Critical Evaluation of Prediction Models for Phosphorus Partition between CaO-based Slags and Iron-based Melts during Dephosphorization Processes

ording to the experimental results of hot metal dephosphorization by CaO-based slags at a commercial-scale hot metal pretreatment station, the collected 16 models of equilibrium quotient k_P or phosphorus partition L_P between CaO-based slags and iron-based melts from the literature have been evaluated. The collected 16 models for predicting equilibrium quotient k_P can be transferred to predict phosphorus partition L_P . The predicted results by the collected 16 models cannot be applied to be criteria for evaluating k_P or L_P due to various forms or definitions of k_P or L_P . Thus, the measured phosphorus content [pct P] in a hot metal bath at the end point of the dephosphorization pretreatment process is applied to be the fixed criteria for evaluating the collected 16 models. The collected 16 models can be described in the form of linear functions as y = c_0 + c_1 x , in which independent variable x represents the chemical composition of slags, intercept c_0 including the constant term depicts the temperature effect and other unmentioned or acquiescent thermodynamic factors, and slope c_1 is regressed by the experimental results of k_P or L_P . Thus, a general approach to developing the thermodynamic model for predicting equilibrium quotient k_P or phosphorus partition L P or [pct P] in iron-based melts during the dephosphorization process is proposed by revising the constant term in intercept c_0 for the summarized 15 models except for Suito’s model (M3). The better models with an ideal revising possibility or flexibility among the collected 16 models have been selected and recommended. Compared with the predicted result by the revised 15 models and Suito’s model (M3), the developed IMCT– L_P model coupled with the proposed dephosphorization mechanism by the present authors can be applied to accurately predict phosphorus partition L_P with the lowest mean deviation δ_L_P of log L_P as 2.33, as well as to predict [pct P] in a hot metal bath with the smallest mean deviation δ_[% P] of [pct P] as 12.31.