Architecture for a neural expert system for condition-based maintenance of blanking

This paper describes a proposal for a hybrid architecture for monitoring and Condition-based Maintenance (CBM) of punching/blanking of sheet metal. Previous work shows that it is possible for certain applications (in which process parameters are sufficiently stable) to detect tool wear and other important aspects of the blanking process from monitoring force-displacement measurements. In order to be able to apply this principle into practice, typical functionality of both a neural network and an expert system will be required. After a review of existing literature, this paper expands on a basic architecture for such application and discusses translating functionality into a practical solution.