The Prediction Of Mechanical Properties Of High Speed Wire Based On Neural Networks

It is always the aim for the metallurgists to improve the properties of the products. In order to obtain this target, it is necessary to predict the properties of the product to be produced. Then, the metallurgists can design the composition of steel and the producing technology according to the results which they predict beforehand. So they can reduce the cost of the production and shorten the time for research and development.The high speed wire of high carbon steel of some companies of our country has the phenomenon of unstable properties and drawing rupture, which has brought the difficulty for drawing. The steel's chemical composition and the process parameters influence the mechanical properties of high speed wire. But there are many disturbing factors during the producing process, and there are no traditional formulas to describe the relations between the mechanical properties and the chemical composition or process parameters. In other words even if we simplify the model by sacrificing somewhat precision and get a model for the real time, confronted with some uncertain factors such as the change of the climate, we show no ability to overcome the difficulty. However, artificial neural network has the superiorities in the modeling of the complicated systems. Its ability for any non-linear mapping's approach, adaptation for the uncertain factors, self-study ability and its ability that it can memorize and dispose a great deal of information make it very useful in many fields such as building model or identifying in the complicated systems and predicting for the dynamic output. Therefore, I have made some research on the artificial neural network and the influence on the properties of high speed wire and obtained some prediction models for the production of high speed wire based on BP neural network.There are three chapters in this thesis, in which the first section is the description of research situation and some documents, including the origin of the task, the actuality of research, the research and development on artificial network in or out of our country and the application of artificial network in material processing; The second section describes neural network' s basic knowledge, including BP network and its studyingmethods, BP network' s designing, MATLAB and the tools box of neural network and the development of artificial network and its qualitative conclusion. The third section includes the basic research of the prediction model and the mechanical property prediction of high speed wire by using neural network. In this section, I analyzed every influencing factor on the properties of high speed wire, including chemical composition and producing technology condition, and through experiment comparison with each other I constructed the prediction models for high speed wire. In addition, this section also includes the application of the models, the discussion of online application technology of property prediction of high speed wire, and the experiment conclusions.We can see that the models have very high precision through the application of the models, and the errors between the prediction and the real measured value are less than 10%, on the other hand, the prediction precisions based on the models are much higher than that based on the formulas of company or regression.