Experimental Assessment and Metaheuristic-Optimized Multi-Output ANN Modeling of Cyclohexane–Safflower-Biodiesel Blends in a Diesel Engine

This study examines engine performance and exhaust concentrations for diesel–safflower biodiesel–cyclohexane (CHx) blends and compares hyperparameter-search strategies for a six-output artificial neural network (ANN). Four fuels were evaluated at 1500 rpm and four nominal load-controller settings, generating 2400 repeated observations across 16 fuel–load conditions. Proton nuclear magnetic resonance (1H NMR) analysis indicated an estimated ester conversion of 95.24%. Across equally weighted load settings, DF50B45CHx5 reduced mean nitrogen-oxides (NOx) concentration from 171.26 to 111.72 ppm, carbon-monoxide (CO) concentration from 156.32 to 146.61 ppm, and smoke absorption from 0.814 to 0.447 m−1 relative to diesel. These reductions coincided with decreases in brake power (BP) from 1.579 to 1.406 kW and brake thermal efficiency (BTE) from 23.063% to 19.706%, while brake-specific fuel consumption (BSFC) increased from 0.476 to 0.580 kg kWh−1. With 30 candidate evaluations per optimizer, Covariance Matrix Adaptation Evolution Strategy (CMA-ES) achieved the highest six-response mean coefficient of determination (R2), 0.9926 versus 0.9810 for particle swarm optimization (PSO), 0.9794 for Random Search, and 0.9465 for the genetic algorithm (GA). In separate condition-held-out validation, mean target-normalized root mean squared error (RMSE) was 0.427 for ridge regression, 0.294–0.330 for a joint ANN, and 0.255–0.262 for separate ANNs. The results identify an emissions-performance trade-off and distinguish internal model-selection accuracy from prediction at withheld fuel–load combinations.

Authors

Publication Details

Journal
Energies
Published
2026-11-16
DOI
https://doi.org/10.3390/en19225539
Primary Topic
Biodiesel Production and Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Experimental Assessment and Metaheuristic-Optimized Multi-Output ANN Modeling of Cyclohexane–Safflower-Biodiesel Blends in a Diesel Engine

ZEYNEP İLKILIÇ AYTAÇ
Energies
Biodiesel Production and Applications
article

Experimental Assessment and Metaheuristic-Optimized Multi-Output ANN Modeling of Cyclohexane–Safflower-Biodiesel Blends in a Diesel Engine

ZEYNEP İLKILIÇ AYTAÇ
article en

Abstract

This study examines engine performance and exhaust concentrations for diesel–safflower biodiesel–cyclohexane (CHx) blends and compares hyperparameter-search strategies for a six-output artificial neural network (ANN). Four fuels were evaluated at 1500 rpm and four nominal load-controller settings, generating 2400 repeated observations across 16 fuel–load conditions. Proton nuclear magnetic resonance (1H NMR) analysis indicated an estimated ester conversion of 95.24%. Across equally weighted load settings, DF50B45CHx5 reduced mean nitrogen-oxides (NOx) concentration from 171.26 to 111.72 ppm, carbon-monoxide (CO) concentration from 156.32 to 146.61 ppm, and smoke absorption from 0.814 to 0.447 m−1 relative to diesel. These reductions coincided with decreases in brake power (BP) from 1.579 to 1.406 kW and brake thermal efficiency (BTE) from 23.063% to 19.706%, while brake-specific fuel consumption (BSFC) increased from 0.476 to 0.580 kg kWh−1. With 30 candidate evaluations per optimizer, Covariance Matrix Adaptation Evolution Strategy (CMA-ES) achieved the highest six-response mean coefficient of determination (R2), 0.9926 versus 0.9810 for particle swarm optimization (PSO), 0.9794 for Random Search, and 0.9465 for the genetic algorithm (GA). In separate condition-held-out validation, mean target-normalized root mean squared error (RMSE) was 0.427 for ridge regression, 0.294–0.330 for a joint ANN, and 0.255–0.262 for separate ANNs. The results identify an emissions-performance trade-off and distinguish internal model-selection accuracy from prediction at withheld fuel–load combinations.

EnergiesVol. 19(22)
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