METHODS SYSTEMS AND APPARATUS FOR ANALYZING COMPLEX SYSTEMS VIA PROGNOSTIC REASONING

• US 20110118905A1
• Filed: 11/16/2009
• Published: 05/19/2011
• Est. Priority Date: 11/16/2009
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Mylaraswamy, Dinkar, Hadden, George Daniel
• Granted Patent: US 8,285,438 B2
• Time in Patent Office: Days
• US Class Current: 701/3
• CPC Class Codes: G05B 23/0251 Abstraction hierarchy, e.g. "complex systems", i.e. system is divided in subsystems, subsystems are monitored and results are combined to decide on status of whole system
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G07C 5/085 using electronic data carriers

Abstract

Methods and apparatus are provided for analyzing a complex system that includes a number of subsystems. Each subsystem comprises at least one sensor designed to generate sensor data. Sensor data from at least one of the sensors is processed to generate binary evidence of a sensed event, and complex evidence of a sensed event. The complex evidence has more sophisticated mathematical properties than the binary evidence. The complex evidence comprises one or more of: a condition indicator (CI), a health indicator (HI), and a prognostic indicator (PI). A system fault model (SFM) is provided that defines statistical relationships between binary evidence, complex evidence, and an underlying failure mode (FM) that is occurring in the complex system. The binary evidence and the complex evidence are processed to identify failure modes taking place within one or more of the subsystems. Based on the binary evidence and the complex evidence and the SFM, diagnostic conclusions can be generated regarding adverse events that are taking place within the complex system, and prognostic conclusions can be generated regarding adverse events that are predicted to take place within the complex system.

Claims

1. A method for analyzing a complex system comprising a number of subsystems, wherein each subsystem comprises at least one sensor designed to generate sensor data, the method comprising:

processing sensor data from at least one of the sensors to generate binary evidence of a sensed event;

processing sensor data from at least one of the sensors to generate complex evidence of a sensed event, wherein the complex evidence has more sophisticated mathematical properties than the binary evidence, wherein the complex evidence comprises at least one of;

a condition indicator (CI), a health indicator (HI), and a prognostic indicator (PI);

providing a system fault model (SFM) that defines statistical relationships between binary evidence, complex evidence, and an underlying failure mode (FM) that is occurring in the complex system;

processing the binary evidence and the complex evidence to identify failure modes taking place within one or more of the subsystems; and

generating, based on the binary evidence and the complex evidence and the SFM, diagnostic conclusions regarding adverse events that are taking place within the complex system, and prognostic conclusions regarding adverse events that are predicted to take place within the complex system.

2. A method according to claim 1, wherein the step of processing sensor data from at least one of the sensors to generate binary evidence of a sensed event, comprises:

receiving a univariate signal comprising sensor data generated by a particular sensor for a particular subsystem; and

executing a built-in-test (BIT) on the sensor data to generate the binary evidence, wherein each instance of binary evidence comprises;

BIT evidence generated by a particular built-in-test (BIT).

3. A method according to claim 2, wherein the BIT evidence implicates an ambiguity group of failure modes, and wherein the SFM captures a relationship between the FM and the BIT evidence as an exhibit.

4. A method according to claim 1, wherein the step of processing sensor data from at least one of the sensors to generate complex evidence of a sensed event, comprises the steps of:

receiving a multivariate signal comprising sensor data; and

performing signal processing on the sensor data of the multivariate signal to generate the CI, wherein the CI comprises;

derived parametric data that provides partial evidence towards an ambiguity group of failure modes.

5. A method according to claim 4, wherein the mathematical relationship (R1) between the CI and the FM is represented as a low-order polynomial model taking the form:

6. A method according to claim 4, wherein the health indicator (HI) comprises partially-summarized diagnostic conclusions that provide probabilistic evidence for an ambiguity group of failure modes and contain a reference to an ambiguity failure mode set.

7. A method according to claim 6, wherein the mathematical relationship (R2) between the HI and the FM is represented as a probability density function (P(FM|HI)).

8. A method according to claim 6, wherein the prognostic indicator (PI) comprises partially-summarized evidences that indicate a future evolution of the HI over a future time period.

9. A method according to claim 8, wherein the PI comprises:

Prognostic Vectors (PVs) comprising a set of one or more ordered pair of time and probability that indicate complex time to failure.

10. A method according to claim 8, wherein the mathematical format of a relationship (R3) between the PI and the FM is represented as a time series auto regression model taking the form:

11. A method according to claim 1, wherein the SFM further comprises a plurality of other elements comprising:

one or more repair action elements, wherein each repair action element represents corrective repair action for restoring the complex system back to an original state in which the complex system the specified requirements, wherein the relationship between the each repair action element and the FM is represented as a corresponding corrective repair action for restoring the complex system back to the original state;

one or more function elements, wherein each function element represents activities performed by a system or a subsystem;

one or more optional elements, wherein the relationship between the each optional element and the FM is represented by impacts between the FM and the optional element, wherein the mathematical format of the impacts between the optional element and the FM can be represented as a binary number; and

one or more related failure mode elements, wherein each related failure mode element represents a cascade relationship between failure modes that are caused by other failure modes, wherein the relationship between the related failure mode element and the FM is represented as a cascade between FM and related failure mode element.

