An integrated process control and diagnostics system for hot rolling mills based on comparison of physical data and mathematical process models using artificial intelligence

The aim of this multinational research project is to construct a generic, intelligent, sophisticated, portable, modular and configurable monitoring and diagnosis system for HSMs and other similar processes in steel plants with a view to improve the process control strategy. To incorporate as much as possible the multitude of scenarios prevailing in mills and for developing generalised strategies, four research locations were identified to carry out the three main tasks of the collaborative project. The main goal to be achieved at TKS and Corus is the earlier detection of faults and malfunctions of the measuring devices to greatly reduce mill downtime, periods of unnoticed non-optimal production and product quality problems. The general aim of the Arcelor part was to implement an intelligent diagnosing system to improve the width performances in the HSM. At Ilva, regaining the performance of the mill pacing, which had been downgraded by the revamping activities in the past decade, was the main target. The fault detection system developed using neural network algorithms, filter functions and descriptive statistics which has been implemented at TKS and Corus has significantly reduced the routine unnoticed malfunctions of pyrometers and other gauges. Additionally, some application-oriented and case-specific diagnosis routines for the furnace thermocouples were implemented at Corus. At Arcelor, a supervisory system for width gauge performances and several models to improve the width quality of strips were developed and installed. The new mill pacing strategy developed at Ilva using a modified material-flow control scheme reduces the gap time considerably and consequently increases annual production by about 7 %.