公开(公告)号：WO2021197796A1

公开(公告)日：2021-10-07

申请号：PCT/EP2021/056180

申请日：2021-03-11

Applicant: ABB SCHWEIZ AG

Inventor： SCHMIDT, Benedikt ,  AMIHAI, Ido ,  CHIOUA, Moncef ,  KOTRIWALA, Arzam ,  HOLLENDER, Martin ,  JANKA, Dennis ,  LENDERS, Felix ,  SCHLAKE, Jan Christoph ,  KLOEPPER, Benjamin ,  ABUKWAIK, Hadil

IPC: G05B23/02 ,  G06N3/08 ,  G05B13/02 ,  G05B17/02 ,  G05B13/027 ,  G05B23/024 ,  G06N20/00

Abstract: The invention relates to the field of monitoring or controlling an industrial process, particularly by means of an artificial neural net. The method comprises the steps of: training the first control model (10) by means of a first set of input data (31) as first input (11), resulting in a trained first control model (10T); copying the trained first control model (10T) to a second control model (20), wherein, after copying, the second input layer (22) and the plurality of second hidden layers (24) is identical to the plurality of first hidden layers (14), and the first output layer (18) is replaced by the second output layer (28); freezing the plurality of second hidden layers (24); training the second control model (20) by means of the first set of input data (31) as second input (21), resulting in a trained second control model (20T); and running the trained second control model (20T) by means of a second set of input data (32) as second input (21), wherein the second output (29) outputs the quality measure (qm) of the first control model (10).

公开(公告)号：WO2021197782A1

公开(公告)日：2021-10-07

申请号：PCT/EP2021/056093

申请日：2021-03-10

Applicant: ABB SCHWEIZ AG

Inventor： KLOEPPER, Benjamin ,  SCHMIDT, Benedikt ,  AMIHAI, Ido ,  CHIOUA, Moncef ,  SCHLAKE, Jan Christoph ,  KOTRIWALA, Arzam ,  HOLLENDER, Martin ,  JANKA, Dennis ,  LENDERS, Felix ,  ABUKWAIK, Hadil

IPC: G06N20/00 ,  G06N7/00 ,  G06F17/40 ,  G06F16/215 ,  G05B23/00 ,  G05B23/024 ,  G05B23/0283 ,  G06N7/005

Abstract: The invention relates to a computer-implemented method (100) for automating the development of industrial machine learning applications in particular for predictive maintenance, process monitoring, event prediction, or root-cause analysis. The method consists of one or more sub-methods that, depending on the industrial machine learning problem, may be executed iteratively. These sub-methods include at least one of a method to automate the data cleaning in training (S10) and later application (S15) of machine learning models, a method to label (S11) time series (in particular signal data) with help of other timestamp records, feature engineering (S12) with the help of process mining, and automated hyperparameter tuning (S14) for data segmentation and classification.

公开(公告)号：WO2021198356A1

公开(公告)日：2021-10-07

申请号：PCT/EP2021/058476

申请日：2021-03-31

Applicant: ABB SCHWEIZ AG

Inventor： SCHMIDT, Benedikt ,  AMIHAI, Ido ,  KOTRIWALA, Arzam ,  CHIOUA, Moncef ,  JANKA, Dennis ,  LENDERS, Felix ,  SCHLAKE, Jan Christoph ,  HOLLENDER, Martin ,  ABUKWAIK, Hadil ,  KLOEPPER, Benjamin

IPC: G06N3/08 ,  G06N20/00 ,  G05B17/02 ,  G05B19/418 ,  G05B13/02 ,  G05B13/0265 ,  G05B19/41885 ,  G05B2219/32352 ,  G06N3/0454 ,  G06N3/082 ,  G06N3/088 ,  G06N5/022

Abstract: The invention relates to a computer-implemented method of hierarchical machine learning for an industrial plant machine learning system, comprising the steps: receiving (S10), by a machine learning unit (10), a topology model (T), comprising structural information on hierarchical relations between components of the industrial plant (20); wherein the components comprise data signals (S) of sensors of the industrial plant (20) and hierarchical units (A, SU), wherein the hierarchical units (A, SU) comprise assets (A), plant sub-units (SU), plant units and plant sections of the industrial plant (20), determining (S20), by the machine learning unit (10), a representation hierarchy (H) comprising a plurality of levels (Lb, Li, Lt) using the received data signals (S) and the received topology model (T), wherein the representation hierarchy (H) comprises a signal representation (AE1,1) for each of the plurality of received data signals (S) and a hierarchical representation (AEA, AESU) for each of the hierarchical units (A, SU) on different levels; wherein each representation (AEA, AE1,1) on a higher level represents a group of representations (AEA, AE1,1) on a lower level; wherein each of the signal representation (AE1,1) and the hierarchical representation (AEA, AESU) comprise a machine learning model; training (S30), by the machine learning unit (10), an output machine learning model (11) of the machine learning unit (10) using the determined hierarchical representations (AEA, AESU).

