Slope engineering is complex huge open system. Its deformation evolution characters and safety are controlled by many factors that are random and fuzzy and have nonlinear relationship. Because we can't understand mechanism well up date, support vector machine is attractive to these. This paper reports some research progress conducted by the author. The results include: Estimation of slope safety using support vector machine (SM), Slope reliability analysis using support vector machine, Deformation estimation of slope using support vector machine.
Slope engineering is a subject which has strict theory and strong engineering background. During the development of decades, all kinds of solving methods were purposed using conventional methods such as elastic mechanics, elastic-plastic mechanics and mathematics, etc. Because of the complexity, nonlinearity, fuzziness of geomaterials, we don't get the satisfied results by the conventional methods[1]. Intelligent methods such as neural network, support vector machine, etc. are attractive to these. Support vector machine is a new machine learning method; it provides an approach to study the slope engineering. In this paper, we study some problems of slope engineering, and get some important results.
Support vector machine is a new machine learning technical based on statistics learning theory. It was developed at AT&T Bell laboratories by Vania and co-workers. Due to its good characters, it attracted the researcher from all kinds of fields. Originally, support vector machine have been developed for pattern recognition problems. So, it has a lot of successful applications in pattern recognition such as handwrite recognition, face recognition etc [2,3,4]. Recently, with the introduction of Vania's ε-insensitive loss function, support vector machine has been extend to solve nonlinear regression estimation problems, and they exhibit excellent performance. Comparing to neural network, it is based on the structure risk minimization principle; its training is a uniquely solvable quadratic optimization problem, and eventually resulted in better generalization performance. The basic idea of support vector machine is mapping the data into a high dimension feature space, and then finding the nonlinear relationship between input and output in a new space. Support vector machine has strict basement of theory, which is based on the structure risk minimization principle, it's training is a uniquely solvable quadratic optimization problem, and has better generalization performance. It can solve well the real problem with small samples, nonlinearity, and high dimension.
Slope stability analysis based on support vector machine The stability of slope is influenced by many factors. According to the experiment of real case, the influence factors include rock density γ, rock friction angle C, friction coefficient φ, slope angle φr, slope height H, pore water pressure u, type of rock mass structure, etc.
SVM-based reliability analysis is to construct a support vector machine model to approximate and replace the performance function. The value of performance function is easily obtained because of the robust generalization capability. An example is presented for illustrating the applicability of the proposed approach.
