Application of data-based technologies to demonstrate online quality control of mini-mills

The aim of this project was to develop and to implement concepts for the on-line quality control of mini-mills. As the product quality is a result of a wide variety of factors and their combinations, improvements of the product quality shall be achieved by a comprehensive approach in the evaluation of available data in a through process fashion. In order to realize quality control in a mini-mill, the tight interaction of process stages results in special challenges. The layout of the mini-mill at ACB and the principle approach for the quality control is shown in Figure 1. The scrap entering the electric arc furnace runs in about two hours through the whole process chain via ladle furnace, thin slab caster, tunnel furnace and hot rolling mill. In the sequel, data coming from different process units are required to be handled as soon as possible. Crucial for the operation of a through-process quality control system (QCS) is the efficient access to the process, plant and product data along the production chain. After all available data sources were investigated, relevant data have been selected for integration into the new QCS. First a new data acquisition and storing system for collecting relevant data from the whole process chain was integrated. Here the data were collected near-online from the process computers close to the production, so that these data can be used for monitoring of the processes. Different types of data are considered which are mainly existent in the process computers but also highly sampled data from special acquisition systems can be processed properly. To effectively work with the incoming data, they are first aggregated and stored related to length segments of the product. Furthermore, derived features are calculated and used for monitoring the processes with regard to product quality among other purposes. For the efficient near-online handling and storing of all incoming data a special table structure has been developed and successfully put into operation. By the introduction of so-called control-points along the process chain, the availability of data from up-stream process units is ensured for properly defined points in the production chain like the end of steel melt shop and below the cast mould. Here heuristic or data-based models can evaluate the history of the product concerning the expected final product quality. This information can be used in principle in forward or backward direction to either make corrective actions influencing the actual product (e.g. heat or slab) or consecutive products. Within this project the focus for quality control was laid on the surface quality of the hot rolled strip and here in particular on the appearance of shells caused by non-metallic inclusions (NMI). In order to measure the surface quality reliably, an automatic surface inspection system (SIS) needed to be installed in the hot rolling mill. This involved modifications in the hot rolling plant to ensure proper operation of the SIS. The hot rolled strip is inspected on the top side after passing the last stand of the hot rolling mill. Due to the detection accuracy of such a system, comprehensive information about an important quality feature of the final product is available. Solutions for the evaluation (grading) of the quality were developed based on several defect classes describing NMI-type defects on the surface.