Improved Ferrite Number Prediction that Accounts for Cooling Rate Effects

INTRODUCTION Stainless steel welds characteristically consist of a two-phase austenite plus ferrite microstructure. Ferrite levels may vary from a few percent in austenitic stainless steel welds to more than 50% in duplex stainless steel welds. The ability to predict the ferrite content in these welds is essential for many reasons. To a large extent, the final ferrite content determines a weldment's properties such as strength, toughness, corrosion resistance, and long-term phase stability. In addition, ferrite content is a useful indicator of the mode of solidification, which strongly influences the hot-cracking propensity during welding. Over the years, various models have evolved to try to accurately predict the ferrite content in stainless steel welds. Constitution diagrams, in which the overall alloy composition is converted into two factors, a chromium equivalent (Cr eq) and a nickel equivalent (Ni eq), have been developed to predict FN in welds. Many diagrams have been proposed since the original diagram of Schaeffler 1-7 , with the WRC-1992 diagram 7 being the most recent and most accurate. The various versions of constitution diagrams differ primarily in the coefficients that are used to convert the alloy composition into the Cr eq and Ni eq ; an extensive review is given in reference 5. In most commonly used constitution diagrams, the weighted coefficients are constant, and this means that a given alloy addition's influence is the same regardless of that element's concentration or the concentration of any other alloying additions. In those cases where non-constant coefficients were proposed, the applicability of the diagram is limited to a restricted composition range. Clearly, constant coefficients cannot represent real behavior very well. For example, the effect of carbon should be very different depending on whether carbide forming elements are present or not. This limitation has been removed with the development of predictive models based on an artificial neural network analysis 8-11. Artificial neural networks are ideally suited for predicting ferrite content because they offer improved flexibility, robustness, and accuracy as a consequence of their use of non-linear regression methods. Two recently developed neural network models predict FN as a function of the concentration of 13 elements 8-10. These models have been shown to be more accurate than the WRC-1992 constitution diagram. Furthermore, they account for interactions among alloying elements so that the predicted impact of a given alloying addition depends on the actual alloy composition. In addition to the effect of alloy composition, …