Artificial neural network–modified heat transfer model for furnace coil rolling of magnesium alloy plate

Abstract

Through the experimental temperature measurement and numerical simulation of veneer and stacked plates with different thicknesses, the heat transfer behavior of magnesium alloy wide plate furnace coil rolling was studied. It found that the radial thermal radiation heat transfer rate of the magnesium alloy coil was greater than the thermal conduction heat transfer rate between the coil layers, and the average heating rate of the outer layer of the coil was about twice that of the middle and inner layers. A dimensionless correction factor λ was introduced to the traditional heat transfer model of magnesium alloy coil rolling process based on artificial neural network (ANN) model. After the modification, the reduced Chi-Spr of the heat transfer models of the coil rolling process was reduced significantly from 6.9195 to 1 nearby. In addition, there was a negative correlation between the plate thickness and the heating rate, and the heating rate of the sheet with a small Biot (Bi) value to different temperatures varied greatly. The heating rate of the 3.6mm sheet augments with the increase of temperature. The cooling time of the magnesium plate was proportional to the thickness of the plate, and the cooling rate of the plate with a large Bi value varied greatly from distinct initial temperatures. The cooling rate of 6.0mm plate at high temperature was greater than that at low temperature. According to the temperature distinction of different parts of the sheet, the precise cooling model was constructed.