Abstract: Predicts the endpoint in ladle refining process by way of developing a relevant model through modified BP neural network in combination with the characteristics of ladle refining process. The prediction is based on such a concept that the parameters of the model are analyzed and determined in terms of the in site data recorded previously with other existing neural network models taken as reference, then the prediction model is set up. Other related data can therefore be predicted according to the model. Simulation results showed that the endpoint predicted by applying the modified BP neural network benefits much the ladle refining process.