Advanced methodology of determination water jet cooling intensity during the casting process

Abstract

Companies which produce aluminum alloy ingots seek a final product without structural defects. One crucial factor is the cooling during the semi continual casting process. In the beginning of the process, most cracks are made with lengths up to 300 mm, and then, by selecting a suitable method of water cooling, the cracks are closed. A major influence on defect generation is the superheat extraction from the incoming liquid metal by the secondary water-cooling system due to direct water impingement on the ingot surface. The temperature distribution during the casting process can be simulated numerically with known boundary conditions (cooling intensity along the surface). Boundary conditions are obtained by experimental investigation and subsequent evaluation. A special experimental device was designed for measurement. The device’s main function is to ensure that the position of the mold and the sample during measurement is as it would be during the real casting process. The aluminum sample was equipped with a set of thermocouples placed along the cooling surface. The hot vertical surface was cooled down during the experiments by a flat water jet. The impact area is located in the upper part of the cooling surface. The rest of surface is cooled by water flow down along the surface. This article deals with the evaluation of this type of experiment. The boundary conditions (heat transfer coefficients) are estimated as a function of temperature and vertical position. Unfortunately, the results obtained by standard methods for solving the inverse heat conduction problem (for example, using the 2D sequential function specification method) are blurred. This is caused by the Leidenfrost effect and this special type of cooling. A sharp border between the transient and film boiling modes is created and moves down during the experiment. This article illustrates an applicable solution based on shifting computation element borders during the inverse computations. The method was tested on measured data.