DEVELOPMENT OF NEURAL NETWORK AND FRAME WORK TO OPTIMIZE CUT LENGTH IN CONTINUOUS CASTING SHOP USING VARIOUS PARAMETERS

Continuous casting of steel is a process in which liquid steel is continuously solidified into semi-finished or finished product. In continuous casting, one of the major operational irregularities encountered is wrong cutting of slab (long or short) and grade mix of two different steel during casting of slab which leads to productivity loss, capital loss, and safety hazards etc. In the present work, a neural network model is developed to track the tundish filling and casting speed. The model is fully transient and consists of three sub models, which account for mixing in the tundish, mixing in the liquid core of the strand, and solidification. For which two-layer feed forward back-propagation model is developed for predicting the existence of primary intermix region that might lead to a grade mix slab. Firstly, the networks is trained with four input parameter i.e. slab width, casting time, casting speed, tundish weight). Networks training were performed using the Levernberg Marquard training algorithm. The output from these neural networks is logical 1 (if a wrong cut of slab) and a logical 0 (if no wrong cut of slab occur). The neural networks model training, validation, testing is done by MATLAB