STOCHASTIC EVIDENCE AGGREGATION SYSTEM

• US 20080195572A1
• Filed: 06/19/2007
• Published: 08/14/2008
• Est. Priority Date: 02/09/2007
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Guralnik, Valerie, Mylaraswamy, Dinkar
• Granted Patent: US 7,778,943 B2
• Time in Patent Office: Days
• US Class Current: 706/59
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A system for obtaining diagnostic information, such as evidence about a mechanism, within an algorithmic framework, including filtering and aggregating the information through, for instance, a stochastic process. The output may be an overall belief value relative to a presence of an item such as, for example, a fault in the mechanism.

Claims

1. An information fusion system comprising:

at least two diagnostic algorithms for providing belief values about a failure mode of a mechanism;

a filter for receiving the belief values from the diagnostic algorithms and providing smoothed belief values; and

an aggregator for aggregating the smoothed belief values into at least one diagnostic state.

2. The system of claim 1, wherein the diagnostic state indicates an overall belief value about the failure mode.

3. The system of claim 1, wherein the belief values are provided for particular instances of past and/or present time by the diagnostic algorithms.

4. The system of claim 1, wherein the filter is for smoothing noise in the belief values provided by diagnostic algorithms.

5. The system of claim 4, wherein the filter is for filling a gap in belief values resulting from occasional drop-outs in the belief values provided by the diagnostic algorithms.

6. The system of claim 4, wherein the filter is for accounting for a propagation of faults within the mechanism.

7. The system of claim 4, wherein the filter uses a diagnostic state at time t-1 resulting from the aggregating.

8. The system of claim 4, wherein the filter uses a diagnostic state at time t-1 to smooth out noise in the belief value provided by a diagnostic algorithm at time t.

9. The system of claim 8, wherein the filter is a Kalman filter.

10. The system of claim 1, wherein the aggregating the smoothed belief values from the filter takes into account a conflict among evidence provided by diagnostic algorithms.

11. The system of claim 10, wherein the conflict among the evidence is resolved by generating multiple failure mode hypotheses.

12. The system of claim 11, wherein the multiple failure mode hypotheses cover two or more failures occurring simultaneously.

13. The system of claim 11, wherein some multiple failure mode hypotheses are restricted relative to prior knowledge about failure mode probabilities.

14. The system of claim 10, wherein multiple failure mode hypotheses are generated by taking a pair-wise union of elements in the conflict and iteratively taking pair-wise unions of multiple failure hypotheses generated during a previous iteration.

15. The system of claim 10, wherein a diagnostic state for single and/or multiple failure mode hypotheses set is calculated by a weighted averaging of smoothed belief values.

16. The system of claim 1, wherein the aggregating is of data from a stochastic process.

17. An evidence aggregation system comprising:

a diagnostic module;

a filter connected to an output of the diagnostic module; and

an aggregation module connected to an output of the filter.

18. The system of claim 17, wherein:

the diagnostic module comprises;

a delay operator; and

a demultiplexer connected to the delay operator and the filter;

the delay operator has an input for receiving an initial diagnostic state;

the delay operator has an output for providing a time delayed diagnostic state to the demultiplexer; and

the demultiplexer has an output for providing selected diagnostic state information to the filter.

19. The system of claim 18, wherein:

the filter is for providing belief values to the aggregation module;

the aggregation module is for providing diagnostic states; and

a diagnostic state is related to a failure mode of an observed mechanism.

20. A method for fusing information comprising:

algorithmically obtaining belief values about one or more failure modes of a stochastic system;

filtering the belief values into smoothed belief values; and

aggregating the smoothed belief values into at least one diagnostic state about the system.