Gaussian Process Of Deformation Based Algorithm

With the rapid economic development, construction is increasing.The construction ofthe building, from the start of construction to be completed, and the completion of theentire period of operation should be continuous monitoring, in order to grasp thedeformation, to identify problems and to ensure the safety of the engineering andconstruction. Accordingly, deformation monitoring for large buildings and processing thedata is particularly important.At present,there are many domestic and foreign deformation analysis modelalgorithm,especially in the intelligent algorithm of neural network model. In recent years,the " nuclear learning " is a issue in the field of machine learning, the most representativeis the support vector machine,gaussian process. Gaussian process, as a new machinelearning methods,providing a principled, probabilistic approach to learning in kernelmachines.This gives advantages with respect to the interpretation of model predictions andprovides a well founded framework for learning and model selection. The oretical andpractical developments of over the last decade have made gaussian process a seriouscompetitor for real supervised learning applications. In this paper, the angle as a point tostudy gaussian process in deformation monitoring data processing, main contents andresults are as follows:First, we systematically expound the theory, principles and ideas of the gaussianprocess, using gaussian process theory to analyze the deformation monitoring data,examples show that gaussian process regression in the deformation monitoring dataprocessing have high precision and simple procedure.Second,the hyper-parameters of the gaussian process regression is obtained bytraditional optimization methods (conjugate gradient method),Conjugate gradient methodin the optimization process dependent on the initial value,which is difficult to determinethe number of iterations and drawbacks such as local optimization.Defects in thetraditional methods, we use Gaussian process regression algorithm based on particleswarm. At last, we use Gaussian process model, particle swarm gaussian process model, BPmodel on the deformation monitoring data of a tunnel project.Evaluating the accuracy ofthe model by a certain error,the particle swarm algorithm gaussian process modelprocessing gets better results, it is applicability.