The Application Of Ann In The Soft Ground Of Settlement Prediction And Intelligent Evaluation In Improvement Of The Expressway

Ground is the most important structure to expressway, and directly influences the construction quality of highway. The post-construction settlement of ground is one of the key factors influencing the quality of ground. In the process of subgrade construction, the control of ground filling speed and each layer construction time of road is based on roadbed settlement prediction. So, it is very important and significant, objectively, accurately and timely Predict the final settlement, and take reasonable scheme, for ensuring the stability in ground construction and post-construction settlement. The settlement of soft ground is affected by multiple factors, so it is a very complicated technical problem to precisely predicted and take reasonable reinforcing scheme. Giving above, this paper focuses on settlement prediction and evaluation method of reinforcement scheme of soft ground of expressway.According to the research contents and sequences of chapters and sections, the main achievements of the dissertation are concisely mentioned as follows:(1)In recent years, although the research and application of neural networks in the prediction of soft ground settlement has been a lot, mostly for the BP neural network or combination methods of research and application, can not be fundamentally overcome some of the BP neural network inherent defect. This paper presents a new neural network—GRNN neural network, and be used to predict the final post-construction settlement of soft ground, this modeling requires less parameters、shorter training time、network more stable and network more operational, relatives to the BP network. At the same time, it is adaptable to small sample, and approximation ability and extrapolation ability are also strong advantages. At the same time, it has a strong ability to adapt to small sample, approximation ability and extrapolation ability are also stronger.etc advantages.(2)In view of the soft ground reinforcement scheme less intelligent evaluation system, this paper puts forward a new evaluation method, PNN neural network evaluation method. This evaluation method, its modeling needs less parameters^training process sample、fast convergence rate, less human factors, higher prediction accuracy、stronger additional ability of the sample of additional ability and other advantages, compared with BP evaluation method, it can be used in the future engineering application.