Application Of Machine Learning In Safety Management Of Civil Engineering Construction

With the concept of artificial intelligence and the rapid development of computer software and hardware,machine learning as a powerful tool to achieve artificial intelligence has brought tremendous impetus and far-reaching influence to the development of various fields.Finding the appropriate way to apply it is the key to breaking the bottleneck of traditional industries and achieving intelligence.Today,accidents occur frequently during the construction in our country,and cause huge economic or life losses.Therefore,safety management is particularly important,even vital.However,at present,major construction safety management methods used by managers such as pre-job training,manual supervision,and video surveillance cause huge waste of labor,time and money,and it is impossible to achieve efficient prevention or timely warning.In order to promote the informatization,automation and intelligent development of construction safety management,this paper proposes a way to apply the related technologies of machine learning in conjunction sites.The main application exploration includes the following three aspects:(1)Based on sensors embedded in smartphones and machine learning algorithms to classify workers' actions and identify dangerous behaviors under specific conditions.An intelligent behavior recognition system for construction workers,especially for those who work in high altitude was established.It can automatically determine whether the worker use the seat belt correctly or not,so the motions of workers would be recognized and classified into the safe one or the dangerous one,whereby reducing the occurrence of high altitude falling accidents.(2)Based on deep learning theory,an image recognition model and a detection model were trained.The former could automatically recognize the bridges' types and the latter one could detect and locate the main components in images.Both of them made it into reality that monitoring equipment could truly “know” and “understand” structures.(3)Based on deep learning theory,multi-target recognition and positioning algorithms were used in the monitored hoisting scene.according to the recognition and tracking results,the spatial interaction between the work and the lifted block could be described and the dangerous level of the construction scene could also be evaluated.Once pre-defined dangerous scene occurred,timely warning would be sent to prevent lifting.Through the validation of experiments and practice,the application of proposed methods in this paper has achieved satisfactory results.The rational application of machine learning algorithms makes the intelligence of construction safety management no longer out of reach.