Using Mobile Crowd Sensing Network To Detect And Identify Construction Workers' Near-Miss Falls Based On ANN

Fall accidents?trips,slips,and falls from high elevations?are one of the leading causes of fatalities and injuries in the construction industry.The factors contributing to fall accidents are numerous and complex.However,a growing body of evidence indicates the strong correlation between posture instability?loss-of-balance?and fall accidents.This paper aims to reduce injury and promote safety management at construction sites through iterative technique and improved management.At iterative technique level,the smartphone was utilized as a data acquisition tool as well as the basic unit of crowd-sensing.Method with Artificial Neural Network?ANN?,its alternative performance in addressing complex problems,which usually need subjective judgment and identification,has been proved.At improved management level,since under certain circumstances,near-miss falls can evolve into fall accidents,insight into near-miss falls offers an efficient way to better understand the causes of fall accidents and helps to initiate proactive actions to prevent accidents before they occur.In training experiments,a loss-of-balance environment was artificially established by means of a balance board to simulate the scenarios in near-miss falls,which efficiently and economically provided training samples for ANN.Through a transition model between static and dynamic near-miss falls,the similarity between simulated and actual scenes of near-miss falls was improved.Data analysis indicated that the mean absolute difference in acceleration features between compensated static and dynamic near-miss falls of all experiment subjects was less than 13%.In evaluation experiments,the evaluation system framework of ANN was built based on TP,FP and FN.Thus,the feasibility of adopting ANN to correctly identify near-miss falls was verified.The results showed that the average precision,recall,and F1 score were 90.02%,90.93%,and 90.42%,respectively,with an average error-detection rate of 16.26%.In test cases,the thresholds H_{2 0%}and H_10%were acquired and illustrated from the perspective of probability.Using 1 min as a time window to amplify thresholds H_20%andH_10%,this paper obtained H_20%'and H_10%',whose were 4.62 and 3.64,respectively.Because the threshold was the ratio of the number of near-miss falls and time,it had the potential to measure the quantitative relationship between near-miss falls and fall accidents.Furthermore,a mobile crowd sensing network based on smartphone was built to accomplish a primary explore of extension from one single construction worker to 5-workers construction team.It explored the crowd-sensing construction safety management,which took workers as producers of sensing data and managers as consumers of sensing data.This approach will help detect hazardous elements,hidden dangers and vulnerable workers.In addition,it provides a new perspective for further holding the safety situations of construction sites in a timely and efficient way.