Research On Data Fusion Technology Of Main Parameters In Sugar Boiling Process

In the sugar boiling process, the detection of super saturation has become a difficult problem.This paper tries to use data fusion technology to solve this problem. Among many algorithms of data fusion, artificial neural networks receive increasing attention in the applicaton of data fusion because of its strong generalization ability and other advantages.This paper uses artificial neural networks and the data measured at scene to establish a structure of 3×3×1 BP neural network model, which can forecast the super saturation and achieve the integration of testing parameters, so as to solve the detection problem much better.Contrapose to BP's shortage, this paper utilizes PSO to optimize BP neural network's connection weights, and propose an improved particle swarm optimization algorithm-BWPSO (Best and Worst Particle Swarm Optimization). In addition, contrapose to LMBP's problem in the process of determinating parametersμand V, it uses PSO algorithm to guide the selection parameters, which do not have to rely on experience or repeated tests. ACO is taken in to optimize BP neural network's connection weights, at the same time, PSO, ACO and BP are put in various groups to propose different training methods of the super saturation forecasting model.Based on the true test data, the best combination algorithm is selected to forecast the super saturation by comparing the error of various integration algorthm.Finally, using Lab VIEW as the development platform and in conjunction with MATLAB, this paper builds a data acquisition and analysis system in sugar boiling process based on virtual instrument technology and data fusion technology.