You are on page 1of 18

Applied Mathematical Modelling 37 (2013) 611–628

Contents lists available at SciVerse ScienceDirect

Applied Mathematical Modelling
journal homepage:

The present state of modeling entrainment defects in the shape
casting process
C. Reilly a,c,⇑, N.R. Green b, M.R. Jolly a

School of Mechanical Engineering, University of Birmingham, UK
School of Metallurgy and Materials, University of Birmingham, UK
School of Materials Engineering, University of British Columbia, Canada

a r t i c l e

i n f o

Article history:
Received 27 September 2011
Received in revised form 3 April 2012
Accepted 21 April 2012
Available online 27 April 2012

a b s t r a c t
The entrainment of oxide films into the bulk material has been shown to have a detrimental effect on casting integrity. A number of mechanisms have been shown to initiate the
entrainment of oxide films, including: returning waves, plunging jets, bubble trails and
fountains. Therefore, the assessment of the casting system for these features by the foundry
engineer is critical in improving casting quality.
The use of computational fluid dynamics software packages, which are now widely available to the foundry engineer, has allowed the foundry engineer to improve casting system
design by using qualitative parameters. Optimization software is now an economically viable option for many foundries. However, optimization for casting integrity requires a quantitative casting integrity assessment technique, which allows the modeling and
quantification of defects. Therefore, modeling and quantification of defects is becoming
an ever more important research area to allow the optimization software manufacturers
to meet the needs of industry.
The current methods found in published literature for the modeling of casting defects
have been described and critically reviewed, shedding light on the qualities and issues currently associated with the present available methods. However it is clear that further
investigations and developments are still required to allow the accurate and efficient modeling of casting defects. The topics of research relating to the modelling of casting defects
which require further investigation have been highlighted.
Ó 2012 Elsevier Inc. All rights reserved.


Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Computational modelling background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Indiscrete modeling of entrainment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cumulative entrained free surface area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Vorticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cumulative scalar technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
MAGMAsoft air entrainment model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
FLOW-3D air entrainment model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dimensionless number criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .


⇑ Corresponding author. Address: Materials Engineering, The University of British Columbia, Frank Forward Building, 309-6350 Stores Road, Vancouver,
BC, Canada V6T 1Z4. Tel.: +1 778 996 3026; fax: +1 604 822 3619.
E-mail address: (C. Reilly).
0307-904X/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.

. . . . . . . . . . . . . . The entrainment of oxide films into the bulk fluid has been shown to have detrimental effects on cast component integrity [4]. . . . . . . . . . . . . 5. . . . . . . . . . . . . . This has given further insight onto this important topic. . . . . . . . . . Reilly et al. . . . . . . . . . . . . . . . . . . . . . . . . would be computationally expensive and outside of reasonable simulation times as the adaptive meshing procedure would be computationally very expensive. . . Conclusions. . . food and cosmetics. . . . . . . . . . . . Many defect modeling topics relating to the casting process have been researched. . . . . . including solidification and thermal modelling (porosity). . . . . . . . . When the surface oxide film is entrained into the bulk fluid the film may be broken up by turbulence flow into numerous single entities known as oxide films. . . . . die/mould based modelling (burn on. . . . Many liquid metals form an oxide film upon their free surface due to the reaction with the oxygen in the atmosphere. . . . . . . If a single mesh capable of resolving the lm thick film was utilized the runtimes for a casting fluid flow simulation would be many orders of magnitude above what is reasonable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In recent times the economic viability and increased ease of use has encouraged many larger foundries to use computational optimization software. . . allowing a finer mesh to be implemented for the film and a coarser mesh for the bulk fluid. . . . . . . . . . However. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . previously proposed methods of modeling defects. . . . . . . . forms almost instantaneously and thickens with time. . . . . The modelling of a thin film in the order of lm in thickness on top of a volume of fluid yields significant modelling difficulties from the meshing perspective. . . . . . . . . . . . . . . . . . . . . .3. . . . . . . paints. The modeling of defects is essential to allow the optimization of casting systems for component integrity. . . . . . . . . . . . . . . . . . . . . . . . . . .612 4. . . . . . . . . . . . . . . . Therefore. . . . . . . . . . . . . . . . . . . The direct modelling of the physics of the entrainment process is currently seen as unrealistic giving the processes complexity and current computational power. . . . . . . Oxide film entrainment model (OFEM) . . . . . 5. . . . . . . . . . 5. . . . . Recent work has both proposed new methods and further developed. . . . . . . . . . . . . . . . . Optimization can only occur if ‘‘the right optimization criteria to formulate the objective functions are available’’ [2]. . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . die soldering. . . . . . . . . . . . . . . . This suggests that the accurate modeling of entrainment defects could potentially yield benefits in the accurate modeling of porosity. . . . . . . to optimise a casting system for casting integrity. . . . . and assessed. . . . . . . . . . . Modeling of oxide entrainment . .1. . . . . . . . . . . . . . . . . ‘‘the use of entrainment models to optimise filling systems designs for castings has huge commercial potential that has so far being neglected by modellers’’ [5] has still not been adequately heard as it appears that significant research is still required in to this topic. . . . . . Campbell’s 2006 paper [5] summarised most of the methods researched for the modeling of defect entrainment in castings thus far developed. . . Introduction With competition within the foundry industry becoming fiercer and customers demanding higher quality components. . . . in the range of nm to lm in thickness. . . A discussion of the currently available methods for assessing casting integrity both quantitatively and qualitatively is presented below. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . shorter development times and more complex geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . the discrete and the indiscrete modeling of entrainment defects. . . This extremely thin solid film. . . . . . . . . . . . . . Modeling of oxide film deformation . . . . . . . . . . . . . . . . . . knowledge of defect formation. but few teams are presently active. die life prediction) and stress strain modelling (distortion. . . . . . . . . . . . . . . . . . . . . Even using an adaptive meshing technique. . . . Modeling of discrete defects . . . . . . . . . . . . . . . . .4. . whereby the mesh is dynamically modified during simulation. . . Multi-phase modeling . . . . . It appears that Campbell’s final. . . Acknowledgments . . . . These oxide films act as a crack initiation sites upon solidification as they do not bond within the metallic structure and are therefore weaknesses. . . . . . 5. . . . . . . . in this paper only the modeling of filling related defect assessment of the casting process is reviewed. . . . . . . . . . Bubble entrainment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . These can be largely split into two groups. . . . . . . . . . . . . / Applied Mathematical Modelling 37 (2013) 611–628 3. . . . . . . . . . . . . . . . . . . . . . . . . distribution and quantity is required. 619 619 620 621 622 623 624 624 626 626 626 626 1. . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . This is the challenge facing modellers. . . . . . . . . . .5. . . . it should be noted that published research has shown porosity to be linked to oxide film defects [7–10]. . C. . . . . . . . . sand erosion. . . . . . . . .2. . . . . . . . . . . . . where entrainment of surface films or gasses is often a process limiting parameter. . . . . . . . . . . . . 7. . . . . . . can be found here [6]). . . . . . . but possibly most important conclusion. hot tearing). . . . . . . . . It should be remembered that the accurate modeling of entrainment also has the potential to yield significant benefits in optimizing the manufacturing processes of chemicals. . . . . . . . . . . . . . . . . . . . . . Although work on modeling porosity is not reviewed here (one such review of porosity modeling by Lee et al. . . . . A variety of approaches into the methods of modeling of entrainment defects have been investigated. . . . . . . . . . . . . . . . . . . . . . . . . . . the use of computational simulation has become essential to stay competitive [1]. . . . . . . . . . . . . . . . . . . . . . . . . . Modeling of oxides in steels . . . . As these optimization software such as MAGMAfrontier [3] become more user friendly and the performance of computer hardware increases the requirement for accurate and quantitative defect assessment criteria will become even more acute. . . . . . . . . . . . .7. . . . . . . . . which acts as an initiation sites for cracks to propagate from. . . . . . . . . . . . . . . . . . . Modeling the folding mechanism. . . . . . .

