Automation method of AI-based diagnostic technology for equipment application

Patent metadata

Field	Value
Publication number	US-11941923-B2
Application number	US-202117508255-A
Country	US
Kind code	B2
Filing date	Oct 22, 2021
Priority date	May 17, 2021
Publication date	Mar 26, 2024
Grant date	Mar 26, 2024

Abstract

An automation method of an artificial intelligence (AI)-based diagnostic technology for equipment application includes receiving one or more pieces of data among vibration data, noise data, and controller area network (CAN) data, a data input processing operation of trimming the input data, an operation of extracting features from the trimmed data, setting a setting value of a hyper-parameter with respect to the one or more pieces of data thereamong, and generating a total of N models to include both of machine learning (ML) and deep learning (DL) as N individual models and generating ensemble prediction model structures for the N individual models. As a parameter updating is being proceeded due to the hyper-parameter so as to minimize values of cost functions of the N individual models, a reward for model accuracy performance is optimized and the ensemble prediction model structures of the N individual models change.

First claim

What is claimed is:

1. An automation method of an artificial intelligence (AI)-based diagnostic technology for equipment application, the automation method comprising:
receiving one or more pieces of data inputted from among vibration data, noise data, and controller area network (CAN) data, which are collected from a rotating body in a vehicle;
a data input processing operation of trimming the input one or more pieces of data;
an operation of extracting features from the trimmed one or more pieces of data;
setting a setting value of a hyper-parameter with respect to the input one or more pieces of data among the vibration data, the noise data, and the CAN data; and
generating a total of N models to include both of machine learning (ML) and deep learning (DL) as N individual models and generating ensemble prediction model structures with respect to the N individual models,
wherein, as a parameter updating is being proceeded due to the hyper-parameter so as to minimize values of cost functions of the N individual models, a reward with respect to model accuracy performance is optimized and the ensemble prediction model structures of the N individual models change.

2. The automation method of claim 1 , wherein, in the data input processing operation, the input one or more pieces of data is trimmed according to a problem frequency band and a data time length.

3. The automation method of claim 2 , wherein the trimmed one or more pieces of data is classified into a training dataset, a validation dataset, and a test dataset.

4. The automation method of claim 3 , wherein, in the operation of extracting, one algorithm or two or more algorithms for extraction of independent features are used according to a classification performance determination index, and an ensemble prediction model is selectively additionally applied.

5. The automation method of claim 4 , wherein:
when the two or more algorithms for extraction of the independent features are used, each of the two or more algorithms for extraction of the independent features has a weight value of 1:1; and
when the ensemble prediction model is selectively additionally applied, a sum of the weight values is one.

6. The automation method of claim 1 , wherein:
optimizing the hyper-parameter is performed by a grid search, a random search, or a random Latin hypercube automation algorithm; and
as the hyper-parameter is updated, an Auto ML/DL model structure is optimized.

7. The automation method of claim 6 , wherein, when the Auto ML/DL model structure, to which a final hyper-parameter is applied, is optimized, model verification is performed using a validation dataset, and evaluation of a final model is performed using a test dataset.

8. The automation method of claim 7 , wherein cost functions of the N individual models are confirmed, and then a robust model configuration is obtained by applying the cost functions of the N individual models to the ensemble prediction model structures, respectively.

9. The automation method of claim 8 , wherein a weight value is assigned to an individual cost function constituting a corresponding one of the cost functions applied to the ensemble prediction model structures.

10. The automation method of claim 1 , wherein the rotating body is a rotating body for power generation or power transmission.

11. An equipment to which the automation method of an artificial intelligence (AI)-based diagnostic technology for equipment application according to claim 1 is applied.

Assignees

• Hyundai Motor Co Ltd
• Kia Corp

Inventors

• Jung In-Soo
• Lee Seung-Hyun
• Jin Jae-Min
• Lee Dong-Chul

Classifications

• G07C5/0808 Primary

  Diagnosing performance data (testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005) · CPC title

• G06N3/042

  Knowledge-based neural networks; Logical representations of neural networks · CPC title

• H04L12/40 Primary

  Bus networks · CPC title

• H04L2012/40215

  Controller Area Network CAN · CPC title

• H04L2012/40273

  the transportation system being a vehicle · CPC title