The present work aims to simulate the solidification process during beam blank continuous casting, to determine the causes of defects, and to optimise the process parameters. A transient finite element model has been developed to compute the temperature and stress profile in beam blank continuous casting. It has been found that the surface temperature fluctuated greatly and that the equivalent stress and strain have their highest values at the flange tip and the web. By comparing the calculated values with metallurgical constraints, the key factor causing surface cracks on the web was determined and an optimum cooling regime was proposed. Online verification of this optimisation project has led to a reduction in the incidence of web surface cracks. List of symbolsc(T) specific heat , J kg 21 uC 21 f s solid fraction h heat transfer coefficient, W m 22 K 21 H(T) enthalpy, J m 23 k(T) thermal conductivity, W m 21 uC 21 L latent heat of fusion, J kg 21 q heat flow density, kW m 22 q latent heat source, W m 23 t casting time, s T temperature, uC T 0 pouring temperature, uC T a air temperature, uC T s surface temperature of blank, uC T w cooling water temperature, uC x,y,z rectangular coordinates, m e strain e' emission ratio, e'50?8 r(T) density, kg m 23 r(T) s density of solid, kg m 23 s stress, MPa s' Boltzman constant, s'55?765610 28 , W m 22 uC 24

Based on a coupled heat and stress model, an artificial intelligence optimisation program was developed to optimise the process parameters in the continuous casting of steel. The program can be used to identify the key factors for the cracks of the strand and obtain the optimum process parameters of continuous casting. The optimisation program contains two models, one is the thermomechanical model developed using the finite difference method, and the other is the optimisation model developed using the subproblem approximation method. The whole program works by automatic iteration between the two models. This study has taken the stress constraint in the solidified shell into consideration when studying the metallurgical constraints. The application of the achieved optimum parameters would make it possible to run the caster at maximum productivity, with minimum cost and fewer defects. After manufacturing verification of this optimisation project, the incidence of cracks has reduced from 8% to 2%, and water consumption in the secondary cooling zone has been decreased by 25%.List of symbols c(T) specific heat, J kg 21 uC 21 f s solid fraction g gravity acceleration, m s 22 H(T) enthalpy, J m 23 k thermal conductivity, W m 21 uC 21 L latent heat of fusion, J kg 21 obj objective function q : latent heat source, W m 23 s m shell thickness at the mould exit, mm t time, s T temperature, uC T centre temperature at metallurgical length point, uC T pouring pouring temperature, uC T surface surface temperature in secondary cooling zones, uC T unbending temperature at unbending point, uC DT max reheating between zones, uC v casting speed, m min 21 WR water flow rate, L kg 21 W sprays water consumption, L min 21 W spray0 mould water consumption, L min 21 W spray1 water consumption in first secondary cooling, L min 21 W spray2 water consumption in second secondary cooling, L min 21 W spray3 water consumption in third secondary cooling, L min 21 x, y, z rectangular coordinates, mGreek symbols r(T) density, kg m 23 r(T) s density of solid, kg m 23 e strain s stress, MPa