Brown University, Joint Materials/Solid Mechanics Seminar Series
General Motors R&D Center, Materials and Processes Laboratory | |
Abstract: Hemming and trimming of age-hardenable aluminum (e.g. AA6111) sheet metal components presents significant manufacturing difficulties due to hem cracking, flange splitting, and sliver generation. These problems often result in lower throughput, higher scrap rates, and increased metal finishing for aluminum assemblies. This presentation describes patented processes developed by General Motors to prevent splitting during hemming and eliminate slivers during trimming. Both processes involve selective application of retrogression heat treatments to aluminum components using induction heating. The kinetics of retrogression heat treatments will be discussed, as well their effect on formability and fracture. Retrogression heat treatments are currently being applied to flanged outer panels of the General Motors Suburban liftgate to prevent edge cracking and flange splitting during rope hemming of AA6111.
Brown Applied Mathematics Pattern Theory and Vision Seminar
e-mail: norbert@cn.stir.ac.uk | |
Abstract: Vision faces the problem of an extremely high degree of vagueness and uncertainty in its low level processes such as edge detection, optic flow analysis and stereo estimation. However, the human visual systems acquires visual representations which allows actions with high precision and certainty within the 3D world under rather uncontrolled conditions. The human visual system can achieve the needed certainty and completeness by integrating visual information across modalities and across frames. This integration is manifested in the huge connectivity between brain areas in which the different visual modalities are processed as well as in the large number of feedback connections between higher and lower cortical areas.
The essential need for integrating visual information across modalities in addition to optimising single modalities has been recognised in the vision community after a long period of work on improving single modalities. The power of modality fusion arises from the huge intrinsic regularities in visual scenes. Two important regularities in visual data with distinct properties are (1) motion (most importantly rigid body motion, RBM) and (2) statistical interdependencies between features such as collinearity and symmetry. In contrast to RBM, the statistical interdependencies between features are much harder to describe analytically. Accordingly, developmental psychology shows strong evidence that visual experience plays an important role to achieve the ability to use these interdependencies in visual processing (e.g., the effect of illusionary contours appears after 5 month).
Collinearity and parallelism do not describe a deterministic relation between features but probabilistic relation, (e.g., the occurrence of a line segment in visual data has a distinct impact on the likelihood of the occurrence of a line segments at a different position with different orientation (see, e.g. Kruger (1998). Collinearity and Parallelism are Statistically Significant Second Order Relations of Complex Cell Responses. Neural Processing Letters 8(2)). In my talk I address the statistics of natural scenes regarding additional modalities such as color or optic flow. As a main result it turns out that statistical interdependencies in visual scenes become significantly stronger when multiple modalities are taken into account. In my talk I will also sketch how these statistical regularities can be combined with the regularity 'Rigid Body Motion' for a stable tracking of objects which, recursively, can be used to stabilize the extraction of object representations from stereo sequences.
<--- 2002 Index