Integrated framework for diagnosis and prognosis of components

• US 7,577,548 B1
• Filed: 03/01/2007
• Issued: 08/18/2009
• Est. Priority Date: 03/04/2006
• Status: Expired due to Fees

• Current Assignee: HRL Laboratories LLC (The Boeing Co.)
• Original Assignee: HRL Laboratories LLC (The Boeing Co.)
• Inventors: Przytula, Krzysztof W, Kadambe, Shubha, Srinivasa, Narayan
• Primary Examiner(s): Cosimano; Edward R
• Application Number:
  US11/713,561
• Time in Patent Office: 901 Days
• Field of Search: 340/500, 340/540, 340/679, 340/680, 340/870.01, 340/870.07, 340/870.16, 702/33, 702/34, 702/127, 702/182, 702/183, 702/184, 702/185, 702/186, 702/187, 702/188, 702/189, 702/1
• US Class Current: 702/182
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Described is a system for diagnosis and prognosis of a component. The system is configured to receive a signal from a component. The signal is representative of a current health observation of the component. The system also computes a present likelihood of the component failure based on the signal. Additionally, the system computes a future likelihood of failure of the component for a given future mission. Through diagnosis, a user can determine the present health of the component, and based on the present health and future mission, determine whether or not the component will fail in the future mission.

Claims

1. A system for diagnosis and prognosis of a component, comprising:

a diagnosis module, the diagnosis module being configured to;

receive a signal from a component, the signal being representative of a current state observation of the component;

compute a present state of the component based on the signal; and

a prognosis module, the prognosis module being configured to compute a likelihood of a future state of the component for a given future mission, whereby through diagnosis, a user can determine the present state of the component, and based on the present state and future mission, determine the likelihood of the future state in the future mission.

2. A system for diagnosis and prognosis as set forth in claim 1, wherein the likelihood of a future state is a future likelihood of failure of the component for a given future mission, and wherein the current state and present state are the current health and present health respectively, whereby through diagnosis, a user can determine the present health of the component, and based on the present health and future mission, determine whether the component will fail in the future mission.

3. A system for diagnosis and prognosis as set forth in claim 2, wherein the diagnosis module and the prognosis module are collectively configured to perform operations of:

computing present prior probabilities of component failure;

computing present posterior probabilities of component failure given present health observations;

computing prior probabilities of component failure at an end of the future mission; and

computing future posterior probability of component failure at the end of future the mission, where the future posterior probability indicates the likelihood of failure for the component at the end of the future mission.

4. A system as set forth in claim 3, wherein when computing present prior probabilities, the system is further configured to perform operations of:

generating a failure histogram of component failures as a function of a component'"'"'s usage variables, where each component has usage variables that are indicative of usage of the component, and the histogram having a past bin, a present bin, and a future bin;

using the histogram, determining the prior probability of component failure for each component;

identifying a present value of the usage variable from the failure histogram, the present value being the present bin; and

updating the failure histogram for each component to reflect that the component did not fail until the present bin, with the probability of the present bin in the updated failure histogram being the present prior probability of component failure.

5. A system as set forth in claim 4, wherein when updating the failure histogram, the system is further configured to perform operations of:

setting the values of probabilities for the past bins to zero; and

normalizing the probabilities for the present and future bins so that their sum is equal to one, with the probability for the present bin being the present prior probability of component failure.

6. A system as set forth in claim 5, wherein when updating the failure histogram, if a histogram is unknown for a component, setting the prior probabilities for that component to a failure rate across the components.

7. A system as set forth in claim 6, wherein when computing the present posterior probabilities of component failure, the system is further configured to perform operations of:

setting the prior probabilities of the components to probability values computed in claim 5;

determining the failure state of present-health observations by applying an observation algorithm to an observation selected from a group consisting of sensor outputs, manual measurements, and component symptoms; and

computing posterior probability of failure of the components given prior probabilities and observations, where the posterior probability of failure indicates which components have likely already failed.

8. A system as set forth in claim 7, wherein when computing the posterior probability of failure, the posterior probability of failure is computed using a Bayesian Network inference algorithm.

9. A system as set forth in claim 8, wherein when computing prior probabilities of component failure at the end of the future mission, the system is further configured to perform operations of:

renormalizing the failure histograms for each component by replacing the prior probability of failure for the present bin with the posterior probability of failure and updating the probability values for the future bins so that a sum of the probabilities for the present bin and future bins remains equal to a predetermined value, thereby creating a discrete density function;

computing a distribution function from the discrete density function, with each bin of the distribution function being a sum of preceding bins in the renormalized failure histogram; and

determining a bin that corresponds to an end of the future mission, where the value in the bin that corresponds to the end of the future mission is the prior probability of component failure at the end of the future mission.

