Time and volume-ratio effect on reusable polybenzoxazole nanofiber oil sorption capacity investigated via machine learning

Abstract

Diesel oil sorption capacities (DOSCs) of polybenzoxazole/polyvinylidene fluoride nanofiber mats with four different groups (-O-, -S-S-, phenylene and diphenylene) in the main chain structures were investigated. Different experimental duration and diesel-oil/tap-water volume ratio pairs were used for diesel oil sorption. No degradation was observed in the nanofiber mat structures after diesel oil sorption. The characterizations of polybenzoxazole (PBO) nanofibers with high diesel oil selectivity were performed by scanning electron microscopy, atomic force microscopy, Fourier transform infrared spectroscopy, x-ray diffraction, thermal gravimetric analysis, differential scanning calorimetry, Brunauer-Emmett-Teller (BET), and contact angle measurement analysis. According to the result of characterizations, superoleophilic and superhydrophobic nanofiber mats show high water contact angle value in the range of 132-140(circle) and show high separation efficiency. In this study, we integrated ensemble gradient boosting model (XGBoost) to predict the DOSC of sorbent nanofiber and obtain an optimal set of conditions to maximize the DOSC. The predicted PBO-E sorbent at the 0.5 ratio of diesel-oil/tap-water measured at the end of the 3rd minute showed the most reliable and stable diesel oil sorption with at least 9.39 and at most 12.33 g/g sorbent with 95% of confidence.