commonly as they are proprietary or they are left out of product manuals in order to simplify the document for the user. and not to evaluate free surface turbulence such as that which would entrain defects. (1) (for fluid momentum) and (2) (for mass conservation). o ðqcTÞ þ r:ðqcuTÞ ¼ r:ðkrTÞ þ Sh : ot ð3Þ Eqs. There have been various versions of the marker-and-cell (MAC) method developed for FV methods. A variety of meshes can be used for both of these methods. p is the pressure and leff is the effective viscosity. However. These turbulence models are used to aid the modeller in producing a model that correctly predicts fluid flow. the approach has now been further developed for use in FE methods [13]. and therefore readers are encouraged to review the original publications which are referenced. The second and most popular method of simulating free surface flows in shape casting is to use a fixed grid approach. The first method effectively tracks the evolution of the free-surface as the mould fills [12]. it should be stated that in many cases the even the most mathematical formulations are not published. MAC methods have not been used widely in the CDF modelling of casting filling as MAC methods only calculate within the region of the mesh where liquid metal is present.C. Below. either structured or unstructured. the ability to define complex geometries and the computational speed of the approach. Reilly et al. The defect prediction tool can then be applied to this model. in order to predict other phenomena. The fluids free surface is allowed to deform and the mesh to deform with it using arbitrary Lagrangian–Eulerian (ALE) techniques. The volume of . It maybe that future developments resolve some or all of these issues. where c is the fluids specific heat capacitance. The choice of the most appropriate method is largely dependent upon the compromise made between the ability to define and capture the free surface. Eqs. oðqui Þ oðquj ui Þ o þ  ot oxj oxj  leff oui oxj  ¼ Sj  oq oðqui Þ ¼ 0: þ oxj ot oq . in order to accurately predict defects created by free surface turbulence using CFD modelling techniques the ability of the model to define and capture the free surface is critical. For other techniques the authors felt that it was beyond the scope of this paper to present all the equations required to gain a meaningful insight into some of the complex mathematical models described. and. It should be noted that the function of these turbulence models is to evaluate energy loss due to bulk fluid turbulence. Many of the computational models described below use a turbulence model to account for energy losses cause by internal fluid turbulence. A fixed grid. it is outside the scope of this paper to be able to review and fully describe all of the methods and techniques that have and are used in the creation of these CDF models. oxi ð1Þ ð2Þ In casting simulation software mould filling is normally accompanied by heat transfer and solidification which is described by Eq. However. k the conductivity and Sh includes the latent heat on solidification. Also if the free surface profile becomes complex then it is no longer possible for the free surface to be defined as a single function. such as many types of defects. (3). where q is the fluid density. In the majority of cases the liquid metal is assumed to be a Newtonian fluid which is represented by the Navier-Stokes equations. such as those caused by splashing. The method essentially seeds the initial free surface with mass less particles and follows their position with time using simple velocity calculations. These CDF models have been developed and applied to the metal casting scenario over the last 40 years [11]. Obviously. using both different structures and element types. However. (1)–(3) can be solved using either a finite element (FE) or finite volume (FV) discretization methods. 2. waves or other forms of free surface turbulence as they require frequent remeshing of the domain in order to avoid numerical degradation caused by large element ratios. This approach is attractive as the coupled behaviour of the free surface with an expanding flow domain is represented directly. Computational modelling background In order to produce accurate defect predictions for a casting an accurate computational fluid dynamics (CDF) model of the moulds filling is first required. Once this initial fluid flow model has been created one or more of the models described below can then be applied as additions in an attempt to predict defect entrainment. within the model domain. information throughout the whole domain is required. often by mould geometry. a brief description of the most popular modelling techniques and methodologies is given. Two main approaches have been utilised to predict and define the free surface interface in the CFD software used in the work related to modelling casting defects reviewed below. / Applied Mathematical Modelling 37 (2013) 611–628 613 Wherever possible the equations salient to the techniques presented have been included in this article. most other researchers have avoided this approach because of accuracy and computational efficiency issues associated with tracking the free surface. final defect location. These techniques encounter problems when dealing with large surface distortions. is defined. however. A number of workers have published methods for solving free-surface problems in this manner for casting [14]. Another approach to free surface tracking for casting related software is to discretize the flow domain and make the fluids free surface a calculation domain boundary. and in some cases.

1. for models of identical geometry). due to commercial sensitivities. a stable solution is ensured only by using a small-enough time step to satisfy the smallest Courant–Friedrich–Levy (CFL) [17] number in the mesh. Cumulative entrained free surface area Work by both Lai et al. Alternatively an equation for the advection of the scalar is solved numerically. Since both methods are explicit. 3.614 C. although conservative. To develop this technique an efficient algorithm to calculate the minimum free surface area is required. There are a host of workers who have essentially employed the VOF concept within algorithms using all kinds of combinations of FE and FV methods. [26] investigated using the difference in free surface area to describe the magnitude of entrainment. the approach has been shown to yield results in accord with more sophisticated models and experimental data. The VOF method consists of three ingredients: a scheme to locate the surface. For very simple geometries comparison between differing geometries is possible. A study of this literature with the intension of developing a casting based vorticity model is outside the scope of this article. is computationally more demanding. CEI and Tecplot. However. Research undertaken in other fields has used vorticity to characterise air entrainment. The work by Lai et al. and a means of applying boundary conditions at the free surface. The lack of ability to track the motion of the entrained defects also proved detrimental to the usefulness of this technique. This is then compared to the proposed instantaneous free surface area assuming the mould had been filled in a tranquil manner. this technique is unsuitable to use for optimization except for instances where direct comparison can be made between two or more models (i. although easily understood and requiring minimal computational power has one major drawback. namely. The VOF tracking method leads to a sharp interface. This technique [25]. to allow the excess of free surface area to be calculated. although the authors reported positive results using the technique. The CFL number defines the smallest time step which can be used to gain correct results for a given mesh.2. This scalar once entrained into the bulk fluid at the free surface is allowed to gradually diffuse throughout the fluid and advect with the flow of the bulk fluid. an algorithm to track the surface as a sharp interface moving through a computational grid. The problem arises however in filtering of the data. but presents future research. These analysis tools allow the vortex core location and axis and vortex magnitude to be defined. though time consuming. this could prove near impossible. In contrast.. 3. how to define the minimum free surface area should the mould fill quiescently. and it suffers from numerical diffusion. This is a simple and robust approach which neglects the physics involved in bi-film entrainment. [24. producing many vortices. Indiscrete modeling of entrainment 3. 3. The work showed that the largest excess free surface area was during pouring from the furnace into the crucible and from the crucible into the mould.3. Cumulative scalar technique A cumulative near surface scalar technique has been developed by a number of the commercial casting software manufacturers [29–31]. the surfacecapturing approach. Little is known of the work by Sun et al. this oxide accumulation is described by a scalar parameter. [25] takes the instantaneous free surface area and plots it against time. The technique works by assuming that oxide defects accumulate upon the fluids free surface at a constant rate. but it is not guaranteed to conserve mass.e. / Applied Mathematical Modelling 37 (2013) 611–628 fluid (VOF) concept is then used to track the free surface interface and to reconstruct geometrically its location [15]. There are now many useful tricks for ensuring that this can be achieved efficiently [18–23].25] and Sun et al. This technique gives no distribution of defects but is felt to be nevertheless highly informative as it is a strong indicator of which stage of mould filling is likely to generate the most significant number of defects. The authors are currently unaware of any work which has been undertaken relating vortex assessment using a computational model to defect entrainment or casting integrity. The bulk fluid flow in the casting scenario is usually in a highly turbulent regime. As stated by the Barkhudarov and Hirt [30]. This highlights the fact that the quality of the metal entering the running system is of extreme importance to casting integrity. This allows a final defect probability to be obtained. Therefore. This function is also available in CFD post processing software such as Field View. Reilly et al. and then the interface is captured as a discontinuity in the solution field via an appropriate non-diffusive scheme [16]. Vorticity Both MAGMAsoft [27] and FLOW-3D [28] have developed techniques to identify and assess vortices within the bulk fluid during flow simulation of mould filling. the cumulative scalar technique does have some drawbacks in the casting scenario. namely: . For complex geometries however. which the authors feel could yield highly beneficial results. This area requires significant further research in order to understand whether it is viable to develop a free surface vorticity based entrainment model. filtering the data to only show those relevant to free surface entrainment can be highly problematic.

