Modeling FTIR Spectra of Ramie Non-Woven Fabrics with Corona Plasma Treatment: ANN, SVR, Random Forest, XGBoost

This study used machine learning techniques such as artificial neural networks, support vector regression, random forest, and XGBoost to examine the effect of Corona plasma treatment parameters on non-woven fabrics and evaluate the model's ability to predict Fourier Transform Infrared Spectroscopy (FTIR) spectral information from FTIR measurements. We examined the FTIR spectra information of non-woven fabrics based on FTIR measurements by varying the variables associated with Corona plasma treatment. FTIR percent transmission (T%) was predicted based on three input parameters: wavenumber, electrode distance, and exposure time. Our study is unique in that we used an artificial neural network (ANN) to model the plasma treatment on non-woven materials made from ramie fiber and the FTIR spectra accurately for the first time. According to this study, the random forest (RF), XGBoost, and ANN with three hidden layers were appropriate for determining FTIR percent transmission (T%). Based on this research, the values of R2 for determining the percent transmission (T%) for random forest, XGBoost, and ANN were 0.97, 0.99 and 0.999 respectively. This study leverages artificial intelligence to simulate FTIR spectra of plasma-treated textiles, demonstrating the potential of machine learning–driven modeling to advance plasma treatment research for non-woven materials.