AUTOMATED QUALITY CONTROL

Author: SALAMANCA MIÑO SANTIAGO.
Year: 2004.
University: NACIONAL DE EDUCACIÓN A DISTANCIA [www.uned.es].
Place of defense: E.T.S. INGENIEROS INDUSTRIALES.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIEROS INDUSTRIALES..

Summary: The computer vision aims to extract and interpret information in various levels of sophistication using the image of the scene. One way of achieving this goal is through the creation of models that enable them effectively analyze the characteristics of the scene. This thesis proposes new techniques for creating spherical models in partial views and for free-form objects with the aim of being applied to problems of recognition and positioning. First, it has conducted a review of the various modeling techniques of 3D data through polygonal mesh. Then you define a porcedimiento to adapt existing models to spherical models for partial views, defining new topological properties on the field teselada. As a final step in the modeling technique, a method of partial melting of the models in order to obtain a complex model of the object. This phase will include a technique which enables an improvement in the regularization of the models totals. As implementation of the proposed models develops an algorithm for recognizing partial views based on the coupling of these on the complete model. This algorithm applied sequentially different invariant global ranging reducing potential candidates and the various docking areas around them. All the techniques proposed in this thesis, both modeling and application recognition are validated experimentally on the basis of data obtained with a range of sensor range.

Author: DIAZ FERNANDEZ ANA MARIA.
Year: 2006.
University: OVIEDO [www.uniovi.es].
Place of defense: DPTO. INGE. ELEC. ELECTRO. DE COMPU. Y SISTE..
Place of preparation: DPTO. INGENIERIA ELECTRICA, ELECTRONICA DE COMPUTADORES Y SISTEMAS.

Summary: Steel is an alloy iron carbon whose properties depend on two factors: the chemical composition and quality. Both factors are critical for practical applications for a particular type of steel is valid. Speaking of quality steel is meant mainly presence of defects such as cracks, inclusions, or flaws in a segregation center, which is the problem studied in this thesis. Segregation core is a phenomenon associated with the solidification of metals and alloys, consisting of a non-uniformity of the chemical composition. In the case of steel, alloy contains elements such as chromium, phosphorus, sulfur, vanadium .... some of whom are forming solutes which are more soluble in strong in the liquid, so that during the solidification of steel in the casting process, they tend to leave the newly formed solid and enrich the remaining liquid. In the case studied in this thesis, steel solidifies forming slabs of steel in that the effect of segregation, there is a backlog of these solutes in the centerline of roughing it can cause failure of the final product. The objective of this thesis is to develop a model based on algorithms of automatic learning capability to predict the severity of the segregation in those Central slabs of steel. This has been selected four modeling techniques: regression líneal, neural networks, topographic maps autoorganizados and fuzzy algorithm GAP which is a combination of genetic algorithms and genetic programming in designing a fuzzy system. The results suggest that the problem of segregation is modeable through the techniques employed, offering the best results in terms of capacity SOM prediction algorithms and fuzzy GAP in terms of ability to explain the phenomenon. In addition, there is dependency on the phenomenon of sulfur steel, being necessary to sort data entry on the basis of this parameter and develop different models for each group with different input parameters.