Search

The presence of a heavily deformed surface layer (a few microns thick) on rolled aluminium alloy is understood to be one of the main reasons contributing to the Filiform Corrosion (FFC) susceptibility of the alloy. The surface layer is formed during the thermo-mechanical processing of the sheet metal, in the rolling mills. The surface layer characteristics are process dependent, i.e. depends on thermo-mechanical processing and the composition of the alloy in addition to post-production surface treatment steps. While removal of the surface layer solves the issue of FFC, it leads to economical consequences. In order to improve the FFC resistance of the aluminium alloy it is necessary to investigate the surface layer itself, its formation and evolution along the rolling mills. This thesis develops a detailed understanding of the formation and evolution of a FFC susceptible surface layer on a recycled aluminium alloy, along the rolling mills. The current trend in aluminium industry is to indulge more into recycling for cost and environmental reasons. The study of a recycled alloy, instead of conventionally investigated model and primary alloys, and the investigation of the surface layer at each and every step of the process route makes this thesis interesting and of commercial relevance. Finally some suggestions have been made to improve the FFC resistance of a rolled recycled aluminium alloy (AA5050).

In the thesis two different models of solidification of aluminum alloys are presented and analyzed. The first 1--D solidification model is derived from the conservation of solute, heat and mass. With numerical experiments it is shown that simulations with the Finite Difference discretization must adhere a relation similar to the Courant-Friedrichs-Lewy relation in order to get numerical stable solutions. Furthermore, it is demonstrated that depending on the chosen discretization of the interface conditions an additional singularity which can destroy stability occurs. The occurrence of this singularity is not related to the Finite Difference discretization. Singularities are also found with application of the Collocation Method. An example of a solidifying aluminum slab is successfully solved with application of the model equations and avoiding stability infringements. The possibility to apply the model in an extended simulation, i.e. especially in the second solidification model, is discussed. The second new 2--D solidification network model is used to simulate fluid flow and solidification within the mush. It is based on the assumption that the liquid parts of the mush can up to some degree be represented by a network of geometrical simplified liquid channels. Numerical simulations with this model indicate that conduction and convection of heat and solute affects directly only the permeability of a small region of the mush. Self induced effects and the formation wormholes have not been found. Finally, a detailed study of the influence of various model parameters of the network model is presented. A more detailed abstract is included within the thesis.