公开(公告)号：WO2020030543A1

公开(公告)日：2020-02-13

申请号：PCT/EP2019/070878

申请日：2019-08-02

Applicant: ABB SCHWEIZ AG

Inventor： COHEN, Andrew ,  HOLLENDER, Martin ,  LI, Nuo ,  CHIOUA, Moncef ,  LUCKE, Matthieu

IPC: G05B23/02

Abstract: The present invention relates to an apparatus (10) for alarm information determination. It is described to provide (110) a processing unit with historical process control data from an input unit, wherein the process control data comprises a plurality of data signals, a plurality of alarm data and data relating to an event of interest. The processing unit determining (120) a plurality of correlation scores for the plurality of data signals paired with the plurality of alarm data, wherein a correlation score is determined for a data signal paired with an alarm data and wherein a high score indicates a higher degree of correlation than a low score. The processing unit identifies (130) at least one first alarm data from the plurality of alarm data, the identification comprising utilization of the data relating to the event of interest. At least one first data signal is identified (140) from the plurality of data signals, the identification comprising the processing unit utilizing the correlation scores for the identified at least first one alarm data paired with the plurality of data signals. An output unit outputs (150) the at least one first data signal.

公开(公告)号：WO2017191253A1

公开(公告)日：2017-11-09

申请号：PCT/EP2017/060651

申请日：2017-05-04

Applicant: ABB SCHWEIZ AG

Inventor： HOLLENDER, Martin ,  KLÖPPER, Benjamnin ,  CHIOUA, Moncef

IPC: G05B23/02

CPC classification number: G06N20/00 ,  G05B19/41885 ,  G05B23/0232 ,  G05B2219/34457

Abstract: Alarm handling method and system in plant process automation Alarm handling method and system in plant process automation including a data processing device comprising • · at least one interface (10), accessing or processing one or more process signals and determining corresponding process variables (24), • · an alarm configuration device (20), accessing or providing alarm configuration information comprising at least one setpoint for one or more determined process variables, • · a prediction device (30) determining and processing the current rate of change of at least one process variable to predict how long it will take or predict at which time a provided setpoint or threshold, in particular a predefined setpoint or threshold and in particular a consequence threshold (28), is reached or crossed, or determines whether and when at least one of the monitored or determined process variable values will cross the respective setpoint, in particular the alarm setpoint (26), for example when indicating a return-to-normal scenario.

Abstract translation: 工厂过程自动化中的报警处理方法和系统包括数据处理设备的工厂过程自动化中的报警处理方法和系统包括： 至少一个接口（10），访问或处理一个或多个过程信号并确定对应的过程变量（24），·中间过程 警报配置设备（20），访问或提供包括用于一个或多个所确定的过程变量的至少一个设定点的警报配置信息，其中， - 预测装置（30），其确定并处理至少一个过程变量的当前变化率，以预测将在何时花费或预测提供的设定值或阈值，特别是预定的设定值或阈值，特别是后果 阈值（28），或者确定是否以及何时监测或确定的过程变量值中的至少一个将跨越相应的设定值，特别是警报设定值（26），例如当指示返回到阈值 正常情况。

公开(公告)号：WO2021209432A1

公开(公告)日：2021-10-21

申请号：PCT/EP2021/059529

申请日：2021-04-13

Applicant: ABB SCHWEIZ AG

Inventor： CHIOUA, Moncef ,  DIX, Marcel ,  KLOEPPER, Benjamin ,  LYMPEROPOULOS, Ioannis ,  JANKA, Dennis ,  RODRIGUEZ, Pablo

IPC: G05B23/02 ,  G05B13/02

Abstract: The invention relates to the field of intelligent alarm management, particularly in industrial processes (50). The method comprises the steps of: • training a machine learning model (10) by means of input data (20) and score data (30), wherein the input data comprises a first time-series of at least one observable process variable and wherein the machine learning model (10) is an artificial neural net, ANN; • running the trained machine learning model (10) by applying the first time-series (21) to the trained machine learning model (10); and • outputting, by the trained machine learning model (10), an output value (40), comprising at least a second criticality value (42) of at least one predicted observable process-value, PPV, indicative of the abnormal behaviour of the industrial process (50) in a predefined temporal distance (T1).

公开(公告)号：WO2021198354A1

公开(公告)日：2021-10-07

申请号：PCT/EP2021/058474

申请日：2021-03-31

Applicant: ABB SCHWEIZ AG

Inventor： SCHMIDT, Benedikt ,  AMIHAI, Ido ,  KOTRIWALA, Arzam ,  CHIOUA, Moncef ,  JANKA, Dennis ,  LENDERS, Felix ,  SCHLAKE, Jan Christoph ,  HOLLENDER, Martin ,  ABUKWAIK, Hadil ,  KLOEPPER, Benjamin

IPC: G06N3/08 ,  G06N5/02 ,  G05B13/02 ,  G05B17/02 ,  G05B19/418 ,  G06N20/00 ,  G06N3/04 ,  G05B13/0265 ,  G05B19/41885 ,  G05B2219/32352 ,  G06N3/0454 ,  G06N3/082 ,  G06N3/088 ,  G06N5/022

Abstract: The invention relates to an industrial plant machine learning system, comprising a machine learning model, providing machine learning data, an industrial plant providing plant data and an abstraction layer, connecting the machine learning model and the industrial plant, wherein the abstraction layer is configured to provide standardized communication between the machine learning model and the industrial plant, using a machine learning markup language.

