Statistical Design and Evaluation of Experiments in Steel Production

Abstract

Making experiments under industrial production conditions gives many times difficulties in varying the important parameters in a relevant way in order to significantly Influence the process and thus make it possible to measure the result. The common way of varying one parameter at a time, gives rise to many experiments and mostly, large difficulties in the evaluation.The statistical design method, on the other hand, where the parameters are varied systematically and simultaneously, gives possibilities to influence the process in a more efficient way, thus reducing the necessary number of experiments, and at the same time to obtain more reliable results. In spite of the linear approach presented here, it is always possible to evaluate any kind of influence on a process, by choosing the studied parameters in a relevant way. This is simply done by means of a suitable variable transformation before the experimental matrix is set.Analytical evaluation of the experiments often raises a voluminous mathematical treatment of the results. However due to the linearity of the primary functions, the experiments can be evaluated by means of a simplified linear procedure, and even large experimental matrices can easily be evaluated only by means of a simple spread sheet program.By application of the method to industrial problems, a calculation model was developed for the prediction of process parameters for quenching and self tempering of rebar, as a function of requirements and chemical composition of the steel. Alternately the required chemical composition can be predicted when required mechanical properties and quenching parameters are given, or the mechanical properties can be predicted when composition and quenching parameters are given. Another use of the method has been to optimize the finish rolling and coiling temperatures for on line treated high carbon steel wire rod.