Development of algorithms for manufacture informatics and its evaluation in steel industry
| Project/Area Number |
19H04176
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | Kyushu University |
|---|
Principal Investigator |
saigo hiroto 九州大学, システム情報科学研究院, 准教授 (90586124)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
齊藤 敬高 九州大学, 工学研究院, 准教授 (80432855)
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Project Status |
Completed (Fiscal Year 2022)
|
| Budget Amount *help |
¥16,120,000 (Direct Cost: ¥12,400,000、Indirect Cost: ¥3,720,000)
Fiscal Year 2022: ¥3,380,000 (Direct Cost: ¥2,600,000、Indirect Cost: ¥780,000)
Fiscal Year 2021: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2020: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2019: ¥5,200,000 (Direct Cost: ¥4,000,000、Indirect Cost: ¥1,200,000)
|
|---|
| Keywords | 機械学習 / 異常検知 / 鉄鋼生産 / 外挿予測 / 転移学習 / ガウス過程 / 多相融体の粘度 / 深層学習 / 鉄鋼製造 / 疎性非線形モデル / 操業データ / 特徴選択 / 回帰 / ベイズ最適化 / CNN / 解釈可能性 |
|---|
| Outline of Research at the Start |
本研究計画では次の3つのテーマに取り組む;【1】操業時の異常検知問題、【2】マルチタスク学習による 高温状態の粘度予測、【3】高コストな問題における実験計画。
このうち、【1】は代表者が単独で取り組むものであるが、【2】と【3】については、分担者が実験で得たデータを基に代表者の研究を進めるものである。このため、数ヶ月に１回程度、あるいは必要に応じてミーティングを行う予定である。
|
| Outline of Final Research Achievements |
In the "Anomaly Detection in Blast Furnaces" problem, we have developed approaches using unsupervised learning based on the work of Itakura et al. (IBIS2022), and supervised learning based on the work of Kizaki (IBIS2021). In the supervised learning approach using CNN, we have confirmed that utilizing data from 5 to 15 minutes prior leads to improved accuracy.
We have also developed a method for "Viscosity Prediction of High-Temperature States through Multi-Task Learning" as described in the study by Saigo et al. (Scientific Reports, 2022). In addition to robust extrapolation prediction, we have proposed a transfer learning method that leverages room temperature experimental data for high-temperature experiments.
|
| Academic Significance and Societal Importance of the Research Achievements |
教師なし学習は人手により教師ラベル作成の労力を減らすことを可能とする。現実世界の多くのデータは教師ラベルがないか、もしくはそのラベル付けに多大なコストが必要な場合が多いため、現実社会での実装において重要なテーマである。
一方で、機械学習手法の多くは過去のデータから学習し、その評価を交差検証に頼っているため、ロバストな外挿予測問題への取り組みは学術的に重要である。本研究では流体力学という現実の問題への解決策を示したものであり、同種の問題に適用可能である。
|
Report
(5 results)
Research Products
(24 results)
-
-
-
-
-
-
-
-
-
[Journal Article] DeepSuccinylSite: a deep learning based approach for protein succinylation site prediction2020
Author(s)
Thapa, N., Chaudhari, M., McManus, S., Roy, K., Newman, R.H., Saigo, H., KC, D.B.
-
Journal Title
BMC Bioinformatics
Volume: 21(Suppl 3)
Related Report
Peer Reviewed / Open Access / Int'l Joint Research
-
[Journal Article] DeepRMethylSite: a deep learning based approach for prediction of arginine methylation sites in proteins2020
Author(s)
Chaudhari, M., Thapa, N., S., Roy, K., Newman, R.H., Saigo, H., KC, D.B.
-
Journal Title
Molecular Omics
Volume: 16
Related Report
Peer Reviewed / Open Access / Int'l Joint Research
-
[Journal Article] RF-MaloSite and DL-Malosite: Methods based on random forest and deep learning to identify malonylation sites2020
Author(s)
Al-barakati, H.J., Thapa, N., Saigo, H., Roy, K., Newman, R.H., Bahadur, K.C.D.
-
Journal Title
Computational and Structural Biotechnology Journal
Volume: 18
Pages: 852-860
DOI
Related Report
Peer Reviewed / Int'l Joint Research
-
-
-
-
-
-
-
-
-
-
[Presentation] DeepSuccinylSite: a deep learning based approach for protein succinylation site prediction2019
Author(s)
Thapa, N., Chaudhari, M., McManus, S., Roy, K., Newman, R.H., Saigo, H., KC, D.B.
Organizer
Joint GIW/ABACBS-2019 Bioinformatics Conference
Related Report
Int'l Joint Research
-
-
-