10. A system as set forth in claim 9, wherein when computing future posterior probability of component failures at the end of the future mission, the system is further configured to perform operations of:

setting the prior probabilities of the components to values computed in claim 9;

setting health-trend observations as values obtained from a trending algorithm; and

computing the future posterior probabilities of component failure using a Bayesian Network inference algorithm, where the future posterior probability indicates the likelihood of failure for a particular component at the end of the future mission.

11. A method for diagnosis and prognosis of a component, comprising acts of:

defining a usage variable for a component, the usage variable being a variable that is indicative of usage of the component;

defining a future mission in terms of future usage of the component in completing the future mission, as applied to the usage variable;

defining present health observations, the present health observations being indicative of a present health state of the component;

defining future health observations by trending the present health observations up to the future usage; and

utilizing a Bayesian Network model to determine the likelihood of failure for the component at the end of the future mission from the present health observations and the future health observations.

12. A method for diagnosis and prognosis of a component, comprising acts of:

receiving a signal from a component, the signal being representative of a current state observation of the component;

computing a present state of the component based on the signal; and

computing a likelihood of a future state of the component for a given future mission, whereby through diagnosis, a user can determine the present state of the component, and based on the present state and future mission, determine the likelihood of the future state in the future mission.

13. A method for diagnosis and prognosis as set forth in claim 12, wherein in the acts of receiving a signal, computing a present state, and computing a likelihood of a future state, the likelihood of a future state is a future likelihood of failure of the component for a given future mission, and wherein the current state and present state are the current health and present health respectively, whereby through diagnosis, a user can determine the present health of the component, and based on the present health and future mission, determine whether the component will fail in the future mission.

14. A method for diagnosis and prognosis as set forth in claim 13, further comprising acts of:

computing present prior probabilities of component failure;

computing present posterior probabilities of component failure given present health observations;

computing prior probabilities of component failure at an end of the future mission; and

computing future posterior probability of component failure at the end of future the mission, where the future posterior probability indicates the likelihood of failure for the component at the end of the future mission.

15. A method as set forth in claim 14, wherein the act of computing present prior probabilities further comprises acts of:

generating a failure histogram of component failures as a function of a component'"'"'s usage variables, where each component has usage variables that are indicative of usage of the component, and the histogram having a past bin, a present bin, and a future bin;

using the histogram, determining the prior probability of component failure for each component;

identifying a present value of the usage variable from the failure histogram, the present value being the present bin; and

updating the failure histogram for each component to reflect that the component did not fail until the present bin, with the probability of the present bin in the updated failure histogram being the present prior probability of component failure.

16. A method as set forth in claim 15, wherein the act of updating the failure histogram further comprises acts of:

setting the values of probabilities for the past bins to zero; and

normalizing the probabilities for the present and future bins so that their sum is equal to one, with the probability for the present bin being the present prior probability of component failure.

17. A method as set forth in claim 16, wherein in the act of updating the failure histogram, if a histogram is unknown for a component, setting the prior probabilities for that component to a failure rate across the components.

18. A method as set forth in claim 17, wherein the act of computing the present posterior probabilities of component failure further comprises acts of:

setting the prior probabilities of the components to probability values computed in claim 16;

determining the failure state of present-health observations by applying an observation algorithm to an observation selected from a group consisting of sensor outputs, manual measurements, and component symptoms; and

computing posterior probability of failure of the components given prior probabilities and observations, where the posterior probability of failure indicates which components have likely already failed.

19. A method as set forth in claim 18, wherein in the act of computing the posterior probability of failure, the posterior probability of failure is computed using a Bayesian Network inference algorithm.

20. A method as set forth in claim 19, wherein the act of computing prior probabilities of component failure at the end of the future mission further comprises acts of:

renormalizing the failure histograms for each component by replacing the prior probability of failure for the present bin with the posterior probability of failure and updating the probability values for the future bins so that a sum of the probabilities for the present bin and future bins remains equal to a predetermined value, thereby creating a discrete density function;

computing a distribution function from the discrete density function, with each bin of the distribution function being a sum of preceding bins in the renormalized failure histogram; and

determining a bin that corresponds to an end of the future mission, where the value in the bin that corresponds to the end of the future mission is the prior probability of component failure at the end of the future mission.

