Method and device for fault auto-detection in an on-board system
View Patent ↗A method for automatic fault detection with data representing a fault in an on-board system, the method including: determining, with a processor, at least two physical variables represented by said data, said at least two physical variables including at least one time-dependent variable representative of a persistence of an event in time, and at least one instantaneous variable; representing, with the processor, said data by points of a space whose coordinates are values of said at least two physical variables; an detecting, with the processor, a fault by employing a frontier defining a subspace of said space in which representation of a datum is associated with the detection of a fault.
1. A method for automatic fault detection with data representing a fault in an on-board system, the method comprising:
determining, with a processor, at least two physical variables represented by said data, said at least two physical variables including at least
one time-dependent variable representative of a persistence of an event in time, and
at least one instantaneous variable;
representing, with the processor, said data by points of a space whose coordinates are values of said at least two physical variables; and
detecting, with the processor, a fault by employing a frontier defining a subspace of said space in which representation of a datum is associated with the detection of a fault.
2. The method according to claim 1 , further comprising:
estimating, with the processor, the reactivity of detection of said fault as a function of said frontier and of a position of a representation of a first datum associated with said fault in the said subspace.
3. The method according to claim 2 , wherein the estimating the reactivity includes using a function of a distance between the first datum in the said subspace and an orthogonal projection of said first datum on a hyperplane formed by the frontier.
4. The method according to claim 3 , further comprising:
determining, with the processor, said frontier as a function of values of said at least two physical variables representative of learning data distributed in different classes as a function of whether or not they are associated with the fault.
5. The method according to claim 4 , wherein the determining said frontier includes obtaining the learning data from a simulation of a model of dysfunctioning, each point in the space being associated with an information item of whether or not maintenance is necessary.
6. The method according to claim 4 , wherein the determining said frontier includes obtaining the learning data from bench or flight tests.
7. The method according to claim 4 , wherein the determining said frontier includes determining the frontier on a basis of least squares between points of the different classes.
8. The method according to claim 4 , wherein in the determining said frontier, support vector machines are employed.
9. The method according to claim 2 , further comprising:
determining, with the processor, said frontier as a function of values of said at least two physical variables representative of learning data distributed in different classes as a function of whether or not they are associated with a fault.
10. The method according to claim 9 , wherein the determining said frontier includes obtaining the learning data from a simulation of a model of dysfunctioning, each point in the space being associated with an information item of whether or not maintenance is necessary.
11. The method according to claim 9 , wherein the determining said frontier includes obtaining the learning data from bench or flight tests.
12. The method according to claim 9 , wherein the determining said frontier includes determining the frontier on a basis of least squares between points of the different classes.
13. The method according to claim 1 , further comprising:
determining, with the processor, said frontier as a function of values of said at least two physical variables representative of learning data distributed in different classes as a function of whether or not they are associated with a fault.
14. The method according to claim 13 , wherein the determining said frontier includes obtaining the learning data from a simulation of a model of dysfunctioning, each point in the space being associated with an information item of whether or not maintenance is necessary.
15. The method according to claim 13 , wherein the determining said frontier includes obtaining the learning data from bench or flight tests.
16. The method according to claim 13 , in which in the determining said frontier, the frontier separates two classes of events of
events necessitating maintenance, and
events not necessitating maintenance.
17. The method according to claim 13 , wherein the determining said frontier includes determining the frontier on a basis of least squares between points of the different classes.
18. A device for automatic fault detection with data representing a fault in an on-board system, the device comprising:
a means for determining at least two physical variables represented by said data, said at least two physical variables including at least
one time-dependent variable representative of a persistence of an event in time, and
at least one instantaneous variable;
a means for representing said data by points of a space whose coordinates are values of said at least two physical variables; and
a means for detecting a fault by employing a frontier defining a subspace of said space in which representation of a datum is associated with the detection of a fault.
19. The device according to claim 18 , further comprising:
means for estimating a reactivity of detection of said fault as a function of said frontier and of a position of a representation of a first datum associated with the fault in the said subspace.
20. The device according to claim 19 , in which the means for estimating the reactivity uses a function of a distance between the first datum in said subspace and an orthogonal projection of said first datum on a hyperplane formed by the frontier.
21. A non-transitory computer readable storage medium encoded with instructions, which when executed by a computer causes the computer to execute a method for automatic fault detection with data representing a fault in an on-board system, the method comprising:
determining, with the computer, at least two physical variables represented by said data, said at least two physical variables including at least
one time-dependent variable representative of a persistence of an event in time, and
at least one instantaneous variable;
representing, with the computer, said data by points of a space whose coordinates are values of said at least two physical variables; and
detecting, with the computer, a fault by employing a frontier defining a subspace of said space in which representation of a datum is associated with the detection of a fault.
22. The non-transitory computer readable storage medium of claim 21 , wherein the method further comprises:
estimating, with the computer, a reactivity of detection of said fault as a function of said frontier and of a position of a representation of a first datum associated with the fault in said subspace, wherein the estimating the reactivity uses a function of a distance between the first datum in said subspace and an orthogonal projection of said first datum on a hyperplane formed by the frontier.