Using Process Data for Finding Self-optimizing Controlled Variables

In the process industry it is often not known how well a process is operated, and without a good model it is difficult to tell if operation can be further improved. We present a data-based method for finding a combination of measurements which can be used for obtaining an estimate of how well the process is operated, and which can be used in feedback as a controlled variable. To find the variable combination, we use past measurement data and fit a quadratic cost function to the data. Using the parameters of this cost function, we then calculate a linear combination of measurements, which when held constant, gives near-optimal operation. Unlike previously published methods for finding self-optimizing controlled variables, this method relies only on past plant measurements and a few plant experiments to obtain the process gain. It does not require a model which is optimized off-line to find the controlled variable.