Research On Intelligent Evaluation Technology Of Dynamic Characteristics,Stability And Disaster Scope Of Accumulative Slope Under Earthquake Action

Earthquake is one of the most important factors that cause landslides.Every year,there are countless landslides caused by earthquakes in China,especially in the western mountainous areas,with the annual loss of billions.However,the research results at this stage mainly focus on model test,numerical simulation and theoretical derivation.With the rapid development of artificial intelligence technology,the research on slope stability through artificial intelligence algorithm has gradually become a trend.However,there are few studies on the whole chain from earthquake occurrence—dynamic characteristics of slope—slope stability evaluation—disaster scope caused by landslide,the specific research results are as follows1)Based on the research results of the method of earthquake magnitude evaluation and prediction,the index system of earthquake magnitude prediction including PGA,PGV,PGD,peak power spectrum and epicenter distance is established.Combined with support vector machine artificial intelligence algorithm,the key parameters such as penalty parameter C and Gaussian kernel function parameters ? are optimized by particle swarm optimization algorithm.A method of earthquake magnitude prediction based on particle swarm optimization and support vector machine(PSO-SVM)is proposed,and the robust effect is studied to verify the stability of the algorithm.2)A large-scale shaking table test with geometric similarity ratio of 1:10 was designed and completed,covering 150 groups of test data of acceleration amplification effect with different earthquake intensity,slope angle and seismic wave type.An intelligent prediction index system of slope acceleration amplification effect is established,which covers the types of seismic waves,peak ground acceleration and physical parameters(density,cohesion,internal friction angle,slope angle and volume ratio)of slope accumulation.The factors and rules of peak acceleration amplification effect of piled slope are analyzed.An intelligent prediction model of peak acceleration amplification effect of piled slope is proposed based on limit learning machine,compared with the traditional neural network algorithm,the advantages of this algorithm are shown from the calculation speed,root mean square error and prediction fit.The superiority of the algorithm is verified,through the analysis of its robust effect,the algorithm has strong stability.3)Based on the shaking table test results and test model data,the slope seismic safety factors under various conditions are calculated by Lizheng slope stability analysis software according to the geometric similarity ratio.The correctness of the calculation results is verified by comparing with the shaking table test results.On this basis,an intelligent prediction index evaluation system for seismic stability of piled slope is constructed,which includes slope,soil weight,cohesion,internal friction angle and seismic intensity.The weights and thresholds of BP neural network are optimized by genetic algorithm,Furthermore,a prediction model of seismic safety coefficient of piled slope based on genetic algorithm and BP neural network is proposed.The results of robust effect analysis show that the model has strong stability.4)Based on the measured data of landslide points in the published literature at home and abroad,the influencing factors of earthquake induced landslide disaster range are studied.The results show that the horizontal sliding distance is positively correlated with logarithm lg V of landslide volume and slope height.The accuracy effect of inverse distance weight method,radial basis function interpolation method and Kriging interpolation method in determining PGA of landslide disaster points is compared and analyzed.The results show that Kriging interpolation method has the highest accuracy in calculating PGA of landslide disaster points.On this basis,an artificial intelligence prediction index system covering landslide volume,slope height,slope gradient and three-dimensional seismic peak acceleration is constructed,and the smooth factor parameters of generalized regression neural network are optimized by using k-fold crossover algorithm,Then,an artificial intelligence prediction model of the disaster area of the accumulation slope based on k-fold cross algorithm and generalized regression neural network is proposed.The results of robust effect analysis show that the model has strong stability.5)Through the investigation and analysis of Xiejiadianzi landslide in Pengzhou City after Wenchuan earthquake,the algorithm model established in this paper is used for the intelligent prediction of slope peak acceleration amplification factor,slope safety factor and its horizontal sliding distance.The prediction result is good,which verifies the practical feasibility of the whole chain artificial intelligent prediction model.