Systems and methods for detection and prediction of cobble events
AssigneeEverguard, Inc.
InventorsYong Wu, Changsoo Jeong
Life of this patent
Estimating the term…
Abstract
Systems and methods for monitoring an industrial process are disclosed. The method may include receiving image information conveying a time-ordered series of images of the industrial process; generating segmentation maps for the images; identifying individual sets of the images that depict cobble event candidates; determining whether the cobble event candidates are actual cobble events; and generating control signals to remediate actual cobble events.
Claims20
A system configured to detect and predict occurrences of cobble events in an industrial environment in which hot metal is being processed, the system comprising:
- an image sensor configured to capture image information conveying a time-ordered series of images of the industrial environment that depict at least one area of the industrial environment through which hot metal is being moved;
- and
- one or more hardware processors configured by machine-readable instructions to:
- receive the image information captured by the image sensor;
- generate segmentation maps for individual ones of the images with a first machine-learning model, wherein the individual segmentation maps indicate the portion or portions of the individual images that depict hot metal moving through the industrial environment;
- identify individual sets of one or more images that depict cobble event candidates such that a first set of one or more images is identified as a cobble event candidate, wherein the individual sets are identified by analyzing the portion or portions of the individual images indicated by the segmentation maps as depicting hot metal moving through the industrial environment;
- responsive to identification of the first set of one or more images as a cobble event candidate, provide a first series of one or more images to a second machine-learning model configured to determine whether the first series of one or more images depicts an actual cobble event or a false positive, wherein the first series of one or more images includes the first set of one or more images;
- and
- responsive to determining the first series of one or more images depicts an actual cobble event, generate control signals for one or more pieces of equipment present in the industrial environment that cause the one or more pieces of equipment to modify operation to remediate the actual cobble event and/or provide notification of the actual cobble event.
A method for detecting and predicting occurrences of cobble events in an industrial environment in which hot metal is being processed, the method comprising:
- receiving image information captured by an image sensor, the image sensor being configured to capture image information conveying a time-ordered series of images of the industrial environment that depict at least one area of the industrial environment through which hot metal is being moved;
- generating segmentation maps for individual ones of the images with a first machine-learning model, wherein the individual segmentation maps indicate the portion or portions of the individual images that depict hot metal moving through the industrial environment;
- identifying individual sets of one or more images that depict cobble event candidates such that a first set of one or more images is identified as a cobble event candidate, wherein the individual sets are identified by analyzing the portion or portions of the individual images indicated by the segmentation maps as depicting hot metal moving through the industrial environment;
- responsive to identification of the first set of one or more images as a cobble event candidate, providing a first series of one or more images to a second machine-learning model configured to determine whether the first series of one or more images depicts an actual cobble event or a false positive, wherein the first series of one or more images includes the first set of one or more images;
- and
- responsive to determining the first series of one or more images depicts an actual cobble event, generating control signals for one or more pieces of equipment present in the industrial environment that cause the one or more pieces of equipment to modify operation to remediate the actual cobble event and/or provide notification of the actual cobble event.
Dependencies are inferred from the claim text.