Geotechnical Engineering Parameter Inversion Based On Improved SVM Based On Swarm Intelligence Algorithm

The geotechnical mechanical index is very important in geotechnical engineering.Whether it is reasonable or not has a great influence on the design and construction of geotechnical engineering.The acquisition of geotechnical mechanical parameters in engineering is mainly through field tests and laboratory tests.Due to the complexity,non-linearity,uncertainty and ambiguity of geotechnical media,traditional methods sometimes have certain limitations and it is difficult to achieve satisfactory results.With the development of science and technology,various intelligent methods have emerged,such as neural networks,genetic algorithms and support vector machine methods.Among them,displacement back analysis based on support vector machine(SVM)is a relatively new inversion method of geotechnical engineering parameters.Due to a relatively complete theoretical basis and good learning performance,it can solve small samples Practical problems such as linearity,high dimensionality,and local minima make it a new research hotspot after neural networks.Although the performance of the method has been verified in many practical applications,there are still some problems in the calculation,including slow training speed,complex algorithms that are difficult to implement,and low algorithm efficiency.In order to improve the calculation accuracy and efficiency of the geotechnical engineering parameter inversion,the main problems in the parameter inversion of the support vector machine are analyzed-the impact of the support vector machine penalty parameter C,the core width g,and the insensitive parameter ? on the final calculation result Larger,correct selection of support vector machine parameters is the key to accurate calculation.By introducing swarm intelligence algorithm to optimize support vector machine parameters.The swarm intelligence algorithm is an optimization method designed by simulating natural population foraging and other activities.It has the characteristics of high search efficiency,fast convergence speed and simple logic.This paper studies three swarm intelligence algorithms of gray wolf,artificial bee colony,and differential evolution gray wolf to improve the research ofsupport vector machine model parameter selection from the perspective of improving search efficiency and model calculation accuracy,and has made innovations in the field of geotechnical engineering.The main research contents are summarized as follows:(1)Support vector machine is a machine learning method based on statistical theory.It needs to construct learning samples for training.The quality of learning samples directly affects the performance of model calculation.New sample construction method.(2)Study the fusion of differential evolution algorithm and gray wolf algorithm to obtain a differential evolution gray wolf hybrid algorithm.Through the preparation of Matlab program,the gray wolf algorithm,artificial bee colony algorithm,differential evolution gray wolf hybrid algorithm and support vector machine are combined,and the performance analysis of the above three support vector machine models is carried out through actual engineering.The analysis shows that,compared with the gray wolf algorithm and the artificial bee colony algorithm,the differential evolution gray wolf hybrid algorithm significantly improves the parameter search efficiency and the prediction accuracy of the support vector machine model.(3)Based on the previous work,combined with a subway tunnel project in Kunming,to carry out engineering application research.The Midas / GTS finite element method was used to construct a three-dimensional simulation model to simulate the construction process of the tunnel,and the displacement deformation in different construction stages was studied as the inversion method to calculate the displacement change amount in another construction stage.Based on the measured deformation of the surface of a tunnel section,the support vector machine model optimized by the differential evolution gray wolf hybrid algorithm is used to invert the rock mass mechanical parameters,and finally the inversion parameters are used to predict the amount of displacement change of the segment of the tunnel section floating Calculated and compared with actual segment deformation monitoring values,the results show that the prediction is accurate and the method is feasible.It provides a new way for the acquisition,verification and deformation prediction of geotechnical mechanical parameters.