Modeling On The Characteristics Of Vermicular Graphite Cast Iron And It's Application Based On The BP Neural Network

With increasing of the diesel engine power density, the HT250 cylinder head which made by Guangxi Yulin Diesel Engine Co.Ltd could not perform well enough for the usually application. It is well known that vermicular graphite cast iron is fit for the cylinder head of highly strengthened diesel engine, and it can bear the impact of higher temperature and higher pressure, also it has excellent heat conduction ability and higher strength. So the development of the cylinder head made by vermicular graphite cast iron must be conducted immediately.The engine cylinder head made by vermicular graphite cast iron is not easy to control because of many factors influenced on the production quantity and its optarion limits are very narrow.We must find out the most important influencing factor of them and control it firstly in order to develop the cylinder head made by vermicular graphite cast iron .And this needs a lot of experiments to do.For the sake of reducing the developing cost of cylinder head made by vermicular graphite cast iron and accomplishing this task within smaller time, a model based on neural network theory with less testing data is presented in this paper. This model can be used to predict the properties of vermicular graphite cast iron. In addition it was used to analyze the influencing factors on vermicular graphite cast iron using the orthonormal planing method. The main works are shwon as follows:(1)According to vermicular graphite cast iron production of Guangxi Yulin Diesel Engine Co.Ltd, 14 influencing factors from its production process on vermicular graphite cast iron were analyzed.(2) On the basis of analyzing the algorithm, nonlinear mapping ability, generalization capability, the existing problem and its corresponding improving methods, and applicability of modeling on the properties of vermicular graphite cast iron using BP neural network were discussed;(3)The properties of G2000 vermicular graphite cast iron cylinder head were measured using the foundry production line and instrumentations of Guangxi Yulin Diesel Engine Co.Ltd. And 23 learning samples and 8 test samples were obtained;(4)On the baisi of the analyzing of all kinds of algorithms, Bayesian Regularization algorithm was uesed to training modeling of the vermicular graphite cast iron.And higher learning and prediction precision were ontained; (5)A model for vermicular graphite cast iron properties based on the BP neural network was established. And then, it was used to forecast model of the vermicular cast iron properties successfully;(6)A forecast model of the vermicular graphite cast iron properties was used to analyze the influencing extent of all factors by ways of orthonormal planing method.And the most important influencing factor has been found out;The basis of this prediction method is the test data, its prediction results and accuracy are not dependent on the vermicular graphite cast iron mathematic model. Due to the less experiment, it can save lots of research costs and period. Additionally it was applied to analyze the effecting factors on vermicular graphite cast iron and finding out the primary influencing factor of them by ways of orthonormal planing method.So this is a really newly and and accurate method.It follows that this new method with lower requirements can facilitate research of the cylinder head made by vermicular graphite cast iron and own higher theory meaning and wider engineering practical value.