Casting integrity of high-pressure die-castings

Abstract

High-pressure die casting (HPDC) is widely used in the manufacture of automobiles and aeroplanes; however, the mechanical properties of die-castings are notoriously inconsistent. This inconsistency leads to high scrap rates and increased safety factors for component design. Although the mechanical properties of die-castings have been linked to various microstructural heterogeneities, the underlying cause of variability remains somewhat enigmatic First, a Baseline HPDC process is established that is representative of commercial foundry practice. Porosity and non-metallic inclusions are identified as the main sources of variability in tensile ductility, for specimens produced under these conditions. It is proposed that these non-metallic inclusions form during the pyrolysis of commercial plunger lubricants, and that these large pores derive from dilatational strains introduced during semi-solid deformation. The ensuing series of experiments explore ways of reducing defect size using conventional HPDC equipment. Changing the kinematics of the plunger can greatly reduce the scatter in tensile ductility, which is attributed to the reduced size of pores observed under these conditions. The breakup of defect-forming suspensions during the transportation of liquid metals is then considered. Increasing the dissipation rate of turbulent kinetic energy leads to a refinement of non-metallic inclusions and primary 𝛼-Al crystals nucleated in the shot chamber. This refinement enhances the tensile ductility of the castings. Grain refinement is attributed to the fragmentation of incipient grains following turbulent oscillations of the surrounding liquid. Finally, a novel technique—based on X-ray tomography and digital image processing—is presented to predict the areal fraction of porosity involved during tensile failure. By coupling this technique with an existing analytical model, the tensile fracture strain and tensile fracture stress are predicted to within 10.9 % and 8.1 % error, respectively. This fares well against its predecessor for which maximum errors of 242 % and 33.5 % were reported, respectively.