    The adhesion of oxide film to mould walls is not accounted for. MAGMAsoft air entrainment model An air entrainment model has been developed by MAGMAsoft as a mechanism to track small air bubbles transported by the bulk flow. The venting model is the main mechanism for tracking air pockets. this tracks changes in topology of air pockets and calculates their thermodynamic parameters. and wall contact time and are defined below:  Flow length – distance the metal has flown since entering the cavity  Material age – length of time the material has been in the cavity  Wall contact time – length of time the material has been in contact with the wall 3.4. The model is valid for small bubbles. along with their density. 1) unlike bi-film defects which remain as single entities unless torn. 1.5. This is to try and quantify the mass of oxide entrained. Air entrapment operates only on the air volume transporting it with the bulk melt velocity field. Reilly et al.C. the number of discrete air pockets and each pockets location is known. These include flow length. Ox is oxide and kl is a rate constant (10. The air entrapment models give the user a contour map of air distribution within the melt volume. volume. (although at the time of writing not all are available to customers) several scalar quantities aimed at helping the foundry engineer assess the likelihood of entrainment. mass. These particles are then given properties and allowed to move within the constraints of a set of governing equations. The cumulative scalar technique operates in the same way as that described previously but with the exception that the quantification is not constant.9  103 kg m2 s was used). material age. The venting (permanent moulds/dies) or permeability (consumable moulds) of the mould is modelled to allow accurate modeling of vented regions. (4). An almost identical technique is also utilised in smoothed particle hydrodynamics (SPH) [32–34]. Diffusion of oxide scalar can clearly be seen [30]. but defined by a relationship proposed by Backer et al. FLOW-3D air entrainment model FLOW-3D have developed an algorithm to model the turbulent entrainment of air at a free surface [36]. The venting model can operate only on air pockets that are resolved by at least several mesh elements. mould and liquid cannot be modelled. Flow 3D defect tracking scalar example. temperature and pressure. Alongside their main air entrainment model MAGMAsoft have also implemented in their code. Air entrapment is a model that enables tracking air pockets that are too small to be tracked by the venting model. However the defect location patterns are still valid. where t is time. The model works by assessing whether the turbulent energy at the free surface is enough to overcome the restraining effects of the surface . No oxide film strength is modelled No buoyancy of oxide film is modelled Without any experimental results the significance of the absolute values of the scalar are meaningless. SPH is a gridless technique where the particles can move anywhere within the domain and interact with each other following a set of defined physical rules. Therefore representing individual defects as a scalar quantity is always going to be problematic as the interactions between the defects. Air pocket can collide with other air pockets and can merge or split. dOx ¼ kl : dt ð4Þ It must be remembered that these scalars diffuse within the liquid (Fig. The criteria MAGMAsoft use to define the quantity and threshold of air entrainment at the free surface into the bulk fluid is proprietary. SPH is a technique whereby the bulk fluid is divided into a series of discrete elements known as particles. Eq. / Applied Mathematical Modelling 37 (2013) 611–628 615 Fig. a venting model and an air entrapment model. Using this. [35]. The model is made of two main constituents. 3.

The We and Fr are defined in Eqs. The use of dimensionless numbers does not enable the tracking of defects.5 as entrainment is assumed to occur on average over half the surface area. / Applied Mathematical Modelling 37 (2013) 611–628 tension and gravity. The volume of air entrained per unit time (Va) is calculated by Eq.616 C. (10). q is density (kg m3). scaled down models of large civil engineering structures have been used to assess the flow and entrainment characteristics before construction commences. when compared to full scale.2) the gravitational acceleration normal to the fluid surface. where T is the width of the free surface and A is the flow area.e. Previous studies utilising the Weber Number (We). The bulking of the fluid with the volume of entrained air can be modelled. whose default value is 0. The effects of experimental conditions. in an open channel this is taken to be the hydraulic mean depth (Dh) as defined by Eq. fountains. Lt ¼ cnu qffiffiffiffiffiffi 2 ð23ÞQ 3 D ð5Þ : This turbulent length scale (Lt) is then used to calculate the disturbance kinetic energy per unit volume (i. This is given in Eq. i. r is the surface tension pressure (N m). bubble trails and colliding fluid fronts. The entrainment threshold for a hydraulic jump has been shown experimentally to lie over the huge range of Fr numbers of 1–4 [41–43]. 2 We2 ¼ 2 Inertial pressure ql v 2 q lv We ¼ . pressure) associated with a fluid element raised to a height Lt with the surface tension energy based upon the curvature of Lt. Pd (surface tension and gravity).45]. of the casting system. now has to be questioned due to recent research findings. However.. underestimate the energy dissipation and entrain fewer bubbles of a greater size for similar inflow conditions [39]. The accuracy of this data however. and l is the characteristic length (m). size and quantity of bubbles entrained. The model does show an excellent correlation with the location of entrainment [36]. (8) and (7) respectively. namely: plunging jets. thus still making it an extremely valuable modeling tool.6.085. (ratio of surface tension to inertial pressure) [48] and Froude Number (Fr). where v is the characteristic velocity (m s1). spill ways and plunging jets were used for validation of the model [36]. D is a dissipation function and cnu is a constant which is also used in the calculation of turbulent viscosity. The volume of air entrained experimentally at hydraulic jumps. Reilly et al. (7). it does have the potential to allow the quantification of entrainment. P d ¼ qg n Lt þ r Lt ð6Þ : For air entrainment to occur the turbulent kinetic energy per unit volume (Pt) must be large enough to overpower the surface stabilising forces. Results from Chanson show that small scale models. This value is derived explicitly for the re-normalised group (RNG) turbulence model. dissipate and escape at the free surface. where Cair is a coefficient of proportionality and As is proportional to the surface area. It must be remembered that the greatest limitations for the use of dimensionless numbers for quantifying entrainment in casting systems is firstly their inability to differentiate between the many types of entrainment mechanisms.50] and Isawa [51] respectively. Secondly. Recent research in the hydraulic engineering field has shown that the scale of an experiment has an effect on the entrainment threshold (critical condition upon which entrainment commences). where Q is the turbulent kinetic energy. demonstrates quantitatively that dynamic similarity cannot be achieved with either the Fr or We numbers as has traditionally been assumed [37–41]. ¼ Surface tension pressure lr r ð8Þ . even if the questions can be raised about the magnitude of the entrained gas. A characteristic size of turbulent eddies is defined by Eq. This is the value was used for validation of the model. Recent work however. they are not able to assess entrainment in all regions. This air is then entrained into the fluid and allowed to advect.. For a given jet velocity the volume of air entrained is proportional to the jet disturbance [44]. If the magnitude of surface turbulence is able to overcome these restraining effects then a series of equations are used to calculate a quantity of entrained air. inflow conditions and scale can now account for this broad range of results [44]. For a plunging jet entrainment has been shown to only occur when it is perturbed [44. especially in a mould cavity of complex geometry. Cair is usually set at 0. Fluid jets with very high Reynolds numbers can impinge on a volume of fluid without initiating entrainment so long as their surface remains free of perturbations [46]. Dimensionless number criteria The use of dimensionless numbers has been previously proposed for use in assessment of defect entrainment by Campbell [47] among others. It should be noted that all the above research into dynamic similarity was undertaken using water and not liquid metals. (ratio of gravitational pressure to inertial pressure) [48] include that of Cuesta et al.e. (6) where q is the fluid density (kg m3) r the surface tension (Pa s) coefficient and gn (m s . [49. 3. (5). Traditionally. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  ffi ðPt  Pd Þ V a ¼ C air As 2 : ð7Þ q The model was validated on data collected by researchers in the hydraulic engineering fields.

