Predictive Modelling and Optimization of Performance and Emissions of Acetylene Fuelled CI Engine Using ANN and RSM

Abstract

This contemporary work was mainly focused on the application of ANN model and RSM optimization tool for analyzing the performance and exhaust emissions of a single-cylinder Compression Ignition (CI) engine operating with acetylene and diesel on dual fuel mode. Experiments were performed in a 3.5 kW Kirloskar TV1, water-cooled engine with various flow rates (2, 4, and 6 liters per minute) of acetylene gas, variations in the Compression Ratio (CR) (16, 18, and 20), Injection Timing (IT) (19ºbTDC, 23ºbTDC, 27ºbTDC) and Injection Pressure (IP) (200, 220, and 240 bar). Multi-objective optimization was conducted using the RSM, while the central composite design was chosen as the design matrix. The desirability approach was employed to find optimum operating conditions. High flow rate of acetylene injection of 6 lpm, higher IP of 240 bar, CR of 18 and IT of 23ºCA bTDC were arrived as the optimum operating conditions with 0.649 desirability value.The ANN model was developed based on experimentaldata to forecast the brake thermal efficiency (BTH), hydrocarbon (HC), carbon monoxide (CO), oxides of nitrogen (NOx) and smoke. ANN was trained using Levenberg-Marquardt Algorithm for predicting output parameters. To analyze theStatistical error,measuring tools Normalized Root Mean Square Error (NRSME) and Mean Absolute Percentage Error (MAPE) were used in the proposed ANN model.NRSME and MAPE values for the developed ANN model were 0.00908–0.047516 and 0.014502–0.118807 respectively. Regression coefficients (R), Coefficient of determination (R2) were implemented to validate the proposed model. The range of values obtained for R and R2were 0.998742–0.999953, 0.997487–0.999906 correspondingly. The low range of error and closeness of R and R2 values nearest to 1 confirms that the proposed model can predict the output parameters with acceptable limits of error.

Topics & Concepts