Method and apparatus for infrastructure health monitoring and analysis wherein anomalies are detected by comparing measured outputs to estimated/modeled outputs by using a delay
원문보기
IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
UP-0882983
(2007-08-08)
|
등록번호 |
US-7822697
(2010-11-15)
|
발명자
/ 주소 |
- Sobhani Tehrani, Ehsan
- Garabedian, Armenih
- Khorasani, Khashayar
- Bagchi, Ashutosh
- Joshi, Anand
|
출원인 / 주소 |
|
대리인 / 주소 |
|
인용정보 |
피인용 횟수 :
1 인용 특허 :
5 |
초록
▼
There is described herein a method for detecting anomalies in an infrastructure, the method comprising: providing a computationally-intelligent analysis model to model a behaviour of at least one detection instrument in said infrastructure; inputting control instrument data into said analysis model,
There is described herein a method for detecting anomalies in an infrastructure, the method comprising: providing a computationally-intelligent analysis model to model a behaviour of at least one detection instrument in said infrastructure; inputting control instrument data into said analysis model, said control instrument data being provided by control instruments in said infrastructure; outputting an estimated behaviour for said at least one detection instrument from said analysis model; comparing actual data from said at least one detection instrument to said estimated behaviour and generating a set of residuals representing a difference between said actual data and said estimated behaviour; and identifying anomalies when said residuals exceed a predetermined threshold.
대표청구항
▼
The invention claimed is: 1. A method for detecting anomalies in an infrastructure, the method comprising: providing a computationally-intelligent analysis model to model a behavior of at least one detection instrument in said infrastructure; inputting control instrument sensor data into said analy
The invention claimed is: 1. A method for detecting anomalies in an infrastructure, the method comprising: providing a computationally-intelligent analysis model to model a behavior of at least one detection instrument in said infrastructure; inputting control instrument sensor data into said analysis model, said sensor data being provided by sensors in said infrastructure; outputting an estimated behavior for said at least one detection instrument from said analysis model; comparing measured output data from said at least one detection instrument to said estimated behavior and generating a set of residuals representing a difference between said measured output data and said estimated behavior; and identifying anomalies when said residuals exceed a predetermined threshold; and wherein said inputting sensor data into said analysis model comprises: using lag time information to delay corresponding input data; estimating a lag time between a cause and an effect in an infrastructure; identifying a first variable as said cause and a second variable as said effect; specifying a desired time period over which the lag time is estimated; assigning a maximum possible lag time between said cause and effect; calculating a cross-correlation function between said first variable and said second variable over said desired time period; and shifting forward in time said second variable until said maximum lag time is reached while recalculating said cross-correlation function between each shift in time, wherein a total shift needed to reach a maximum absolute cross-correlation corresponds to said lag time. 2. A method as claimed in claim 1, wherein said infrastructure is a dam. 3. A method as claimed in claim 1, wherein said inputting sensor data comprises inputting one or more temperatures explicitly into the analysis model. 4. A method as claimed in claim 1, wherein said providing a computationally intelligent analysis model comprises providing a neural-network. 5. A method as claimed in claim 4, wherein said providing a neural-network comprises providing a data-driven parameterized nonlinear model. 6. A method as claimed in claim 5, wherein said providing a neural-network comprises providing a coupled computationally intelligent model, an output of said model being a joint function of all input variables. 7. A method as claimed in claim 6, wherein said providing a neural-network comprises providing a decoupled computationally intelligent model, a contribution of each input to an output being calculated separately and added together. 8. A method as claimed in claim 1, wherein said providing a computationally intelligent analysis model comprises providing a fuzzy network. 9. A method as claimed in claim 1, wherein said providing a computationally intelligent analysis model comprises providing a neuro-fuzzy network. 10. A method as claimed in claim 1, wherein said providing a computationally intelligent analysis model comprises providing a Bayesian network. 11. The method as claimed in claim 1, wherein said providing said computationally-intelligent analysis model comprises: providing a model learning phase using historical data from at least one of said at least one detection instruments and said sensors within said infrastructure to teach the analysis model; saving optimized parameters into said analysis model; and providing a model execution/testing phase to predict and evaluate said behavior in real-time as sensor data is input therein; and wherein said outputting said estimated behavior comprises outputting a predicted value for said at least one detection instrument. 12. A method as claimed in claim 1, wherein said cause and said effect correspond to said sensors in said infrastructure. 13. A method as claimed in claim 1, wherein said maximum absolute cross-correlation corresponds to a measure of dependency between said cause and said effect. 14. A system for detecting anomalies in an infrastructure, the system comprising: an analysis module comprising a computationally-intelligent model of a behavior of at least one detection instrument in said infrastructure, said model having sensor data sensors sensing said infrastructure as inputs and an estimated behavior for said at least one detection instrument as an output; a comparison module that compares measured output data from said at least one detection instrument to said estimated behavior and generate a set of residuals representing a difference between said measured output data and said estimated behavior; and a detection module that receives said residuals and identifies an anomaly when a predetermined threshold is exceeded; and wherein said inputting sensor data into said analysis model comprises: using lag-time information to delay corresponding input data; estimating a lag time between a cause and an effect in an infrastructure; identifying a first variable as said cause and a second variable as said effect; specifying a desired time period over which the lag time is estimated; assigning a maximum possible lag time between said cause and effect; calculating a cross-correlation function between said first variable and said second variable over said desired time period; and shifting forward in time said second variable until said maximum lag time is reached while recalculating said cross-correlation function between each shift in time, wherein a total shift needed to reach a maximum absolute cross-correlation corresponds to said lag time. 15. A system as claimed in claim 14, wherein said infrastructure is a dam. 16. A system as claimed in claim 14, wherein said model also uses temperature as input. 17. A system as claimed in claim 14, wherein said model is a neural-network. 18. A system as claimed in claim 17, wherein said neural-network is a data-driven parameterized nonlinear model. 19. A system as claimed in claim 18, wherein said neural-network is a coupled computationally intelligent model, an output of said model being a joint function of all input variables. 20. A system as claimed in claim 18, wherein said neural-network is a decoupled computationally intelligent model, a contribution of each input to an output being calculated separately and added together. 21. A system as claimed in claim 14, wherein said computationally intelligent model is a fuzzy network. 22. A system as claimed in claim 14, wherein said computationally intelligent model is a neuro-fuzzy network. 23. A system as claimed in claim 14, wherein said computationally intelligent model is a Bayesian network.
이 특허에 인용된 특허 (5)
-
Kaji, Hirotaka; Yamaguchi, Masashi; Harada, Hiroshi; Matsushita, Yukio, Control system of optimizing the function of machine assembly using GA-Fuzzy inference.
-
Rajamani Ravi ; Chbat Nicolas Wadih ; Ashley Todd Alan, Controller with neural network for estimating gas turbine internal cycle parameters.
-
Rajamani, Ravi; Flodman, David Allen; Garry, Robert Scott; Norman, Bruce Gordon; Tomlinson, Leroy, Gas turbine system operation based on estimated stress.
-
Gross Kenneth C. ; Wegerich Stephan W ; Singer Ralph M. ; Mott Jack E., Industrial process surveillance system.
-
Uluyol, Onder; Nwadiogbu, Emmanuel O., Sensor error detection and compensation system and method.
이 특허를 인용한 특허 (1)
-
Iannone, Michele, Method for performing diagnostics of a structure subject to loads and system for implementing said method.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.