Ultrasound-assisted biodiesel production from Peltophorum pterocarpum oil: A comparative analysis of prediction accuracy between RSM and ANFIS

Background: Biodiesel is recognized as a sustainable alternative to conventional fossil diesel. The use of ultrasound energy in biodiesel production enhances reaction efficiency and reduces costs. This study identifies a new feedstock, Peltophorum pterocarpum (commonly known as copper pod seeds), for biodiesel production. In recent times, machine learning (ML) techniques have been employed to predict the biodiesel yield. Methods: Ultrasound assisted transesterification process was utilized for the production of biodiesel from the extracted Peltophorum pterocarpum (Pp) oil. Probe sonicator was used for FAME production. Calcium oxide catalyst derived from waste Pyrgostylus striatulus shells was used as the catalyst. The functional groups present in the extracted oil was characterized using FT-IR analysis. The fatty acid profiling of extracted Pp oil was performed using Gas Chromatography Mass Spectrometry analysis. The research employed ML algorithm systems, specifically Response Surface Methodology (RSM) and Adaptive Neuro Fuzzy Inference System (ANFIS), to analyze biodiesel production. Central Composite Design (CCD) was utilized to optimize operating parameters, including the methanol to oil ratio (9–15 mol/mol), catalyst loading (3–5 wt%), and ultrasonication time (30–60 min). The biodiesel produced was characterized using FT-IR and ¹H NMR instrumentation techniques. Significant findings: The fatty acid composition rom GC-MS analysis of the Copper pod oil revealed that it contains 42.6% linoleic acid, 21.2% oleic acid, and 19.4% palmitic acid. FT-IR analysis confirmed the presence of functional groups, specifically carboxylic acids. This extracted oil was hence suitable for the transesterification process. The best yield of biodiesel from the extracted oil was observed to be 98.6 wt % at 12 mol/mol methanol to Pp oil molar ratio, 4 wt % of CaO and 45 min of ultrasonication time by ANFIS model. Characterization of biodiesel produced was validated through ¹H NMR and FT-IR analysis. The important physical and chemical properties of the biodiesel were analyzed and were found to be within standard limits, indicating its commercial viability. The interpretation of both RSM and ANFIS models were analyzed statistically based on their predicted data by Coefficient of determination, Root mean square error, Standard error of prediction and mean relative percent deviation. The Goodness of fit R² value calculated for RSM and ANFIS models was 0.954 and 0.999 respectively. Both the models have performed well but comparatively ANFIS model had been more accurate proving ANFIS as a potent tool for modelling and optimization of biodiesel production.