Process planning is defined as sequencing a set of processes to achieve a goal. Quality control inspection process planning involves the translation of a part's dimension and tolerance specifications into a sequence of operations to ensure the accuracy and interchangeability of manufactured parts. Considerable time and effort is generally spent generating manual process plans, even for simple parts.; In this thesis, we describe a case-based quality control inspection planning system. Case-based planning involves the representation of process plans (inspection plans in particular) in such a way as to enable the recognition of similar plans and retrieval of the most similar plan among many plans.; In the course of developing this planning system we have developed a representation model applicable to variant as well as generative approaches to quality control inspection process plans, methods for categorization of such plans into similar groups, and methods for the storage and retrieval of such plans. The inspection process developed is for mechanical parts designed using the Geometric Dimensioning and Tolerancing standard and inspected using Coordinate Measuring Machines.; Process plans are represented in part as m-dimensional vectors (patterns) composed of tolerancing data. A workpiece's process plan is decomposed into smaller plan segments based on its tolerancing specifications. A two-phase similarity process recognizes similarity between workpieces. The first phase uses fuzzy clustering to group the parts based on their m-dimensional tolerancing patterns. The second phase involves numeric comparison of plan features and process sequences in order to recognize and locate the most similar plan among previously planned cases. Storage and retrieval is achieved by indexing plan segments' attributes.; Empirical data attests to the practicality and effectiveness of the representation model and recognition methods. A set of 15 sample designs were planned using a library of 25 representative cases grouped into 14 clusters. All but two sample designs (87%) matched pre-existing cases from the library. Two new categories were created, increasing the average cluster size from 1.8 to 2.5 members per cluster. Sustained growth in the quantity of clusters and their membership signifies addition of new knowledge and improves the probabilities of subsequent matches between a new workpiece and the existing cases in the library of cases.; Variant planning, where existing plan instances are used as "blue prints" for new workpieces to be planned, is implemented using case-based methodology. Although, in general, variant planning may be less efficient than the generative method, in the manufacturing domain, similar parts imply similar process plans; hence, locating closely matched parts greatly enhanced the chances of a near or exact match among (sub)plans of similar parts.; Quality control values correctness of the process plans more than speed of plan generation or of its execution; therefore, the planning outcomes were manually verified for their completeness and correctness. However, efficiency in the planning outcome could be time-optimized by previously published methods.
展开▼
