Aplikasi Statistical Learning untuk Menentukan Parameter Electrospinning yang Signifikan Terhadap Diameter dan Kualitas Serat Nano Polietilen Tereftalat
Muhammad Kevin Alrahmanto, Prof. Yusril Yusuf, S.Si., M.Si., M.Eng., D.Eng.
2023 | Skripsi | FISIKA
Electrospinning (ES) adalah teknik yang banyak digunakan
untuk menghasilkan serat nano untuk berbagai aplikasi. Namun, mengoptimalkan
parameter ES guna mencapai diameter serat dan kualitas serat (bermanik/halus)
yang diinginkan tetap menjadi tantangan yang kompleks karena banyaknya
parameter yang mempengaruhi hasil akhir serat. Pada penelitian ini, digunakan statistical
learning untuk mengidentifikasi parameter-parameter signifikan yang
berpengaruh terhadap diameter serat dan kualitas serat, khususnya serat yang
terbuat dari polietilen tereftalat (PET). Data yang dikumpulkan dalam penelitian ini
diambil dari berbagai studi yang telah dipublikasikan mengenai serat nano PET yang tersedia di Google Scholar. Data mencakup parameter ES, seperti
konsentrasi, tegangan, laju aliran, rasio pelarut asam trifluoroasetat
(TFA)/diklorometana (DCM), jarak jarum dengan kolektor, dan frekuensi putaran
kolektor. Dikumpulkan juga data tentang diameter serat dan kualitas serat yang
dihasilkan. Dengan menggunakan statistical learning, yaitu regresi dan
klasifikasi, data dianalisis untuk menemukan pola dan hubungan antara parameter
ES dengan diameter serat dan kualitas serat. Teknik statistical learning
yang digunakan untuk memodelkan diameter serat adalah regresi linear berganda
dan regresi polinomial dengan regularisasi Lasso. Sementara itu, regresi
logistik digunakan untuk memodelkan kualitas serat. Nilai R-kuadrat
Electrospinning
(ES) is a widely used technique for producing nanofibers for various
applications. However, optimizing ES parameters to achieve the desired fiber
diameter and fiber quality (beads-free) remains a complex challenge due to the
influence of multiple parameters on the final fiber outcomes. In this study,
statistical learning is employed to identify significant parameters that affect
fiber diameter and fiber quality, specifically for fibers made from
polyethylene terephthalate (PET). The data collected for this research was gathered
from numerous published studies on PET nanofibers available in Google Scholar.
The dataset includes ES parameters such as concentration, voltage, flow rate,
trifluoroacetic acid (TFA)/dichloromethane (DCM) solvent ratio,
needle-to-collector distance, and collector rotation frequency. Additionally, data
on fiber diameter and fiber quality generated from these parameters were also
collected. By employing
statistical learning techniques, namely regression and classification, the data
were analyzed to discover patterns and relationships between the ES parameters
and fiber diameter and fiber quality. The statistical learning techniques employed to model fiber diameter were
multiple linear regression and polynomial regression with Lasso regularization.
Meanwhile, logistic regression was utilized to model fiber quality. The R-squared for multiple linear regression
and polynomial regression with Lasso regularization were 0.46 and 0.84,
respectively. According to the analysis of the multiple linear regression model,
concentration and TFA/DCM solvent ratio were identified as the most significant
parameters in explaining nanofiber diameter. Meanwhile, the analysis of the
logistic regression model revealed that concentration, voltage, and
needle-to-collector distance were the most significant parameters in explaining
fiber quality.
Kata Kunci : Electrospinning, Polietilen tereftalat (PET), Statistical Learning
