Function Fitting about Internal Stress of Ceramic Paste Based on BP-NN and SVM

Abstract

With the development of modern intelligence technology, BP Neural Network (BP-NN) and Support Vector Machine (SVM) have become hot topics of current international machine learning community. In order to solve the function fitting problem about the internal stress of ceramic paste, a fitting method based BP-NN and SVM is proposed in this paper. By introducing the structures and characteristics of two methods briefly, two methods can rationally solve the problem of multi-input and single-output function fitting during the soft measurement process about internal stress of ceramic paste. The simulation results show that BP-NN and SVM methods can both make up the limitations of "cftool" function in MATLAB which only solves the problem of the single-input single-output, In addition, SVM is better than BP-NN on approximation and generalization ability, and the simulation speed of SVM is also faster than the ones of BP-NN.