12. A method according to claim 1, wherein the complex system comprises a vehicle and wherein the analysis system comprises an Integrated Vehicle Health Management (IVHM) system.

13. A method according to claim 12, wherein the vehicle comprises:

an aircraft or a spacecraft.

14. A system, comprising:

a complex system comprising a number of subsystems each having at least one sensor designed to generate sensor data; and

an analysis system, comprising;

a simple diagnostic and prognostic monitor (SDPM) being designed to process sensor data from at least one of the sensors and generate binary evidence of a sensed event;

an advanced diagnostic and prognostic monitor (ADPM) being designed to process sensor data from at least one of the sensors and generate at least one form of complex evidence of a sensed event, wherein the complex evidence comprises at least one of;

a condition indicator (CI), a health indicator (HI), and a prognostic indicator (PI); and

a maintenance computer (MC), communicatively coupled to the SDPM and the ADPM, and comprising;

an advanced diagnostics and prognostics reasoner (ADPR) module designed to process the binary evidence and the complex evidence to identify failure modes taking place within one or more of the subsystems of the complex system, wherein the ADPR module comprises;

a system fault model (SFM) that defines probabilistic relationships between binary evidence, complex evidence, and an underlying failure mode (FM) that that is occurring in the complex system;

a diagnostic processing module (DPM) designed to generate, based on the binary evidence and the complex evidence and the SFM, diagnostic conclusions regarding adverse events that are taking place within the complex system; and

a prognostic processing module (PPM) designed to generate, based on the binary evidence and the complex evidence and the SFM, prognostic conclusions regarding adverse events that are predicted to take place within the complex system.

15. A system according to claim 14, wherein each SDPM is designed to receive a univariate signal comprising sensor data generated by a particular sensor for a particular subsystem, and to execute a built-in-test (BIT) on the sensor data to interpret the sensor data and generate the binary evidence, wherein each instance of binary evidence comprises:

BIT evidence generated by a particular built-in-test (BIT), wherein the BIT evidence implicates an ambiguity group of failure modes, and wherein the SFM captures a relationship between the FM and the BIT evidence as an exhibit.

16. A system according to claim 14, wherein the SDPM is designed to receive a multivariate signal comprising sensor data, and to perform signal processing on the sensor data of the multivariate signal to generate the CI, wherein the CI comprises:

derived parametric data that provides partial evidence towards an ambiguity group of failure modes, wherein the mathematical relationship (R1) between the CI and the FM is represented as a low-order polynomial model taking the form;

17. A system according to claim 16, wherein the health indicator (HI) comprises partially-summarized diagnostic conclusions that provide probabilistic evidence for an ambiguity group of failure modes and contain a reference to an ambiguity failure mode set, wherein the mathematical relationship (R2) between the HI and the FM is represented as a probability density function (P(FM|HI)).

18. A system according to claim 17, wherein the prognostic indicator (PI) comprises partially-summarized evidences that indicate a future evolution of the HI over a future time period, wherein the PI comprises:

Prognostic Vectors (PVs) comprising a set of one or more ordered pair of time and probability that indicate complex time to failure.

19. A system according to claim 18, wherein the mathematical format of a relationship (R3) between the PI and the FM is represented as a time series auto regression model taking the form:

20. A system, comprising:

a vehicle comprising a number of subsystems each having at least one sensor designed to generate sensor data; and

an Integrated Vehicle Health Management (IVHM) system, comprising;

a plurality of simple diagnostic and prognostic monitors (SDPMs) each being designed to process sensor data from at least one of the sensors and generate binary evidence of a sensed event;

a plurality of advanced diagnostic and prognostic monitors (ADPMs) each being designed to process sensor data from at least one of the sensors and generate at least one form of complex evidence of a sensed event, wherein the complex evidence comprises at least one of;

a condition indicator (CI), a health indicator (HI), and a prognostic indicator (PI); and

an advanced diagnostics and prognostics reasoner (ADPR) module designed to;

process the binary evidence and the complex evidence based on a system fault model (SFM) that defines statistical relationships between binary evidence, complex evidence, and a failure mode (FM) that represents an underlying fault condition that is occurring in the vehicle;

identify adverse events taking place within one or more of the subsystems of the vehicle; and

generate, based on the binary evidence and the complex evidence and the SFM, diagnostic conclusions regarding adverse events that are taking place within the vehicle, and prognostic conclusions regarding adverse events that are predicted to take place within the vehicle.