Search

Abstract NTP based Ozone injection cascaded with catalysts/adsorbents for diesel exhaust treatment has been demonstrated by various researchers as a competent method for mitigation of nitrogen oxides. In this present study, one more NTP reactor is cascaded instead of catalysts/adsorbents to treat the ozone mixed diesel exhaust. While most of the laboratory studies were carried out using expensive dry air cylinders for ozone generation, in the present study, ozone has been generated from atmospheric air only. The nitrogen oxides removal/conversion efficiency of this process depends on various electrical and physical parameters. A dimensionless relation has been obtained between the ratio of summation of NO and NO 2 concentrations after treatment to that before treatment and other parameters affecting nitrogen oxides removal using dimensional analysis. Prediction of sum of NO and NO 2 concentrations with the variation in six dominant parameters have been done using nonlinear regression technique. Six parameters, i.e. ozone concentration, voltage, frequency, flow rate, temperature and discharge gap have been varied to collect the experimental data required to train and test this model. The predicted sum of NO and NO 2 concentrations are found to be well matching with the experimental results. The RMS error is found to be 3.7 % and 5.9 % respectively for training and test data. [ABSTRACT FROM AUTHOR]

Abstract Non-thermal plasma technique (NTP) for mitigation of NO X from diesel engine exhaust is a laboratory proven technique with a good number of experimental studies. However, there are constraints which are preventing this technique to appear in practical applications. Prior prediction of exhaust characteristics with the NTP treatment according to the variations in its operating parameters can be a step ahead to make this technique applicable in real time. In this present study, experiments are conducted by varying parameters voltage, flow rate, temperature, discharge gap and initial sum of NO and NO 2 concentrations. Six hundred number of datasets are collected to train and test the predictive models which are made using multilayer perceptrons (MLP). The objective of these models is to predict the sum of NO and NO 2 concentrations at the downstream of the reactor during NTP based diesel exhaust treatment. As a part of the study, optimum number of neurons in the hidden layers is also found out. The root mean square error (RMSE) of an MLP with two hidden layers, each of 20 neurons, is found to be 3.29 ppm. Results assure that the prediction of sum of NO and NO 2 concentrations can be achieved with a good accuracy using MLP. [ABSTRACT FROM AUTHOR]