Reverse engineering utilizing domain-specific knowledge

Reverse engineering is the process of defining and instantiating a model based on the measurements taken from an exemplar object. The measurement (or data sensing) process is prone to random and systematic errors and often fails to sense the object in a manner consistent with the intended functionality of the object's design. Therefore a model fit directly to the data will not faithfully capture the geometry of the part (the form), nor the relationships among features of the part (the function) as originally specified by the designer.; Manmade objects are often well defined, following specific rules and structures based on perceived pragmatics. This is especially true in the case of mechanical two and a half dimensional (2.5D) machined parts. Because of the high accuracy needs of this domain, reverse engineering techniques using generic primitives are inappropriate. This dissertation asserts that an understanding of common design practices and manufacturing knowledge specific to 2.5D machining can and should be used to guide the reverse engineering process in order to achieve higher accuracy models.; To this end, reverse engineering is characterized as a constrained optimization problem. Logical laws of form are encoded as constraints in order to coerce new models to emulate the structure common to this genus of parts while best approximating the sensed data. A technique has been created to automatically hypothesize likely constraints that should hold on a hypothesized model. These constraints drive a DOF reduction process on the model and are further encoded as penalty functions during the model optimization. The entire process is formulated in a manner consistent with modern optimization techniques.