LAUSR

Abstract The surface heat transfer coefficients of steel billet play very important roles in the accurate heat transfer model. However, these heat transfer coefficients are difficult to be fixed. The main work of this paper focuses on identifying these heat transfer coefficients by solving the inverse heat transfer problem. In order to estimate these heat transfer coefficients better, this paper presents an improved difference evolution algorithm. The SAE1007 steel billet is used to conform the validity of this algorithm. With the help of the surface temperature measurement, this algorithm can identify these heat transfer coefficients effectively. Comparing with the difference evolution algorithm, this improved algorithm can reduce iterative number and has better convergence. Due to the horrible production environment, there are some outliers in the measured values. To eliminate these outliers, the weight least square method is also introduced. Finally, the corrected heat transfer coefficients are used to enhance the accuracy of the heat transfer model, which is applied to predict the solidified shell thickness of steel billet. The predicted results are confirmed by the actual industrial data.