Rockwell Automation Technologies Incorporated

Gregory J Vance

Francisco Paul Maturana

Mikica Cvijetinovic

Braun C Brennecke

Robert F Trinnes

Described herein are systems and methods for improving defect detection in an automated production process. The system comprises a memory that stores executable components and a processor, operatively coupled to the memory, that executes the executable components. The executable components comprise an automation defect component and a machine learning component. The automation defect component retrieves parametric data associated with the production process. The automation defect component provides the parametric data to a machine learning algorithm. The machine learning component generates common attributes between the defective items. The machine learning component identifies a set of common attributes shared between the defective items and a non-defective item. The machine learning component modifies the set of the common attributes shared between the defective items and the non-defective item. The machine learning component generates defect indicators based on the common attributes. The automation defect component monitors subsequent parametric data to recognize the defect indicators.

Defect detection during an automated production process

A monitor provides run time monitoring of the input values to a signature analyzer having a machine-learning engine to compare statistics measuring the domain of those input values to the domain of input values used in the training set of the machine-learning engine. This comparison yields an indication of whether retraining of the machine-learning engine is required without analysis of the output of the machine-learning engine algorithm or the need for errors to become manifest.

System for detecting data drift in machine-learning process monitoring

Described herein are improvements for identifying defects during automated item production. In one example, a method includes identifying a first defect in a first item. The first defect is associated with a stage of production of the first produced item. The method further includes retrieving first parametric data associated with the stage for the first item and identifying one or more defect indicators based on the first parametric data and second parametric data associated with the stage for one or more second items having defects associated with the stage. The method also includes monitoring subsequent parametric data associated with the stage to recognize the one or more defect indicators in the subsequent parametric data.

A pick and place nozzle performance analytics system streams production data from pick and place machines used in electronic assembly to a cloud platform as torrential data streams, and performs analytics on the production data to track, visualize, and predict performance of individual nozzles in terms of rejects or miss-picks. The analytics system generates a performance vector for each nozzle based on the collected production data, the performance vector tracking both the accumulated rejects and the percentage of rejects as respective dimensions of an x-y plane. The system monitors and analyzes the trajectory of this vector in the x-y plane to predict when performance degradation of the nozzle will reach a critical threshold. In response to predicting that nozzle performance degradation will exceed a threshold at a future time, the system can generate and deliver notifications to appropriate client devices.

Nozzle performance analytics

Described herein are improvements for characterizing performance of a component in an automated industrial process. In one example, a method includes, for each period of a plurality of periods, at least until an alert is triggered, updating a count of failed operations attempted by the first component and a percentage of a count of total operations attempted by the first component represented by the count of failed operations. The method further includes decrementing a trigger value when the count of failed operations does not increase for the period and incrementing the trigger value when the count of failed operations increases for the period. Also, the method includes triggering the alert in response to the trigger value satisfying a trigger value criterion and either the count of failed operations satisfying a count criterion or the percentage satisfying a percentage criterion.

Growing community of inventors

Brunswick, OH, United States of America

Gregory J Vance