Font Size: a A A

Construction Worker’s Dynamic Risk Assessment Of Fall From Height Based On DBN And Computer Vision

Posted on:2021-08-27Degree:MasterType:Thesis
Country:ChinaCandidate:Y M PiaoFull Text:PDF
GTID:2491306107484664Subject:Engineering (Architectural and Civil Engineering)
Abstract/Summary:PDF Full Text Request
The construction industry is a labor-intensive industry.And due to the dynamic changes of various subjects at the construction site and the dangerous construction environment,safety accidents occur frequently,especially accidents of falling from height.The study proposes that risk assessment is an important part of safety management.Appropriate preventive measures can be taken to reduce the possibility of accidents according to the risk assessment.However,the main entities of the construction site,such as workers and equipment,will change dynamically.Consequently,it is necessary to perform dynamic risk assessment of fall from height.More and more studies show that workers working at heights are susceptible to falling due to their unsafe actions.Computer vision can capture information about unsafe actions.Therefore,this study aims to establish a feasible computer vision-based framework of dynamic fall risk assessment.The dynamic risk assessment framework improves the efficiency of risk assessment by automatically identifying the information about risk factors information.And the framework aims to reduce the possibility of falling from height,in other words,the risk of falling.This study first conducts a review of safety risk assessment and fall risk assessment,and summarizes the research characteristics of dynamic risk assessment.In the basic theory section,the study summaries the safety risk assessment methods.Traditional assessment methods cannot directly consider the dynamic changing attributes of risk factors.And it is difficult to update the model during the dynamic change process.In order to overcome the above limitations,the Dynamic Bayesian Network(DBN)method is utilized to perform the dynamic fall risk assessment in this study.Through the method,the fall risk factor system of this study was determined through the literature review,the characteristic analysis of the dynamic risk assessment researches and the consideration of unsafe actions.The DBN was established based on the risk factors.And the DBN reasoning was used to calculate the probability of the fall risk levels.Based on this method,a dynamic risk assessment framework based on computer vision is established.The framework is divided into two parts: the preparation work and dynamic fall risk assessment of the worker who works at height.The detailed elaboration of the workflow is presented.How the preparation part provides information and technical supports for the dynamic risk assessment part is presented in detail.And the application of computer vision in the framework is discussed.Finally,the workflow of the framework is demonstrated through a case study.The case is set up firstly.Then the preparation work in the framework and the implementation process of dynamic fall risk assessment on site are discussed.The case results are discussed and the sensitivity analysis is performed to verify the feasibility and operability of the framework.In addition,combined with the advantages of computer vision which can identify the safety status of each fall risk factor,the advantages are used to warn workers to discuss the early warning measures and prevent fall accidents.Through the above research process,this study determines the risk factors of high fall risk,articulates the DBN-based dynamic fall risk assessment method,and verifies the feasibility of the computer vision-based dynamic fall risk assessment framework.Computer vision can be used to automatically identify the status of falling risk factors to capture the dynamic changes of risk factors.And then dynamic fall risk assessment of the worker can be achieved.Then reasonable preventive measures are adopted to prevent fall accidents.
Keywords/Search Tags:Fall from Height, Dynamic Risk Assessment, Dynamic Bayesian Network, Computer Vision
PDF Full Text Request
Related items