| With the fully construction of the subway,the safety of subway is becoming more and more important,and the monitoring and forecasting analysis to the metro structure deformation has become an important direction for domestic and foreign scholars.This paper focuses on the combined model method to predicting the deformation of the subway structure,so as to improve the prediction accuracy that ensure the safety of the subway.In this paper,the data come from Nanjing subway structural deformation monitored automatically.The main contents and conclusions are as follows:(1)This paper analyzes the causes of the deformation of the subway structure systematically,and introduces several commonly used forecasting models,analyzes the ideas and characteristics of each different model,and elaborates the modeling process of each model in detail.(2)Introduces the basic classification of the combined model,and then introduces the basic idea of combination model modeling based on the error squared sum minimum criterion.Then,it elaborates the method to fix the weight of the combined model.(3)Particle swarm algorithm and BP neural network algorithm are studied emphatically.The weight of the combined model is proposed by using the particle swarm optimization algorithm,and the BP neural network model is used for the first time to compensate the combined model error.Using the Nanjing subway automated monitoring data to verify the case,build a combined mode with time series and regression analysis model.(4)The combination model constructed in this paper is applied to the Nanjing subway monitoring data,and the precision is improved better.The prediction accuracy of the combined model is 10%～22%higher than that of the time series,which is 13%～24%higher than the regression analysis model.(5)By using the BP neural network model constructed in this paper,the combined model is applied to the application of Nanjing subway monitoring data,and the precision is improved obviously.The accuracy of the combined model after BP neural network optimization is 21%～27%higher than that of the combined model,31%-43%higher than the time series model,32%～42%higher than the regression analysis model.The results show that the BP neural network optimal combine model has a good effect on the deformation prediction of the subway structure,It has important theoretical significance and practical application to ensure the safety of subway.(6)The particle swarm algorithm and the linear programming algorithm are used to solve the weights of the combined model respectively.The prediction accuracy of the two models is almost the same.This verifies the correctness of the results based on the error squared sum minimum criterion. |
