STATISTICAL METHODOLOGY APPLIED TO PREDICTION OF MECHANICAL-PROPERTIES OF NIOBIUM MICROALLOYED STEELS PRODUCED AT CLABECQ STEEL PLANT

This paper deals with the statistical modelling of the mechanical prop erties of niobium microalloyed steels produced at Clabecq Steel Plant by accelerated cooling. An appropriate methodology is proposed for pre dicting yield strength, tensile strength, and toughness. This modellin g developed exclusively from available process data, and begins with a n exploratory approach including the use of factor analyses and multiv ariate techniques more oriented towards prediction such as regression and segmentation. In the exploratory approach, the collected data are filtered by one- and multidimensional procedures and subjected to prin cipal components analysis, a technique to describe the interrelationsh ips existing between the numerous potential predictors. A segmentation method, automatic interaction detection, helped to detect interaction s between the important variables and, finally, through a general line ar model prediction, a final expression was derived. All these statist ical treatments have stressed many well known tendencies, such as the major effect of niobium content on mechanical properties and the influ ence of cooling time or precipitation time on yield strength. It has a lso been observed that a reduction of manganese content is recommended to improve both yield strength and impact strength. Except for the im pact strength, application of the statistical models to further data h as shown small residuals between estimated and measured properties. (C ) 1996 The Institute of Materials.