Study On Safety Evaluation Of Highway Slope Based On Online Monitoring System

The safety of highway slopes is an important influence factor that determines the performance of a highway,which also greatly threatens the life and property of road construction workers and road users.During both road construction and operation periods,the highway slope should be monitored and evaluated as a whole to estimate the impact of human activities or environmental conditions act on the slope.Accordingly,disaster prevention measures can be taken in advance once danger is discovered in time.However,although the monitoring work on highway slopes has progressed at home and abroad,it is still in a relatively passive state.The utilization of monitoring data hasn't been enough,while the existed slope safety evaluation systems are not rounded.Aiming at those problems mentioned above,this paper evaluates the safety level of highway slopes using the combination sequences of realtime and forecasted monitoring data of highway slope,and also gives an early warning when danger is identified,which can provide data support and decision making support for highway construction and management.The main works of this paper are listed as below.1.Construction of an online monitoring system for highway slopes.This paper analyzes the sensitivity of slope's influencing factors on the basis of basic types and deformation characteristics of highway slopes.Therefore,data collection requirements of the online monitoring system are determined comprehensively,and the overall frame designing of the slope monitoring system is further realized.Finally,the application principles of monitoring indicators,points and monitoring networks are improved.2.The slope monitoring data is preprocessed by the Improved Unascertained Filtering Method(IUF)based on the Differential Evolution Algorithm(DE).Firstly,this paper takes the original monitoring data of highway slope as the research object to optimize parameters of IUF by differential evolution algorithm,and uses the DE-IUF method to identify the error in original data sequence.Then,the interpolation correction of different interpolation methods is realized using the incomplete original data or preprocessed data sequence,and the applicability and reliability of different interpolation methods is analyzed.3.Multi-source monitoring data of highway slopes are forecasted based on Long ShortTerm Memory Network(LSTM).Firstly,this paper analyzes the basic structure and operation principles of LSTM,which determines the default types of input data and activation functions applicable to highway slope monitoring projects.Secondly,the LSTM forecast model of slope monitoring data is constructed.This model has four structural parameters,such as learning rate,unit value,iteration and nodes of hidden layers.Their specific influences on the prediction model are further analyzed.Next,the optimal parameter combinations of highway slope monitoring data are chosen by comparing the convergence time and average absolute error of models with different parameters.The result shows that the minimum average absolute error can be reduced to 0.120 when the learning rate is 0.040,the batch size is 160,the epoch size is 400,the hidden layers are three layers,the number of hidden layer nodes is 15.and the dropout rate is 0.35,which means the forecasting model for current monitoring displacement data is the best.Finally,the influence of number of monitoring data samples on the prediction results is analyzed,and the prediction results of the model are evaluated comprehensively.The result shows that at least 5680 groups monitoring data are required to meet the prediction accuracy requirements.4.A comprehensive safety evaluation and early warning system for highway slope is established,which is applied to evaluate the real-time safety level of highway slopes.Firstly,this paper uses two dimensionless evaluation indexes,coefficient of stability and coefficient of variation,to judge whether the monitoring point is safe.the highway slope safety evaluation hierarchical model from monitoring index layer to judgment criterion layer and to integrated target layer is constructed based on multi-source monitoring indexes.The AHP-fuzzy comprehensive evaluation method is introduced to construct a fuzzy matrix and calculate the maximum membership degree.Then the real-time safety level of slope is determined on the basis of safety assessment data set.Finally,the early warning principles for safety evaluation of highway slopes are summarized.A comprehensive evaluation and early warning system for highway slope safety is established by combining the microscopic judgment of slope monitoring points and the macroscopic judgment of overall slope safety.