Development of correlation-based pattern recognition and its application to adaptive soft-sensor design

Although linear regression is a simple and useful method to build process models, they do not always function well in practice due to not only changes in process characteristics but differences of specifities between the equipments when multiple equipments are operated in parallel. To cope with them, the correlation between variables should be considered. In the present work, a new pattern recognition method, referred to as nearest correlation (NC) method that can select samples whose correlations are similar to the query point without supervised signal is proposed. The proposed procedures are as follows: 1) Subtract the query point from all the other samples. 2) Calculate the correlation coefficient between all pairs of arbitrary two subtracted samples, and the pairs whose correlation coefficients are close to -1 are selected. 4) Derive the subspace containing the query point from the selected samples. 4) The Q statistics between all samples and the derived subspace are calculated, and the samples whose Q statistic is small are selected as the similar samples to the query point. In addition, a new soft-sensor design method integrating the NC method and just-in-time (JIT) modeling is proposed. This method is referred to as correlation-based JIT (C-JIT) modeling, and it cope with the changes in process characteristics and the differences of specifities between the equipments. The usefulness of the proposed NC method and C-JIT modeling are demonstrated through case studies of CSTR process.