Research On The Earth Pressure Theory And Intelligent Prediction Of Structure Displacement For The Culvert Under High Fill

Culvert is mainly used to discharge water from gulley and stream in two sides of road. Culvert can also traffic without water. Span length is less than 5 meters. Culvert is one of lifeline projects, and frequently occurs in the regions where there is no abundant river water or shallow groundwater, such as traffic, highway, railway, municipal works and war industry, etc. An efficient development of water resources in such regions is an important problem for transportation. Culvert is positive buried structure, and its design is guided by norm when the fill on culvert is low in the past. With the development of China highway, highroad grade criterion increased, and the proportion of culvert with high fill is greater. However, under this situation, High filled culvert designed without the reasonable theory instruction, which leads to some of them fail in the period of construction or after completion of work. So research about vertical earth pressure acting on high filled culvert is significant in engineering and society development. Considering some different influencing factors, modest tests and numerical simulation are performed to analyse vertical earth pressure distribution rule and fill layer displacement field. Using of culvert structure's measure data from model test as sample, programmed AGA-BP (adaptive genetic algorithm-back propagation network) system are developed to forecast culvert section stress. The contents of research in the paper as follow,(1) Based on similarity theory and some experimental results, a renew criterion was firstly put forward to judge rationality whether the test result as to a similarity index. Intermixture with sand, gypsum and diatomite as raw material specimens, tests with different similarity modeling material were performed. The relations among four influencing factors including water quantity, ratio of sand and gypsum, ratio of gypsum and diatomite, curing condition with material mechanical properties were got. According to geometric similarity ratio, similar material was used to make arch and slab culvert model. The model can more objectively reflect culvert force.(2) Adding soil by layers and load by jack to simulate real constructing course, the regularity of vertical earth pressure on the high filled culvert was investigated in the thesis. Vertical earth pressure increased nonlinearly with fill height increase. Subsidiary stress from stiffness differences and soil arching action decided the vertical earth pressure together. Model test and structural mechanics method analysed culvert's high and low stress distribution in order to detect the disease reason. Subsidiary stress causing by temperate construction load and stiffness differences is bigger than lastly load which leads to disease with lower fill. By measurement of index point's displacement, the displacement on top of culvert was protruding plane. With fill height increasing, differential settlement was reduced.(3) Considering some influencing factors including topographies of galley, fill characters, parameters of foundation soil, different foundation treatment and culvert structure form, numerical simulation based on ANSYS software was performed. Different parameters related with vertical earth pressure were found which could offer a reference to deducing a high filled culvert vertical earth pressure formula.(4) On the basis of results from modal test and numeric simulation, distribution regularities of vertical earth pressure got and came to the conclusion that the earth pressure should be different as the height of fill over h0 meters. So, vertical earth pressure formula found respectively according to rigid and flexible foundation processing mode. By contrasting with measurement data and calculating one example with other nine norm formulas, the thesis discussed the results from every formula.(5) Combined with the advantages of genetic algorithm and BP network, programmed AGA-BP intelligence system can forecast culvert displacement. AGA changes cross, mutation probability which can prevent random nomadism in the late of evolution. In order to prevent overtraining effectively, AGA fitness function is defined as standard deviation of accumulation forecast errors. The neural network with optimal architecture determined by AGA is used to train and forecast sample from model test data. Results show that neural network structure determined by AGA-BP system can forecast displacement about culvert with high fill effectively.