作者: Ge, Zhiqiang
来源：IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS卷: 62 期:7 页:4336-4343出版年:Jul.2015
Bayesian regularization mechanism is provided for automatically determining the number of latent variables in the probabilistic principal component regression (PPCR) model. By introducing two hyperparameter vectors, the effectiveness of each latent variable can be well measured and controlled. The Expectation Maximization algorithm is employed for parameter learning of both single and mixture Bayesian regularization models. Two probabilistic soft sensors are then developed for online estimation of key variables in industrial processes, performances of which are evaluated through two case studies.
