Abstract: In order to solve the problem of unsafe factor identification and interactive analysis in the field of behavior based safety, an unsafe factor identification and interaction analysis model based on deep learning was built after building an unsafe factor identification layer from the "human-machine-environment" aspects, which use different deep learning structures to identify the behavior attributes of workers,unsafe factors in work environment scenarios and equipment operating states. Then, through the interaction layer of factors, the correlation and regression multi-value algorithm were used to analyze the unsafe factors. Finally, representation of the analysis results was achieved by the outputting display layer. An interactive analysis was carried out for unsafe factors in a certain coal mine's three types of production activities, fully mechanized coal mining, tunneling and ventilation, by using the optimal deep learning structures selected by Matlab platform,on the basis of the video and audio data provided by the coal mine. The results show that the model can be used to identify and analyze the unsafe factors, such as operation at coal mining surface without support, and feeding materials into a shotcrete machine going wrong, describe unsafe behavior, and classify the risks and the behavior traces.

Key words: deep learning, behavior based safety, human-machine-environment, unsafe factor, interaction analysis