Prediction of the Diameter of Biodegradable Electrospun Nanofiber Membranes: An Integrated Framework of Taguchi Design and Machine Learning

The ability to replicate electrospinning using a computer tool is of critical importance. Even though electrospinning technology has received much attention, there haven't been many simulation studies. Therefore, the present study established a combined design of experiments and machine learning prediction models methodology to offer a sustainable and efficient electrospinning process. To that effect, we built a locally weighted kernel partial least squares regression (LW-KPLSR) model based on Taguchi's statistical orthogonal design to predict the diameter of the chitosan-based electrospun nanofiber membrane. The model's prediction accuracy was assessed using root mean square error (RMSE), mean absolute error (MAE), and the coefficient of determination (R-2). Besides, principal component regression (PCR), locally weighted partial least squares regression (LW-PLSR), partial least square regression (PLSR), least square support vector regression model (LSSVR) and fuzzy modelling were some of the other types of regression models used to verify and compare the results. Our findings reveal that the LW-KPLSR model greatly outperformed competing models predicting the membrane diameter. This is evident by the LW-KPLSR model's much smaller RMSE and MAE values. More so, it provided the highest attainable R-2 values, which reached 0.9996.

Automated summary

The ability to replicate electrospinning using a computer tool is crucial. This study established a combined design of experiments and machine learning prediction models methodology to offer a sustainable and efficient electrospinning process.