Neural network demodulation for an optical sensor

• US 7,603,004 B2
• Filed: 06/12/2007
• Issued: 10/13/2009
• Est. Priority Date: 06/12/2006
• Status: Active Grant

• Current Assignee: Curators of the University of Missouri (University of Missouri)
• Original Assignee: Curators of the University of Missouri (University of Missouri)
• Inventors: Dua, Rohit, Watkins, Steve E., Wunsch, Donald C.
• Primary Examiner(s): Wong; Tina M
• Assistant Examiner(s): Chu; Chris H
• Application Number:
  US11/761,814
• Publication Number:
  US 20070297714A1
• Time in Patent Office: 854 Days
• Field of Search: None
• US Class Current: 385/12
• CPC Class Codes: G01L 1/24 by measuring variations of optical properties of material when it is stressed, e.g. by photoelastic stress analysis using infrared, visible light, ultraviolet
• CPC Class Codes: G01M 11/08 Testing mechanical properties G01M11/005 takes precedence
• CPC Class Codes: G01M 5/0041 by determining deflection or stress
• CPC Class Codes: G01M 5/0091 by using electromagnetic excitation or detection

Abstract

Methods and systems of neural network demodulation for an optical sensor. An optical sensor may be coupled to a structure and be capable of reflecting a reflected optical signal. A wavelength of the reflected optical signal may be spread based on a strain being applied to the structure. A replication device may receive the reflected optical signal from the optical sensor and produce a plurality of optical signals. A filter may be coupled to the replication device to receive an optical signal from the plurality of optical signals and filter the received optical signal. A detector may receive the filtered optical signal and provide a voltage output proportional to an amount of the filtered optical signal received. A neural network may receive the voltage output and determine the strain on the structure.

Claims

1. A system comprising:

an optical sensor coupled to a structure, the optical sensor to receive a source optical signal and reflect a reflected optical signal, at least one wavelength of the reflected optical signal being spread based on a strain being applied to the structure;

a replication device coupled to the optical sensor, the replication device to receive the reflected optical signal from the optical sensor and produce a plurality of optical signals;

at least one filter coupled to the replication device to receive an optical signal from the plurality of optical signals and filter the received optical signal;

at least one detector, coupled to the at least one filter, to receive the filtered optical signal and provide an output signal representing an amount of the filtered optical signal received; and

a neural network coupled to the at least one detector, the neural network capable of learning and performing pattern recognition, the neural network to receive the output signal from the at least one detector and determine the strain on the structure based on the learning and the performing of the pattern recognition and the output signal.

2. The system of claim 1, further comprising:

an analysis subsystem to receive a further signal from the neural network for further processing and analysis.

3. The system of claim 1, wherein the optical sensor includes at least one of a fiber optical sensor or Bragg optical grating sensor.

4. The system of claim 1, wherein the number of filters of the at least one filter is equal to a number of detectors of the at least one detector.

5. The system of claim 1, wherein a number of filters of the at least one filter is one less than a number of detectors of the at least one detector.

6. The system of claim 1, wherein the neural network has characteristics of the at least one filter and uses the characteristics when determining the strain on the structure.

7. The system of claim 1, wherein the at least one filter includes a plurality of filters and the at least one detector includes a plurality of detectors, at least some filters of the plurality of filters having a different frequency range, at least some detectors of the plurality of detectors having a different output signal reflecting the different frequency range of a coupled filter, and the neural network to receive a plurality of different output signals from the plurality of detectors and determine the strain on the structure based on the learning and the performing of the pattern recognition and the different output signals.

8. The system of claim 1, wherein the at least one filter includes at least three filters, a first filter of the at least three filters having a first filter profile that is monotonically non-decreasing with a particular wave length of the at least one wavelength, a second filter of the at least three filters having a second filter profile that is monotonically non-increasing with the particular wavelength, and a third filter of the at least three filters having a third filter profile that is approximately centered with a zero strain point of the optical sensor.

9. The system of claim 1, wherein the neural network includes a plurality of interunit connection strengths, and the learning and the performing of the pattern recognition is at least performed by the plurality of interunit connection strengths.

10. The system of claim 1, wherein the neural network generates functional mapping from non-linear learning from data, and wherein capability of the learning and the performing of the pattern recognition is based on the functional mapping.

11. The system of claim 1, wherein at least some of characteristics of the at least one filter are not sufficiently known to have a functional mapping.

12. the system of claim 1, wherein the one or more filters include a low profile filter and a high profile filter, the low profile filter to provide a nearly linear change in a first linearly changed output signal to the neural network and the high profile filter to provide the nearly linear change in a second linearly changed output signal to the neural network, wherein the neural network receives the first linearly changed output signal and the second linearly changed output signal and determines the strain on the structure based on the learning, the performing of the pattern recognition, and the output signal.

13. the system of claim 1, wherein the at least one detector includes a proportional integration spectrum detector, the proportional integration spectrum detector to output a proportional output integration signal that is proportional to the integration of a spectrum of the reflected optical signal multiplied by a respective filter of the at least one filter, and wherein the neural network determines the strain on the structure based on the learning and the performing of the pattern recognition and the proportional output integration signal.

14. A method comprising:

receiving a reflected optical signal from an optical sensor, the optical sensor coupled to a structure, at least one wavelength of the reflected optical signal being spread based on a strain being applied to the structure;

replicating the reflected optical signal to produce a plurality of optical signals;

filtering at least one of the plurality of optical signals to produce at least one filtered optical signal;

detecting the at least one filtered optical signal to determine an output signal representing an amount of the filtered optical signal received; and

using a neural network to perform a neural network analysis on the output signal to determine an amount of strain applied to the structure, the neural network capable of learning and performing pattern recognition.

15. The method of claim 14, wherein detecting further comprises:

detecting an unfiltered optical signal from the plurality of optical signals to determine the voltage output proportional to an amount of the unfiltered optical signal.

16. The method of claim 14, wherein performing the neural network analysis comprises:

preprocessing the output signal; and

performing the neural network analysis on the preprocessed output to determine the amount of strain applied to the structure.

17. The method of claim 16, wherein preprocessing the output signal comprises at least one of scaling the output signal or normalizing the output signal.

18. The method of claim 14, wherein the filtering is performed by a plurality of filters having a different frequency range, the detecting is performed by a plurality of detectors having a different output signal reflecting the different frequency range of a coupled filter, and the neural network to perform the neural network analysis on a plurality of different output signals from the plurality of detectors to determine the strain applied to the structure.

19. A method comprising:

receiving a reflected optical signal from an optical sensor, the optical sensor coupled to a structure, at least one wavelength of the reflected optical signal being spread based on a strain being applied to the structure;

decomposing the reflected optical signal to produce a plurality of different components of the reflected optical signal;

filtering the plurality of different components to produce a plurality of filtered components;

detecting the plurality of filtered components to determine an output signal representing an amount of at least one of the plurality of filtered components; and

using a neural network to perform a neural network analysis on the output signal to determine an amount of strain applied to the structure, the neural network capable of learning and performing pattern recognition.

20. The method of claim 19, further comprising:

performing further processing and analysis on the amount of strain applied to the structure.

21. The method of claim 19, wherein the at least one wavelength of the reflected optical signal being spread based on the strain is a center wavelength with a linear shift.

22. The method of claim 19, wherein the filtering is performed by a plurality of filters having a different frequency range, the detecting is performed by a plurality of detectors having a different output signal reflecting the different frequency range of a coupled filter, and the neural network to perform the neural network analysis on a plurality of different output signals from the plurality of detectors to determine the strain applied to the structure.