Research On Prediction Of NO_x Emission Characteristics Of Marine Diesel Engine Based On BP Neural Network

With the development of the world shipping business, a large number of vessels are put into service. And the air pollution caused by the harmful gases which are emitted by the marine diesel engines becomes more and more serious. In view of this situation, IMO(International Maritime Organization) formulated the revelant conventions to limit the emission of these harmful gases. So, knowing well emission characteristics of these harmful gases is of vital importance. In this paper, the neural network theory is introduced into the research on the prediction of NOX emission characteristics, making use of MATLAB software to establish the BP neural network model to predict and analyze the NOX emission characteristics. BP neural network model is based on the general test equipments, less test datas and not relevant to the engine's mathematical model. This method needs smaller size of experiment and can save the test cost. So,it is a quick, accurate and effective method. The main work is as follows:Firstly, it discusses the methods to research the emission characteristics of marine diesel engines and the current research both at home and abroad, introducing the basic knowledge relevant to BP neural network modeling.Secondly, it designs the experimental program to choose the learning samples in Uniform Design method to choose the representative sample working condition points. Relevant datas got from W6L20C diesel engine tests are changed to be the NOX mass flow rate of every working condition point by Carbon\Oxygen balance method, and the training and test samples for modeling are gotten. The relevant structure and training parameters like hidden layer neurons, frequency of training,learning rate, etc. are confirmed through the comparison of the net errors and property. The marine diesel engine NOX emission characteristics prediction model is finally confirmed, utilizing MATLAB software to train the net and analyze the generalization ability of the net.Finally, the NOX emission characteristics of W6L20C diesel engine is predicted and analyzed by the established BP neural network model.This established model has a higher predictable accuracy. The NOX emission prediction coincides with the samples well. The relative errors of the train samples simulation outputs are within 0.4%;the relative errors of the test samples are within 5%.This model can predict the NOX emission characteristics of marine diesel engine effectively.