Research On GRNN Model Based On Improved FOA And Its Application In Deformation Monitoring

Deformation monitoring is an effective way to monitor the deformation of the deformed body.By processing the deformation information,the deformation state of body is analyzed,and the future state is predicted.For there are so many factors affecting the deformation,which could lead to nonlinearity,ambiguity and uncertainty.These cause to a large difference between the traditional mathematical model prediction results and the actual situation,so other suitable models are need to be introduced.General Regression Neural Network(GRNN)is a variation of Radial Basis Function Neural Network,and is well suited to solve nonlinear problems.The smoothing factor of the GRNN model determines the accuracy of the prediction data.There are generally two methods for selecting the smoothing factor: the experimental method and the Swarm Intelligence Optimization Algorithm.The experimental method needs to blindly select different smoothing factors over and over again,then by verification,the best smoothing factor is selected among the known candidate values,that waste lots of time and not with high accuracy.The Fruit Fly Optimization Algorithm(FOA)is a novel Swarm Intelligence Optimization Algorithm that helps fruit flies quickly find food and balance between olfactory memory and visual memory.FOA is very good for parameter optimization and also suitable for global optimization.The research content of this paper is summarized as follows:(1)On the basis of discussing the theory of BP neural network model and RBF neural network model in Artificial Neural Network model,the algorithm theory,algorithm structure,model characteristics,network training method and application of GRNN model in deformation monitoring data processing are analyzed in detail.(2)Through the theoretical analysis of Genetic Algorithm(GA)and Differential Evolution algorithm(DE),the comparison of GA shows that DE is better than GA.Based on the experimental analysis and the theory and algorithm steps of FOA,the effects of the parameters such as population size,initial position of population,search step,determination value of taste concentration and iteration times on the optimal value search of the model are analyzed.(3)FOA research: Analyze existing improved FOA algorithms from the aspects of search radius,optimization solution generation mechanism,fly strategy,multi-group cooperative search and combination algorithm.It is proposed to improve FOA with DE,add a escaping coefficient to get a new improved model(DEFOA),which is verified by experimental function.DEFOA is superior to standard FOA in global optimization.(4)As for the problem of GRNN smoothing factor value,propose use FOA model to optimize the smoothing factor,thus obtain FOA-GRNN model.However,in the global optimization process,FOA fly to the optimal taste concentration due to the search mechanism,so it is easy to fall into the local extremum,so we use DEFOA to optimize the GRNN model to obtain the DEFOA-GRNN model.The engineering example shows that the prediction accuracy of DEFOA-GRNN model is better than of both FOA model and FOA-GRNN model,which proof suitable for deformation monitoring data processing and data prediction.