Fan behavior anomaly detection using neural network

Patent metadata

Field	Value
Publication number	US-12396126-B2
Application number	US-202117554495-A
Country	US
Kind code	B2
Filing date	Dec 17, 2021
Priority date	Dec 17, 2021
Publication date	Aug 19, 2025
Grant date	Aug 19, 2025

Abstract

Generally discussed herein are devices, systems, and methods for predicting fan failure. A method includes providing, to a first NN that models nominal behavior of a first fan and provides a value indicating a first amount of deviation from nominal as an output, first parameters of fan operation of the first fan, receiving, from the first NN and responsive to the first parameters, first data indicating the first amount of deviation from nominal, providing, to a second NN that models nominal behavior of a second fan and provides a value indicating a second amount of deviation from nominal as an output, second parameters of fan operation of the second fan, receiving, from the second NN and responsive to the first parameters, second data indicating the second amount of deviation from nominal, and estimating that the first or second fan has failed by comparing the first and second data.

First claim

What is claimed is:

1. A method comprising:
providing, to a first neural network (NN) trained to model nominal behavior of a first fan and provide a value indicating a first amount of deviation from nominal as an output, first parameters of fan operation of the first fan;
receiving, from the first NN and responsive to the first parameters, first data indicating the first amount of deviation from nominal;
providing, to a second NN trained to model nominal behavior of a second fan and provide a value indicating a second amount of deviation from nominal as an output, second parameters of fan operation of the second fan;
receiving, from the second MN and responsive to the first parameters, second data indicating the second amount of deviation from nominal; and
estimating that the first or second fan has failed or will fail based on a comparison of the first and second data.

2. The method of claim 1 , further comprising:
receiving, from a fan system including the first and second fans, a message; and
deriving the first and second parameters based on data in the message.

3. The method of claim 2 , wherein the data in the message includes two or more of time, ambient temperature about a corresponding fan, motor temperature of a motor of the corresponding fan, and rotations per unit time of the motor of the corresponding fan.

4. The method of claim 3 , wherein deriving the first and second parameters includes determining a time difference between the message and a second time in an immediately previous message and providing the time difference as one of the first or second parameters.

5. The method of claim 3 , wherein deriving the first and second parameters includes determining a difference between the rotations per unit time and a second rotations per unit time in an immediately previous message resulting in a different in rotations and providing the rotations per unit time and the difference in rotations as one of the first or second parameters.

6. The method of claim 3 , wherein deriving the first and second parameters include providing the ambient temperature from the message and a motor temperature from an immediately previous message as parameters of the first or second parameters.

7. The method of claim 3 , wherein the output of the first and second NNs is an estimate of a motor temperature of the respective fan and the first and second deviation from nominal are determined based on a corresponding motor temperature from the message and the estimate of the motor temperature.

8. A non-transitory machine-readable medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:
providing, to a first neural network (NN) trained to model nominal behavior of a first fan and provide a value indicating a first amount of deviation from nominal as an output, first parameters of fan operation of the first fan;
receiving, from the first NN and responsive to the first parameters, first data indicating the first amount of deviation from nominal;
providing, to a second NN trained to model nominal behavior of a second fan and provide a value indicating a second amount of deviation from nominal as an output, second parameters of fan operation of the second fan;
receiving, from the second NN and responsive to the first parameters, second data indicating the second amount of deviation from nominal; and
estimating that the first or second fan has failed or will fail based on a comparison of the first and second data.

9. The non-transitory machine-readable medium of claim 8 , wherein the operations further comprise:
receiving, from a fan system including the first and second fans, a message; and
deriving the first and second parameters based on data in the message.

10. The non-transitory machine-readable medium of claim 9 , wherein the data in the message includes two or more of time, ambient temperature about a corresponding fan, motor temperature of a motor of the corresponding fan, and rotations per unit time of the motor of the corresponding fan.

11. The non-transitory machine-readable medium of claim 10 , wherein deriving the first and second parameters includes determining a time difference between the message and a second time in an immediately previous message and providing the time difference as one of the first or second parameters.

12. The non-transitory machine-readable medium of claim 10 , wherein deriving the first and second parameters includes determining a difference between the rotations per unit time and a second rotations per unit time in an immediately previous message resulting in a different in rotations and providing the rotations per unit time and the difference in rotations as one of the first or second parameters.

13. The non-transitory machine-readable medium of claim 10 , wherein deriving the first and second parameters include providing the ambient temperature from the message and a motor temperature from an immediately previous message as parameters of the first or second parameters.

14. The non-transitory machine-readable medium of claim 10 , wherein the output of the first and second NNs is an estimate of a motor temperature of the respective fan and the first and second deviation from nominal are determined based on a corresponding motor temperature from the message and the estimate of the motor temperature.

15. A system comprising:
processing circuitry;
a memory coupled to the processing circuitry, the memory including instructions that, when executed by the processing circuitry, cause the processing circuitry to perform operations comprising:
providing, to a first neural network (NN) trained to model nominal behavior of a first fan and provide a value indicating a first amount of deviation from nominal as an output, first parameters of fan operation of the first fan;

receiving, from the first NN and responsive to the first parameters, first data indicating the first amount of deviation from nominal;
providing, to a second NN trained to model nominal behavior of a second fan and provide a value indicating a second amount of deviation from nominal as an output, second parameters of fan operation of the second fan;
receiving, from the second NN and responsive to the first parameters, second data indicating the second amount of deviation from nominal; and
estimating that the first or second fan has failed or will fail based on a comparison of the first and second data.

16. The system of claim 15 , wherein the operations further comprise:
receiving, from a fan system including the first and second fans, a message; and
deriving the first and second parameters based on data in the message.

17. The system of claim 16 , wherein the data in the message includes two or more of time, ambient temperature about a corresponding fan, motor temperature of a motor of the corresponding fan, and rotations per unit time of the motor of the corresponding fan.

18. The system of claim 17 , wherein deriving the first and second parameters includes determining a time difference between the message and a second time in an immediately previous message and providing the time difference as one of the first or second parameters.

19. The system of claim 17 , wherein deriving the first and second parameters includes determining a difference between the rotations per unit time and a second rotations per unit time in an immediately previous message resulting in a different in rotations and providing the rotations per unit time and the difference in rotations as one of the first or s

Assignees

• Raytheon Co

Inventors

• Warner Eric
• Salpukas Michael R
• Noushin Arjang J
• Tipaldo Daniel J
• Sheahan Daniel D

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/045

  Combinations of networks · CPC title

• F05D2270/709

  with neural networks · CPC title