Fig. the Re number has been applied by foundry engineers to assess running systems for turbulence. [49. There are some questions over the work on the We number by Cuesta et al. has shown both experimentally and using modelling that the Fr and Re numbers can be used to characterise the flow form of water entering a vertical rectangular die.C. Cuesta et al. Fig. Four discrete flow forms were defined: transition. which is entraining both air and oxide film into the bulk fluid is advantageous. the shorter the time for the disappearance of the ‘vena contracta’ and that this is desirable for an optimised running system. it is widely acknowledged that a value of approximately 160° is appropriate for most liquid metals.50]. which lead the authors to question its validity. Gravitational pressure lg ql g lg l ¼ Dh ¼ A : T ð9Þ ð10Þ Work by Isawa using the Froude number in the casting arena involved water modeling in a supposedly impermeable mould showed a vena contracta forming at the sprue to runner junction [51]. mound. This work. Hernandez-Ortega et al. Secondly.5–1. which is higher than the theoretical value of 1. The flow conditions through the ingate were assessed using the Weber number. state that assessment of the conditions took place once the numerical modeling ‘‘proved to be accurate enough’’. 2. modelled results shaded in grey. and further research on this topic is required. and not free surface turbulence and that flow with high levels of bulk turbulence often exist which do not show free surface entrainment and therefore would induce entrainment of oxide film defects. and is therefore unknown. Cuesta et al. It seems that the use of dimensionless numbers for assessment of in-gate flows is an area of research which requires a more detailed investigation.5 for all channel shapes. 2. By calculating the dimensions of the volume of air present in this vena contracta and the Fr number of the incoming flow. However further research is required to validate the technique for liquid metals and more complex die geometry. However. as the author states that after twenty-four hours of applying the surface coat. The main findings of the work were that both the size and shape of the in-gate has an effect on the critical velocity at which entrainment will occur. palm and shell in order of increasing probability of entrainment occurring. investigated the influence of geometry on the critical velocity for free surface entrainment of aluminium. water hardly penetrated the mould walls. The permeability of the mould walls to air was not measured. this assumption is incorrect. It would appear that the author is thus incorrectly recommending that the presence of a hydraulic jump. geometries and materials. This is not to say that the Re number has no place in predicting casting defects.7 and (b) 2. The paper then goes on to propose the entrainment threshold lying in the range of We number of 0. (a) has a We number of 4. However it is this author’s opinion the models shown in the paper are in some cases inaccurate. just that current research has not clearly defined its application. this work contains no experimental validation for these threshold values. Historically. Reilly et al. Isawa concludes that the higher the Fr number of the system. This work suggested an entrainment threshold We number of 1. Modelling versus experimental results from Cuesta et al. [52] used a combination of both the Fr and Reynolds (Re) numbers (ratio of viscous to inertial forces) to characterise the filling patterns of a vertical rectangular die using low and medium pressure die casting. and validating against previously published data. The permeability of the mould to air is questionable. It is not clear whether this air was transported into the mould cavity by the hydraulic jump or escaped through the mould walls. under the assumption that it directly correlated with free surfaces turbulence. . The results matched closely with the experimental data. Using a commercial CFD software. / Applied Mathematical Modelling 37 (2013) 611–628 617 2 Fr2 ¼ Inertial pressure ql v 2 v 2 v ¼ 3 ¼ Fr ¼ pffiffiffiffiffi . an empirical relationship was then used to calculate the time a flow of that Fr number would take to remove the air in the vena contracta. This technique could be used to allow the foundry engineer to assess the likelihood of entrainment by calculating the Fr and Re numbers of the flow entering a vertical rectangular die and see whether it is likely to be entraining. However. that the Re number is a measure of bulk. Firstly the choice of a contact angle of 10° between the mould wall and liquid metal [49] seems unrealistic. It should be remembered however. Experimental results are outlined in white.4 for the entrainment of oxide in liquid aluminium in square inlet channels.3. they modelled both round and rectangular cross section vertical in gates to assess the critical conditions at which free surface entrainment is initiated. although not directly modeling defect entrainment.

