Title:
PREDICTION MODEL TRAINING METHOD, COMPOSITE STRUCTURE EVALUATION METHOD, COMPOSITE STRUCTURE MANUFACTURING METHOD, AND COMPOSITE STRUCTURE EVALUATION DEVICE
Document Type and Number:
WIPO Patent Application WO/2022/158447
Kind Code:
A1
Abstract:
This prediction model training method uses a training device to train a prediction model for predicting a physical quantity in a target region when a load is added to a composite structure obtained by combining a plurality of members. The training method causes the training device to execute: a step for acquiring, as teacher data, a stress distribution or a strain distribution for each member for the target region, and the physical quantity of the target region corresponding to the stress distribution or the strain distribution; and a step for training the prediction model such that the physical quantity of the teacher data becomes the output when the stress distribution or the strain distribution of the teacher data is the input.
Inventors:
OKA KOSUKE (JP)
Application Number:
PCT/JP2022/001600
Publication Date:
July 28, 2022
Filing Date:
January 18, 2022
Export Citation:
Assignee:
MITSUBISHI HEAVY IND LTD (JP)
International Classes:
G06F30/23; G06F30/27
Foreign References:
| JP2019128815A | 2019-08-01 | |||
| JP2004110793A | 2004-04-08 | |||
| JP2019106188A | 2019-06-27 | |||
| JP2019032768A | 2019-02-28 |
Attorney, Agent or Firm:
SAKAI INTERNATIONAL PATENT OFFICE (JP)
Download PDF: