Research on the development of a machine learning systems that simultaneously improve operator skills and inspection accuracy.

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K13471
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 07080:Business administration-related
• Research Institution: Ibaraki University
• Principal Investigator: HARAGUCHI HARUMI 茨城大学, 理工学研究科(工学野), 講師 (70796325)
• Project Period (FY): 2022-04-01 – 2024-03-31
• Keywords: 検品支援 / 作業者訓練 / 機械学習 / 製造業 / 品質管理

Outline of Final Research Achievements

The objective of this research is to develop a system that simultaneously improves the inspection accuracy of visual inspection by humans and automatic inspection by machines.The system is designed to improve the inspection accuracy of drill-like instruments used in dentistry.The completed system consists of three tools: (1) a labeling tool, (2) an inspection training tool, and (3) an automatic identification tool.The system operates in the following steps: (1) The products to be inspected are classified as quality or defective by the labeling tool. Based on the criteria, the operators are trained to inspect the products using (2) the inspection training tool.The classification model for the (3) automatic identification tool, incorporated in the automatic inspection machine, is also created.An improvement in identification accuracy was confirmed by reflecting the results obtained with the (2) inspection training tool in the (3) automatic identification tool.

Free Research Field

経営工学

Academic Significance and Societal Importance of the Research Achievements

本研究の成果は主に2つに要約される．ひとつ目が実現場に即した作業者による目視検品と機械による自動検品の双方を統合した検品システムのモデルを考案したこと．ふたつ目が目視検品の訓練に使用する検品訓練ツールで得られた目視判断の傾向を自動検品用の判別モデル作成に適用することによって，従来判別モデルを作成しづらいと言われてきた特徴量を抽出しづらいうえ不良品の絶対数が少ない工業製品に対して精度の高い判別モデルの作成を実現したことである．