Quantifying a condition with a neural network

Patent metadata

Field	Value
Publication number	US-9239983-B1
Application number	US-201213434072-A
Country	US
Kind code	B1
Filing date	Mar 29, 2012
Priority date	Sep 2, 2011
Publication date	Jan 19, 2016
Grant date	Jan 19, 2016

Abstract

Quantification of a condition of a selected item using a neural network is disclosed. A device defines a good hypertube in a neural state space based on good item state points obtained from one or more items that exhibit desired operating characteristics, and a bad hypertube in the neural state space based on bad item state points obtained from one or more items that exhibit undesirable operating characteristics. A current item state hyperpoint is determined in the neural state space based on a current item state point of the selected item. A condition of the selected item is quantified as a function of a location of the current item state hyperpoint with respect to at least a portion of the good hypertube and with respect to at least a portion of the bad hypertube.

First claim

What is claimed is:

1. A method for determining a condition of a selected item, comprising:
defining, by a computing device comprising a processor, a good hypertube in a neural state space, the good hypertube learned based on a plurality of good item state points obtained from at least one good item that exhibits desired characteristics, wherein the good hypertube comprises a plurality of good hypercubes, each good hypercube based on at least one good item state hyperpoint, the at least one good item state hyperpoint based on a good item state point of the plurality of good item state points;
defining, by the computing device, a bad hypertube in the neural state space, the bad hypertube learned based on a plurality of bad item state points obtained from at least one bad item that exhibits undesirable characteristics, wherein the bad hypertube comprises a plurality of bad hypercubes, each bad hypercube based on at least one bad item state hyperpoint, the at least one bad item state hyperpoint based on a bad item state point of the plurality of bad item state points, and wherein the good hypertube and the bad hypertube occupy less than a totality of the neural state space;
determining, by the computing device, a current item state hyperpoint in the neural state space based on a current item state point of the selected item; and
quantifying the condition of the selected item as a function of a location of the current item state hyperpoint with respect to at least a portion of the good hypertube and with respect to at least a portion of the bad hypertube.

2. The method of claim 1 , wherein quantifying the condition of the selected item as the function of the location of the current item state hyperpoint with respect to the at least the portion of the good hypertube and with respect to the at least the portion of the bad hypertube comprises quantifying the condition of the selected item as the function of the location of the current item state hyperpoint with respect to a first set of good hypercubes and with respect to a second set of bad hypercubes.

3. The method of claim 2 , wherein the first set of good hypercubes comprises each good hypercube in the plurality of good hypercubes, and wherein the second set of bad hypercubes comprises each bad hypercube in the plurality of bad hypercubes.

4. The method of claim 2 , wherein the first set of good hypercubes comprises each good hypercube within a hyperradius of the location, and wherein the second set of bad hypercubes comprises each bad hypercube within the hyperradius of the location.

5. The method of claim 2 , wherein quantifying the condition of the selected item as the function of the location of the current item state hyperpoint with respect to the first set of good hypercubes and with respect to the second set of bad hypercubes comprises:
determining an aggregate good resonance of the current item state hyperpoint with respect to the first set of good hypercubes;
determining an aggregate bad resonance of the current item state hyperpoint with respect to the second set of bad hypercubes; and
determining a difference resonance between the aggregate good resonance and the aggregate bad resonance, the difference resonance quantifying the condition of the selected item.

6. The method of claim 5 , further comprising:
determining which particular resonance range of a plurality of resonance ranges the difference resonance falls within, each resonance range associated with a different condition; and
indicating the condition of the selected item based on the particular resonance range.

7. The method of claim 1 , wherein the plurality of good item state points is obtained from a plurality of good items that exhibit the desired characteristics over a period of time, and wherein the plurality of bad item state points is obtained from a plurality of bad items that exhibit the undesirable characteristics over the period of time.

8. The method of claim 1 , wherein the good item state point represents a state of the at least one good item at a particular instant in time.

9. The method of claim 8 , wherein the good item state point comprises at least one sensed attribute value.

10. The method of claim 9 , wherein the good item state point comprises a plurality of sensed attribute values, and wherein the at least one good item state hyperpoint comprises a multi-dimensional vector, each dimension of the multi-dimensional vector corresponding to a particular sensed attribute value of the plurality of sensed attribute values.

11. The method of claim 1 , further comprising providing a visual indicator of the condition of the selected item.

12. The method of claim 1 , wherein each of the plurality of good item state points is based on a plurality of sensed attribute values obtained from the at least one good item, each sensed attribute value quantifying one sensed attribute of a plurality of sensed attributes, and each of the plurality of bad item state points is based on a plurality of sensed attribute values obtained from the at least one bad item, each sensed attribute value quantifying one sensed attribute of the plurality of sensed attributes, wherein determining the current item state hyperpoint comprises:
receiving, from a plurality of different sensors that sense the plurality of sensed attributes of the selected item, a plurality of sensed attribute values obtained from the selected item; and
determining the current item state hyperpoint based on the plurality of sensed attribute values of the selected item.

13. A device, comprising:
a memory; and
a processor coupled to the memory, the processor configured to:
define a good hypertube in a neural state space, the good hypertube learned based on a plurality of good item state points obtained from at least one good item that exhibits desired characteristics, wherein the good hypertube comprises a plurality of good hypercubes, each good hypercube based on at least one good item state hyperpoint, the at least one good item state hyperpoint based on a good item state point of the plurality of good item state points;
define a bad hypertube in the neural state space, the bad hypertube learned based on a plurality of bad item state points obtained from at least one bad item that exhibits undesirable characteristics, wherein the bad hypertube comprises a plurality of bad hypercubes, each bad hypercube based on at least one bad item state hyperpoint, the at least one bad item state hyperpoint based on a bad item state point of the plurality of bad item state points, and wherein the good hypertube and the bad hypertube occupy less than a totality of the neural state space;
determine a current item state hyperpoint in the neural state space based on a current item state point of the selected item; and
quantify a condition of the selected item as a function of a location of the current item state hyperpoint with respect to at least a portion of the good hypertube and with respect to at least a portion of the bad hypertube.

14. The device of claim 13 , wherein to quantify the condition of the selected item as the function of the location of the current item state hyperpoint with respect to the at least the portion of the good hypertube and with respect to the at least the portion of the bad hypertube, the processor is further configured to:
determine a first resonance associated with a set of good hypercubes in proximity to the location;
determine a second resonance associated with a set of bad hypercubes in proximity to the location; and
quantify the condition of the selected item based on a function of the first resonance and the second resonance.

Assignees

• Harrison Gregory A
• Das Sreerupa
• Bodkin Michael A
• Hall Richard M
• Worden Eric W
• Herzog Jr Stefan
• Lockheed Corp

Inventors

• Harrison Gregory A
• Das Sreerupa
• Bodkin Michael A
• Hall Richard M
• Worden Eric W
• Herzog Jr Stefan

Classifications

• G06N3/043 Primary

  based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS] · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/04 Primary

  Architecture, e.g. interconnection topology · CPC title