View in EDS

A machine learning approach for predicting sealing strength in medical device packaging: model development and evaluation.

Saved in:
Bibliographic Details
Title: A machine learning approach for predicting sealing strength in medical device packaging: model development and evaluation.
Authors: Orłowski, Robert1 (AUTHOR), Burduk, Anna1 (AUTHOR) anna.burduk@pwr.edu.pl, Zyblewski, Paweł1 (AUTHOR)
Source: Archives of Civil & Mechanical Engineering (Elsevier Science). Jul2026, Vol. 26 Issue 4, p1-19. 19p.
Subjects: Machine learning, Sealing (Technology), Boosting algorithms, Regression analysis, Experimental design, Artificial neural networks, Packaging equipment
Abstract: In this article, we present a methodology for employing nonlinear machine learning (ML) models to predict the sealing strength of disposable packaging for medical devices. The proposed approach serves as an alternative to traditional Design of Experiments (DoE) methods, which require conducting numerous series of experiments according to a predetermined plan. The performance of selected regression models was compared based on the repeated cross-validation evaluation protocol, supported by statistical analysis (combined 5 × 2 CV F-test). The database utilized in the study comprises key process set-up parameters as well as detailed data regarding the materials and tools employed, thereby enabling a comprehensive analysis of the factors affecting the technological process under investigation. The results indicate that the XGBoost model performs best under industrial conditions. Furthermore, the analysis revealed that the most critical factors determining sealing strength are dwell time, sealing pressure, and film thickness. The developed methodology not only facilitates the prediction of sealing strength but also enables the determination of process parameter values for new configurations of tools and materials, thereby contributing to a reduction in experimental costs and expediting the validation of new processes. The paper also outlines potential directions for further research, such as expanding the database and applying deep neural networks. [ABSTRACT FROM AUTHOR]
Copyright of Archives of Civil & Mechanical Engineering (Elsevier Science) is the property of Springer Nature 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 this article, we present a methodology for employing nonlinear machine learning (ML) models to predict the sealing strength of disposable packaging for medical devices. The proposed approach serves as an alternative to traditional Design of Experiments (DoE) methods, which require conducting numerous series of experiments according to a predetermined plan. The performance of selected regression models was compared based on the repeated cross-validation evaluation protocol, supported by statistical analysis (combined 5 × 2 CV F-test). The database utilized in the study comprises key process set-up parameters as well as detailed data regarding the materials and tools employed, thereby enabling a comprehensive analysis of the factors affecting the technological process under investigation. The results indicate that the XGBoost model performs best under industrial conditions. Furthermore, the analysis revealed that the most critical factors determining sealing strength are dwell time, sealing pressure, and film thickness. The developed methodology not only facilitates the prediction of sealing strength but also enables the determination of process parameter values for new configurations of tools and materials, thereby contributing to a reduction in experimental costs and expediting the validation of new processes. The paper also outlines potential directions for further research, such as expanding the database and applying deep neural networks. [ABSTRACT FROM AUTHOR]
ISSN:16449665
DOI:10.1007/s43452-026-01570-x