Intelligent evaluation of melt iron quality by pattern recognition of thermal analysis cooling curves

The quality of an iron melt which refers to the soundness of melting and subsequent treatments of the melt can be identified and recognized with its thermal analysis cooling curve. To compare two cooling curves, both the separating distance of the two curves and the shape similarity of the curves should be considered. A comprehensive parameter Omega can be used to identify the difference of the two cooling curves. When Omega is at minimum, the two curves must be the closest couple among all cooling curves. It is found that the difference of every feature related to melt iron quality converges to zero when the value of Omega approaches zero. Two databases of grey and nodular cast irons have been set up, in which the thermal analysis cooling curves, composition, microstructure and mechanical properties are included. For the prediction of nodularity of ductile irons, an accuracy of 5% is realized if the value of Omega of the two matching cooling curves is less than 2 degrees C. This method is self-adaptive to the production condition and has been adopted in several foundries. (c) 2004 Elsevier B.V. All rights reserved.