The We number showed a large difference in magnitude between the filtered and unfiltered conditions but incorrectly differentiated between the two filtered conditions and the two unfiltered condition. The Hsu data correlated with experimental data in the 4 and 6 mm mesh condition but not that of the 2 mm condition. To enable this to be achieved the flow regime was first characterised as one of four types as described in Fig. This allows the quantitative comparison of running systems. Weber number [48] and Hsu number [56] were used for the assessment of returning wave forms in horizontal runner bars. Upon completion of the model the instantaneous dimensionless number could be integrated with respect to time to calculate a single quantitative ‘total damage’ value for each model.55]. Examination and comparison of the experimental results and modelled results showed that they were consistent with one another. Once characterised the appropriate assessment technique allowed the extraction of the relevant velocities and length parameters to allow calculation of the dimensionless numbers within the runner bar. the models included the modeling of the pour. whereas the We number was not found to accurately predict casting quality. [53–55] have used dimensionless numbers to create criterion functions with which to interrogate computational models for quantification of entrainment. Reilly et al. i. The high energy flows usually present within a runner bar often overcome the surface tension forces.59]). / Applied Mathematical Modelling 37 (2013) 611–628 Fig. However due to being single phase flow it was not possible for the software to accurately model the large numbers of bubbles seen in some experimental conditions. The high head head condition mould created tensile specimens with so great a degree of entrainment present (multiple visible bubble defects within the test specimen) that tensile testing was deemed as inappropriate.55]. The Fr data was tested for mesh sensitivity and was found to correlate with experimental data in the 2 and 4 mm mesh condition but not that of the 6 mm condition. bubbles and micro-porosity [53. 4 and 6 mm side lengths was also under taken [53. two head heights (high and low). where v is the velocity and l is a length [89]. where H (m) is the fluid depth and c (Pa s) is the surface tension (N m1). 3. If the surface tension forces were sufficiently great then there would be no entrainment even for the . returning waves and rising jets. Analysis of the model suggested that this was indeed due to the sensitivity of the model to mesh size rather than a sensitivity of the Fr or Hsu criterion. The Froude number [48].618 C. the modelled flow was different in the different mesh sizes. The ratio of inertial to gravitational forces (Fr number) appears to provide the best representation of entrainment within the runner bar. Four moulds were cast containing tensile test specimens. This sensitivity to mesh size does however severely limit the use of dimensionless numbers for optimization purposes at this stage of development. namely: oxide films. Flow type schematics for dimensionless number models. The Fr and Hsu numbers were seen to correlate with the experimental data. Scanning Electron Microscope (SEM) analysis was undertaken which showed the cause for specimen failure was consistent with those associated with entrainment defects. The difference between the two filtered conditions and between the two unfiltered conditions was much smaller in magnitude than that between the filtered and unfiltered conditions. A mesh sensitivity study using a regular Cartesian mesh of 2. low head with filter and low head conditions [53. 32 and 8 respectively for the high head with filter. both with and without reticulated foam filters.55] (the higher the Weibull modulus the greater the integrity [4. 3. The experimental procedure was modelled in FLOW-3D. The experimental data gave Weibull moduli of 38. Reilly et al. This technique was validated against experimental work. The Hsu number is defined as the ratio of inertial to gravitational plus surface tension pressures as shown by Eq. Hsu ¼ Inertial pressure Gravitational pressure þ Surface tension pressure Hsu ¼ qv 2 ðqgHÞ þ ð4HcÞ : ð11Þ A user customisation was programmed in FLOW-3D to extract each of the dimensionless numbers at a determined frequency from the runner bar of the model [57]. (11). The integrity of each system was assessed using the Weibull modulus [58]. An example of fluid energy overcoming surface tension forces can be seen in flow structures including: plunging jets.e. showing the same flow types..

However further development is still required before these codes are viable as commercial software packages. Many of the techniques described below have had to make assumptions about: both physical characteristics of the defects and their behaviour.67] which will allow the possibility of accurately assessing a simulation software particle tracking model. It is only current work by Griffith et al. In this investigation the theoretical entrainment threshold of 1 and the oxide entrainment rate being linearly proportional to the dimensionless number was used. Many of the following techniques used to model discrete oxide film entrainment utilise particles to represent entrained defects. the technique requires further developments. undertook investigation [69] into the adherence of re-oxidisation inclusions onto mould walls in steel. Work by Ohl et al. Whilst these results appear encouraging. Reilly et al. and establishment of a relationship between the dimensionless numbers and degree of oxide entrainment. coalescence and separation with experimental data. Based upon these qualitative findings they allowed re-oxidisation inclusions to adhere . their entrainment. This is obviously very challenging and usually requires greater computational expense than the indiscrete methods described previously. [46] and Chanson et al. The particle-fluid coupling has only been assessed qualitatively thus far [68]. / Applied Mathematical Modelling 37 (2013) 611–628 619 free surface profile of a return wave as the surface tension would restrain the free surface preventing the entrainment of air packets. [66. Carlson et al. computationally highly intensive when compared to single phase flow modeling due to the substantial additional complexities of modeling the second phase. At the current time the authors are unaware of any two-phase-focused software incorporating quantitative defect prediction models. advection and coalescence of bubbles within the melt [62. The definition of these entrainment thresholds may be possible by further developing the work of Pita and Felicelli [60]. There are however some current issues which require further investigation. These software are. However. By modeling both the bulk fluid and surrounding gas (two phase modeling) it has been possible to describe the entrapment. It should be remembered that the surface tension of water is approximately 10% that of liquid aluminium. modellers will have to work closely with experimentalists for effective progress to be made within this field. this does not mean that they have not been successfully validated [64] and applied qualitatively in the application of defect prediction and process optimization [65]. critical entrainment thresholds and interaction of defects with both mould materials and each other.7. This makes it difficult therefore to quantify the magnitude of entrainment without assessing these factors alongside the dimensionless number. The use of any two-phase flow software to quantify or track the defects produced by the entrained gas has yet to be under taken. these include the quantification of the entrainment threshold in liquid metal as opposed to water. The properties of oxide adhesion to mould walls is not fully understood. has shown that the magnitude of undulations upon the fluids surface and the physical scale of the entraining phenomena greatly impact the magnitude of entrainment. Therefore. Further research is required into the following topics. However. Initial results show correlation of bubble motion. Further development of discrete film entrainment techniques would investigate the possibility of modeling oxide film defects as sites for porosity nucleation.C. namely: Many of the particle models within the software have had no experimental validation. but determining exactly what mechanism or physical situation to model is the greatest challenge facing modellers. It appears that currently the developers of two-phase-focused software are concentrating on developing the flow modeling rather than the addition of models for the quantitative modeling of casting defects. The effect of the changing surface tension with age of the oxide file could also make defining the entrainment threshold problematic as it is likely to be extremely sensitive to surface tension. the ratio of inertial to surface tension forces (We number) does not well represent the likelihood of entrainment in this scenario. Modeling of discrete defects Methods have been developed to model the entrainment and advection of discrete defects. It is often not the practical modeling. [39–41. Obviously. advection and coalescence is an important element of the modeling of casting entrainment. Multi-phase modeling Bubble trails have been shown to be highly detrimental to casting integrity. as expected. although the addition of one or more of the techniques described to model discrete defects within this paper has the potential to yield good results. However their minimal computational overhead makes this technique attractive to industrial and optimization applications should they gain further development to resolve these issues. The accurate modeling of bubbles. For this reason.63]. 3. The models for predicting porosity to use in heterogeneous nucleation have thus far concentrated on bubbles as sites for porosity formation. the adherence of oxides to mould walls can greatly affect the models results. 4. often caused by not having understanding of the physical behaviours of oxide films in the real world. namely: entrainment mechanisms can often be identified and the final defect location can be obtained. This comes with some currently inherent issues.61]. However there are some considerable advantages associated with this approach. modeling both the liquid and gas phase appears to be the only way to correctly model highly aerated flows. This work has simulated the movement and breakup of a thin solid film in a volume of fluid and as is discussed later in this review article in the section entitled ‘‘Modeling of Oxide Film Deformation’’.

