IP Library › Granted Patent US 11,500,370
Granted Patent B2
US 11,500,370 · App. 16/547,374 · Granted Nov 15, 2022

System for predictive maintenance using generative adversarial networks for failure prediction

Inventors: Shuai Zheng (San Jose, CA); Ahmed Khairy Farahat (Santa Clara, CA); Chetan Gupta (San Mateo, CA)
Assignee: HITACHI, LTD.
G05B23/0283G05B23/0275G06K9/6259G06N3/08
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,500,370
App. No.
16/547,374
Granted
Nov 15, 2022
Kind
B2
Abstract

Example implementations involve a system for Predictive Maintenance using Generative Adversarial Networks for Failure Prediction. Through utilizing three processes concurrently and training them iteratively with data-label pairs, example implementations described herein can thereby generate a more accurate predictive maintenance model than that of the related art. Example implementations further involve shared networks so that the three processes can be trained concurrently while sharing parameters with each other.

Claims (24)

1. A method for training a predictive maintenance model, comprising:

executing a first process configured to generate sample data from real time sensor data and noise that is trained against a first loss function in comparison to the real time sensor data;

executing a second process configured to generate labels from the real time sensor data that is trained against a second loss function in comparison to actual labels;

executing a third process configured to generate labels for the generated sample data using a same network as the second process that is trained against a third loss function in comparison to the real time sensor data and the actual labels;

generating the predictive maintenance model derived from a convergence of the first process, the second process, and the third process; and

performing predictive maintenance based on the generated predictive maintenance model.

2. The method of claim 1 , further comprising applying a shared layer between parameters of the first process and the second process.

3. The method of claim 1 , wherein the first process and the third process are general adversarial networks (GAN).

4. The method of claim 1 , wherein the first process, the second process, and the third process are repeatedly iterated together until the convergence is reached.

5. The method of claim 4 , wherein the convergence is determined from minimizing the first loss function, the second loss function, and the third loss function.

6. The method of claim 1 , wherein the first process is configured to generate the sample data from a latent code vector and a noise vector.

7. The method of claim 1 , wherein the predictive maintenance model is configured to provide one or more of failure prediction, failure isolation, and failure detection from an input of data.

8. A non-transitory computer readable medium storing instructions for training a predictive maintenance model, the instructions comprising:

executing a first process configured to generate sample data from real time sensor data and noise that is trained against a first loss function in comparison to the real time sensor data;

executing a second process configured to generate labels from the real time sensor data that is trained against a second loss function in comparison to actual labels;

executing a third process configured to generate labels for the generated sample data using a same network as the second process that is trained against a third loss function in comparison to the real time sensor data and the actual labels;

generating the predictive maintenance model derived from a convergence of the first process, the second process, and the third process; and

performing predictive maintenance based on the generated predictive maintenance model.

9. The non-transitory computer readable medium of claim 8 , further comprising applying a shared layer between parameters of the first process and the second process.

10. The non-transitory computer readable medium of claim 8 , wherein the first process and the third process are general adversarial networks (GAN).

11. The non-transitory computer readable medium of claim 8 , wherein the first process, the second process, and the third process are repeatedly iterated together until the convergence is reached.

12. The non-transitory computer readable medium of claim 11 , wherein the convergence is determined from minimizing the first loss function, the second loss function, and the third loss function.

13. The non-transitory computer readable medium of claim 8 , wherein the first process is configured to generate the sample data from a latent code vector, and a noise vector.

14. The non-transitory computer readable medium of claim 8 , wherein the predictive maintenance model is configured to provide one or more of failure prediction, failure isolation, and failure detection from an input of data.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Aug 21, 2019
From: ZHENG, SHUAI; FARAHAT, AHMED KHAIRY; GUPTA, CHETAN
To: HITACHI, LTD.
Reel/Frame 050123/0282 →
Continuity (1)
Related Publication 20210055719A1 · Feb 25, 2021