System and method for detecting change over in manufacturing field
AssigneeHitachi, Ltd.
InventorNobuyuki Muranaka
Life of this patent
Estimating the term…
Abstract
Example implementations described herein are directed to a change over detection system and method. Such example implementations facilitate updating the information of the production line in an Internet of Things (IoT) system to the latest version. Through the example implementations described herein, device rearrangement in manufacturing field can thereby be detected.
Claims11
A method for detecting a changeover of an asset in a manufacturing field, executed by a process, comprising:
- for a detection of a stoppage and a restart of the programmable logic controller (PLC) within a threshold period of time:
- retrieving first metrics from the PLC from before the stoppage and second metrics from the PLC from after the restart;
- extracting first features from the first metrics and second features from the second metrics;
- and
- for a difference between the first features and the second features exceeding a threshold, transmitting a notification to an asset management server indicative of the changeover of the asset of the PLC;
- retrieving, from operating signals of the PLC, series data corresponding to the PLC;
- for a score of a changeover detection model executed on the series data of the PLC exceeding a threshold, transmitting the notification to an asset management server indicative of the changeover of the asset of the PLC;
- and
- applying reinforcement learning on the changeover detection model based on a portion of the series data corresponding to the PLC before the notification indicative of the changeover of the asset, with indications regarding whether the changeover has occurred or not.
A non-transitory computer readable medium, storing instructions for executing a process for detecting a changeover of an asset in a manufacturing field, the instructions comprising:
- for a detection of a stoppage and a restart of the programmable logic controller (PLC) within a threshold period of time:
- retrieving first metrics from the PLC from before the stoppage and second metrics from the PLC from after the restart;
- extracting first features from the first metrics and second features from the second metrics;
- and
- for a difference between the first features and the second features exceeding a threshold, transmitting a notification to an asset management server indicative of the changeover of the asset of the PLC;
- retrieving, from operation signals of the PLC, series data corresponding to the PLC;
- for a score of a changeover detection model executed on the series data of the PLC exceeding a threshold, transmitting the notification to an asset management server indicative of the changeover of the asset of the PLC;
- and
- applying reinforcement learning on the changeover detection model based on a portion of the series data corresponding to the PLC before the notification indicative of the changeover of the asset, with indications regarding whether the changeover has occurred or not.
An apparatus for detecting a changeover of an asset in a manufacturing field, comprising:
- a processor, configured to, for a detection of a stoppage and a restart of the programmable logic controller (PLC) within a threshold period of time:
- retrieve first metrics from the PLC from before the stoppage and second metrics from the PLC from after the restart;
- extract first features from the first metrics and second features from the second metrics;
- and
- for a difference between the first features and the second features exceeding a threshold, transmit a notification to an asset management server indicative of the changeover of the asset of the PLC;
- retrieve from operation signals of the PLC, series data corresponding to the PLC,
- for a score of a changeover detection model executed on the series data of the PLC exceeding a threshold, transmit the notification to an asset management server indicative of the changeover of the asset of the PLC;
- and
- apply reinforcement learning on the changeover detection model based on a portion of the series data corresponding to the PLC before the notification indicative of the changeover of the asset, with indications regarding whether the changeover has occurred or not.
Dependencies are inferred from the claim text.