MULTIPLE-REGRESSION ANALYSIS OF JOMINY HARDENABILITY DATA FOR BORON TREATED STEELS

The relations between chemical composition and their hardenability of boron treated steels have been investigated using a multiple regressio n analysis method. A linear model of regression was chosen. The free b oron content that is effective for the hardenability was calculated us ing a model proposed by Jansson. The regression analysis for 1261 Stee l heats provided equations that were statistically significant at the 95% level. Ail heals met the specification according to the nordic cou ntries producers classification. The Variation in chemical composition explained typically 80 to 90% of the variation in the hardenability, In the regression analysis elements which did not significantly contri bute to the calculated hardness according to the F test were eliminate d. Carbon, silicon, manganese, phosphorus and chromium were of importa nce at all Jominy distances, nickel, vanadium, boron and nitrogen at d istances above 6 mm. After the regression analysis it was demonstrated that very few outliers were present in the data set, i.e. data points outside four times the standard deviation. The model has successfully been used in industrial practice replacing some of the necessary Jomi ny tests.