Research On Diesel NO_x Transient Emissions Predictive And Optimization Based On Support Vector Machine

With the strengthening of environmental awareness, vehicle emission regulations become more stringent; engine transient emissions also become the focus of the study of engine emissions. Establish NOx transient emissions forecasting model will have a good guide to NOx transient emissions Experimental study. However, there is no clear theory to the law of the NOx emissions under the engine transient conditions. In this paper, supporting vector machine technology, using the experimentally measured transient emission data, establishing and validating the predictive model to forecast the NOx transient emissions.Finding the optimal method and the effective forecast using a finite number of samples of experimental data is important to build a forecast model. With the extensive application of artificial intelligence, machine learning is widely used in the prediction. The more mature theory in machine learning is statistical learning theory. Support Vector Machine (SVM) is the most practical applications in Statistical Learning Theory (SLT), and it is the new tools to solve the problem of machine learning with the optimization method.Support Vector Machine is mainly used in classification and regression problems. In this paper, we use the engine transient emissions data measured by experiment, and select the appropriate kernel function to establish the regression model to predict. Modeling will be optimized to model main parameters using a grid search method, genetic algorithm, particle swarm optimization, and then compare the model forecasting effect, and select the best forecasting model. the relative error of transient conditions emissions prediction is less than3%. The results show the model can be used to predict transient NOx emission. As the input was mostly control parameters, the model for in-depth study can be carried out. Finally, using the obtained best model, changing the value of the model input parameters appropriately to predict NOx emissions, and eventually get the optimum calibration of the reducing NOx transient emissions.