INTEMA   05428
INSTITUTO DE INVESTIGACIONES EN CIENCIA Y TECNOLOGIA DE MATERIALES
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Self organizing neural networks for direct piercing mandrels wear study
Autor/es:
CARR, GUSTAVO EDUARDO; MESCHINO, GUSTAVO; CHAPETTI, MIRCO DANIEL; CONDE, RAÚL HORACIO
Lugar:
Rosario, Argentina
Reunión:
Conferencia; 17ma Conferencia de Laminación y 4ta Conferencia sobre usos del acero; 2008
Institución organizadora:
Instituto Argentino de Siderurgia
Resumen:
An important topic in industry is to improve useful life of mandrels for direct piercing seamless tube making. In the present work a technique for the study of variables influence on mandrels wear is proposed. A laboratory scale scale Mannesmann rotary piercing prototype was designed and built. Piercing sequences using this prototype were performed, from which process data was collected. Mandrels profile variation due to wear was obtained from digital photography. The collected data were used to train an artificial neural network. Using the aforementioned equipment and technique, the group of factors involved in mandrels lifespan could be determined. Artificial Neural Networks in the form on Self Organizing Maps was very useful to identify important variables and to implement different wear detection criteria without problems. Some combination of process parameters variables values that determined minima of wear indicative variables were found.Due to its high adaptability, the aforementioned technique is proposed as an alternative and complement to different wear study criteria (based on production variables, etc.)  as well as full scale implementation.