Study On Smart Prediction Method Of Slop Stability Based On Hybrid Kernel Extreme Learning Machine Trained And Optimized By Particle Swarm Optimization

Under the background of the rapid development of the national economy in recent years,the construction of infrastructure such as road traffic and railways is testing the builders’ sensitivity to safety accident prevention.The stability research of slope plays an important role in disaster prevention and reduction,and the stability assessment of slope is an important basis for slope harness],reinforcement and above-ground structure design,and an important subject in the field of geotechnical engineering.The traditional slope stability assessment is difficult to rely on a single discipline for research.There are such problems as high nonlinearity,high uncertainty of the relationship between the structure and the large composition of experience assessment.With the development of computer science,the complexity and nonlinear problems of slope stability analysis and evaluation can be solved to some extent.Compared with the calculation of traditional physical mechanics,the establishment of different intelligent prediction models can often play a better effect,providing new ideas and new methods for the study of slope stability.This paper is based on the hybrid kernel extreme learning machine(HKELM)optimized by particle swarm optimization(PSO)such as a new optimization machine learning algorithm.The scheme of slope stability evaluation is proposed,including direct prediction solution of safety factor,search of non-arc slide surface and short-term prediction of cumulative landslide displacement.Firstly,the basic principles of the extreme learning machine algorithm and its application process are summarized.This paper explores the applicable characteristics of different nuclear functions and puts forward the improvement scheme of applying hybrid nuclear function.The method has both local learning and global generalization.In order to improve the defects of multi-parameter input and nonlinearity of slope,the particle swarm optimization algorithm is used as the decision algorithm of the nuclear parameter and output weight.Then,for the relatively simple homogeneous slope and the single characteristic of data input dimension,it is suitable for the PSO-HKELM model to predict the safety factor directly.Through the prediction experiment of slope safety factor of 13 groups of Three Gorges reservoir area,the prediction performance of the model and genetic algorithm optimization,bat algorithm optimization and simple ELM algorithm is analyzed,and the reliability and accuracy of the homogenous slope coefficient are verified.Thirdly,in view of the complex situations such as non-homogeneous slope,the most dangerous sliding surface is searching by particle swarm optimization,and the stability of slope is indirectly predicted.The APSO prediction is close to the standard answer by analyzing the four classic slope questions of the ACADS.The method has a good representation of the search reliability of three typical slope sliding surfaces,such as homogeneous slope,non-uniform medium and weak mezzanine.Finally,the analysis of the time range and the prediction method of cumulative displacement in the process of landslide rock earth creep by WA-PSO-HKELM are proposed.The cumulative displacement signal values are deconstructed by the wavelet function for different influence syllables,including trend items,period items,and random items.The sample of phase space reconstruction and false proximity reduction of learning difficulty,training difficulty and redundancy are reduced.The PSO-HKELM model successfully fitted the cumulative displacement value of the monitoring points related to the tree-level landslide in the Three Gorges reservoir area.This method further verifies that the PSO-HKEM model has excellent applicability and performance for slope problems.