Journal of University of Science and Technology Beijing
Volume 13, Number 6, December 2006, Page 1
Corresponding author: Ling Wang, E-mail: linda-gh@sina.com
Materials
Application of support vector machine in the prediction of
mechanical property of steel materials
Ling Wang, Zhichun Mu, and Hui Guo
Information Engineering School, University of Science and Technology Beijing, Beijin 100083, China
(Received 2006-02-11)
Abstract: It is a key point for the mechanical property predictive system to investigate the influences of the important parameters in-
cluding steel chemical compositions and hot rolling parameters on the mechanical properties of steel. To improve the prediction ac-
curacy, support vector machine was used to predict the mechanical properties of hot rolled plain carbon steel Q235B. Support vector
machine is a novel machine learning method, which is powerful for the problem characterized by small sample, nonlinearity, high
dimension with a good generalization performance. On the basis of the data collected from the hot rolling supervisor, support vector
regression algorithm was used to build prediction models, the off-line simulation indicates that predicted and measured results are in
good agreement.
Key words: mechanical property; support vector machine; support vector regression; chemical compositions; hot rolling parameters
1. Introduction
The mechanical property of steel materials [1] is
important to steel quantity, which is influenced by the
microstructure, chemical content and all kinds of fac-
tors in manufacturing. The main compositions of the
mechanical property include elongation rate, yield
point and tensile strength. The hot rolling process is a
complicated process. As we have known, it is difficult
to build an exact model of the mechanical property
based on the basic principle, so the practice of the
mechanical property prediction can reflect steel’s
property more overall, meanwhile reduce the time of
steel property test, shorten the cycle of production
supplying.
It is a key point for the mechanical property predic-
tive system to investigate the influences of the impor-
tant parameters including steel chemical compositions
and hot rolling parameters on the mechanical proper-
ties of steel. The regression equations for the effect of
the microstructure and chemical content on the me-
chanical property of steel have been obtained using
traditional regression analysis method under some test
conditions [2], but it has poor prediction ability in
complicated and nonlinear systems. To improve the
prediction accuracy, the artificial neural network
(ANN) is available especially to apply in nonlinearity
system for its better learning precision and summariz-
ing ability. ANN has been applied to the prediction of
mechanical properties of steel materials [3-4, 8]. De-
spite many of these advances, there still remain a
number of weak points such as the existence of many
local minima solutions, poor generalization capability
outside the range of training data and how to choose
the number of hidden units. Major breakthroughs are
obtained at this point with a new class of neural net-
works called support vector machines (SVM), devel-
oped within the area of statistical learning theory and
structural risk minimization [5]. Established on this
principle, SVM will achieve an optimum network
structure and better generalization performance than
other neural networks. Moreover, unlike other ma-
chine learning methods, the number of free parameters
in the SVM does not depend explicitly on the input
dimensionality of the problem, which suggests that
SVM be especially useful in problems with a large
number of inputs. Currently, SVM has many applica-
tions in pattern recognition, function estimation, signal
procession, control and other fields [5-6]. The method
to solve regression problems using SVM is called as
support vector regression (SVR), which is one of the
most important application of function approximation.
In this article, SVR was adopted to predict the me-
chanical property of steel materials during hot rolling
process. A brief summary of SVR was given. The
adequate mechanical property data and manufacturing
process control data were applied to the model and
good prediction results were obtained.
资源评论
weixin_38609732
- 粉丝: 8
- 资源: 963
