Modeling And Simulation Of Shield Sealed Cable's Earth Pressure Based On LS-SVM

The earth pressure of the sealed cabin is a very important controlling parameter and is intimately related to the stability of excavated surface and deformation of ground. However, at present, most constructions are manual operations and controls, which lead to low control accuracy. Thus, the realization of automatic control during the excavation of Shield is a hot topic in the present automation research.At present, most of the researches are focus on some relatively independent areas, such as Propulsion System or Casting System. While research on the whole characteristic of EPB shield system is relatively less and exclude the consideration of affect of the speed of cutter head. The result is that it can't reflect the overall performance during the excavation of shield. In the other side, the research methods, such as Artificial Neural Networks or Fuzzy Algorithm, which can't get a specific form of the model. In addition, the situation of the construction of shield is complex and there are many factors that could affect the earth pressure. Moreover, it generally requires lots of investment in equipments to perform on site test or experiments of physical model which set back the process of research. Therefore, most of the analysis and designs of Earth Pressure Balance Shield System are based on engineering experience. For these reasons, this paper raise a method of establishing the earth pressure model of EPB shield, and do a research of the character of the construction process of EPB shield combined with construction-site data, which could provide some instructions and references for the construction of shield.It is difficult to establish the mechanism model of shield EPB system directly. Therefore, the thesis identify the earth pressure value may be relative with cutter head speed, main thrust, screw conveyor speed and advancing speed based on the mechanism during the excavation of shield and the study on shield theory at home and abroad. Furthermore, the use of principal component analysis to estimate the value of earth pressure and the strength of the correlation of these factors lead to a conclusion that the earth pressure has four principal elements as follows: advancing speed, cutter head speed, screw conveyor speed and main thrust. For these reasons, first, preprocess the construction-site data. Secondly, obtaining the specific model of earth pressure by using Least Squares Support Vector Machine (LS-SVM) method combined with the construction-site data, and using six kinds of kernel function to construct the earth pressure model. With the introduction of RMSE(Root Mean Square Error) and MAPE(Mean Absolute Percent Error), then estimating the precision of prediction of the constructed model, and get the conclusion that the constructed model with a quadratic polynomial kernel function has the optimal precision of prediction. At last, verifying the validity of the mathematical model by simulation experiments, and comparing the estimates capacity of LS-SVM model and BP Neural Network model with MAPE as an estimated standard. The simulation represents that the LS-SVM model gets an excellent predictive ability and generalization ability.