.png){kind=icon}
.png){kind=icon}
Return
Comparative optimization of PMMA nanofiber production using response surface methodology and machine learning algorithms
.png){kind=icon}
PRE
AI
DOI：10.1080/00405000.2026.2668802.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Nanofibers enhance the properties of materials and have a wide range of application areas because of their large surface area per unit mass. The solution blow spinning technique is a new and highly productive way to fabricate nanofibers. This study presents a comparative optimization analysis of Response Surface Methodology (RSM) and Machine Learning (ML) for PMMA nanofiber production using solution blow spinning, based on a central composite design (CCD) and considering different independent parameters such as flow rate (3-7 mL/h), polymer concentration (17-25 wt%), and air pressure (3-5 bar). The flow rate remained ineffective between 3-7 mL/h, but analysis of variance (ANOVA) showed that concentration was the main factor affecting the diameter of PMMA nanofibers. When the concentration is decreased from 25% (w/w) to 17%, the fiber diameter decreases from an average of 928 nm to 235 nm. Also, a comparison of four different machine learning algorithms, including Decision Tree (DT), Random Forest Regressor (RFR), Support Vector Machine (SVM), and Gaussian Process Regression (GPR) was provided. GPR was identified as the most suitable algorithm with a high regression coefficient (R-2=0.99) between observed and predicted values. As a result of evaluating the performances of RSM and GPR with determined target points, the experimentally obtained results in both strategies provided a high agreement. The GPR was more reliable because of the higher correlation. This study presented parametric research to obtain the required PMMA fiber diameter by adjusting system parameters.
Keywords:
PMMA nanofiber
solution blow spinning
central composite design
response surface methodology
decision tree
machine learning
Gaussian process regression
Journal
J
IF:
1.5
Papers:
116
Citations:
5.4K