公开(公告)号：WO2021197783A1

公开(公告)日：2021-10-07

申请号：PCT/EP2021/056099

申请日：2021-03-10

Applicant: ABB SCHWEIZ AG

Inventor： SCHMIDT, Benedikt ,  AMIHAI, Ido ,  KOTRIWALA, Arzam ,  CHIOUA, Moncef ,  LENDERS, Felix ,  JANKA, Dennis ,  HOLLENDER, Martin ,  SCHLAKE, Jan Christoph ,  ABUKWAIK, Hadil ,  KLOEPPER, Benjamin

IPC: G06N20/00 ,  G06N3/08 ,  G06N3/04 ,  G05B23/02 ,  G05B19/418 ,  G05B23/024 ,  G06N3/0454 ,  G06N3/0472

Abstract: A computer-implemented method of generating a training data set for training an artificial intelligence module (10), AI module, is provided. The method comprises providing, on a data storage (102), a first data set (12) and a second data set (14), wherein the first data set 5 includes one or more first data elements (13) indicative of a first operational condition of an industrial system, wherein the second data set includes one or more second data elements (15) indicative of a second operational condition of the industrial system, wherein the first operational condition substantially matches the second operational condition. The method further comprises determining a data transformation for transforming the one or more first 10 data elements (13) of the first data set (12) into the one or more second data elements (15) of the second data set (14), applying the determined data transformation to the one or more first data elements (13) of the first data set and/or to one or more further data elements of one or more further data sets, thereby generating a transformed data set, and generating a training data set for training the AI module (10) based on at least a part of the transformed 15 data set.

公开(公告)号：WO2020064309A1

公开(公告)日：2020-04-02

申请号：PCT/EP2019/073957

申请日：2019-09-09

Applicant: ABB SCHWEIZ AG

Inventor： CHIOUA, Moncef ,  LUCKE, Matthieu ,  KOLB, Emanuel ,  HOLLENDER, Martin ,  LI, Nuo ,  COHEN, Andrew

IPC: G05B23/02

Abstract: System (100), methods and computer program products are provided for determining an abnormal technical status of a technical system (200). The computer system (100) receives, from the technical system (200), a plurality of signals wherein each signal (S1 to Sn) reflects the technical status of at least one system component. The system further retrieves, from an alarm management system (300), high alarm thresholds (H1 to Hn) and low alarm thresholds (L1 to Ln) associated with respective received signals (S1 to Sn). Signal values in a range between the associated high alarm threshold and the associated low alarm threshold reflect normal operation of the respective system component. For each signal (S1) a univariate distance to its associated alarm thresholds (H1/L1) is computed to quantify a degree of abnormality for the respective system component.Based on the univariate distances an aggregate abnormality indicator (AAI) is computed which reflects the technical status of the entire technical system (200). A comparison of the aggregate abnormality indicator (AAI) with a predetermined abnormality threshold (AAT) is provided to an operator (10).

公开(公告)号：WO2018172166A1

公开(公告)日：2018-09-27

申请号：PCT/EP2018/056445

申请日：2018-03-14

Applicant: ABB SCHWEIZ AG

Inventor： HOLLENDER, Martin ,  KLÖPPER, Benjamin ,  LUNDH, Michael ,  CHIOUA, Moncef

IPC: G05B23/02

Abstract: Computer system (100), computer-implemented method and computer program product are provided for monitoring the technical status of an industrial process system (300) being under control of an advanced process controller (APC). The industrial process system (300) has an operation (330) for processing flow materials. The advanced process controller (APC) is responsive to one or more sensor signals (320). The computer system (100) includes an interface module (110) configured to receive technical status data (321) describing the current technical state of the industrial process system (300) with regards to a respective processing component or the processed material wherein the technical status data (321) corresponds to or is derived from the one or more sensor signals (320). Further, the interface outputs an anomaly alert (AA) in case of an anomaly detection for the industrial process system to enable deactivating of the advanced process controller (APC). The computer system further includes an anomaly detection module (120) to apply one or more Machine Learning Models (MLMn) to the received technical status data (321) to analyze the technical status data for detecting one or more indicators of an abnormal technical status prevailing in the industrial process system. The one or more Machine Learning Models (MLMn) are trained on historic raw or pre-processed sensor data. The anomaly detection module generates the anomaly alert (AA) based on the one or more indicators.