LEARNING DATA GENERATION METHOD, METHOD FOR ESTIMATING PHYSICAL PROPERTY VALUE OF MATERIAL, LEARNING DATA GENERATION PROGRAM, PROGRAM FOR ESTIMATING PHYSICAL PROPERTY VALUE OF MATERIAL, AND LEARNING DATA GENERATION DEVICE

• Main Authors: NISHIKAWA TOMOHIRO, YAMAMOTO UMI, SHIGEMOTO ISAMU
• Format: Patent
• Language: English; Japanese
• Published: 08.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTEDFOR SPECIFIC APPLICATION FIELDS; PHYSICS

To provide a learning data generation method, a method for estimating the physical property value of materials, a learning data generation program, a program for estimating the physical property value of materials, and a learning data generation device with which it is possible to suppress a decrease in accuracy when estimating physical property values.SOLUTION: A learning data generation method according to the present invention includes: a grouping step for reading a plurality of datasets from a storage unit, each including a material manufacturing condition and multiple types of physical property values, and grouping the plurality of datasets for each manufacturing condition; a selection step for selecting the physical property value to be corrected and a physical property value that is used as a reference for correction; a statistical quantity calculation step for calculating the statistical quantity of the physical property value; a correction model generation step for generating a correction model using the statistical quantity of the physical property value; a correction step for correcting the statistical quantity of the physical property value to be corrected, by using the correction model; and a generation step for generating learning data using the datasets after correction.SELECTED DRAWING: Figure 1 【課題】物性値を推定する際の精度低下を抑制することができる学習用データ生成方法、材料の物性値推定方法、学習用データ生成プログラム、材料の物性値推定プログラムおよび学習用データ生成装置を提供すること。【解決手段】本発明に係る学習用データの生成方法は、材料の製造条件、および複数種の物性値をそれぞれ含む複数のデータセットを記憶部から読み出し、該複数のデータセットを製造条件ごとにグループ化するグループ化ステップと、補正対象の物性値、および、補正の基準とする基準用物性値を選択する選択ステップと、物性値の統計量を算出する統計量算出ステップと、基準用物性値の統計量を用いて補正モデルを生成する補正モデル生成ステップと、補正モデルを用いて補正対象の物性値の統計量を補正する補正ステップと、補正後のデータセットを用いて学習用データを生成する生成ステップと、を含む。【選択図】図１