Research projects

Ongoing projects

NEUROLL, Task 2, Task 3: Knowledge extraction and Process monitoring

The study has concentrated in finding correlations between quality parameters (e.g., width and thickness deviation and surface defects) and other process parameters using traditional correlation analysis and the SOM-based methods. The results are utilized in improving the process in the general level and finding the variables for the process modeling. Several, mostly SOM-based, approaches for process state visualization and monitoring are being considered.

Contact: Jussi Ahola, Juha Vesanto

NEUROLL, Task 10: Surface quality

Studying the surface quality of hot rolled strip. Finding a model for surface defects by examing process data. Creating off-line model for pointing risks in the process.

Contact: Jukka Parviainen, Esa Alhoniemi

Continuous casting

Continuous casting is a large scale industrial method for producing steel slabs for further refinement, e.g. for hot rolling. There may come up surface ruptures on the steel slab during the casting. The ruptures lower the product quality, and sometimes they may cause a breakthrough of liquid steel that leads to a long maintenance pause in the casting plant. The casting process is, however, monitored by thermocouples inside the mold. The obtained temperature data is analyzed using SOM-based visualization tools in order to get efficient features for rupture and breakthorugh warnings in the automation system.

Contact: Johan Himberg

Metsäteho

Data collected by forest harvesters is analysed. Data consists of measurements from individual trees and forest areas as a whole. The main goal is to apply unsupervised clustering techniques to group similar forest areas together. This clustering could prove to be useful when controlling the resource management for Finnish sawmills.

Contact: Pekka Hippeläinen, Johan Himberg

Closed projects

Sellunkeitto

Contact: Esa Alhoniemi

ENTIRE

Contact: Juha Vesanto

Postipankki

Contact: Johan Himberg

Umbrella projects

MENES

MENES (Methods for Demanding Neurocomputing Applications) is a part of Finnish technology program "Applications of Intelligent and Learning Systems" funded by TEKES. Look more at MENES project on MENES web pages. Our part of MENES was monitoring and modelling of complex processes. ENTIRE was a project for forest industry.

IMPRESS

Partners UPM-Kymmene, Metsäteho and Perusmetalli.

NEUROLL

NEUROLL (Application of neural network based models for optimization of the rolling process) is a large EU project from February 1997 lasting three years. The project coordinator is a Finnish steel manufacturer Rautaruukki Oy. Partners are Siemens AG, Krupp Hoesch Stahl AG and Freiburg University of Mining and Technology from Germany, MEFOS from Sweden and our laboratory. Our purpose is produce data analysis methods useful for improving quality in steelmaking process.

This page is maintained by juuso@mail.cis.hut.fi.
Last modified Oct 12th 1998.
URL http://www.cis.hut.fi/projects/ide/projects.html