Optimization of rubber seed oil extraction: adaptive neuro-fuzzy inference-based yield prediction model by studying polarity and moisture content

Due to the high oil content, rubber seed oil (RSO) has the potential to be used as a raw material in biodiesel manufacturing. The present study was aimed at oil extraction on rubber seed composed of 47–51% kernel by seed weight. The effects of extraction time (0.75–120 h), particle size (0.14 ± 0.047 µm, 0.35 ± 0.003 µm, and 0.68 ± 0.1 µm), dryness, the type of solvent (polar and non-polar), and different methods of extraction were studied in this work. RSO extracted with a non-polar solvent, hexane, yielded 46.8 ± 0.92%, while a polar solvent, methanol, yielded a lower yield (28.5 ± 0.36%). The average sample size weakly influenced the extract yield with a weak correlation (r = − 0.213), but a strong negative relationship was found between the extract’s yield and the samples’ moisture (r = − 0.974). Therefore, this study used an Adaptive Neuro-Fuzzy Inference System (ANFIS) model to evaluate the prediction of the extract yield performance versus particle size and moisture of the samples. Results confirm consistency between the compositional results of the unsaturated fatty acids obtained by the gas chromatography–mass spectrometry (GC–MS) method (linoleic [39.72%], oleic [23.02%], linolenic [11.91%], and saturated [25.32%]). RSO’s high extraction yield as a low-cost feedstock offers a cost-effective and ecologically acceptable alternative to conventional feedstock for biodiesel synthesis.