Condition-Based Monitoring System For Machinery And Associated Methods

• US 20100114806A1
• Filed: 10/19/2009
• Published: 05/06/2010
• Est. Priority Date: 10/17/2008
• Status: Active Grant

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Hall, Richard, Das, Sreerupa, Bodkin, Michael A., Worden, Eric W., Herzog, Stefan, Harris, Michelle L., Harrison, Gregory A.
• Granted Patent: US 8,332,337 B2
• Time in Patent Office: Days
• US Class Current: 706/14
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/043 based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]

Abstract

Real-time condition-based analysis is performed on a machine for providing diagnostic and prognostic outputs indicative of machine status includes a signal processor for receiving signals from sensors adapted for measuring machine performance parameters. The signal processor conditions and shapes at least some of the received signals into an input form for a neural network. A fuzzy adaptive resonance theory neural network receives at least some of the conditioned and shaped signals, and detects and classifies a state of the machine based upon the received conditioned and shaped signals, and upon a predetermined ontology of machine states, diagnostics, and prognostics. The neural network can also determine from the machine state a health status thereof, which can comprise an anomaly, and output a signal representative of the determined health status. A Bayesian intelligence network receives the machine state from the neural network and determines a fault probability at a future time.

Claims

1. A system for performing real-time condition-based analysis on a machine for providing diagnostic and prognostic outputs indicative of machine status comprising:

a signal processor for receiving signals from a plurality of sensors positioned and adapted for measuring a plurality of machine performance parameters, and for conditioning and shaping at least some of the received signals into a form for inputting into a neural network;

a neural network adapted to receive at least some of the conditioned and shaped signals, to detect and classify a state of the machine based upon the received conditioned and shaped signals and upon a predetermined ontology of machine states, diagnostics, and prognostics, to determine from the machine state a relative health status thereof, and to output a signal representative of the determined relative health status; and

a Bayesian intelligence network adapted to receive the machine state from the neural network and to determine therefrom and output a probability of a fault at a predetermined future time.

2. The system recited in claim 1, wherein the signal processor is further adapted for selecting from the received signals a subset thereof based upon pattern recognition and optimizing the selected subset of signals.

3. The system recited in claim 2, wherein the subset-selecting is accomplished with the use of correlation and previously learned machine state data.

4. The system recited in claim 3, wherein the previously learned machine state data comprise data input from a subject matter expert having knowledge of the machine.

5. The system recited in claim 3, wherein the machine comprises a subject machine, and the previously learned machine state data comprise data on a second machine in a same class as the subject machine.

6. The system recited in claim 2, wherein the optimizing is accomplished with the use of a genetic algorithm and programming system.

7. The system recited in claim 1, wherein the signal processor comprises a filter for removing noise and accent signals from the received signals.

8. The system recited in claim 7, wherein the filter comprises a self-organizing map for reducing a dimensionality of the received signals.

9. The system recited in claim 1, wherein the neural network comprises a fuzzy adaptive resonance theory neural network.

10. The system recited in claim 1, wherein the machine state comprises a plurality of machine states, each machine state representative of a component of the machine.

11. The system recited in claim 1, wherein the relative health status comprises at least one of an anomaly and an indication of a closeness of the machine state to a predetermined normal operation state.

12. A method for performing real-time condition-based analysis on a machine for providing diagnostic and prognostic outputs indicative of machine status comprising:

conditioning and shaping signals received from a plurality of sensors positioned and adapted for measuring a plurality of machine performance parameters into a form for inputting into a neural network;

inputting at least some of the conditioned and shaped signals into a neural network adapted to detect and classify a state of the machine based upon the received conditioned and shaped signals and upon a predetermined ontology of machine states, diagnostics, and prognostics;

using the neural network to determine from the machine state a relative health status of the machine;

outputting a signal representative of the determined relative health status;

inputting the machine state into a Bayesian intelligence network adapted to determine therefrom a probability of a fault at a predetermined future time; and

outputting the determined fault probability.

13. The method recited in claim 12, wherein the conditioning and shaping comprises selecting from the received signals a subset thereof based upon pattern recognition and optimizing the selected subset of signals.

14. The method recited in claim 13, wherein the subset-selecting is accomplished with the use of correlation and previously learned machine state data.

15. The method recited in claim 14, further comprising, prior to the conditioning and shaping, receiving data from a subject matter expert having knowledge of the machine, and incorporating the subject matter received data into the previously learned machine state data.

16. The method recited in claim 14, wherein the machine comprises a subject machine, and further comprising, prior to the conditioning and shaping, receiving data on a second machine in a same class as the subject machine and incorporating the data on the second machine into the previously learned machine state data.

17. The method recited in claim 13, wherein the optimizing comprises using a genetic algorithm and programming system.

18. The method recited in claim 12, wherein the conditioning and shaping comprises filtering the received signals to remove noise and accent signals therefrom.

19. The method recited in claim 18, wherein the filtering comprises reducing a dimensionality of the received signals with the use of a self-organizing map.

20. The method recited in claim 12, wherein the neural network comprises a fuzzy adaptive resonance theory neural network.

21. The method recited in claim 12, wherein the machine state comprises a plurality of machine states, each machine state representative of a component of the machine.

22. The method recited in claim 12, wherein the relative health status comprises at least one of an anomaly and an indication of a closeness of the machine state to a predetermined normal operation state.

23. The method recited in claim 22, wherein the relative health status comprises an anomaly, and further comprising incorporating at least one of the determined anomaly and the determined fault probability into the ontology of machine states, in order to continue training the neural network for subsequent use.