Machine learning-based optimization, characterization and CI engine assessment of biodiesel synthesised from waste soybean cooking oil

Abstract

Waste cooking oil as a feedstock for biodiesel production is a sustainable choice which resolves the environmental concerns related to its disposal. The present paper focusses on biodiesel production from waste soyabean oil via transesterification using ethanol and potassium hydroxide (KOH) as catalyst. To estimate the yield of biodiesel, Central Composite Design (CCD) technique used adjustments of reaction temperature, catalyst concentration, alcohol-oil ratio, and reaction time. The created response surface model was further optimized by machine learning-inspired metaheuristic algorithms like Dragonfly Algorithm (DA), Genetic Algorithm (GA), and Grasshopper Optimization Algorithm (GOA). Of these techniques, GOA showed best results, achieving a maximum predicted yield of 80.057%, whereas DA and GA provided 76.384%, and 76.593% respectively. Experimental validation at best conditions (100 °C, 1.3% KOH, 50 min reaction time, and 8:1 alcohol-oil ratio) showed yields between 78.95% and 81.68%, thereby supporting the optimization results. Fatty acid Ethyl esters production was confirmed via FTIR, GC-MS, and NMR techniques and, also. Biodiesel characteristic ester, carbonyl, and hydrocarbon functional groups presence were confirmed. The produced biodiesel was mixed with diesel (B10D90 – B50D50) and were tested in a single-cylinder Kirloskar TV1 compression-ignition (CI) engine. With reference to neat diesel, B20D80 and B30D70 resulted in lower levels of CO emissions, by upto an average of 42–48%, HC emissions by 41–45%, and also higher brake thermal efficiency by 6–11%. At medium to high loads, the specific fuel consumption was further reduced by 8–18% without any need for engine modifications. The findings visibly indicate that a combination of machine learning-assisted optimization and waste cooking oil utilization offers a promising and reliable route for biodiesel production and use in CI engines.