| Deformation monitoring includes continuous observation on the metamorphism of deformation body, analyzing the deformation behavior of the body, forecasting the development status of the body, safety assessment, and so on. With the development of informationization surveying and mapping, observation methods of deformation monitoring have shifted from the traditional manual measurement to modern monitoring methods that converged networking paradigm based on satellite positioning, multi-sensor, the Internet, mobile communications and many other technologies, but the lack of deformation analysis evaluation in practical work results in having no idea about how to assess the security status of the project in the massive monitoring data. Especially the gas tunnel constructed limited by the economic conditions and limited technical level, its security operation matters great to the smooth operation of the west-east national gas transmission project. Its security should be paid serious attention due to bad tunnel site, the complex geological conditions, and the existing security risks. However, there is still no systematical codes of practice on tunnel safety evaluation in our country, it can only be a simple evaluation framework in practice with very limited operability. Therefore, a scientific study on tunnel deformation prediction and safety assessment is very important.Taking Xinjiang East Gas Pipeline-Yanshuigou Tunnel deformation monitoring projects as the basis, based on the theory of deformation monitoring, taking data analysis as the core, with the advantages of artificial neural networks and model portfolio, this paper has studied the application of intelligent neural network research in the operation of the tunnel deformation prediction analysis and safety evaluation, analyzed and evaluated the large number of monitoring data obtained during deformation monitoring Yanshuigou tunnel to ensure safe operation of the tunnel. In the process of paper research, the author has in-depth studied all aspects of key technologies, mainly as follows:1. Aiming at the defect of BP algorithm, the author has researched momentum-adaptive learning rate algorithm, LM algorithm and genetic algorithm to improve the standard BP algorithm, conducted simulation experimental comparison between the three improved methods and standard BP algorithm, the results show that the improved algorithms are significantly better than the standard BP algorithm in both convergence rate and fitting accuracy.2. The model combination is divided into tandem, parallel type and hybrid-type, design BP neural network combination to build an optimal intelligent neural network model- double BP neural network by the hybrid combination of the three improved models and combiners. The case studies show that the intelligent neural network combination model can effectively improve the fitting precision single model.3. Things mode combined with multi-sensor fusion tunnel deformation monitoring of engineering characteristics and difficulties, researched on the application of intelligent neural networks in deformation prediction of tunnel operation period based on time series and affecting factors of tunnel deformation, to solve the problem of tunnel deformation of high accuracy in the multivariate association of micro-system multi-space gray system.4. The author has studied on the construction of tunnel safety evaluation index system and multi-level indicators measure, designed five evaluate sets, used intelligent neural network to learn expert knowledge, carried out comprehensive evaluation on tunnel operation safety evaluation step by step recursive, determine the security level of operation monitoring section according to the evaluation results. |
PDF Full Text Request