Abstract:
A classification neural network for piecewise linearly separating an input space to classify input patterns is described. The multilayered neural network comprises an input node, a plurality of difference nodes in a first layer, a minimum node, a plurality of perceptron nodes in a second layer and an output node. In operation, the input node broadcasts the input pattern to all of the difference nodes. The difference nodes, along with the minimum node, identify in which vornoi cell of the piecewise linear separation the input pattern lies. The difference node defining the vornoi cell localizes input pattern to a local coordinate space and sends it to a corresponding perceptron, which produces a class designator for the input pattern.
Abstract:
A pattern of material is continuously deposited onto a substrate. The substrate and a mask are continuously brought together over a portion of a drum where a deposition source emits material. The mask includes apertures that form a pattern, and the material from the deposition source passes through the pattern of the mask and collects onto the substrate to form the pattern of material. The elongation and the transverse position of the substrate and the mask may be controlled. Pattern elements of the substrate and of the mask may be sensed in order to adjust the elongation and/or the transverse position of the substrate and/or mask to maintain a precise registration. Furthermore, the apertures may have a least dimension on the order of 100 microns or less to thereby create features on the substrate having least dimensions on the order of 100 microns or less.
Abstract:
A web-handling machine frame, at least partially built from reconfigurable interconnected blocks having precision alignment points. From such blocks, modular equipment web-handling lines for web-handling operations can be constructed ad hoc with their shaft mounted web-contacting devices in tram without the need for separate alignment procedures.
Abstract:
Methods and systems for determining longitudinal position of an elongated web are described. Sensors are used to detect one or more substantially continuous fiducial marks disposed longitudinally on the web. The sensors generate signals associated with the fiducial marks. A position detector receives the signals and determines a longitudinal position of the web using the sensor signals. The fiducial marks may be periodic fiducial marks, such as sine and/or cosine marks on the web and/or may be piecewise continuous marks. The coarse longitudinal position of the web can be determined based on periodically recurring features of the fiducial marks. The fine longitudinal position can be determined based on continuous portions of the fiducial marks between the periodically recurring features.
Abstract:
Methods and systems for forming layered electronic devices on a flexible, elongated substrate are described. The layered electronic devices include at least one electronically or optically active layer. Deposition of one more layers of the electronic devices occurs as the flexible substrate is moved through one or more deposition stations. At each deposition station the substrate is aligned with an aperture mask having apertures arranged in a pattern. The aperture mask and the substrate are brought into proximity over a portion of a circumference of a rotating drum. A layer of the layered electronic devices is formed by deposition of material through the apertures of the aperture mask. At each deposition station, registration between at least two layers of the layered electronic devices is maintained.
Abstract:
A microreplicated article is disclosed. The article includes a web including first and second opposed surfaces. The first surface includes a first microreplicated structure having a plurality of first features. The second surface includes a second microreplicated structure having a plurality of second features. Corresponding opposed features cooperate to form a lens features.
Abstract:
Web tension in web material passing through a web transport system is determined in real time using position sensors coupled to driven rollers that define a beginning and an end of a tension zone. The position sensors on the rollers provide information related to the amount of strained web material that has been added and subtracted from the web material present in the tension zone. The amount of web material added to, subtracted from and present in the tension zone in a sample time period is then converted to an unstrained amount of web material that when combined provides an estimate for the present amount of unstrained web material present in the tension zone. Because the length of the tension zone is both fixed and known, the tension in the web material is determined from the present amount of unstrained web material in the tension zone.