Abstract: Because of the modeling problems of the ladle furnace (LF) desulfurization process that are nonlinear， intensive dynamic and characterized of multiple conditions， end point sulphur content prediction model was proposed based on local model migration algorithm， where a simplified principle model was first established to capture the main process behavior and fine corrected by local model migration method to compensate the remain prediction error caused by mechanism simplification process and condition changes. A new local model migration algorithm was developed to automatically rectify the process nonlinearity deviation. The new method works by integrating several local migration models that are established in several local regions of the input space. The presented predictor shows better performance with respect to existing intelligent predictors due to the full exploitation of first principles， which is validated by the practical data.

Key words: ladle furnace, end point sulphur content prediction model, model migration, clustering analysis, fuzzy TS