21. A method as set forth in claim 20, wherein that act of computing future posterior probability of component failures at the end of the future mission further comprises acts of:

setting the prior probabilities of the components to values computed in claim 20;

setting health-trend observations as values obtained from a trending algorithm; and

computing the future posterior probabilities of component failure using a Bayesian Network inference algorithm, where the future posterior probability indicates the likelihood of failure for a particular component at the end of the future mission.

22. A computer program product for diagnosis and prognosis, the computer program product comprising computer-readable instruction means encoded on a computer-readable medium that are executable by a computer for causing the computer to:

receive a signal from a component, the signal being representative of a current state observation of the component;

compute a present state of the component based on the signal; and

compute a likelihood of a future state of the component for a given future mission, whereby through diagnosis, a user can determine the present state of the component, and based on the present state and future mission, determine the likelihood of the future state in the future mission.

23. A computer program product for diagnosis and prognosis as set forth in claim 22, wherein the likelihood of a future state is a future likelihood of failure of the component for a given future mission, and wherein the current state and present state are the current health and present health respectively, such that the computer program product further comprises instruction means to cause a computer to perform operations of:

receive a signal from a component, the signal being representative of a current health observation of the component;

compute a present health of the component based on the signal; and

compute a likelihood of a future health of the component for a given future mission, whereby through diagnosis, a user can determine the present health of the component, and based on the present health and future mission, determine the future likelihood of failure of the component.

24. A computer program product for diagnosis and prognosis as set forth in claim 23, further comprising instruction means for causing a computer to perform operations of:

computing present prior probabilities of component failure;

computing present posterior probabilities of component failure given present health observations;

computing prior probabilities of component failure at an end of the future mission; and

computing future posterior probability of component failure at the end of future the mission, where the future posterior probability indicates the likelihood of failure for the component at the end of the future mission.

25. A computer program product as set forth in claim 24, further comprising instruction means for causing a computer to perform operations of:

generating a failure histogram of component failures as a function of a component'"'"'s usage variables, where each component has usage variables that are indicative of usage of the component, and the histogram having a past bin, a present bin, and a future bin;

using the histogram, determining the prior probability of component failure for each component;

identifying a present value of the usage variable from the failure histogram, the present value being the present bin; and

updating the failure histogram for each component to reflect that the component did not fail until the present bin, with the probability of the present bin in the updated failure histogram being the present prior probability of component failure.

26. A computer program product as set forth in claim 25, further comprising instruction means for causing a computer to perform operations of:

setting the values of probabilities for the past bins to zero; and

normalizing the probabilities for the present and future bins so that their sum is equal to one, with the probability for the present bin being the present prior probability of component failure.

27. A computer program product as set forth in claim 26, wherein when causing a computer to update the failure histogram, if a histogram is unknown for a component, further comprising instruction means for causing a computer to perform an operation of setting the prior probabilities for that component to a failure rate across the components.

28. A computer program product as set forth in claim 27, further comprising instruction means for causing a computer to perform operations of:

setting the prior probabilities of the components to probability values computed in claim 26;

determining the failure state of present-health observations by applying an observation algorithm to an observation selected from a group consisting of sensor outputs, manual measurements, and component symptoms; and

computing posterior probability of failure of the components given prior probabilities and observations, where the posterior probability of failure indicates which components have likely already failed.

29. A computer program product as set forth in claim 28, further comprising instruction means for causing a computer to compute the posterior probability of failure using a Bayesian Network inference algorithm.

30. A computer program product as set forth in claim 29, further comprising instruction means for causing a computer to perform operations of:

renormalizing the failure histograms for each component by replacing the prior probability of failure for the present bin with the posterior probability of failure and updating the probability values for the future bins so that a sum of the probabilities for the present bin and future bins remains equal to a predetermined value, thereby creating a discrete density function;

computing a distribution function from the discrete density function, with each bin of the distribution function being a sum of preceding bins in the renormalized failure histogram; and

determining a bin that corresponds to an end of the future mission, where the value in the bin that corresponds to the end of the future mission is the prior probability of component failure at the end of the future mission.

31. A computer program product as set forth in claim 30, further comprising instruction means for causing a computer to perform operations of:

setting the prior probabilities of the components to values computed in claim 30;

setting health-trend observations as values obtained from a trending algorithm; and

computing the future posterior probabilities of component failure using a Bayesian Network inference algorithm, where the future posterior probability indicates the likelihood of failure for a particular component at the end of the future mission.