“…In addition, with the rapid development of distributed control systems, many data-driven soft sensor modeling techniques have been extensively studied and applied successfully to many domains with time savings and low cost, particularly in process monitoring. 5,6 Many data-driven soft modeling methods, such as artificial neural networks (ANNs), 7,8 principal component analysis (PCA), 9 and partial least squares (PLS), 10 have extended the popularity for soft sensors. However, they can only utilize labeled data that contain both input and output samples, and extensive unlabeled data are discarded.…”