Study On Approaches For Estimation Of Ternary Solution Physical Parameters

As one of developing strategies of sensing technology, multifunctional sensing technology has been gradually employed in real industries. Multifunctional sensing technology consists of two aspects which are multifunctional sensor and its signal reconstruction. Multifunctional sensor simultaneously can sense several parameters and, signal reconstruction realize measurands reconstruction based on outputs of that sensor. In real industries, signal reconstruction is very important to realize detecting some significant parameters on line.Focus on ternary solution (mixture liquid of sodium chloride and sucrose) used in osmotic dehydration process in food engineering, this dissertation studies on signal reconstruction of multifunctional sensor for sensing concentrations of ternary solution to realize estimation of parameters (concentration, density, viscosity) of this solution and make some preparation for on-line detection of parameters.This dissertation orderly proposes three approaches of signal reconstruction to realize estimation of parameters of ternary solution, including B-Spline Least-Squares (B-Spline LS), Least-Squares Support Vector Machine (LS-SVM) and Adaptive Network-Based Fuzzy Inference System (ANFIS). This dissertation deeply researches mathematical principle of these approaches, gives details of realizing estimation of parameters of ternary solution with these approaches, and gives detailed analysis and comparison of estimated results of parameters.High dimensional B-Spline could be constructed by special method. Theory and reality indicate that B-Spline LS is suit for processing grid data but not scattered data. The reconstruction of concentrations of ternary solution is a problem of processing scattered data in 4-dimension space. LS-SVM makes up limitation of B-Spline LS and realizes reconstruction of concentrations with high accuracy. But with mass data, for losing sparseness, the structure of LS-SVM will be complicated. ANFIS avoids that and still realizes reconstruction of concentrations with high accuracy. In the course of studying ANFIS, the author sufficiently utilizes characteristics of Subtractive Clustering and Fuzzy c-means Clustering and aggregates them to realize structure identification of ANFIS. That makes good effect.This dissertation provides basis of detecting parameter of ternary solution on line, and establishes definite theoretics foundation for application of multifunctional sensing technology.