Bio-oil production from ripe pawpaw seeds and its optimal output: Box-Behnken Design and Machine Learning approach

Abstract

The purpose of this study is to use ripe pawpaw seeds for a rapid Soxhlet extractor process with n-hexane to increase the yield of the bio-oil and to make the bio-oil produced a beneficial addition to the field of bio-oil production parameter optimization and utilization for industrial applications. Thus, the produced bio-oil from kwale ripe pawpaw seeds was modeled and optimized (Ripe Carica papaya seeds) via Soxhlet extractor using Box-Behnken Design from Response Surface Methodology and Machine Learning (Python code) techniques. The result from Box-Behnken Design shows that the highest yield obtained was 23.93 wt.% at 55 g of Sample weight, 50 min of Extraction time, and 250 ml of Solvent volume while the highest value of the bio-oil yield from the Machine Learning approach is 23.97 wt.%, which is closely related to the value (23.93 wt.%) obtained from Box-Behnken Design. The R2 of the model from Box-Behnken Design was 0.9786 while the R2 from Machine Learning was 1.0. Also, the visualization in Machine Learning was more appealing than in Box-Behnken Design. Thus, Machine Learning via python coding was more reliable and effective than Box-Behnken Design in terms of prediction and accuracy of bio-oil production. Both models, however, delivered a reliable reaction under the operational conditions considered. The physicochemical characteristics of the bio-oil from ripe pawpaw seeds also meet the requirements for bio-oil. Thus, this study concluded that green biowaste oil obtained from Kwale ripe pawpaw seeds waste might be used to produce biodiesel as well as to cool spinning equipment parts.