Remaining useful life and wear estimation of the refractory bricks of the ladle lining by Artificial Intelligence

The steel industry is a key driver of new developments in the refractory industry due to its high market share (>70%) and the harsh conditions involved in the steelmaking processes. In addition, the annual refractory consumption in a special steel plant that produces more than 750.000 ton of steel per year, can reach around 10.000 ton. The RFCS European E-CO-LadleBrick project aims to develop an ecological and economic waste management of the ladle refractory bricks by implementing circular economy criteria (Reduce, Reuse, Re-manufacture and Recycle). Due to the higher impact, it is very important to promote the reduction in the waste generation by optimizing the use of the refractories. Could AI help, in this sense, to the first of these R’s and reduce the refractory consumption in the ladles? To answer this question, this work proposes a methodology by using two sources of information: First, refractory remaining thickness measurements obtained through a 3D laser equipment, and second, the information of the operation variables with the greatest influence on refractory degradation, collected heat-by-heat by a steelmaker. By means of using this information, this paper presents a Machine Learning tool with the main purpose of optimizing the refractory consumption: 1) by estimating the Remaining Useful Life with the information of the operation variables and 2) by estimating the wear of the MgO-C refractory lining that a specific heat can suppose. The work has been validated in a real application with data from a relevant steel company of the Basque Country (Spain).