"DATA MINING APPLIED TO EXTRUSION PROCESS IMPROVEMENT OF ELASTOMERS"

Author: Claudia Barreto Cabrera.

Abstract:

This work has focused on the extrusion process of rubber profiles for automotive industry. Rubber profiles are
long products and a variety of forms that are placed in cars to provide air tightness and sealing doors and 
windows.Using profiles avoids several drawbacks in the car, such as unwanted noise in the external environment, 
vibrations and the entry of water and air.

The production line of rubber profiles is composed of several processes such as extrusion, knurling, the 
vulcanization, the flock and cutting the size of the product mentioned, according to car model. From the above 
processes the investigation focused on the extrusion, as though its use dates back to 1800 [Bhowmick et al.(1994),
Rauwendaal (2002)], there is still much to learn from him.

According to the National Consortium of Rubber Industry in Spain, in 2005, 40% of the sales of rubber products in
this country were aimed at the automotive sector [CNIC (2005)]. Furthermore, Spain is the third largest European 
manufacturer of rubber products since 2001.

In order to facilitate both quality control and production control product is carried out the detailed log of the
important parameters of the profiles manufacturing. That is, it makes product traceability. What can provide 
knowledge of the production process to determine, at any given time, the causes of a possible defect, even after 
leaving factory and go back, if necessary, the quality of the raw materials used. It was assumed that problems 
occur during the extrusion process rather than after it.

In reviewing the literature, have found successful applications in industry [Rodríguez et al. (2003), Martínez de 
Pisón et al. (2003), Martínez de Pisón et al. (2005)]. As it has been assumed that by applying Data Mining to 
information that is stored, particularly the parameters obtained from the production of rubber profiles, valuable 
information can be obtained, which would propose strategies to improve the quality of final product.

In particular, in this work, were used the Support Vector Machines to model the best fit of the curves extruders
and identify best practices, creating a support system for the operation. It has been developed a methodology 
used to classify and model semi-automatically the best curve fit of an extrusion process of elastomers for the 
automotive industry. These boot machine, due to their difficulty, are done manually and depend on operator 
experience and type of product to extrude. If the product does not get properly extruding produces a considerable
loss of time and raw material. The ultimate goal of work presented in this work focuses on finding and modeling, 
from the historical database of manufacturing, the best starting curves made for each of the profiles produced. 
With new models obtained can automate the process and reduce time spent in those starts with a consequent 
increased production, improved quality, reduced stress and defective material for production personnel.