Method and system for automatically developing a fault classification system by segregation of kernels in time series data
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06E-001/00
G06F-015/18
출원번호
UP-0755893
(2007-05-31)
등록번호
US-7814034
(2010-11-01)
발명자
/ 주소
Eklund, Neil Holger White
Yan, Weizhong
Anil, Varma
Bonissone, Piero Patrone
출원인 / 주소
Lockheed Martin Corporation
대리인 / 주소
Bracewell & Giuliani LLP
인용정보
피인용 횟수 :
0인용 특허 :
2
초록▼
A method and system for automatically developing a fault classification system from time series data. The sensors need not have been intended for diagnostic purposes (e.g., control sensors). These methods and systems are functionally independent of knowledge related to a particular equipment system,
A method and system for automatically developing a fault classification system from time series data. The sensors need not have been intended for diagnostic purposes (e.g., control sensors). These methods and systems are functionally independent of knowledge related to a particular equipment system, thereby allowing seamless application to multiple systems, regardless of the suite of sensors in each system. Because this algorithm is totally automated, substantial savings in time and development cost can be achieved. The algorithm results in a classification system and a set of features that might be used to develop alternative classification systems without human intervention.
대표청구항▼
What is claimed is: 1. A method for automated segregation of kernels, comprising: collecting sensor data in a time series format; labeling sensor data in a time series format as either normal, or one of one or more possible faults; segmenting sensor data in a time series format into blocks having a
What is claimed is: 1. A method for automated segregation of kernels, comprising: collecting sensor data in a time series format; labeling sensor data in a time series format as either normal, or one of one or more possible faults; segmenting sensor data in a time series format into blocks having a substantially uniform slope with respect to time; labeling the blocks as having a rising, falling, or flat slope; joining adjacent blocks with slopes having a same sign; identifying candidate kernels; convoluting the sensor data in a time series format with candidate kernels; applying a feature selection method to determine which kernels have discriminatory power; and training a fault classification system based on kernels having discriminatory power. 2. The method of claim 1, wherein collected sensor data comprises pressure, temperature, and speed information. 3. The method of claim 1, wherein labeling the blocks comprises applying statistical criteria, manually set thresholds, or fuzzy logic. 4. The method of claim 1, wherein Kernels comprise regions containing at least one region of non-zero slopes between regions of a substantially zero slope. 5. The method of claim 1, wherein the feature selection method comprises variable importance statistic from the Random Forests (RF). 6. The method of claim 1, wherein training the fault classification system comprises applying Random Forests, neural networks, or support vector machines. 7. A method operable to develop a fault classification system based on the extracted features, comprising: collecting sensor data in a time series format; labeling sensor data in a time series format as either normal, or one of one or more possible faults; segmenting sensor data in a time series format into blocks having a substantially uniform slope with respect to time; labeling the blocks as having a rising, failing, or flat slope; joining adjacent blocks with slopes having a same sign; identifying candidate kernels; convoluting the sensor data in a time series format with candidate kernels; applying a feature selection method to determine which extracted features have discriminatory power; and training a fault classification system based on extracted features having discriminatory power. 8. The method of claim 7, wherein collected sensor data comprises pressure, temperature, and speed information. 9. The method of claim 7, wherein labeling the blocks comprises applying statistical criteria, manually set thresholds, or fuzzy logic. 10. The method of claim 1, wherein Kernels comprise regions containing at least one region of non-zero slopes between regions of a substantially zero slope. 11. The method of claim 7, wherein the feature selection method comprises variable importance statistic from the Random Forests (RF). 12. The method of claim 7, wherein training the fault classification system comprises applying Random Forests, neural networks, or support vector machines. 13. A method operable to generating classifier systems for univariate or multivariate equipment sensor data, comprising: collecting univariate or multivariate equipment sensor data in a time series format; labeling univariate or multivariate equipment sensor data in a time series format as either normal, or one of one or more possible faults; segmenting univariate or multivariate equipment sensor data in a time series format into blocks having a substantially uniform slope with respect to time; labeling the blocks as having a rising, falling, or flat slope; joining adjacent blocks with slopes having a same sign; identifying candidate kernels; convoluting the univariate or multivariate equipment sensor data in a time series format with candidate kernels; applying a feature selection method to determine which extracted features have discriminatory power; and training a fault classification system based on extracted features having discriminatory power. 14. The method of claim 13, wherein collected univariate or multivariate equipment sensor data comprises pressure, temperature, and speed information. 15. The method of claim 13, wherein labeling the blocks comprises applying statistical criteria, manually set thresholds, or fuzzy logic. 16. The method of claim 13, wherein extracted features comprise regions containing at least one region of non-zero slopes between regions of a substantially zero slope. 17. The method of claim 13, wherein the feature selection method comprises variable importance statistic from the Random Forests (RF). 18. The method of claim 13, wherein training the fault classification system comprises applying Random Forests, neural networks, or support vector machines. 19. A system operable to develop a fault classification system based on the extracted features, comprising: a data collection system operable to gather and store sensor data in a time series format; and a processing system operable to: receive the sensor data in a time series format; label the sensor data in a time series format as either normal, or one of one or more possible faults; segment sensor data in a time series format into blocks having a substantially uniform slope with respect to time; label the blocks as having a rising, falling, or flat slope; join adjacent blocks with slopes having a same sign; identifying candidate kernels; convolute the sensor data in a time series format with candidate kernels; apply a feature selection method to determine which extracted features have discriminatory power; and train a fault classification system based on extracted features having discriminatory power. 20. A system operable to generating classifier systems for univariate or multivariate equipment sensor data, comprising: a data collection system operable to gather and store univariate or multivariate equipment sensor data in a time series format; and a processing system operable to: receive the univariate or multivariate equipment sensor data in a time series format; label the univariate or multivariate equipment sensor data in a time series format as either normal, or one of one or more possible faults; segment univariate or multivariate equipment sensor data in a time series format into blocks having a substantially uniform slope with respect to time; label the blocks as having a rising, falling, or flat slope; join adjacent blocks with slopes having a same sign; identifying candidate kernels; convolute the univariate or multivariate equipment sensor data in a time series format with candidate kernels; apply a feature selection method to determine which extracted features have discriminatory power; and train a fault classification system based on extracted features having discriminatory power.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (2)
Weekley, Richard A.; Goodrich, Robert K.; Cornman, Lawrence B., Feature classification for time series data.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.