Modeling and forecast of glazing thickness deposition rate using artificial neural network

Abstract

Glazing deposition rate model is a key issue of the off-line trajectory planning for robotic spray glazing. In order to perform the automatic trajectory planning, achieve the accuracy control of glaze film thickness, a modeling method of the glazing thickness deposition rate fitted by the artificial neural network is presented. Based on the experimental data of the glazing thickness, the model is fitted by using the Bayesian normalization and LM optimization algorithm respectively. In contrast with two kinds of simulated results, it shows two models are consistent with the experimental data. However, compared with LM optimization algorithm, Bayesian normalization algorithm converges faster and more accurate. So Bayesian normalization algorithm is better than LM optimization algorithm in fitting the model. The method is feasible to control the precision of glazing thickness. This paper provides a specific theoretical and methodological support for robotic offline programming in ceramic spray glazing manufacturing.

Numerical simulation and experimental study of dip angle spray film thickness under elliptical double Gaussian sum model

“…This study uses the Levenberg-Marquardt optimization algorithm 15 to solve for the spray model parameters, the basic steps of which are as follows:…”