3m Innovative Properties Company

Carl J Skeps

Steven P Floeder

James A Masterman

Brandon T Berg

Steven R Wageman

Kenneth G Brittain

Jeffrey P Adolf

Wenyuan Xu

Jason P Smith

Matthew V Rundquist

Shane T Van Kampen

Terry A Okonek

Steven R Dreger

Wayne R Roller

Matthew P Peick

Xi Yu

Samuel S Schreiner

An example method for selecting product images for training a machine-learning model includes obtaining product images to include in an image population; receiving an indication of an image selection strategy for determining if a product image is to be included in a set of images of interest; determining image transforms based on configuration data for the indicated image selection strategy, wherein the image transforms perform image manipulation operations to obtain transformed image data for each of the product images in the image population; selecting a subset of images from the image population for inclusion in the set of images of interest based on the indicated image selection strategy and the transformed image data; determining one or more descriptive labels and applying the one or more descriptive labels to the respective sets of images; and training an inspection model for a product inspection system based on the labeled images.

Active learning management system for automated inspection systems

Techniques are described for maintaining synchronization of inspection data when a web roll is converted into intermediate smaller rolls prior to cutting the web into individual parts. A system comprises a database that stores anomaly data acquired from a manufactured web. The anomaly data specifies positions anomalies within a manufactured web relative to a set of fiducial marks on the manufactured web. A conversion processing line comprises a fiducial mark reader to output position information for the set of fiducial marks on the manufactured web, a slitter that cuts the manufactured web into slit rolls, and a fiducial mark printer to print a set of fiducial marks on each slit roll. A position monitoring system maintains spatial synchronization of the anomaly data by computing an updated position for the anomalies relative to the set of fiducial marks printed on the slit rolls.

Maintaining registration of spatially synchronized data for web products

Techniques are described for inspecting a web and controlling subsequent conversion of the web into one or more products. A system, for example, comprises an imaging device, an analysis computer and a conversion control system. The imaging device images the web to provide digital information. The analysis computer processes the digital information to identify regions on the web containing anomalies. The conversion control system subsequently analyzes the digital information to determine which anomalies represent actual defects for a plurality of different products. The web inspection system may preferentially apply different application-specific defect detection recipes depending on whether a given anomaly is a repeating or random anomaly.

Application-specific repeat defect detection in web manufacturing processes

A system for the characterization of webs that permits the identification of anomalous regions on the web to be performed at a first time and place, and permits the localization and marking of actual defects to be performed at a second time and place.

Apparatus and method for the automated marking of defects on webs of material

A conversion control system is described for spatially synchronizing data gathered from a plurality of operations performed on a web. The conversion control system applies a set of fiducial marks to a web, performs a plurality of operations on the web, generates a first and a second set of digital information for first and second operations, respectively, in accordance with respective first and second coordinate systems using the set of fiducial marks such that the each of the sets of digital information includes position data for respective first and second sets of regions on the web. The conversion control system may then register the position data of the first set of regions and the position data for the second set of regions to produce aggregate data and outputting a conversion control plan.

Multi-unit process spatial synchronization

A system for preferentially marking defects on a web is described. The system includes a web of material to be converted into individual sheets of a plurality of different grade levels, a database storing anomaly data of anomalies on the web, wherein an anomaly is a potential defect in at least one of the plurality of different grade levels, a marker that associates a unique mark with at least one of the grade levels, and a controller to retrieve the anomaly data from the database and to signal the marker as to where to make a mark, wherein the marker applies the mark associated with at least one of the grade levels for which the anomaly may cause a defect. The system may provide advantages, such as that a converter of various products from a single web roll may determine which regions of the web satisfy each grade level.

Preferential defect marking on a web

A manufacturing system includes rollers having synchronization marks to indicate complete rotations. Synchronization mark readers read the synchronization marks of the plurality of rollers and output roll synchronization signals. An encoder outputs a position signal indicative of a down-web distance of the web. An inspection system inspects the web and outputs anomaly data identifying positions of anomalies on the web. A synchronization unit receives the position signal from the encoder and the plurality of roll synchronization signals from the synchronization mark readers and converts the occurrence of each of the roll synchronization signals into down-web positions within a coordinate system associated with web process line. An analysis computer processes the anomaly data and the synchronization signals to identify repeated anomalies and to determine which of the rollers caused the repeated anomalies.

Multi-roller registered repeat defect detection of a web process line

A conversion control system is described that includes a database to store data defining a set of rules and an interface to receive local anomaly information from a plurality of different analysis machines associated with a plurality of manufacturing process lines that perform a plurality of operations on a web of material, and each of the manufacturing process lines includes position data for a set of regions on the web containing anomalies. The system also includes a computer that registers the position data of the local anomaly information for the plurality of manufacturing process lines to produce aggregate anomaly information. The system further includes a conversion control engine that applies the rules to the aggregate anomaly information to determine which anomalies represent actual defects in the web for a plurality of different products.

Multi-unit process spatial synchronization of image inspection systems

Techniques are described for inspecting a web and controlling subsequent conversion of the web into one or more products. A system, for example, comprises an imaging device, an analysis computer and a conversion control system. The imaging device images the web to provide digital information. The analysis computer processes the digital information to identify regions on the web containing anomalies. The conversion control system subsequently analyzes the digital information to determine which anomalies represent actual defects for a plurality of different products. The conversion control system determines a value for at least one product selection parameter for each of the products, and selects one of the products for conversion of the web based on the respective determined value. Exemplary product selection parameters include web utilization, unit product produced, estimated revenue or profit, process time, machine capacity and demand for the different products.

Maximization of yield for web-based articles

Techniques are described for inspecting a web and controlling subsequent conversion of the web into one or more products. A system, for example, includes an imaging device, an analysis computer and a conversion control system. The imaging device images a sequential portion of a web to provide digital information. The analysis computer processes the digital information with an initial algorithm to identify regions on the web containing anomalies. The conversion control system selects one of a plurality of converting sites to convert the web into a product, and analyzes at least a portion of the digital information with at least one subsequent algorithm to determine which anomalies represent actual defects in the web for the products.

Inventory control for web-based articles

A method of inspecting a moving web. The method includes imaging a sequential portion of the continuously moving web to provide digital information. The digital information is then processed with an initial algorithm to identify any regions on the web containing anomalies. Image information corresponding to any identified region within the digital information is then selected. The selected image information is then analyzed with at least one subsequent algorithm to distinguish actual defects from among the anomalies.

Apparatus and method for automated web inspection

An imaging device for sequentially imaging a portion of a continuously moving web to provide a digital data stream which is then analyzed by a single computer without the used of dedicated signal processing hardware. Techniques for operating on the data stream from an imaging device are disclosed, particularly including operations based on blob information stored in terms of starting position and segment run lengths in a crossweb direction. These allow definitions of blobs to be accumulated in a line-by-line fashion, and allow classes of defects commonly found in continuous web manufacturing to be identified with far less computing power than was previously required. In particular, in the challenging application of inspecting flexible circuits, data rates in excess of 10 mega-pixels/second are achieved and successfully processed.

Growing community of inventors

Lakeville, MN, United States of America

Carl J Skeps