TWO-STAGE CASE-BASED REASONING FRAMEWORK WITH TREE-BASED ENSEMBLE METHODS FOR PRINTED CIRCUIT BOARD YIELD PREDICTION.

Saved in:

Bibliographic Details
Title:	TWO-STAGE CASE-BASED REASONING FRAMEWORK WITH TREE-BASED ENSEMBLE METHODS FOR PRINTED CIRCUIT BOARD YIELD PREDICTION.
Authors:	Park, Jun-Ho¹, Baek, Seung-Hyun² sbaek4@hanyang.ac.kr
Source:	International Journal of Industrial Engineering. 2026, Vol. 33 Issue 1, p15-46. 32p.
Subjects:	Case-based reasoning, Ensemble learning, Automatic optical inspection, Prediction models, Clustering algorithms, Printed circuits industry, Manufacturing process management
Abstract:	In printed circuit board (PCB) manufacturing, yield is a critical indicator that directly impacts production costs and delivery schedules. For new products, the industry currently relies on empirical predictions by engineers based on historical yields of similar products. However, this approach suffers from limitations in accuracy and consistency due to subjective and nonsystematic assessment criteria, issues that are compounded by the complex product structures and multi-stage processes used in the PCB industry. To address these challenges, this study proposes a yield prediction methodology that leverages Case-Based Reasoning (CBR) and tree-based prediction models. This methodology consists of two stages. The first stage is to predict the yield of model units before the new product is put into the first process. In this stage, we utilize SHapley Additive exPlanations (SHAP)-based variable importance to reflect the weight of each variable and determine the optimal number of clusters based on silhouette analysis to search for similar cases. Then, data from retrieved cases is analyzed with a tree-based prediction model to predict yield, which evaluates the relationship between dates and yield to detect whether a statistically significant change has occurred at a particular point in time. The second stage utilizes additional defect information from Automated Optical Inspection (AOI) reports to make more precise yield forecasts on a lot-by-lot basis. This further increases the reliability of the forecast by reflecting the quality variability that occurs during production. Based on these predictions, more accurate material input calculations can be performed, thereby improving delivery compliance rates and minimizing excess inventory. [ABSTRACT FROM AUTHOR]
	Copyright of International Journal of Industrial Engineering is the property of International Journal of Industrial Engineering and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	In printed circuit board (PCB) manufacturing, yield is a critical indicator that directly impacts production costs and delivery schedules. For new products, the industry currently relies on empirical predictions by engineers based on historical yields of similar products. However, this approach suffers from limitations in accuracy and consistency due to subjective and nonsystematic assessment criteria, issues that are compounded by the complex product structures and multi-stage processes used in the PCB industry. To address these challenges, this study proposes a yield prediction methodology that leverages Case-Based Reasoning (CBR) and tree-based prediction models. This methodology consists of two stages. The first stage is to predict the yield of model units before the new product is put into the first process. In this stage, we utilize SHapley Additive exPlanations (SHAP)-based variable importance to reflect the weight of each variable and determine the optimal number of clusters based on silhouette analysis to search for similar cases. Then, data from retrieved cases is analyzed with a tree-based prediction model to predict yield, which evaluates the relationship between dates and yield to detect whether a statistically significant change has occurred at a particular point in time. The second stage utilizes additional defect information from Automated Optical Inspection (AOI) reports to make more precise yield forecasts on a lot-by-lot basis. This further increases the reliability of the forecast by reflecting the quality variability that occurs during production. Based on these predictions, more accurate material input calculations can be performed, thereby improving delivery compliance rates and minimizing excess inventory. [ABSTRACT FROM AUTHOR]
ISSN:	10724761
DOI:	10.23055/ijietap.2026.33.1.11287