Date of Award

5-2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Mechanical Engineering

Committee Member

Dr. Gregory M. Mocko, Committee Chair

Committee Member

Dr. Georges Fadel, Committee Member

Committee Member

Dr. Laine Mears, Committee Member

Committee Member

Dr. Michael Porter, Committee Member

Abstract

The objective of this research is to couple product and process design knowledge to enable continuous improvement of assembly processes. Specifically, the use of assembly solid model similarity to mine databases and retrieve assembly process information is investigated. Nine techniques of computing solid model similarity from literature are investigated for their correlation with human interpretation of assembly model similarity. A method of computing solid model similarity by using frequency distributions of tessellation areas is developed and investigated. For each of the nine solid model similarity methods, the results from use of component solid model similarity in conjunction with assembly model similarity are compared to the results when only assembly model similarity is used. A survey is conducted to gather human interpretation of assembly solid models from the perspective of assembly process similarity. From the tests conducted it is found that the method of using tessellation area distributions has weak correlation to human interpretation of assembly solid model similarity from the perspective of assembly processes. The D1 method, which uses distance between centroid and random points on the surface of solid models, was found to have highest correlation to survey results. The use of component model similarity in conjunction with similarity of the assembly model was found to improve the precision of the solid model similarity methods. Text similarity techniques from literature are investigated for their correlation with human interpretation of assembly work instruction similarity. Through testing, Latent Semantic Analysis is chosen as the method of computing assembly work instruction similarity since it has moderately positive correlation with respect to survey results and is less sensitive to the use of synonyms than the three other methods of text similarity investigated in this research. The Jaccard method of computing similarity is inherently a measure of consistency in the terminology used between the two texts being compared and this can be used to provide decision support while engineers author assembly work instructions. This will allow authors to understand the level of consistency between their work instructions and the other work instructions within the specific enterprise. The D1 method of computing solid model similarity and Latent Semantic Analysis to compute assembly work instruction similarity are used to compare assembly solid models and assembly work instructions obtained from a survey. In this survey, participants were presented with assembly solid models and asked to author assembly work instructions. The correlation between the solid model similarity scores and assembly work instruction similarity scores (within and across participants) indicates that regardless of assembly work instruction authors, assembly solid models and assembly work instruction share a moderately positive correlation. These results, coupled with the understanding that the causation between assembly work instructions and solid models is limited to those work instructions which describe handling of components and mating of components, can be used for process design knowledge retrieval and reuse. The results from this research can be used to mine databases by using solid model similarity and retrieve assembly work instructions. This will couple product design and assembly process design and allow for continuous improvement of the latter.

Share

COinS