Particulate matter and nitrogen oxide emissions prediction using artificial neural network for diesel engine running on biodiesel-diesel fuel with nano-additive

Fossil fuel combustions from automotive industries and vehicles causes second highest emission of gases influencing global warming and climate change. Biofuels and biodiesels are renewable energy sources and alternative candidates to fossil fuel but have limitations creating requirement for blending and application of additives to biodiesel-diesel fuels. Nano-additives is promising due to higher atomic level and surface area to volume ratio; however, higher cost of nano-additives makes random selection for testing many varieties difficult, also; nitrogen oxide (NOx) emissions and particulate matter (PM) from unburnt nanoparticles is a major challenge. This work therefore uses artificial neural network (ANN) feed forward back propagation as learning algorithm to predict PM and NOx emissions using experimental data from test conducted on a single cylinder diesel engine running on palm oil biodiesel blended with conventional diesel and Iron (II) oxide (Fe₂O₃) nano-additive stabilized in isopropyl as surfactant at three engine loads (25%, 50%, 75%). Levenberg-Marquardt was used for training data with 6 input, two hidden layers of 5 set (10 total) and 2 output layers. The target parameters (NOx and PM) were accurately predicted by ANN training, the highest performance denoted by R and R² of values 0.99999 and 0.9999 respectively. Based on experimental results and weight of input parameters, it is conclusive that higher percentage by volume of nano-additive reduces PM until optimal level before ‘excess’ dose Fe₂O₃ nano-additive causes higher PM emitted; lower nominal NOx resulted with continuous nano-additive increment for all load conditions. A satisfactory ANN application for prediction was achieved.