Virtual Sensing Technology in Process Industries: Trends and Challenges Revealed by Recent Industrial Applications

Abstract

Virtual sensing technology is crucial for high product quality and productivity in any industry. This review aims to clarify the trend of research and application of virtual sensing technology in process industries. After a brief survey, practical issues are clari ed by introducing recent questionnaire survey results: 1) changes in process characteristics and operating conditions, 2) individual di erence of equipment, and 3) reliability of soft-sensors. Since input variable selection is crucial for high estimation performance, conventional methods and new group-wise variable selection methods are introduced, and the usefulness of the group-wise variable selection methods is demonstrated through industrial case studies. Just-in-time (JIT) modeling is dealt with as a promising virtual sensing technology that can cope with changes in process characteristics as well as nonlinearity. Recent developments leading to successful industrial applications are introduced: correlation-based JIT (CoJIT) modeling and locally weighted regression (LWR), especially LW-PLS, with modi ed similarity measures. Manufacturing processes in di erent industries are quite di erent in appearance, but they have very similar problems from the viewpoint of quality issue. There remain practical issues requiring further research e orts to realize high-performance, maintenance-free virtual sensing technology.

Adaptive soft sensor for quality prediction of chemical processes based on selective ensemble of local partial least squares models

“…The partial least squares (PLS) is adopted to construct local models, due to its previously discussed merits and wide population [Kano and Fujiwara, 2013;Kim et al, 2013a]. The problem of PLS in dealing with nonlinear processes can be tackled by the local learning strategy, i.e., developing locally valid PLS (LPLS) models.…”

“…For comparison purpose, the performance of several state-of-the-art and commonly used PLS based adaptive soft sensing methods was also provided and analyzed. These benchmark methods consist of the recursive PLS (RPLS) [Qin, 1998], the locally weighted PLS (LWPLS) [Kano and Fujiwara, 2013], the moving window PLS (MWPLS) [Liu et al, 2010] and the localized adaptive soft sensor (LASS) [Ni et al, 2014]. The estimation accuracy is evaluated on the test dataset by several commonly used error indexes including the root mean squares error (RMSE), the relative RMSE (RRMSE) and the maximum absolute error (MAE), which are defined as …”

“…Another critical problem needs to be solved in soft sensor development is the performance deterioration of soft sensors after they are deployed into real-life operation due to drifts of process characteristics Kim et al, 2013a;Kano and Fujiwara, 2013]. These drifts may result from catalyst deactivation, mechanical ageing, change of operating conditions, variation of feed properties, or even climatic change, et al Therefore, developing adaptive soft sensors to adapt them to new process dynamics automatically is necessary for prolonging their life time in industrial applications.…”

Prediction of Molten Steel Temperature in Steel Making Process with Uncertainty by Integrating Gray-Box Model and Bootstrap Filter

“…Grbić et al (2013) used the confidence region corresponding to ±2 standard deviations to evaluate the reliability of adaptive soft-sensors. In addition, the reliability of soft-sensors was discussed in the review paper by Kano and Fujiwara (2013).…”

High-Performance Prediction of Molten Steel Temperature in Tundish through Gray-Box Model

“…11) In PLS with one output variable, input data and output data are decomposed as follows: T is a loading vector of y, and E and f are errors. N, M, and R denote the number of samples, that of input variables, and that of adopted latent variables, respectively.…”