Application of Self
Organizing Map (SOM) to model
a machining process
Mohamad Saraee
School of Computing, Science and Engineering,
University of Salford, Salford, UK, and
Seyed Vahid Moosavi and Shabnam Rezapour
Industrial Engineering Department,
Amirkabir University of Technology, Tehran, Iran
Abstract
Purpose – This paper aims to present a practical application of Self Organizing Map (SOM) and
decision tree algorithms to model a multi-response machining process and to provide a set of control
rules for this process.
Design/methodology/approach – SOM is a powerful artificial neural network approach used for
analyzing and visualizing high-dimensional data. Wire electrical discharge machining (WEDM)
process is a complex and expensive machining process, in which there are a lot of factors having
effects on the outputs of the process. In this work, after collecting a dataset based on a series of
designed experiments, the paper applied SOM to this dataset in order to analyse the underlying
relations between input and output variables as well as interactions between input variables. The
results are compared with the results obtained from decision tree algorithm.
Findings – Based on the analysis of the results obtained, the paper extracted interrelationships
between variables as well as a set of control rules for prediction of the process outputs. The results of
the new experiments based on these rules, clearly demonstrate that the paper’s predictions are valid,
interesting and useful.
Originality/value – To the best of the authors’ knowledge, this is the first time SOM and decision
tree has been applied to the WEDM process successfully.
Keywords Self Organizing Map, Decision trees, Artificial neural nets
Paper type Research paper
1. Introduction
The Self Organizing Map (SOM) is a prominent unsupervised neural network model
providing a topology preserving mapping from a high-dimensional input space onto
a 2D map space. This capability allows an intuitive analysis and exploration of
interesting and previously unknown knowledge (Tomsich et al., 2000).
The SOM is useful in many applications where the dimensionalities of the feature
spaces to be analyzed and the amount of data encountered are too large to allow a fast,
interactive training of the neural network. The SOM provides a form of cluster analysis
by producing a mapping of high-dimensional input data onto a usually 2D output
space while preserving the topological relationships between the input data items
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1741-038X.htm
The authors would like to thank the Machining Workshop Centre of Isfahan University of
Technology for their collaborations and providing the dataset for this experiment.
JMTM
22,6
818
Received September 2010
Revised April 2011
Accepted April 2011
Journal of Manufacturing Technology
Management
Vol. 22 No. 6, 2011
pp. 818-830
q Emerald Group Publishing Limited
1741-038X
DOI 10.1108/17410381111149666