Process stabilizing process controller

Abstract

A process controller is disclosed for controlling a process. The process controller includes a process supervisor for accessing process control data clusters from a library. Each data cluster provides control data, e.g., setpoints, deadbands, execution frequencies and alarm ranges for process actuator controllers. Each actuator controller maintains a related process attribute within a deadband by controlling at least one process actuator. Data clusters for the library are provided in substantially real-time by a learning unit. Data continuously output from process related sensors are transformed and supplied to the learning unit as training data for learning at least one relationship between process control data and data for at least one substantially uncontrollable process influencing factor (e.g., processing site environmental factors such as temperature and barometric pressure). Prior to inserting a newly provided data cluster into the library, the data cluster is evaluated for predicted successful performance when accessed by the supervisor. Data clusters are also evaluated after each use by the supervisor, with non-performing clusters becoming inaccessible for further use. The supervisor can also fine tune or adjust control data to more closely fit the present behavior of the process. The process controller is embodied in controlling a coal drying process using a combustion gas heat transfer technique.