These fluid droplets have an oxide skin around them preventing recalescence with the melt. such as the ‘rising jet or fountain effect’. Therefore. Most available commercial packages could describe the initial entrapment of gas or void regions caused by macroscopic fluid flow scenarios. This can cause . However. However. shape and density. most of the currently available casting software only consider the bulk fluid to be present in the model (single phase modeling). Should a bubble breech the free surface of the fluid during the filling of the cavity. the use of this method proves difficult as it is not possible to specify a mould material permeable to the entrained gas and atmospheric gas within the mould cavity. In the low energy flow fields found in the test bars between filling of the mould and solidification. the greater the probability of a highly damaging defect which initiates failure being present. small droplets may be produced which either adhere to the mould walls or re-enter the melt. 5. Work by Carlson et al. or varying density and constant size. However. Entrapped bubbles (bubbles which do not escape from the bulk fluid) are commonly incorrectly assumed to be created through the rejection of gas upon solidification. thus becoming more numerous and smaller [70]. A bubble can only be modelled if the bubble size is greater than the size of a mesh cell. splash defects and entrapped bubbles. although this work is not conclusive. However the lack of knowledge of the film strength obviously complicated matters. This obviously has a huge effect on the minimum bubble size it is possible to model before simulation time becomes unjustifiable due to the increase in computational time caused by the use of a fine mesh. Defects caused by the entrainment of air into the bulk fluid include bubble trails. these are known as splash defects. It is incorrect to assume that the number of particles directly correlates with the number of oxide films which would be created experimentally from the same flow phenomena. This can obviously have a huge effect on the final location of defects. it is possible to interpret the networks of highly tangled oxide films [47] and large dross defects [69] in published work as evidence for the scenario that oxide films adhere to one another should two films collide. however it is unreasonable to add vents to every mould cell. Bubble entrainment A major omission from many casting software is the ability to model correctly the entrainment of air into the bulk fluid during casting. Further detailed investigations are required to confirm this hypothesis. Therefore. Experimental work has shown that when a filter is used there is the possibility of the oxide films becoming shredded. many models are not capable of modeling this. this assumption has no experimental validation. Although many discrete particle entrainment techniques do define and track entraining events they currently do not quantify the amount of entrained oxide. For the case of sand moulds where it can be assumed that the mould is permeable to gas within the mould cavity. mould material. The particles are commonly specified as spheres of either constant density and varying size. Full physical modeling of bubble entrainment would require the modeling of bubble entrapment. However. However. gas pockets which are either entrapped against the mould walls or entrained within the fluid cannot be vented through mould walls and are therefore often incorrectly trapped within the mould volume (unless a pre-placed vent is present within the entrapped volume). Agglomeration of entrainment defects is a difficult subject as to date no experimental work has been published into the adhesion of oxide films to one another. It is anticipated that future development of the code could define the particle size as a function of the area of entrained oxide film. / Applied Mathematical Modelling 37 (2013) 611–628 to the mould walls in their model. [69] (dealing with re-oxidation rather than oxide film inclusions) allowed particles to agglomerate as a way of easing the computational load and to more accurately describe the characteristics of oxidisation inclusions in steel. the buoyancy and drag force of each particle will determine their final position. it is felt that the mould surface. The use of spherical particles and the generic properties to represent individually unique oxide films is currently not able to be validated due to the lack of experimental data. The entrapped air volumes can be specified as adiabatic and are initially trapped at atmospheric pressure. this affect is unaccounted for in the current models. the greater the numbers of defects present. it is not possible to categorically state that the number of particles entrained correlates with the area of oxide film entrained and thus damage to the material. Although many commercial software codes have the ability to model bubbles with pseudo two-phase flow as discussed above. where a volume of air is encapsulated by the fluid. therefore. Instead vents have to be added. its advection including the drag forces placed on the bubbles motion by its oxide trail the bubble trail and bubble agglomeration. currently commercially available software capable of modeling the two-phase flow (which is required to model large scale bubble entrainment and detrainment effectively) for complex casting shapes often struggle to meet the industrial users requirement for short runtimes. The morphology of the particle will have an effect on its motion due to a change in drag forces. Bubble trails are hollow cracks (tubes) which create leak paths through the casting [71–73]. If the bubble is near the surface it is assumed that it is the result of some reaction with the mould or mould coating. The coefficient of restitution used in the models determines whether particles adhere to mould walls or rebound with an energy loss. folded or creased) between its formation and its final form in the solidified material is not accounted for. Reilly et al. velocity (both magnitude and direction) and defect properties will all affect the defects adherence to the mould wall. The use of pseudo two-phase flow for permeable moulds can be highly problematic. The direct simulation of the bubble entrapment and subsequent effects of the bubble have been avoided.620 C. The changing the film’s morphology (for example a large thin film becoming ‘screwed up’. There has been no conclusive research into the adhesion of oxide films to mould walls. Currently this is not accounted for in the models described below in this section. varying in size. Oxide films are individually unique.

However the technique was found to be extremely sensitive to the particles buoyancy force (this is related to the particle density and size). The final location of these particles and their size give probabilistic representation of the likelihood of entrainment defects being present. The particle motion is tracked until solidification and agglomeration of colliding particles (oxides) is permitted. These are then used to define the location of heterogeneous nucleation sites for of gas porosity [74–77]. by the relationship n0 ¼ l0 . is defined by Eq. or placed upon the free surface so as to give a minimum free surface particle density. i. 5. The number of inclusions nucleated. an extremely fine mesh is required. This technique has been developed to remove the need to use extremely small mesh element sizes to track small bubbles. This is most likely due to the perceived difficulty of even obtaining experimental data on the physical properties of the bubble trail and creating a bubble trail model which has practical use. their growth is not permitted. and Nbirth inclusions are added to the free surface cell. overall mass transfer coefficient (cm/s). Where Abirth (cm2) is the free surface area available in the cell. The modeling of the bubble trail with all its intricacies has yet to be attempted. It is highly likely that if bubble trails were to be modelled the extreme computational expense would be too great to allow its introduction into commercial software or to run casting models of even moderately complex geometries and scale.e the minimum number of particles upon a free surface is defined by the user and the code adds particles if required to achieve this value. This number is then rounded up to the nearest integer. (12). These single phase software require. The release mechanism allows the user to release particles at the fluid inlet of a user defined size and at a defined inclusion density (number of inclusions per unit area). This produces runtimes many magnitudes longer than is acceptable in most industrial scenarios. Reilly et al. which indicates the desired initial spacing between inclusions that nucleate on the free surface. developed a technique to place particles when the void region becomes too small to be defined by the mesh. which occurs only when they are at the free surface. Single phase modeling software such as that used in the foundry industry are limited to the size of mesh they can use due to the requirements for fast results [11]. and b is an effective. is determined by Eq. l0 (cm). To allow the tracking of these small bubbles Ohnaka et al. Therefore.80] introduced particles into the melt and allowed them to grow when upon the free is the area (cm2) of the melt free surface that is contributing oxide to the growing inclusion. The birth mechanism creates inclusions of negligible size on the free surface of the melt during casting. which can be used separately or together. Nbirth. When the particles are sub-surface. (13). The nucleation spacing is related to the number of nuclei per unit 2 free surface area. The model allows two mechanisms for the creation of oxide films. Modeling of oxides in steels A method based upon the formation of oxides from nuclei was used as a methodology to model defects in steels. Work by Ohnaka et al. where Vincl is the inclusion volume (cm3). Obviously using the bubble model is not unrealistic for impermeable dies where vents can be added in the same location as the real vents in the die.C. These particles are then tracked and their final locations defined. AFS. spaced a distance of l0 apart. n0 (cm–2). V incl ¼ AFS. The particles were either added to the incoming fluid stream. which is the cell free surface area minus the free surface area taken up by existing inclusions in the cell. 2 Nbirth ¼ n0 Abirth ¼ l0 Abirth : ð12Þ The growth of the inclusions. only their advection is modelled. lcrit rffiffiffiffiffiffiffiffiffiffi dmax ¼ 0:084 : 2 ð15Þ . This technique is an adaptation of that developed previously by Tomiyama et al. This allows the tracking of the bubbles without the computational expense of small mesh cell sizes. d¼ rffiffiffiffiffiffiffiffiffiffiffiffi 3 6V incl : p ð14Þ To determine if two inclusions agglomerate the relationship described in Eq. Therefore to track small bubbles such as those entrained by a returning back wave.1. The user specifies a ‘‘nucleation’’ spacing. / Applied Mathematical Modelling 37 (2013) 611–628 621 the flow fields to be incorrectly modelled and sometimes gives pressure convergence issues. release and birth. [78]. A team from Iowa University lead by Beckermann [79. at minimum a single cell devoid of fluid to be able to define a bubble. The modelling of mould venting of consumable moulds such as sand and investment moulds is a topic that would benefit from further research and development. (14). They found the technique gave results which correlated well with experimental results. All inclusions are assumed to be spheres of diameter d and constant density b: dt ð13Þ The inclusion diameter after growth is then calculated using Eq. it is felt likely that other models approximating the bubble trail may be the solution to this problem. has used particles to represent and track entrained bubbles for the prediction of porosity [74–77]. (15) was used where dmax (cm) is the diameter of the larger of the two inclusions lcrit is defined as the critical distance under which two inclusions with agglomerate.

