Adaptive optimal control for repetitive processes

The DOP-capable PSO is employed to adapt controller gains to changing repetitive process parameters.

Repetitiveness of a process to be controlled is innate to many industrial systems. Its presence creates new possibilities in developing control algorithms that take into account errors from previous passes to shape the control signal in current pass. Nevertheless, many repetitive processes can and are effectively controlled using non-repetitive controllers, i.e. quality requirements can be met without implementing repetitive controllers. However, it is not rare that such systems are subject to dynamics variations and in turn adaptation of controller gains could be a desirable feature. This submission illustrates concept of employing a particle swarm optimizer (modified to enable dynamic optimization problem solving) to adapt non-repetitive controller gains in the gradient-free manner. The performance is continually evaluated -- during the regular operation of the system -- using user-defined performance index and the PSO tracks the moving optimum as the plant is assumed to be time-variant (is not stationary). For more information please see comments inside the m-file.