Research On Early-Warning Method Of Construction Safety Of Deep Foundation Of Subway Based On Multiple Information Fusion

According to statistics,in the past 15 years,there have been 246 deep foundation pit accidents in China.The frequent occurrence of safety accidents in deep foundation pit is directly related to the characteristics of "deep,large and tight",but it also exposes the shortage of safety early warning at the present stage.The key to the early warning of the deep foundation pit is the analysis and processing of the early warning information,but at the present stage,in the analysis and processing of the early warning information,there are many deficiencies: The randomness and fuzziness of single index early-warning + artificial comprehensive decision making,the early warning failure caused by the misleading combination of early warning information with different directivity,and the untimely handling of comprehensive information when facing a lot of information.On the other hand,the continuous development and extensive application of multi information fusion technology provide a new opportunity for the comprehensive treatment of the safety early warning of deep foundation pit.Firstly,according to the deficiencies,based on the characteristics of deep foundation pit construction and the theoretical basis of safety warning,the paper identifies five principles of identification of information sources: systematicness,universality,scientificity,independence and goal orientation.Under the guidance of the above principles,the selection and identification of information sources for security early-warning are determined from two aspects: one is divided by spatial object of the deep foundation,and the other is divided by police situation according to construction safety accident of deep foundation pit(collapse accident for example).Secondly,based on the research of multi information fusion theory,T-S fuzzy neural network is applied to this research to deal with nonlinear and large number of mathematical analysis of multi information fusion model,and the model is implemented by MATLAB software.Through a large number of samples training and detection of the model,the results show that all models have better accuracy and generalization ability.Finally,this paper takes the deep foundation pit project of Huangpu metro station as the research object,early warning information is adopted to each multi information fusion model which have been trained.The fusion results of the corresponding spatial location model show that the risk of the east half side of the south end of the foundation pit is particularly concentrated,especially the southern end of the foundation pit which has reached the state of alarm.What’s more,the result of the fusion results of the corresponding model of the alarm situation based on the collapse accident show that the risk of inrushing at the bottom of foundation pit by the pressurized water is particularly prominent,and the tendency of the occurrence is great.Corresponding to these results,the actual situation in the field is that the inrushing happens at the bottom of the south end of foundation pit after the day when the information is collected.The result of the fusion coincides with the actual situation of the site,which not only verifies the feasibility of the method,but also lays a solid foundation for the future application of this method in the early warning of foundation pit.