A graph grammar based approach to automated manufacturing planning
In this thesis, a new graph grammar representation is proposed to reason about the manufacturability of solid models. The knowledge captured in the graph grammar rules serves as a virtual machinist in its ability to recognize arbitrary geometries and match them to various machine operations. Firstly, a novel convex decomposition algorithm has been developed to decompose a given part into multiple sub-volumes, where each sub-volume is assumed to be machined in one operation or to be non-machinable. Then the decomposed part is converted into a graph so that graph grammar rules can determine the machining details. A candidate plan is a feasible sequence of all of the necessary machining operations needed to manufacture this part. If a given geometry is not machinable, the rules will fail to find a complete manufacturing plan for all of the sub-volumes. As a result of this representation, designers can quickly get insights into how a part can be made and how it can be improved based upon the feedback of the rules. A variety of tests of this algorithm on both simple and complex engineering parts show its effectiveness and efficiency.