Distributed Loop-shaping for Iterative Control of Batch Processes

The idea of iterative control of batch processes has been discussed in the literature for some time. Many related concepts have been published under the names iterative control, learning control, run-to-run control, and others. This article uses the name iterative learning control (ILC). The main idea of ILC is illustrated in Fig. 1. An ILC controller provides feedforward control input to a plant through a batch run. The plant is shown in Fig. 1 in combination with a feedback controller. The feedforward control history for an entire batch run is stored in a memory buffer, and the plant output in the batch is stored in another memory buffer. The ILC update is computed between the batches and uses the output history buffer data to calculate an update to the feedforward control history buffer data. Conceptually, pure feedback control algorithms consider the tracking error to be completely unknown and continuously attempt to compensate for this error as the process goes through a batch run. Feedback control ignores the fact that the error is mostly reproducible from one batch run to another. An ILC approach uses this knowledge to update the feedforward control sequence. Some relevant ideas of the prior work are briefly summarized below. A more detailed bibliography on ILC can be found in [1], [2] and in the recent special issue on the topic [3]. Most of the ILC approaches in the literature consider updating the feedforward control as U (k) = U (k−1) −AY (k−1), (1)