Discrimination Analysis Used with Optical Computing Devices

• US 20150205000A1
• Filed: 06/07/2013
• Published: 07/23/2015
• Est. Priority Date: 06/07/2013
• Status: Active Grant

• Current Assignee: Halliburton Energy Services Incorporated (Halliburton Company)
• Original Assignee: Halliburton Energy Services Incorporated (Halliburton Company)
• Inventors: Perkins, David L., Chen, Dingding, Jones, Christopher Michael, Shen, Jing
• Granted Patent: US 10,684,388 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G01N 21/31 Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry G01N21/72 takes precedence
• CPC Class Codes: G01N 2201/1296 using neural networks
• CPC Class Codes: G01V 8/10 Detecting, e.g. by using light barriers by reflection from the object G01S17/00

Abstract

Disclosed are systems and methods that use discriminant analysis techniques and processing in order to reduce the time required to determine chemical and/or physical properties of a substance. One method includes optically interacting a plurality of optical elements with one or more known substances, each optical element being configured to detect a particular characteristic of the one or more known substances, generating an optical response from each optical element corresponding to each known substance, wherein each known substance corresponds to a known spectrum stored in an optical database, and training a neural network to provide a discriminant analysis classification model for an unknown substance, the neural network using each optical response as inputs and one or more fluid types as outputs, and the outputs corresponding to the one or more known substances.

Claims

1. A method, comprising:

optically interacting a plurality of optical elements with one or more known substances, each optical element being configured to detect a particular characteristic of the one or more known substances;

generating an optical response from each optical element corresponding to each known substance, wherein each known substance corresponds to a known spectrum stored in an optical database; and

training a neural network to provide a discriminant analysis classification model for an unknown substance, the neural network using each optical response as inputs and one or more fluid types as outputs, and the outputs corresponding to the one or more known substances.

2. The method of claim 1, wherein at least one of the plurality of optical elements is an integrated computational element.

3. The method of claim 1, further comprising:

optically interacting the plurality of optical elements with an unknown substance, thereby generating an unknown substance optical response;

applying the unknown substance optical response to the classification model; and

outputting a predicted substance type corresponding to the unknown substance.

4. The method of claim 3, wherein the predicted substance type corresponds to a highest probability as calculated by the classification model.

5. The method of claim 3, wherein the predicted substance type corresponds to a mathematical linear combination of at least two substance types of the one or more known substances, the mathematical linear combination providing a new fluid type.

6. The method of claim 5, further comprising:

adding the new fluid type to the one or more known substances such that the new fluid type becomes one of the one or more known substances;

generating a new optical response from each optical element corresponding to each known substance; and

training a new neural network using each new optical response as input and thereby providing a new classification model.

7. The method of claim 3, further comprising:

generating at least one quantitative model corresponding to at least one of the one or more known substances;

using the predicted substance type corresponding to the unknown substance to select the at least one quantitative model;

applying the selected at least one quantitative model to provide an unknown substance property prediction; and

validating the unknown substance property prediction by referencing the optical database.

8. The method of claim 1, wherein optically interacting the plurality of optical elements with the one or more known substances comprises generating a plurality of classification outputs that compare a prediction from the classification model with a random noise corrupted target value corresponding to each of the one or more outputs.

9. The method of claim 1, further comprising quantifying with the one or more outputs a concentration of the unknown substance with reference to the one or more known substances.

10. A method, comprising:

training a neural network having one or more outputs corresponding to one or more known substances, thereby generating a multivariate classification model for discriminant analysis of an unknown substance;

introducing into the multivariate classification model an unknown substance optical response corresponding to the unknown substance; and

outputting a predicted substance type corresponding to the unknown substance from the one or more outputs.

11. The method of claim 10, wherein training the neural network comprises:

optically interacting a plurality of optical elements with the one or more known substances, each optical element being configured to detect a particular characteristic of the one or more known substances;

generating an optical response from each optical element corresponding to each known substance, wherein each known substance corresponds to a known spectrum stored in an optical database; and

applying each optical response to the neural network.

12. The method of claim 11, wherein at least one of the plurality of optical elements is an integrated computational element.

13. The method of claim 11, wherein introducing into the multivariate classification model the unknown substance optical response comprises:

optically interacting the plurality of optical elements with the unknown substance, thereby generating the unknown substance optical response; and

applying the unknown substance optical response to the multivariate classification model.

14. The method of claim 10, wherein the predicted substance type corresponds to a highest probability as calculated by the multivariate classification model.

15. The method of claim 10, wherein the predicted substance type corresponds to a mathematical linear combination of at least two substance types of the one or more known substances, the mathematical linear combination providing a new fluid type.

16. The method of claim 15, further comprising:

adding the new fluid type to the one or more known substances such that the new fluid type becomes one of the one or more known substances;

generating a new optical response from each optical element corresponding to each known substance; and

inputting each new optical response into the multivariate classification model, thereby providing a new multivariate classification model.

17. The method of claim 10, further comprising classifying the predicted substance type with respect to at least one of the one or more known substances.

18. The method of claim 10, further comprising quantifying with the one or more outputs a concentration of the unknown substance with reference to the one or more known substances.