INDUSTRIAL ASSET TEMPORAL ANOMALY DETECTION WITH FAULT VARIABLE RANKING

• US 20200133253A1
• Filed: 10/25/2018
• Published: 04/30/2020
• Est. Priority Date: 10/25/2018
• Status: Active Grant

• Current Assignee: GE Digital Holdings LLC (GE Aerospace)
• Original Assignee: General Electric Company
• Inventors: HUANG, Hao
• Granted Patent: US 11,320,813 B2
• Time in Patent Office: Days
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G05B 23/0272 Presentation of monitored results, e.g. selection of status reports to be displayed; Filtering information to the user

Abstract

A method of temporal anomaly detection includes accessing sensor data readings obtained at a monitored industrial asset, performing a data cleanup operation on at least a portion of the accessed sensor data readings, transforming at least the cleaned-up portion of the accessed sensor data readings to time series feature space sensor data, applying a multi-kernel-based projection algorithm to the time series feature space sensor data, computing a respective anomaly score and a respective ranking for one or more variables of the sensor data readings, and providing at least the computed respective anomaly score or the respective ranking for at least one of the one or more variables to a user. Ranking the anomaly score includes comparing each anomaly score to a threshold and then assigning a ranking to scores with a magnitude greater than the threshold based on its magnitude. A system and a non-transitory computer-readable medium are also disclosed.

Claims

1. A method of temporal anomaly detection, the method comprising:

accessing sensor data readings obtained at a monitored industrial asset;

performing a data cleanup operation on at least a portion of the accessed sensor data readings;

transforming at least the cleaned-up portion of the accessed sensor data readings to time series feature space sensor data;

applying a multi-kernel-based projection algorithm to the time series feature space sensor data;

computing a respective anomaly score and a respective ranking for one or more variables of the sensor data readings; and

providing at least the computed respective anomaly score or the respective ranking for at least one of the one or more variables to a user.

2. The method of claim 1, including the data cleanup operation including filling missing values and removal of outlier data points.

3. The method of claim 1, including transforming the cleaned-up portion by performing temporal feature engineering techniques.

4. The method of claim 1, the multi-kernel-based projection algorithm including:

accessing a training data set and a testing data set;

selecting one or more kernel construction options;

constructing a similarity matrix for the training data set based on each selected kernel option;

calculating a projection matrix for each similarity matrix element;

calculating projected embeddings for each projection matrix element by applying the testing data set;

computing an anomaly score matrix for each projection matrix element; and

computing the respective anomaly score from corresponding elements of the projection score.

5. The method of claim 1, computing the respective ranking including:

comparing the respective anomaly score to a predefined threshold;

for each respective anomaly score having a magnitude greater than the predefined threshold assigning a ranking to each respective anomaly score based on its respective magnitude.

6. A non-transitory computer-readable medium having stored thereon instructions which when executed by a control processor cause the control processor to perform a method of temporal anomaly detection, the method comprising:

accessing sensor data readings obtained at a monitored industrial asset;

performing a data cleanup operation on at least a portion of the accessed sensor data readings;

transforming at least the cleaned-up portion of the accessed sensor data readings to time series feature space sensor data;

applying a multi-kernel-based projection algorithm to the time series feature space sensor data;

computing a respective anomaly score and a respective ranking for one or more variables of the sensor data readings; and

providing at least the computed respective anomaly score or the respective ranking for at least one of the one or more variables to a user.

7. The non-transitory computer-readable medium of claim 6, the instructions further configured to cause the control processor to perform the data cleanup operation by filling missing values and removal of outlier data points.

8. The non-transitory computer-readable medium of claim 6, the instructions further configured to cause the control processor to perform transforming the cleaned-up portion by performing temporal feature engineering techniques.

9. The non-transitory computer-readable medium of claim 6, the instructions further configured to cause the control processor to perform the multi-kernel-based projection algorithm by including:

accessing a training data set and a testing data set;

selecting one or more kernel construction options;

constructing a similarity matrix for the training data set based on each selected kernel option;

calculating a projection matrix for each similarity matrix element;

calculating projected embeddings for each projection matrix element by applying the testing data set;

computing an anomaly score matrix for each projection matrix element; and

computing the respective anomaly score from corresponding elements of the projection score.

10. The non-transitory computer-readable medium of claim 6, the instructions further configured to cause the control processor to perform computing the respective ranking including:

comparing the respective anomaly score to a predefined threshold;

for each respective anomaly score having a magnitude greater than the predefined threshold assigning a ranking to each respective anomaly score based on its respective magnitude.

11. A system for temporal anomaly detection, the system comprising:

a control processor in communication with a data store across an electronic communication network, the control processor including a processor unit;

the data store including executable instructions and sensor data records representing monitored conditions of one or more components of an industrial asset;

the executable instructions when executed by the processor unit cause the processor unit to perform a method, the method comprising;

accessing sensor data readings obtained at a monitored industrial asset;

performing a data cleanup operation on at least a portion of the accessed sensor data readings;

transforming at least the cleaned-up portion of the accessed sensor data readings to time series feature space sensor data;

applying a multi-kernel-based projection algorithm to the time series feature space sensor data;

computing a respective anomaly score and a respective ranking for one or more variables of the sensor data readings; and

providing at least the computed respective anomaly score or the respective ranking for at least one of the one or more variables to a user.

12. The system of claim 11, the executable instructions further configured to cause the processor unit to perform the data cleanup operation by filling missing values and removal of outlier data points.

13. The system of claim 11, the executable instructions further configured to cause the processor unit to perform transforming the cleaned-up portion by performing temporal feature engineering techniques.

14. The system of claim 11, the executable instructions further configured to cause the processor unit to perform the multi-kernel-based projection algorithm by including:

accessing a training data set and a testing data set;

selecting one or more kernel construction options;

constructing a similarity matrix for the training data set based on each selected kernel option;

calculating a projection matrix for each similarity matrix element;

calculating projected embeddings for each projection matrix element by applying the testing data set;

computing an anomaly score matrix for each projection matrix element; and

computing the respective anomaly score from corresponding elements of the projection score.

15. The system of claim 11, the executable instructions further configured to cause the processor unit to perform computing the respective ranking including:

comparing the respective anomaly score to a predefined threshold;

for each respective anomaly score having a magnitude greater than the predefined threshold assigning a ranking to each respective anomaly score based on its respective magnitude.