Product quality model based on wavelet relevance vector machine

According to the fact that a common method for product quality modeling has not very high modeling accuracy and its prediction intervals can not be given, a model of product quality based on wavelet relevance vector machine was proposed. The simulation data and the real field data of zinc coating mass from strip hot-dip galvanizing were used for validation. The results show that the model based on wavelet relevance vector machine has a higher prediction precision than those based support vector machine and relevance vector machine, and its prediction intervals can he given. The zinc coating mass forecasting model based on wavelet relevance vector machine for multi-group data has an average of the relative prediction error of 4.52%; thus for the quality control, it provides the necessary decision supports and analysis tools.