System and method for advanced condition monitoring of an asset system

• US 7,756,678 B2
• Filed: 05/29/2008
• Issued: 07/13/2010
• Est. Priority Date: 05/29/2008
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Hu, Xiao, Taware, Avinash Vinayak, Mitchell, Robert James Jr., Subbu, Rajesh Venkat Jr., Bonissone, Piero Patrone, Hershey, John Erik
• Primary Examiner(s): TSAI, CAROL S W
• Application Number:
  US12/129,632
• Publication Number:
  US 20090300417A1
• Time in Patent Office: 775 Days
• Field of Search: 702/182, 702/183, 702/56, 702/190, 702/122, 702/187, 702/185, 702/184, 700/26, 700 28- 31, 700 48- 50, 700/109, 700/104, 700/110, 700/108, 706/6, 706/15, 706/23, 714/37, 714 46- 48, 714/26
• US Class Current: 702/182
• CPC Class Codes: G05B 2219/33295 Fuzzy expert system for diagnostic, monitoring
• CPC Class Codes: G05B 2219/33296 ANN for diagnostic, monitoring
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles

Abstract

A method for advanced condition monitoring of an asset system includes using a plurality of auto-associative neural networks to determine estimates of actual values sensed by at least one sensor in at least one of the plurality of operating regimes; determining a residual between the estimated sensed values and the actual values sensed by the at least one sensor from each of the plurality of auto-associative neural networks; and combining the residuals by using a fuzzy supervisory model blender; performing a fault diagnostic on the combined residuals; and determining a change of the operation of the asset system by analysis of the combined residuals. An alert is provided if necessary. A smart sensor system includes an on-board processing unit for performing the method of the invention.

Claims

1. A method for advanced condition monitoring of an asset system, the method comprising the steps of:

segmenting an operating space of an asset system into a plurality of operating regimes;

using a plurality of auto-associative neural networks (AANNs) to determine an estimated value of an actual value sensed by the at least one sensor in a plurality of operating regimes;

determining a plurality of residuals based on a difference between the estimated value and the actual value from each of the plurality of auto-associative neural networks;

using a fuzzy supervisory model blender to generate a weighting factor used in determining an integrated overall residual from the plurality of residuals;

determining a change of an operating condition of the asset system by analysis of the integrated overall residual; and

providing an alert if a change of the operation of the asset system has been determined.

2. The method of claim 1, wherein the AANN comprises hidden nodes of the tan-sigmoid (tansig) type.

3. The method of claim 1, wherein the AANN comprises a central bottleneck layer with embedded linear transformation functions.

4. The method of claim 1, wherein the alerting step comprises distributing a message via an information transport medium selected from the group consisting of an electrical cable, power line conduction, an intranet, the Internet, and a wireless transmission.

5. The method of claim 1, wherein the step of determining the change in the operating condition is performed when the asset system is transiting between two of the plurality of operating regimes.

6. The method of claim 1, wherein the alerting step is performed when the asset system is transiting between two of the plurality of operating regimes.

7. The method of claim 1, wherein the step of determining the change in the operation condition is performed when the asset system is operating in one of the plurality of operating regimes.

8. The method of claim 1, wherein the alerting step is performed when the asset system is operating in one of the plurality of operating regimes.

9. The method of claim 1, wherein the fuzzy supervisory model blender implements a global model over the plurality of operating regimes of the asset system.

10. The method of claim 1, wherein the at least one sensor comprises a smart sensor.

11. The method of claim 10, wherein the smart sensor is connected to a smart sensor system.

12. The method of claim 11, wherein the smart sensor system comprises an on-board processing unit for advanced condition monitoring of the asset system.

13. A smart sensor system, comprising:

a smart sensor for monitoring an operating condition of an asset system, the smart sensor being connected to an input port via a cable; and

an on-board processing unit for advanced condition monitoring of the asset system using the method of claim 1.

14. The system of claim 13, further comprising a wireless communications unit for wirelessly transmitting signals from the smart sensor system to a peripheral processing device.

15. The system of claim 13, further comprising a display and alarm unit for displaying computational results from the processing unit.

16. The method of claim 1, wherein the integrated overall residual is determined using a weighted average of the plurality of residuals.