Developing Soft Sensors Based on Data-Driven Approach

Considering the time-varying nature of an industrial process, a soft sensor based on the fast moving window partial least squares (FMWPLS) is developed. The proposed approach adapts the parameters of the inferential model with the dissimilarities between the new and oldest data and incorporating with the kernel algorithm for the PLS, therefore, the computational loading of the model adaptation is independent on the window size. Since a moving window approach is sensitive to outliers, the confidence intervals for the primary variables are created based on the prediction uncertainty. In addition, the prediction performance of a soft sensor is not only dependent on the capability of the inferential model, but also relies on the data quality of the input measurements. In this paper, the input sensors are validated before performing a prediction. The deterioration of the prediction performance due to the failed sensors can be removed by the sensor validation approach.