Study On Deep Deformation Forecasting Model Based On Gene Expression Programming

With the construction of large municipal facilities and the development of undergroundspace, a lot of deep excavation engineering appeared. Once the deformation value exceeds thedesign value, it may damage or even cause huge economic disasters and casualties. Prediction ofdeep excavation deformation has become an important topic. The traditional prediction methodsof deep excavation deformation have some shortcomings. Therefore, we need to seek a newprediction method for deep excavation deformation. GEP has a very strong capability of functiondiscovery and high learning efficiency. It is able to dig out the more accurate formula without anyprior knowledge and understanding the internal mechanism of things. Therefore, it is necessary tostudy the prediction model of deep excavation deformation based on gene expressionprogramming.Firstly, the paper describes the background and significance of gene expressionprogramming method in the prediction of deep excavation deformation. It explores the researchstatus of deep excavation deformation and gene expression programming. It analyzes thedeformation mechanism of deep excavation and summarizes the advantages and disadvantages ofgeneral prediction methods.Secondly, based on the theory of grey system and BP neural network, the GM(1,1)modeland BP neural network model have been established with MATLAB. Through the research ongene expression programming, the modeling process is the selection of function set and terminalset, population initialization, chromosome decoding, fitness evaluation, genetic operations. Thepaper has completed the modeling work of gene expression programming with the Javaprogramming language of the Eclipse platform.Finally, GM (1,1) model, BP neural network model and gene expression programmingmodel have been used to make a prediction for a certain deep excavation in Guang Zhou. Theaverage relative error of pile inclinometer of enclosure wall is2.7771%,0.8881%and0.5606%.The average relative error of the horizontal displacement of the top of the pile is2.3739%,1.7183%and0.5809%. The results show that gene expression programming which is moresignificant than the GM (1,1)model and the BP neural network model in learning efficiency canimprove the prediction precision of deep excavation deformation. Therefore, gene expressionprogramming model can be applied to the prediction of deep excavation deformation.