karem Asthrid

Abstract

Construction companies are responsible for managing safety within their facilities and work shifts. The use of artificial vision (AV) in civil construction has provided project executors with numerous benefits and opportunities, including extensive data collection, sustainable evaluations, and productivity improvements. The shift toward sustainability in construction is increasingly supported by digital technologies. In this context, this article reviews the literature to analyze the influence of AI in civil engineering. According to findings, the publication trend peaked among researchers in 2020. Risk management in construction is crucial for maintaining a safe work environment free from threats that could harm both project progress and worker productivity. However, this aspect often receives insufficient attention, partly due to the reliance on traditional risk identification methods, which can be inefficient or slow. For this reason, this study aims to automate the risk management process by identifying and counting such situations in structural tasks in the city of Lima, emphasizing activities involving height risks or falling objects. The methodology followed includes: (A) data collection and analysis through expert judgment, (B) assessment of the traditional risk identification process across four evaluated projects, (C) development of an automated risk identification process using artificial vision, and (D) implementation of this process. The results demonstrate an increase in risk situation identification and improved evaluation of the causes behind these risks for subsequent mitigation. The main conclusion is that artificial vision technology automates the risk identification process, enabling real-time detection and significantly reducing the time compared to traditional methods.