Artificial intelligence based-prediction of energy efficiency and tailpipe emissions of soybean methyl ester fuelled CI engine under variable compression ratios

Abstract

Biodiesel, a widely adopted substitute for conventional mineral diesel, is often derived from raw vegetable oils. The use of non-edible oils for the production of biodiesel has been favored for an extended period of time due to the potential impact on the availability of edible oil as a food source. The generation of soybean methyl ester biodiesel using the transesterification process. The primary objective of this study is to examine the potential use of soybean methyl ester biodiesel (referred to as S) as a substitute for conventional diesel fuel in a standard light-duty compression ignition four-stroke, single-cylinder, water-cooled engine. The present study aimed to evaluate the impact of increased compression ratio (ranging from 14 to 18) on several combustion, performance, and emissions characteristics. This investigation was conducted using biodiesel blends consisting of 5%, 10%, 20%, and 40% soybean methyl ester, and the results were compared with those obtained using conventional diesel fuel. Additionally, in order to enhance the test results, a training process is being conducted on an artificial intelligence network (ANN). The analysis of the data indicated a positive correlation between the compression ratio (CR) and the in-cylinder pressure values at maximum load conditions. Additionally, a negative relationship was seen between the CR and the brake-specific fuel consumption (BSFC), suggesting a drop in BSFC as the CR rose. Upon evaluating the discharge numbers, it was seen that there was a notable reduction in other pollutants, a fall in NOx emissions, and an elevation in haze levels. Nevertheless, surpassing a sulphur content of 20% has an adverse impact on both emissions and the overall functioning of the engine. Based on the findings of the artificial neural network (ANN), the prediction results demonstrate a commendable capacity to accurately predict the experimental data with the R-value changing from 0.81919 to 0.95028 for all engine performance, combustion, and emission metrics.