Moving window adaptive soft sensor for state shifting process based on weighted supervised latent factor analysis

Process nonlinearity and state shifting are two of the main factors that cause poor performance of online soft sensors. Adaptive soft sensor is a common practice to ensure high predictive accuracy. In this paper, the moving window method is introduced to the supervised latent factor analysis model to capture the state shifting feature of the process. To make the moving window strategy more efficient, the weighted form of the supervised latent factor analysis approach is applied. In this method, contributions of training samples are expressed through incorporating the similarity index into the noise variance of the process variable, which renders strong adaptability of the method for describing nonlinear relationships and abrupt changes of the process. A numerical example and a real industrial process are provided to demonstrate the effectiveness of the proposed adaptive soft sensor.