Generative Design-to-Fabrication Automation using Spatial Grammars and Heuristic Search
Author: Ertelt, Christoph Heinrich Walter
Supervisor: Prof. Kristina Shea
Institution: Technische Universität München
Planning for Computerized Numerical Control (CNC) fabrication requires generation of process plans for the fabrication of parts that can be executed on CNC enabled machine tools. To create such plans, a large amount of domain specific knowledge is required to map the desired geometry of a part to a manufacturing process, thus decomposing design information into a set of feasible machining operations. Approaches to automate this planning process still rely heavily on human capabilities, such as planning and reasoning about geometry in relation to machining capabilities. In this thesis, the author presents a new, spatial grammar-based approach for automatically creating fabrication plans for CNC machining from a given part geometry. To avoid the use of static feature sets and their pre-defined mappings to machining operations, the method encodes knowledge of fundamental machine capabilities. The use of spatial grammars as a formalism enables systematic formulation of hard and soft constraints on spatial relations between the volume to be removed and the removal volume shape for a machining operation. Further, a software implementation of the core method is presented and nvalidated using several examples of machining a part on a milling machine including changed tools and tool failures during runtime. Advanced heuristic search methods are investigated to further improve the quality of the generated plans in terms of accuracy and machining time. Overall, the approach and method presented is an enabler for the creation of an autonomous fabrication system and CNC machine tools that are able to reason about part geometry in relation to available capabilities and carry out on-line planning for CNC fabrication.