A model and data-based analysis of longitudinal surface cracking in the continuous casting of middle carbon steel slabs

Quality is a critical consideration in the continuous casting of steel. One quality item that is important in the casting of middle carbon slabs is longitudinal surface cracking. This work is an attempt to understand and predict longitudinal surface cracking, using model quantities which are thought to be relevant to cracking and actual plant data for middle carbon slabs. The model quantities which are used to relate with cracking are steel surface temperatures at the mold exit and in the spray zones, average and variation in shell thickness early in the mold and plastic strain accumulated near the solidification front. Shell thickness variation is calculated on the basis of the variation in mold thermocouple readings. The results of Fourier analysis of these readings and calculations of shell thickness variations, along with a model description for shell deflection, support the idea that shell thickness irregularity occurs as a result of a pulse in the heat flux condition and a response to that pulse by the solid shell. The results of statistical analysis of longitudinal surface cracking suggest that this irregularity in shell thickness, as well as the average shell thickness early in the mold and the surface temperature in the sprays have significant effects on cracking. These results and others fit with a general understanding of longitudinal surface cracking in terms of important aspects such as shell thickness irregularity, stress state and steel ductility. This understanding would be difficult to obtain without the use of modelled quantities in conjunction with operating parameters. Using correlations obtained for cracking, predictions are made for an untested set of data using only operating information in one case and including the modelled quantities in the other. The results of these predictions are primarily dependent on the conservativeness of a given strategy and are not noticeably better for the addition of derived quantities. Apparently, the operating variables are sufficient to provide reasonable prediction of when cracking will occur. Regardless of how the prediction is done, the evaluation of prediction results for complete data over the spectrum of strategies allows for actual operating strategies to be set according to philosophies desired for a given caster.