not available

Modelling of Flow Phenomena during DC Casting Jan Zuidema The production of aluminium ingots, by semi-continuous casting, is a complex process. DC Casting stands for direct chill casting. During this process liquid aluminium transforms to solid aluminium while cooling down. This is not an instantaneous transformation, but occurs in temperature interval. In the casting process the latent heat is moved away by convection and conduction. A number of problems may occur during solidification, because the solidification rate is rather high. The difference in density between liquid and solid aluminium is 7% and because of that solidification porosities may form during solidification when transport of liquid metal is insufficient. If besides this also high enough thermal stresses are present, cracks may be induced from these pores. The cracks, that originate during the solidification of the metal are called hot cracks. This in contradiction with cold cracks. These are formed due to high stress levels in the already solidified aluminium. What happens from the moment that the casting process is started? This is one of the questions that is treated in this thesis. To be able to describe the DC Casting process, it is necessary to have a good model that describes the phenomena at hand adequately and that also enables to do predictions on process changes. The model, that is used , is based on the differential equations that describe the heat- and fluid dynamics transport. In chapter 2 a description of the model is given. Numerical methods for solving these equations are also given there. Without validation, a model is of little use. Model validation can be performed using various methods during DC Casting experiments. Temperature recordings enable to follow the evolution of the temperature during the casting- and solidification process in some important positions. Because part of the heat transport in the DC Casting process is convection driven, it is also important to find information on the flow pattern and velocities in the liquid aluminium. Various methods to evaluate velocities in liquid metals are described in chapter 2. Based on their use for measuring the molten aluminium velocity during DC Casting, a ranking of the methods is made. Good boundary conditions are essential for reliable predictions of the behaviour of aluminium during the mould filling and subsequent solidification. In chapter 3 the determination of the most important boundary conditions is described. Those boundary conditions determine the amount of heat transport in that area of the system, where the heat transfer to the surroundings is the largest. Through sophisticated experiments the heat transfer from a plate of aluminium to (boiling) water could be predicted. This heat transfer model enables the prediction of temperatures close to the surface of the solidifying aluminium. By instrumentation of the experimental casting facility with thermocouples, the temperature close to the surface of the ingot could be measured as function of time. Calculations using the casting- and solidification model with the constant heat transfer coefficients, and calculations with the new model for water cooling were performed. This showed, that in the first centimetres from the surface of the ingot the new model gives a better description of the experiment. Further away from the surface towards the centre of the billet, the constant heat transfer coefficient model is just as good as the new model. The best method to measure liquid metal velocities during DC Casting is a method that is based on magnetic induction. The sensor to measure velocities is based on a ferro magnet, surrounded by a conduction tube with spot-welded thermocouples. Between two opposing thermocouples an inductive voltage is measured, which is proportional with the velocity of the surrounding medium. To test this method, a prototype of this sensor was constructed. In chapter 4 this sensor is described. The sensor was firstly tested in a water setup. Because the electrical conductivity of the water was several orders of magnitude less than that of liquid metals, this gave an unsatisfactory result. It was decided to perform fluid flow experiments in liquid tin. From these experiments results were obtained which gave a velocity dependent signal for the probe output. Unfortunately, electro-magnetical shielding from the surroundings was not good enough. This prevented to quantitatively test the sensor. In chapter 5 the start-up phase of the casting is covered. In the first part of this chapter, a calculation described in the literature was repeated, using the casting- and solidification model. This showed that, due to lack of published data, more validation was necessary to fully validate the model. With the aid of the experimental casting facility, a number of casting experiments were conducted. The experiments were recorded with a video camera and the temperature during casting was recorded by thermocouples at a number of locations in the ingot. The experiments served as the basis for a number of calculations with the casting- and solidification model. A conclusion, that could be drawn from the combination of the results of the casting experiments and the calculations , was that the description of the inlet of the mould in the model as an jet of aluminium was a good description of reality during the first tenths of seconds of the process. From the moment that the aluminium melt level is overflowing the inlet level, the jet description is not accurate anymore and should be replaced with a filling over the complete width of the mould. During casting, unwanted inclusions are filtered from the liquid aluminium by degassers and filters, before the casting table is reached. In order to comply with the increasing demands for the quality of cast products, an additional method to filter out particles from the launder system was evaluated. This is the subject of chapter 6. Manipulation of the flow pattern enables to create wakes, where particles can settle. Two dimensional calculations of the flow pattern in the mould with special flow modifiers, show that this settling is possible in theory. However, it is more likely, that transient velocity fluctuations prevent the settling of particles. To create a better method for settling of unwanted inclusions, a cyclone was designed, which enables the separation of particles 20 micrometers in diameter from a stream of liquid aluminium. The cyclone was used in a series of calculations with the fluid-flow and solidification model. The outcome of these calculations has resulted in a patent for this type of cyclone. Validation measurements using a water model of the cyclone have confirmed the working of the cyclone in separation of denser particles out of a slow flowing medium. The velocity of the particles was therefore tracked using a laser and camera set-up. The laser exposed a two-dimensional area of the water model. By auto-correlation of two subsequent images, it was possible to extract the particulate rate in this area. Velocity patterns indicated, that particles had a tendency of settling down in the lower part of the cyclone. Because there was no possibility to count all particles going in and out of the system, no quantitative comparison with the model was possible.

