Loading...
Development of machine learning models for improving and achieving target fiber diameter of electrospun nanofibers
Subeshan, Balakrishnan
Subeshan, Balakrishnan
Citations
Altmetric:
Files
Loading...
d24039s_Subeshan.pdf
Adobe PDF, 6.58 MB
Authors
Other Names
Location
Time Period
Advisors
Original Date
Digitization Date
Issue Date
2024-12
Type
Dissertation
Genre
Keywords
Subjects (LCSH)
Electronic dissertations
Citation
Abstract
Electrospinning is a widely recognized technique for fabricating nanofibers with tailored properties, essential for applications in fields such as tissue engineering, drug delivery, filtration, energy storage, and sensors. However, the complexity of the electrospinning process, with its various experimental process parameters poses significant challenges in achieving consistent fiber diameters. This study explores the option of integrating machine learning (ML) algorithms to accurately predict and precisely control fiber diameters, thereby enhancing the efficiency of the electrospinning process. The study includes a comprehensive review of current ML applications for electrospun nanofibers. Predictive ML models were developed to train a dataset compiled from published research scientific sources, with eXtreme Gradient Boosting (XGB) achieving a coefficient of determination (R²) value of 0.93 and root mean square error (RMSE) of 127.76 nm on polyacrylonitrile (PAN) nanofibers and an R² value of 0.94 with an RMSE of 79.89 nm on polyvinyl alcohol (PVA) nanofibrous scaffolds for tissue engineering applications. In addition, a broader dataset containing 3000 data points across a range of polymers, solvents, and process parameters was used to refine predictive ML models further. Among the various ML models, the XGB model demonstrated superior performance, achieving an R² value of 0.94 with an RMSE of 275.02 nm. Experimental validation with electrospun polystyrene (PS) nanofibers confirmed the robustness of these predictions, showing strong alignment between predicted and measured fiber diameters. Process optimization was performed using a Genetic Algorithm (GA), achieving target fiber diameters between 100 nm and 4000 nm with low fitness errors. This integrated approach achieves a near-perfect correlation (R² = 1.00) between target and predicted fiber diameters across diverse electrospinning conditions, reducing dependency on trial-and-error experimentation and enabling scalable, data-driven nanofiber fabrication tailored to specific applications.
Table of Contents
Description
Thesis (Ph.D.)-- Wichita State University, College of Engineering, Dept. of Mechanical Engineering
Publisher
Wichita State University
Journal
Book Title
Series
Digital Collection
Finding Aid URL
Use and Reproduction
© Copyright 2024 by Balakrishnan Subeshan
All Rights Reserved