Work

Hybrid RSM-ML framework for data-driven multi-objective optimization of palm biodiesel production via heterogeneous catalytic interesterification

Year: 2026

Type: article

Abstract: This study develops a data-driven hybrid framework integrating RSM-ML for multi-objective optimization of palm biodiesel via heterogeneous catalytic interesterification. The parameters used in this study include catalyst concentration (1.0; 1.5; 2.0%), reaction time (1; 2; 3 hours), and temperature (100; 150; 200 °C) with three responses: yield (%), HHV (MJ/kg), and SEC (MJ/kg biodiesel), using Design Expert (version 13) with a FCCD. Subsequently, using the same data, optimization was performed using machine learning with three regression algorithms, RFR, SVR, and XGBoost, as well as a hybrid RSM-ML approach. The model’s reliability is supported by repeated five-fold cross-validation and experimental validation, with a relative error in the total response of less than 5%. The optimal conditions for the RSM –ML were obtained at a catalyst concentration of 2.0%, a reaction time of 1.857 hours, and a temperature of 164.286 °C, with predicted yields of 88.746%, HHV of 42.236 MJ/kg, SEC of 43.799 MJ/kg of biodiesel, and an overall desirability of 0.912. GC–MS analysis showed FAEE as the dominant fraction at 69.00%, primarily palmitic acid ethyl ester at 34.61%. The product has a density of 0.85 g/mL, a kinematic viscosity of 4.76 cSt, HHV of 42.19 MJ/kg, pour point of -2 o C and a flash point of 152 °C. Compared to the selected experimental conditions, the optimized hybrid validation reduced SEC and electricity-related CO₂ emissions by 15.84%, with a reaction time that was 7.15% shorter. Thus, the RSM–ML hybrid is effective in balancing yield, fuel quality, and energy efficiency.

Source: Results in Engineering

DOI: 10.1016/j.rineng.2026.112740

Authors: Viqhi Aswie, Wipawee Dechapanya, Mahfud Mahfud, Lailatul Qadariyah, Attaso Khamwichit

Institutions: Walailak University, Institut Teknologi Sepuluh Nopember

Language: English

Topic: Biodiesel Production and Applications

Subfield: Fuel Technology

Field: Energy

Domain: Physical Sciences

SDGs: Affordable and clean energy, Responsible consumption and production, Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $1,940