Uncertainty-aware machine learning and NSGA-II framework for multi-objective optimization of third-generation biodiesel production

One of the greatest challenges in the third generation of biofuels is to maximize biodiesel production while minimizing the intensity of the process, in the case of optimization based on limited experimental observations. In this study, an integrated framework is developed by integrating machine learning, physically bounded surrogate modeling, non-dominated sorting genetic algorithm II (NSGA-II), and uncertainty analysis to optimize fatty acid methyl ester (FAME) production from Chlorella protothecoides oil. Five machine learning algorithms, namely Random Forest, Extra Trees, Gradient Boosting, XGBoost, and support vector regression with radial basis function kernel (SVR-RBF), were explored. A physically bounded SVR-RBF formulation with a logit-transformed response was introduced to address the limitations of arbitrary prediction clipping and compared with the original formulation with group-aware validation. The revised model had a test R 2 of 0.8531, test RMSE of 4.1564 percentage points, and test MAPE of 4.5723%, and had an R 2 of 0.9167 and RMSE of 3.3775 percentage points on nested group-aware cross-validation. The results of integration with NSGA-II showed a maximum predicted FAME yield of 98.61% with a methanol-to-oil molar ratio of 6.00, NaOH concentration of 0.800 wt%, a temperature of 51.38 °C, and the reaction time of 165.10 min (PRSI = 0.805). The trade-off of the ideal distance obtained the maximum predicted FAME yield of 83.45% at a significantly lower PRSI of 0.242. This Monte Carlo analysis with 10,000 realizations gave a mean predicted compromise yield of ∼83.75% for the assumed perturbations during operation. Other experimental support and boundary-sensitivity evaluations showed different levels of predictive reliability in the Pareto frontier. The proposed framework can provide an uncertainty-informed method for determining biodiesel production trade-offs and highlights the importance of experimental validation of surrogate-derived operating conditions.

Authors

Institutions

Publication Details

Journal
Case Studies in Thermal Engineering
Published
2026-11-18
DOI
https://doi.org/10.1016/j.csite.2026.108737
Primary Topic
Biodiesel Production and Applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Uncertainty-aware machine learning and NSGA-II framework for multi-objective optimization of third-generation biodiesel production

Mansoor Alruqi, Sanjeev Kumar
1 links
Case Studies in Thermal Engineering
Biodiesel Production and Applications
article

Uncertainty-aware machine learning and NSGA-II framework for multi-objective optimization of third-generation biodiesel production

Mansoor Alruqi, Sanjeev Kumar
article en

Abstract

One of the greatest challenges in the third generation of biofuels is to maximize biodiesel production while minimizing the intensity of the process, in the case of optimization based on limited experimental observations. In this study, an integrated framework is developed by integrating machine learning, physically bounded surrogate modeling, non-dominated sorting genetic algorithm II (NSGA-II), and uncertainty analysis to optimize fatty acid methyl ester (FAME) production from Chlorella protothecoides oil. Five machine learning algorithms, namely Random Forest, Extra Trees, Gradient Boosting, XGBoost, and support vector regression with radial basis function kernel (SVR-RBF), were explored. A physically bounded SVR-RBF formulation with a logit-transformed response was introduced to address the limitations of arbitrary prediction clipping and compared with the original formulation with group-aware validation. The revised model had a test R 2 of 0.8531, test RMSE of 4.1564 percentage points, and test MAPE of 4.5723%, and had an R 2 of 0.9167 and RMSE of 3.3775 percentage points on nested group-aware cross-validation. The results of integration with NSGA-II showed a maximum predicted FAME yield of 98.61% with a methanol-to-oil molar ratio of 6.00, NaOH concentration of 0.800 wt%, a temperature of 51.38 °C, and the reaction time of 165.10 min (PRSI = 0.805). The trade-off of the ideal distance obtained the maximum predicted FAME yield of 83.45% at a significantly lower PRSI of 0.242. This Monte Carlo analysis with 10,000 realizations gave a mean predicted compromise yield of ∼83.75% for the assumed perturbations during operation. Other experimental support and boundary-sensitivity evaluations showed different levels of predictive reliability in the Pareto frontier. The proposed framework can provide an uncertainty-informed method for determining biodiesel production trade-offs and highlights the importance of experimental validation of surrogate-derived operating conditions.

Case Studies in Thermal EngineeringVol. 88
Shaqra University (SA), Delhi Skill and Entrepreneurship University (IN)
Shaqra University
Affordable and clean energy, Industry, innovation and infrastructure
Biodiesel Production and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon