Project : Develop a method of automatically generating controllers for flexible manufacturing systems

A system for modeling the flexible manufacturing system and the parts to be processed in the FMS was developed using graphs. A conversion system was then developed to convert the graphs to a specific type of Petri net. The Petri net functioned as the execution system of the controller (see details here). Based on the Petri net a matrix was created that represented the state of the FMS. The matrix was used to provide the inputs to an artificial neural net where arc weights were restricted to [-1, 0, 1] and special arcs (known as inhibit links) were allowed to prevent the firing of a neuron.

The initial neural network was constructed based on finding paths through the graphs representing the parts and then expanding these paths using the graphs representing the FMS. Each node in the part graph represented a processing step on a specific machine. Each arc in the graph represents a movement of the part from the machine associated with the node at the tail of the arc to the machine associated with the node at the head of the arc. The single movement described in the part graph was expanded to include all of the smaller movements necessary to complete it using the graphs representing the FMS.

After the initial neural net was developed, additional nodes and links were added to prevent simultaneous activation of conflicting outputs. This was done by developing a list of outputs that conflicted based on data from the Petri net and additional data supplied by the user (transportation movements that could not occur simultaneously, but which did not use a common resource). When conflicting outputs were generated, a node representing the conditions that created the conflict and a selection system was added. The selections that could be made were represented using a string and a genetic algorithm was used to select the best string.

For further details

Scott, W.D., "A Flexible Control System for Flexible Manufacturing Systems," Ph.D. Dissertation, Texas A&M University, 2004