公开(公告)号：AU2021286514B2

公开(公告)日：2023-04-27

申请号：AU2021286514

申请日：2021-05-14

Applicant: IBM

Inventor： ZHOU NIANJUN ,  SUBRAMANIAN DHARMASHANKAR ,  GIFFORD WESLEY M

IPC: G06N3/04

Abstract: A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data (504, 514, 702, 802) having forward dependence to a target variable (108, 110, 112, 604, 708, 804), a second time series of downstream sensor data (508, 518, 704, 806) having backward dependence to the target variable (108, 110, 112, 604, 708, 804) and a time series of measured target variable (108, 110, 112, 604, 708, 804) data associated with the target variable (108, 110, 112, 604, 708, 804). The target variable (108, 110, 112, 604, 708, 804) has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data (504, 514, 70, 802) and the downstream sensor data (508, 518, 704, 806). The hybrid sensor can estimate a value of the target variable (108, 110, 112, 604, 708, 804) at a given time, for example, during which no actual measured target variable (108, 110, 112, 604, 708, 804) value is available.

公开(公告)号：AU2021286514A1

公开(公告)日：2022-10-27

申请号：AU2021286514

申请日：2021-05-14

Applicant: IBM

Inventor： ZHOU NIANJUN ,  SUBRAMANIAN DHARMASHANKAR ,  GIFFORD WESLEY M

IPC: G06N3/04

Abstract: A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data (504, 514, 702, 802) having forward dependence to a target variable (108, 110, 112, 604, 708, 804), a second time series of downstream sensor data (508, 518, 704, 806) having backward dependence to the target variable (108, 110, 112, 604, 708, 804) and a time series of measured target variable (108, 110, 112, 604, 708, 804) data associated with the target variable (108, 110, 112, 604, 708, 804). The target variable (108, 110, 112, 604, 708, 804) has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data (504, 514, 70, 802) and the downstream sensor data (508, 518, 704, 806). The hybrid sensor can estimate a value of the target variable (108, 110, 112, 604, 708, 804) at a given time, for example, during which no actual measured target variable (108, 110, 112, 604, 708, 804) value is available.

公开(公告)号：DE112021001953T5

公开(公告)日：2023-03-16

申请号：DE112021001953

申请日：2021-05-14

Applicant: IBM

Inventor： ZHOU NIANJUN ,  SUBRAMANIAN DHARMASHANKAR ,  GIFFORD WESLEY M

IPC: G06Q50/04 ,  G01D1/00 ,  G06F17/18 ,  G06N3/02 ,  G06N20/00

Abstract: Ein Hybridsensor kann durch Schulen eines Maschinenlernmodells, wie beispielsweise ein neuronales Netzwerk, auf Grundlage eines Schulungsdatensatzes generiert werden. Der Schulungsdatensatz kann eine erste Zeitreihe von vorgelagerten Sensordaten (504, 514, 702, 802) mit Vorwärtsabhängigkeit zu einer Zielvariablen (108, 110, 112, 604, 708, 804), eine zweite Zeitreihe von nachgelagerten Sensordaten (508, 518, 704, 806) mit Rückwärtsabhängigkeit zu der Zielvariablen (108, 110, 112, 604, 708, 804), und eine Zeitreihe von gemessenen Zielvariablendaten (108, 110, 112, 604, 708, 804) umfassen, die der Zielvariablen zugehörig sind. Die Zielvariable (108, 110, 112, 604, 708, 804) hat eine Messfrequenz, die niedriger als die Messfrequenzen ist, die den vorgelagerten Sensordaten (504, 514, 702, 802)und den nachgelagerten Sensordaten (508, 518, 704, 806) zugehörig sind. Der Hybridsensor kann zum Beispiel einen Wert der Zielvariablen (108, 110, 112, 604, 708, 804) zu einem bestimmten Zeitpunkt schätzen, während dem kein aktueller Messwert der Zielvariablen (108, 110, 112, 604, 708, 804) verfügbar ist.

公开(公告)号：CA3178043A1

公开(公告)日：2021-12-16

申请号：CA3178043

申请日：2021-05-14

Applicant: IBM

Inventor： ZHOU NIANJUN ,  SUBRAMANIAN DHARMASHANKAR ,  GIFFORD WESLEY M

IPC: G01D21/00

Abstract: A hybrid sensor can be generated by training a machine learning model, such as a neural network, based on a training data set. The training data set can include a first time series of upstream sensor data (504, 514, 702, 802) having forward dependence to a target variable (108, 110, 112, 604, 708, 804), a second time series of downstream sensor data (508, 518, 704, 806) having backward dependence to the target variable (108, 110, 112, 604, 708, 804) and a time series of measured target variable (108, 110, 112, 604, 708, 804) data associated with the target variable (108, 110, 112, 604, 708, 804). The target variable (108, 110, 112, 604, 708, 804) has measuring frequency which is lower than the measuring frequencies associated with the upstream sensor data (504, 514, 70, 802) and the downstream sensor data (508, 518, 704, 806). The hybrid sensor can estimate a value of the target variable (108, 110, 112, 604, 708, 804) at a given time, for example, during which no actual measured target variable (108, 110, 112, 604, 708, 804) value is available.