Study On Deformation Prediction Method Based On EEMD Denoising And FOA-SVM

Deformation monitoring is a technology and methods of collecting the deformation of thedeformable body information, the processing and analysis of the modification information is theultimate purpose of the deformation monitoring. The deformation data preprocessing caneffectively remove data error, and is conducive to improve the accuracy of the deformationanalysis and forecast results. Has the characteristics of nonlinear, ambiguity and uncertainty dueto the deformation of the deformable body, the traditional deformation prediction accuratemathematical model results with the actual situation quite different.Support Vector Machine, based on the statistical learning theory, is proposed a newmachine learning methods by Vapnik et al in the1990s. It can seek the optimal solution of thelimited sample of data, and has stronger theoretical basis and better generalization performancethan neural network learning algorithm based the principle of empirical risk. The parameters ofthe prediction model of support vector machine determine the sample training error and thepromotion of the predicted sample, however, there is no complete theory and method to solvethis problem and only by example simulation algorithm optimization currently. Fruit flyoptimization algorithm (FOA) is based on the characteristics of the fruit fly foraging for food,swarm intelligence manifested olfactory memory and visual memory synergies. It has a goodeffect on the optimal choice of the parameters, and is able to do a global optimization.Firstly, the high-frequency noise signal in the deformation data is isolated by the ensembleempirical mode decomposition (EEMD), and also contains a useful signal for the high-frequencynoise, its threshold quantization process, to keep noise contained useful signal to the completionof the deformation data preprocessing. Then to solve the open-ended question, the selection ofsupport vector machine parameters and also the key to the successful practical application ofsupport vector machine prediction model, fruit fly optimization algorithm (FOA) optimizes theselection with engineering examples, and prove that FOA simplify the parameter selection ofsupport vector machine and avoid the blindness of the hyper parameter selection in the SVMforecasting applications of actual engineering.