Secondary measurement selection and scaling sensitivity

Often in process control, it is necessary to use extra measurements, or so-called secondary outputs for monitoring processes. Traditionally, the selection of these outputs is based on criteria which need to be calculated scheme-by-scheme resulting in a combinatorial problem. In this work, it is shown that these secondary outputs can be selected using an efficient measure, the output effectiveness (OE) when the outputs are scaled in a special way. This method does not need any scheme-by-scheme calculation so that the combinatorial problem is avoided. Since the OE measure is dependent on output scaling, its sensitivity to this scaling is also discussed. A scaling sensitivity matrix is derived for the designer to quickly determine the effect of different choices of scaling factors on the OE measure. This information can then be used for efficient output selection. This is demonstrated on a distillation example