System and method for the detection of faults in a multi-variable system utilizing both a model for normal operation and a model for faulty operation

Patent metadata

Field	Value
Publication number	US-9366451-B2
Application number	US-201113336153-A
Country	US
Kind code	B2
Filing date	Dec 23, 2011
Priority date	Dec 24, 2010
Publication date	Jun 14, 2016
Grant date	Jun 14, 2016

Abstract

A method for detecting faulty operation of a multi-variable system is described. The method includes receiving operational data from a plurality of components of the multi-variable system and processing the operational data in accordance with a plurality of dynamic machine learning fault detection models to generate a plurality of fault detection results. Each fault detection model uses a plurality of variables to model one or more components of the multi-variable system and is adapted to detect normal or faulty operation of an associated component or set of components of the multi-variable system. The plurality of fault detection results are output.

First claim

The claims defining the invention are as follows:

1. A method for detecting faulty operation of a HVAC system, the method including:
receiving operational data from a plurality of components of the HVAC system and feeding the operational data to a plurality of dynamic machine learning fault detection models;
automatically mapping the HVAC system by processing the operational data using the plurality of dynamic machine learning fault detection models;
training one or more of the fault detection models to learn patterns of normal operation of the HVAC system wherein faults are detected as deviations in the normal operation and training one or more of the fault detection models to learn patterns of faulty operation of the HVAC system wherein normal operation is detected as deviations in the faulty operation;
generating a plurality of fault detection results noting the presence or absence of a fault through the detection of deviations in the patterns of normal operation or faulty operation of the HVAC system;
each fault detection model using a plurality of variables to model one or more components of the HVAC system and being adapted to detect normal or faulty operation of an associated component or set of components of the HVAC system; and
outputting the plurality of fault detection results noting the presence or absence of a fault through the detection of deviations in the patterns of normal operation and faulty operation of the HVAC system.

2. The method according to claim 1 , wherein at least two fault detection models are adapted to detect normal or faulty operation of different associated components or sets of components.

3. The method according to claim 1 , wherein the plurality of fault detection models includes:
at least one general fault detection model adapted to detect normal or faulty operation of a general set of components of the HVAC system; and
at least one specific fault detection model adapted to detect normal or faulty operation of a specific component or set of components of the HVAC system, wherein
each specific fault detection model is associated with a general fault detection model, and
the component or set of components associated with the specific fault detection model is a subset of or related to the set of components of the associated general fault detection model.

4. The method according to claim 3 , wherein one of the at least one general fault detection models is adapted to detect normal or faulty operation of the HVAC system as a whole.

5. The method according to claim 3 , wherein the operational data is only processed in accordance with the at least one specific fault detection model if processing the operational data in accordance with the associated general fault detection model indicates the existence of a fault.

6. The method according to claim 3 , wherein if both a general fault detection process and a specific fault detection process associated with that general fault detection process indicate faulty operation of the HVAC system, the fault is diagnosed as a fault in the component or set of components associated with the specific fault detection process.

7. The method according to claim 1 , wherein at least one of the plurality of fault detection models is selected from a group including dynamic Bayesian network based fault detection models and hidden Markov model based fault detection models.

8. The method according to claim 1 , further including:
processing the plurality of fault detection results in accordance with a data fusion process to generate a fused fault detection result, the fused fault detection result including information as to whether a fault is perceived in the operation of the HVAC system; and
outputting said fused fault detection result.

9. The method according to claim 8 , wherein the data fusion process is a Dempster-Shafer based data fusion process.

10. The method according to claim 1 , further including:
for each of the plurality of fault detection results calculating an associated confidence level; and
using the calculated confidence levels to select which of the plurality of fault detection results is or are likely to be correct.

11. The method according to claim 10 , further including fusing said plurality of confidence levels to provide a fused confidence level result; and
using said fused confidence level result along with said fused fault detection result to determine the existence of and diagnose a fault in the HVAC system.

12. The method according to claim 10 , wherein said confidence levels are belief masses calculated according to a Dempster-Shafer based process.

13. The method according to claim 10 , further including:
selecting one or more of the fault detection results based on the confidence level associated with the fault detection result;
generating a set of variables, the set of variables including the variables modelled by the fault detection models which generated the selected fault detection results; and
outputting said set of variables for use in diagnose a fault in the HVAC system.

14. The method according to claim 1 , further including:
processing the operational data in accordance with at least one of the fault detection models multiple times to generate multiple results from said at least one fault detection model, and
clustering the multiple results in accordance with a clustering algorithm in order to increase the reliability of the fault detection result.

15. The method according to claim 14 , wherein the clustering algorithm is a K-means clustering algorithm.

16. The method according to claim 1 , wherein if two fault detection results are contradictory, the method further includes:
flagging one or both of the fault detection models yielding the contradictory results; and
modifying or retraining each flagged fault detection model.

17. The method according to claim 1 , wherein if a fault detection result is identified as being a false alarm, the method further includes:
flagging the fault detection model yielding the result identified to be a false alarm; and
modifying or retraining the flagged fault detection models.

18. The method according to claim 14 , wherein if two fault detection results are contradictory, the method further includes:
flagging one or both of the fault detection models yielding the contradictory results;
modifying or retraining each flagged fault detection model; and
using said flagged fault detection models to modify the clustering algorithm in order to improve the clustering of the multiple search results.

19. A method for detecting faulty operation of a multi-variable system, the method including:
receiving operational data from a plurality of components of the multi-variable system and feeding the operational data to a plurality of dynamic machine learning fault detection models;
automatically mapping the multi-variable system by processing the operational data using the plurality of dynamic machine learning fault detection models;
training one or more of the fault detection models to learn patterns of normal operation of the multi-variable system wherein faults are detected as deviations in the normal operation and training one or more of the fault detection models to learn patterns of faulty operation of the multi-variable system wherein normal operation is detected as deviations in the faulty operation;
generating a plurality of fault detection results noting the presence or absence of a fault through the detection of deviations in the patterns of normal operation or faulty operatio

Assignees

• Guo Ying
• Li Jiaming
• West Sam
• Wall Joshua
• Platt Glenn
• Commw Scient Ind Res Org

Inventors

• Guo Ying
• Li Jiaming
• West Sam
• Wall Joshua
• Platt Glenn

Classifications

• F24F11/38

  Failure diagnosis · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• F24F2140/00

  Control inputs relating to system states · CPC title

• F24F11/63

  Electronic processing · CPC title