Using Campbell’s design rules the maximum sprue exit area for a 175 mm tall sprue is 105 mm2 [47]. used by Lin et al. The model is able to assess the entrainment of air bubbles into the bulk liquid by surface turbulence. This suggests that the technique may be computationally intensive. expansion into three dimensions would severely complicate the programming required and further increase the computational effort. Dai’s work suggests that the oxide entrainment was by the folding of films within the casting volume. It is likely that this initial test opened up crack initiation sites. One major drawback with this model is that it is currently only implemented on the OFET 2D CFD (computational fluid dynamics) in house code. further particles are added as the film is assumed to have torn and immediately new oxide film has been formed. This will mean that there is effectively a plunging jet occurring as the bottom of the down-sprue until there is enough head height from within the casting cavity to back fill the bottom of the down-sprue. Particles are added if the surface is expanding. The methodology used by both Lin and Dai is based upon placing particles on a fluid’s free surface to represent the oxide film. 5. Obviously the results of the latter three point bend test have to be regarded with some caution. Therefore no new particles are added should the film break due to excessive stress. Should the strain exceed the strength of the film. They have created a very elegant and seemingly robust method of modeling oxide inclusions in steels. The broken samples were then subjected to a three point bend test [83]. and Dai et al. The authors would like to make three comments on the work undertaken by Dai. Secondly it should be noted that the comparison of the down-sprues on both the vortex and rectangular runner validation moulds are different [83]. These techniques have also only currently been applied in two dimensions. as was modelled. [81–84] models the entrainment of bi-films through the folding of the free surface. The sprue inlets and pour basins are identical on both. The authors are unaware of any shape casting simulations undertaken with this software. The location and number of these entrained films are tracked. . the code was validated by mechanical testing of samples cut from cast plates. the equations used can be found in [69]. however the sprue exit is 75 mm2 on the rectangular runner and 150 mm2 on the vortex runner. Carlson and Beckermann have further developed and undertaken validation work of this steel inclusion modeling technique showing good correlation with a number of experimental validations and has been successfully used in industrial applications [69]. 4. The particles also have to be replaced onto the fluid’s surface at every time step. Reilly et al. It is therefore only able to model certain entraining phenomena such as returning waves and folding surfaces. / Applied Mathematical Modelling 37 (2013) 611–628 The motion of the inclusions is defined by solving the equations of motion for each spherical inclusion. as the effect of the initial test on the strength of the sample is not quantifiable. There is a large Fig.2. and when a particle is added to the model then all particles are re-labelled. Firstly. This model was then compared to experimental data and deemed to be qualitatively consistent. Dai’s approach varies slightly by assessing the surface normals of the films. Modeling the folding mechanism This method. should the free surface form have changed. Should they be pointing towards each other and their velocity vectors obey a predetermined mathematical rule (meaning that the films would converge) then entrainment is deemed to occur.622 C. Lin assumes the oxide films to be present between neighbouring particles and calculates the strain the film is under by tracking the movement of neighbouring particles. presumably due to its current 2D limitation. potentially severely weakening the samples. and shown it to give reliable results in industrial applications. Example of OFET 2D oxide tracking model used for validation of the technique [81]. It can therefore be seen that the rectangular runner has a choking sprue whilst the vortex runner’s sprue is oversized. It should be noted that the samples were initially tested by subjecting them to a four-point bend test. Once a film is entrained in the bulk material the tracking points are no longer adjusted to fit the free surface and it is assumed that there is no atmosphere for oxidisation within the bulk material.

As little is known about the effect of alloy composition on film strength and tearing properties this becomes difficult to calibrate and validate. meaning |ujs1 + ujs1| is the relative collision velocity. SAverage ¼ M P SM 1 : ¼ Nis Nis s ð17Þ Eqs. The number of entrained oxides per unit area is estimated as a function of collision velocity and alloy composition and is given in Eq. 1. The average surface area of the broken oxides is estimated using Eq. however the nature of this is unknown. Fig. (17). the free surfaces are then assessed using defined physical rules [87] to see if they collide. ð16Þ where a1 and a2 are parameters related to alloy compositions. 5. upon inspection of the models run using the OFET code the flow can be seen to be perfectly symmetrical (Fig. / Applied Mathematical Modelling 37 (2013) 611–628 623 Fig. Eq. Making the assumption that the aluminium surface which is exposed to the atmosphere instantly forms an oxide film. It is known from real time X-ray results that this is unrealistic. (18). distances between particles and the time period. 5 [85]. It is stated only that the values are obtained by experimental work. Real time X-ray example of flow in a cast plate (1995 benchmark test. (as seen in Fig. However the authors feel that further investigations and development are required to exploit its full potential. probability however. more but smaller oxide films are entrained. marker particles are placed to represent the entrained films. The judgment of a free surface collision is classified by assessment of the velocity vectors. This has not been accounted for in the model. The validity of comparing the incorrect computational models to the experimental data is therefore questionable. (13) and (14) mean that at larger collision velocities. that there were a significant number of young oxide films introduced to the plate castings which were entrained in the down-sprue of the vortex runner casting.3. where S is the collision surface area. namely: quantifiable output (number of particles in the model/critical volume can be counted) the motion of the defects can be modelled. It is unclear from the published work however what data was used to define a1 and a2 and also what values were used in this work. 4). Modeling of oxide entrainment The work by Ohnaka et al. 5.C. on modeling bubbles in single phase flow [76] was then further extended to include the modeling of oxide entrainment in aluminium castings [86]. The reason for this flow behaviour is assumed that a pressure boundary was applied to the bottom of the plate as an inlet condition as the 2D OFET code was unable to model the running system. where bd_dw and bd_up are dimensionless distances of downwards flow and upwards flow .75 s) [85]. ujs1 and ujs2 are the collision velocities. Thirdly. This is the experimental data which was modeled by Dia et al. If entrainment occurs. SM the surface area of a broken oxide and Niss is the number of entrapped oxides. 1). (16). It should however be stressed that this method of modeling the entrainment of oxide films through folding of the free surface does have merits. thus entraining oxide films. Reilly et al. Nis ¼ a1 jujs1  ujs2 j þ a2 . Their advection within the flow is then calculated to define their final locations upon solidification.

