Renewable energy approach towards powering the CI engine with ternary blends of algal biodiesel-diesel-diethyl ether: Bayesian optimized Gaussian process regression for modeling-optimization

Abstract

The present study investigates the performance and emissions of a small diesel engine powered with ternary blends of algal biodiesel, diesel, and diethyl ether. The findings indicate that the addition of diethyl ether improved the brake thermal efficiency, peak pressure inside cylinder, and net heat release rate. This is because the additives have a greater oxygen value, higher cetane number, and volatility. However, an increase in NOX emissions was observed. Gaussian Process Regression, a modern supervised machine learning approach, was utilized to develop a prognostic model for the engine's performance and exhaust emission. The Bayesian technique for hyperparameter optimization was used to strengthen the predictive model's training process. As model input parameters, engine load, fuel mixing ratio, and fuel injection pressure were used. The response variables were brake thermal efficiency (BTE), brake-specific fuel consumption (BSFC), NOx, HC, and CO. The created prophetic models predicted engine performance and exhaust emission data with high accuracy, as shown by correlation coefficient values ranging from 0.9746 to 0.9999, low mean squared error up to only 11.023, and low mean absolute error ranging from 0.001 to 2.591. The prognostic model for engine performance and emission developed with Bayesian optimized Gaussian process regression could provide a robust simulation and prognostication efficiency.