Publication: Predictive modeling of engine performance and emissions for castor oil ethyl ester biodiesel blends: A Gaussian process regression approach
1
0
Issued Date
2024-06-01
Resource Type
eISSN
25901230
Scopus ID
2-s2.0-85195321763
Journal Title
Results in Engineering
Volume
22
Rights Holder(s)
SCOPUS
Bibliographic Citation
Results in Engineering Vol.22 (2024)
Suggested Citation
Ariyarit A., Aengchuan P., Wiangkham A., Pumpuang A., Klinkaew N., Theinnoi K., Chuepeng S., Sukjit E. Predictive modeling of engine performance and emissions for castor oil ethyl ester biodiesel blends: A Gaussian process regression approach. Results in Engineering Vol.22 (2024). doi:10.1016/j.rineng.2024.102362 Retrieved from: https://hdl.handle.net/20.500.14740/20197
Corresponding Author(s)
Other Contributor(s)
Abstract
Replacing fossil fuels with cleaner alternatives is essential. This study examines a biodiesel-diesel blend containing 8 % castor oil ethyl ester (COEE8) and its impact on engine performance, combustion characteristics, and exhaust emissions. A single-cylinder diesel engine was tested under consistent conditions: an engine speed of 1500 rpm, varying engine loads (25 %, 50 %, and 75 % of maximum torque), and compression ratios (16, 17, and 18). Engine-out emissions were measured with Testo flue gas analyzers. The results showed that COEE8 combustion significantly decreased HC, CO, and smoke emissions compared to diesel fuel but increased NOx emissions. Additionally, COEE8 exhibited comparable brake-specific fuel consumption (BSFC) and brake thermal efficiency (BTE) to diesel fuel. The optimal engine operating parameters were determined using the Non-dominated Sorting Genetic Algorithm-II (NSGA-II), a multi-objective optimization technique. Due to limited data availability, Gaussian Process Regression (GPR), a machine learning algorithm for small datasets, modeled the multi-objective functions with compression ratio and engine load as input variables. The GPR model demonstrated high prediction performance across all output parameters (BSFC, BTE, HC, smoke, NOx, and CO) for both diesel and COEE8 fuels, with average coefficients of determination (R2) of 0.9896 and 0.9953, respectively, indicating a strong correlation between predicted and actual values. The mean absolute percentage error (MAPE) was also low, averaging 3.11 % and 2.26 %, respectively, demonstrating the model's accuracy. Using the GPR model, the NSGA-II identified the optimal trade-off between NOx and smoke index for both diesel and COEE8 fuels. The optimal compression ratio was 16 for both fuels, while the optimal engine load varied, ranging from 20 % to 30 % for COEE8 and around 30 % (almost 40 %) for diesel fuel.