A Connectionist Expert System Approach for Representing a Press Operator's Machine-Specific Knowledge

An integrated connectionist expert system was developed to replicate the on-line inking adjustments that a press operatof applies during a job run. The inking adjustments correspond to a mapping of the density deviations (between a sample print and the match-print) to the ink key adjustments applied in order to overcome this deviation. The nature of a proficient operator's knowledge is machine-specific, i.e., he has learned to predict that particular machines' reactions, as observed on the print, to his compensatory control actions, consisting fo the inking adjustments. The learing capabilities of connectionist networks (or neural networks) were integrated with the knowledge-based representation of a rule-based expert system in order to represent this machine-specific knowledge.