Aluminum, the most common metallic element in the earth's crust, offers structural materials with high strength and light weight. It is widely used throughout the world economy, particularly in the transportation, packaging, and construction industries. Once extracted from ore, solidification is the most important phase transformation in the development of aluminum products, especially for critical applications. All of these applications rely on the ability of aluminum alloys to form a homogenous granular structure with a relatively small grain size, known as grain refinement. This thesis involves the use of in-situ experimental techniques like neutron scattering, synchrotron radiation, and differential thermal analysis to investigate the mechanism of nucleation and subsequent growth of individual grains during solidification. The aluminium alloys contained grain refiners like TiB2 particles along with solute titanium. The results indicate that the grain refiners have little influence on the overall short-range order in the liquid aluminum alloy prior to the solidification. Grain refinement is only effective when both solute titanium and TiB2 particles are present in the melt during solidification. The results shows that the TiB2 substrates stabilise a TiAl3 phase in a limited temperature range above the solidification temperature of aluminum. This observation pinpoints the highly debated mechanism that is responsible for enhanced grain nucleation during solidification of grain refined aluminum alloys. The results further demonstrate that the nucleation process is limited to the initial stage of the solidification and is complete at a solid phase fraction of about 20% for all the investigated aluminum alloys. The growth curves of the individual aluminum grains exhibit three different stages during solidification. The first stage of grain growth curve is consistent with the model prediction for diffusion-controlled growth of non-interacting grains. The later stages can be explained in terms of a release of latent heat and to a minor extend by the indirect interaction of the growing grains via overlapping diffusion fields.

Powder injection moulding (PIM) is a versatile technology capable of producing complex and near net shape parts. Significant help in optimisation of the PIM process can be obtained from the numerical simulation. To obtain realistic numerical results it is necessary to fulfil certain conditions. The material properties should be correctly determined in the conditions close to those during the real powder injection moulding. The process must be described with the appropriate mathematical model. Also proper initial and boundary conditions must be chosen. In this study a standard feedstock of stainless steel 316 L produced by BASF AG has been characterised for determining the input parameters for the numerical simulation. Viscosity of the feedstock was measured in the wide range of shear rates at four temperatures using the capillary rheometer. Specific heat of the studied feedstock was measured with an advanced DSC apparatus at different cooling rates. The obtained data were found suitable to serve as the input parameters for the simulation of the powder injection moulding process. Simulation of the PIM process using a material model for a Generalised Newtonian Fluid has been performed with three best known commercial packages, C-Mold, Moldflow and ProCAST. The obtained simulation results have been compared with each other and with the experimentally determined ones (melt front advancement, weld line locations, in-cavity temperatures and pressures). It has been proven that the PIM process can be successfully simulated with C-Mold, Moldflow and ProCAST analysis packages. The flow pattern predicted by C-Mold appeared to be less accurate than that predicted by ProCAST and Moldflow. The predictions of the locations of weld lines made by all three analysis packages were in good agreement with the experimental observations, but ProCAST yields the best results. The temperatures of the feedstock predicted in the simulations were in good agreement with the experimentally measured ones. The best predictions of pressure were made by C-Mold. The predicted by C-Mold in-cavity pressure value matched well the experimentally observed one. ProCAST underestimated the in-cavity pressure by 20-30%. The Moldflow pressure results were found to be 3-4 times lower than the experimentally observed ones. Despite minor drawbacks, ProCAST was found to be the best analysis package for the numerical simulation of the PIM process.

When aluminium is protected by an organic coating a special form of corrosion can occur underneath the organic coating; filiform corrosion. This form of corrosion manifests itself as threadlike filaments under the coating, it causes local delamination of the coating and the coating cannot protect the substrate anymore. The properties of the aluminium substrate have a strong influence on the filiform corrosion properties of the end product. It is in the interest of the aluminium producers to understand how the composition and the microstructure of the substrate influence the filiform corrosion properties. There are two mechanisms that can explain filiform corrosion. These are the presence of a surface layer with an extreme low potential and the presence of a large amount of cathodic intermetallics.