Research On Methods For Hazard Assessment And Displacement Early Warning Of Slope Disasters Based On Bayesian Network And Empirical Mode Decomposition

The mountainous area of our country is vast,and all kinds of slope disasters occur frequently.Because of the large number of slope disasters,it is impossible to carry out a detailed study on each disaster point.Therefore,it is very necessary to evaluate the hazard of slope disaster and then take different measures according to different hazard levels.For different application scenarios,the risk assessment needs different quantitative degree;the hazard assessment method with high quantitative degree generally depends on a large number of highquality sample data,which is often difficult to achieve in the actual situation,so the traditional quantitative risk assessment method is prone to over fitting problems in application.For the slope hazard with high risk,it is necessary to know its development situation accurately at any time,so as to make accurate prediction and warning for the major hazard that will happen,in which the slope displacement is the key to realize the process.However,the current methods of slope displacement prediction have some problems,such as weak adaptability,unable to combine a variety of monitoring data and so on,which leads to low accuracy of displacement prediction.In this paper,the methods of slope disaster hazrd assessment and displacement earlywarning are studied.Combined with fuzzy theory,Bayesian network,empirical mode decomposition and other methods,the qualitative,semi quantitative and quantitative assessment methods of slope disaster hazard are proposed,as well as a new displacement earlywarning method which can combine multiple types of monitoring data.In this paper,the following innovative results and conclusions have been achieved:(1)A staged evaluation system of slope hazard is proposed.The evaluation system includes four types of slope hazards and two evaluation stages with different levels of detail.The method can be used for the hazard evaluation of highway slope disasters during the operation period,and it is suitable for the case of a large number of slopes.The application of the system in Chuanjiu road shows that the evaluation effect of the system is good.(2)The fuzzy comprehensive statistical evaluation method of slope disaster hazard is put forward.This method improves the traditional method of solving membership degree in fuzzy comprehensive evaluation,which makes the method more quantitative and objective.The method is applied to the hazard assessment of Wenchuan earthquake slope disaster.The results show that the method is reliable and can consider the regional characteristics of slope disaster.(3)The Fuzzy-SVM method based on Bayesian network for hazard assessment of slope disaster is proposed.This method uses fuzzy theory to quantify the qualitative engineering experience,so as to obtain the prior distribution of network parameters;at the same time,it uses support vector machine to solve the actual sample potenial distribution of network parameters;then it combines the prior knowledge with the actual sample to obtain the posterior distribution of network parameters.The method is applied to the hazard assessment of cutting slope,embankment slope and slope retaining structure in Wenchuan earthquake area.The results show that the method can adapt to the situation of missing attributes,and can easily use the new data to update the model,and the method is suitable for the hazard assessment of small samples.(4)The ARIMA-BP neural network method based on empirical mode decomposition for displacement early warning of slope disaster is proposed.This method uses empirical mode decomposition to decompose the original displacement into trend term,fluctuation term and sensitive term,then uses ARIMA model to predict the trend term,at the same time uses BP neural network to predict the fluctuation term and sensitive term,and finally synthesizes the three to get the displacement prediction value.Using this method to predict the displacement of Baishuihe landslide and to distinguish the disaster early-warning,the results show that the prediction accuracy of this method is generally high,and it can accurately predict the displacement of the steep increase section;moreover,the early-warning results of Baishuihe landslide are consistent with the actual situation.