Two-phase cost-sensitive-learning-based framework on customer-side quality inspection for TFT-LCD industry

Ming-Sung Shih (), James C. Chen (), Tzu-Li Chen () and Ching-Lan Hsu ()
Additional contact information
Ming-Sung Shih: National Tsing Hua University
James C. Chen: National Tsing Hua University
Tzu-Li Chen: National Taiwan University of Science and Technology
Ching-Lan Hsu: National Tsing Hua University

Journal of Intelligent Manufacturing, 2025, vol. 36, issue 6, No 27, 4267 pages

Abstract: Abstract The Covid-19 outbreak in 2020 boosted the stay-at-home economy, causing a surge in electronics industry demand, especially benefiting the LCD panel sector. However, as the pandemic situation improved, countries revised policies, leading to the gradual discontinuation of remote work arrangements in various industries. This resulted in declining dividends for the stay-at-home economy. The decreased demand created intense competition within the TFT-LCD industry, urging panel companies to prioritize product quality enhancement to meet customer expectations. Panel quality inspection heavily relied on manual labor, causing varying inspection levels due to subjective judgments. Understanding and aligning with customer expectations regarding product quality inspections became imperative. Identifying defective products during inspection led to additional costs for the companies. Balancing customer product quality requirements and re-inspection costs became crucial for optimal benefits. This study addresses the binary classification problem of customer-side quality inspection through cost-sensitive learning. The predictive model considers panel process yield, production history, customer feedback, inspection capacity constraints, and cost minimization to predict panel quality as accepted or defective. To tackle the highly imbalanced data, a two-phase cost-sensitive-learning-based framework is proposed, combining data preprocessing methods and models, while considering re-inspection capacity constraints and costs to enhance accuracy. The model’s evaluation uses key performance indicators like AUC and G-mean. Actual inspection cost and defective parts per million (DPPM) are calculated based on the company’s practical assessment. Two products are used for experimentation to validate the proposed model, demonstrating over 50% reduction in inspection cost and over 10% improvement in DPPM.

Keywords: Cost-sensitive learning; Quality inspection; Binary classification; TFT-LCD industry; Capacity constraint (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10845-024-02448-6 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:36:y:2025:i:6:d:10.1007_s10845-024-02448-6

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-024-02448-6

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().