| Today, manufacturing quality control is characterized by a plethora of data, continuously collected across the manufacturing processes and throughout a product's life cycle. The level of details with which products can be measured today allows for extraction of both dimensional and non-dimensional quality characteristics. Traditionally, most of the quality inspection systems can only handle the dimensional information and few addresses the issue of product performance both during assembly and in the field of use. Besides geometric dimensions, the product performance is also determined by other non-dimensional characteristics. For example, the density distribution of a blade plays an important role on its aerodynamic performance. Hence, in order to integrate all the dimensional and non-dimensional information for better quality inspection, process diagnosis and performance analysis, I propose a representation for the concept of "Product DNA".;I have developed a representation of "Product DNA", called as-manufactured CAD model, to encode the information of geometry dimensions, surface texture, and physical attributes such as mass density of a product. Under the framework of as-manufactured CAD model based "Product DNA" concept, this research mainly focuses on the coding process of geometry dimensions, surface texture and physical attributes.;For dimensional geometry genome, a non-rigid registration approach is proposed to encode the geometry information into the as-manufactured CAD model based on inspection points. In this approach, a weighted mutual distance method is utilized to establish the correspondence and then the template object is iteratively aligned to best fit the measured points with affine and free-from deformation transformation. For surface texture genome, a B-spline wavelet-based multi-resolution analysis (MRA) approach has been proposed for surface texture characterization. Compared to traditional surface texture analysis methods, the B-spline wavelet-based MRA method allows finer frequency regimes decomposition and thus achieves more precise diagnosis of process faults. For physical attribute genome, a systematic approach to reconstruct a heterogeneous model based on mass density points is proposed. The decoupled B-spline representation to model geometry and mass density allows more modeling flexibility and save huge storage space. Moreover, constraints and multi-resolution based mass density fitting algorithm guarantees to achieve reasonable mass density range and satisfied accuracy. |
