Development and application of soft computing and data mining techniques in hot dip galvanising

  1. Sanz García, Andrés
Supervised by:
  1. Francisco Javier Martínez de Pisón Ascacíbar Director

Defence university: Universidad de La Rioja

Fecha de defensa: 27 March 2013

  1. Eliseo Pablo Vergara González Chair
  2. Alpha Verónica Pernía Espinoza Secretary
  3. Enrique Alegre Gutiérrez Committee member
  4. Manuel Castejón Limas Committee member
  5. Javier Alonso Ruiz Committee member

Type: Thesis


In a world in which markets are more globalised and continuously evolving, industry need new tools to enhance their flexibility and maintain competitiveness. A key strategy is discovering useful knowledge through the information gathered from production processes. In recent decades, companies have increased investments for improving data storage capacity. The huge volume of information stored by companies and its high complexity render traditional methods of data processing useless. However, the use of tools to extract the information hidden inside databases is still under development. The creation of such methodologies can make the key points of industrial processes more flexible. With the aim of solving this problem, new computer-based methodologies derived from data mining are being developed. By using these methods, researchers are seeking to obtain non-trivial hidden knowledge from historical records of industrial processes. For this reason, data mining has now become a crucial discipline for performing automatic searches inside historical industrial databases, contributing to industrial development and advancement. This thesis focuses on the use of data mining techniques to develop helpful methodologies for tuning industrial production lines. The goal is to increase flexibility in response to the need to meet new consumer expectations. The methodologies developed have been used to improve a continuous galvanising line. Bearing in mind its complexity, our aim is to explore the opportunities that data mining techniques can offer for improving this industrial process. The goal of the first part is to seek novel non-trivial knowledge in the form of patterns to explain failures in production. An overall methodology that integrates both data management and association rule mining is proposed to capture frequent events that coincide when there is a failure in the process. The second part focuses on improving the modelling of non-linear systems using historical information, combining different soft computing techniques. These improved the estimation of temperature set points for the annealing furnace on a galvanising line. The contributions presented in this doctoral thesis provide evidence of the huge potential that data mining has for obtaining useful comprehensible knowledge from industrial processes.