Object-based representation of feature-based part and process knowledge for planning manufacturing applications.

摘要

Intelligent manufacturing systems require an interface between CAD and CAM to permit reasoning about how the designed part will be recognized, transported, manufactured, and inspected. Obstacles to achieving an efficient CAD/CAM interface include: low level geometric CAD representations and planning systems having limited feedback and reasoning at varying levels of abstraction. The objectives of this research are to develop a representation describing multiple abstractions of part design data, develop a representation to support relationships between part design and application specifications, develop a representation of behaviors critical to operations in a manufacturing domain, and develop a planning system capable of adaptation based on in-process constraints and contextual-based requirements, and past experience. This research presents a knowledge representation scheme which supports conceptualization and inference by modeling domain independent ontologies in the form of objects and semantic primitives. The representation schema permits a plan with similar constraints can be used and resulting knowledge transferred and adapted. The research was demonstrated and evaluated in a prototype considering three applications; part recognition, material handling, and surface finishing.;The developed representation of feature-based part geometry and application information promotes the planning capabilities across various domains. Instead of application-specific feature sets, feature-based part geometry and application information are represented in object-oriented class structures and related through semantic nets. The definitions of and relationships between feature-based part and application representations are viewed as constraints by and provide context for the planner.;The planning system uses the developed representation to: (1) generate, validate, and adapt task, path, and trajectory plans, (2) monitor execution, and (3) resolve anomalies given operations plans as inputs. Application cases contain process parameters, pointers to part geometry, equipment specifications, plans, and past plan execution histories. Task plans specify actions such as "Move" and "Recognize." Path plans describe the robot tool location and orientation. The trajectory planner determines joint angles, velocities, and accelerations for a specified tool path. Replanning due to contingencies detected by algorithmic calculations or sensor feedback are assigned to a planner depending on the application, contingency, and system response.