IoT-based real-time wind data prediction for safety monitoring and alerting on construction sites

Abstract

Smart wearable devices are being increasingly used due to their potential to provide a safer working environment for construction workers. Hence, various studies have been conducted to investigate the application of smart wearables to increase the safety of construction sites in terms of human-workplace interactions, as well as monitoring and risk assessment. Among several hazards affecting workers’ safety at work, falling from a height causes the most number of injuries and fatalities. Additionally, more than half of the falls are caused by environmental factors such as snow, ice, extreme cold, and powerful winds. Furthermore, due to continuous climate change, the frequency and intensity of these environmental factors have increased in recent years. High-speed winds are one of the most impactful weather conditions on construction sites and one of the main reasons for incidents and accidents. Traditionally, to monitor high wind speed at the construction site, specific online weather service sites, or in some cases, the weather station located at the construction site are used. However, the traditional wind monitoring methods are not adequate and accurate. The first reason is the fact that real wind speed varies on different locations of the construction site due to the physical shape of the site and does not correspond to the reported wind speed on the websites. The second reason is the lack of real-time monitoring of wind data. In this research, a novel solution to predict real-time wind speed and direction based on IoT sensors mounted on workers’ hardhats and a supervised machine learning algorithm is proposed. The proposed solution corresponds to reducing the existing wind-related risks at construction sites. The main components of the solution are a hardhat equipped with hot-wire sensors and software component that predicts the wind speed and direction using a machine learning algorithm and provides alerts. To create a dataset for wind prediction, a wind tunnel equipped with an anemometer, a rotational platform to provide various wind exposures are used. The gathered data is used to build a regression models for the proposed supervised machine learning algorithm. The accuracy and precision of the prediction algorithm are compared to the data collected by the reference anemometer. The results suggest that the wind’s speed and direction can be predicted with high accuracy in real-time by the proposed solution.

Date	14 Dec 2021
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Ivanka Iordanova (Supervisor) & Ali Motamedi (Co-supervisor)