The modelled results agreed with those found experimentally. The particles motion is then modelled. If the film can be made to deform upon the free surface and break then the possibility of modeling directly the breakup and entrainment of the film will become a reality.7). Dt is the time step and d is the element distance. However. and the effect this solid deformation has upon the fluid motion: i. The average number of particles in the gauge length of the test specimens was compared to that of the Weibull modulus. however the details can be found in [60]. fraction of fluid at both the beginning and end of a time step. Although this technique is currently not aimed at defect entrainment prediction it is included as it has the possibility to be developed into a key constituent in the development of accurate defect entrainment models. / Applied Mathematical Modelling 37 (2013) 611–628 respectively. Once an entraining event has been detected a particle of determined size and density is placed to represent the defect. and solidification elements (phase change and solute transport).7 and 19. and thus incorporates quantitative assessment techniques of the final particle locations. impinging flows and return waves to be modelled. However. Reilly et al. further investigations to give a much larger data set and using a variety of running system designs which emphasise different entraining flow phenomena are required for conclusive validation of the technique. Pita and Felicelli [60] state that they aim to further develop this technique by simulating a more realistic model and including a breakage criterion for the film. n is the inward normal vector to the surface and u is the velocity vector on the element surface. This work consisted of pouring moulds. The technique has in two dimensions shown that it is possible to accurately model the advection and large scale deformation of a solid film within a fluid. If this technique can be developed into a three dimensional model then the possibility of modelling this phenomena will become a reality.. bubbles. colliding fronts. coupling of a fluid and deformable thin film. This is also a code targeted at optimization. The results showed the filtered condition to give test bars of greater integrity than that of the unfiltered condition: Weibull modulus [58] of 37. A significant step in the development of this technique would be the introduction of a free surface within the model domain. With the present state of computational hardware it seems unlikely that this technique can be applied to large scale castings in the near future due to the computational intensity of modeling numerous films (which require the micro flow to be simulated) alongside the macro flow and solidification. [90]. both with and without a 10 ppi reticulated foam filter at the bottom of the down-sprue. The test specimens were then tensile tested. Upon solidification the particles within defined critical volume(s) or the whole casting can be counted to allow quantitative analysis of the casting system. Thus far no evidence has been published to categorically prove or disprove Campbell’s theory. Further validation with respect to the modeling of oxide defects rather than that of porosity which has been so far undertaken [86] would give a greater insight into this methods validity for the modeling of oxide film defects. It is not possible to detail here the exact methodology used as it is numerically complex and beyond the scope of the article. The work shows promise for the applications in modeling the unfurling of oxide films in castings. X ~ n~ uDt P ð1  btd dw  btd up Þd: ð18Þ This methodology allows the entrainment caused by fluid jets. The incorporation and transport of particles within the liquid metal as reported in this work is not unique. and Ohnaka et al. which was shown experimentally to be of greater integrity (Weibull modulus 37. [86]. used a very similar methodology to that developed by Ohnaka et al. [76]. In this work an oxide film entrainment model was developed as a FLOW-3D customisation [88]. however the implementation varied due to factors associated with the software. 5. The addition of the ability to model a breaking film will allow the direct modelling of film entrainment for simple models. the technique may play an important role in gaining insight into the physical behaviour of oxide films. However. Therefore it is known that the failure mechanism is due to the entrainment mechanisms modelled by the OFEM rather than another unaccounted for factor.624 C. Algorithms for doing so having been described previously by Yang et al. it is considered that this work is an initial evaluation and quantitative validation of a promising technique for modelling entrainment defects in shape casting.89]. had an average 1458 particles in the gauge length. of which surprisingly little is definitively known.7). [89]. The experimental work of Green and Campbell [4] was modelled in FLOW-3D and the OFEM applied [88.e. compared to an average of 1945 particles for the unfiltered condition (Weibull modulus of 19.5. orientation of the free surface normal and surface area of free surface cells and defines entraining events by the use of Boolean logic criteria. which is believed to be one of the mechanisms of porosity formation [47]. [83]. The experimental results show the defects ‘‘easily identified as oxides’’ [4] to be the failure mechanism of the test samples. The model assesses the velocity vectors. The filtered condition. Although it has to be expected that this would be extremely computation- . 5. The code is fully 3D implemented.7 respectively.4. It is therefore included here as it is felt to be of the upmost importance. Dai et al. Oxide film entrainment model (OFEM) Research by Reilly et al. Modeling of oxide film deformation Work by Pita and Felicelli [60] has modelled the transport and deformation of a single oxide film within a fluid volume.

Returning waves Low/Med No Cumulative scalar technique Flow-3D MAGMAsoft All Low No Defect transport and probabilistic locations modelled Air entrainment model Flow-3D MAGMAsoft All Med No Entrainment location well defined Dimensionless number criterion Cuesta et al. Hard to define which vortices effect free surface entrainment in highly turbulent flow regime. with assessed parameters including: material properties. density. size. fluid velocities and free surface topology then the accuracy of many current entrainment modeling methods could be greatly improved. adhesion) Difficult to implement in all running systems No entrainment threshold defined for liquid metals Difficult to assess all areas of a casting Oxide trail not modelled Requires an appropriate model to define the initial entrainment of the bubble Computationally intensive Can be computationally intensive at very high degrees of entrainment. size.. density. Benefits Method Significant researchers Entrainment mechanisms Assessed Computational overhead Discreet modelling of oxide films Cumulative entrained free surface area Lai et al.g. Further research required ally intensive.g. adhesion) Models entrainment and transport of defects Models discrete defects and their transportation Film properties modelled (e. under what conditions does entrainment commence? The second significant difficulty. The addition of these entrainment thresholds would solve one of the major questions in entrainment modelling.. No film properties modelled (e. / Applied Mathematical Modelling 37 (2013) 611–628 Table 1 A summary of the benefits and limitations of the major defect modelling techniques which have been explored.. Reilly et al.g. thus ruling out the possibility of full scale casting simulation it would allow the determination of entrainment thresholds. size) Modelling of oxide entrainment Ohnaka et al. strength. Reilly et al.625 C. Isawa Reilly et al. oxide age (thickness). If simple models can be simulated to allow the definition of entrainment thresholds for different entrainment phenomena.g. On this front the current model by Pita and Felicelli is a significant step to being able to assess the accuracy of the current methodology of using spherical particles to represent entrained films which is used by many of .. Sun et al. Dia et al. N/A Low No Vorticity MAGMAsoft Flow-3D Impinging streams. Fountains Low No Computationally simple Easy to quantify Bubble entrainment Ohnaka Bubbles Low No Bubbles sub-mesh cell size can be tracked with low computational overhead Multi-phase modelling CFX Physica All High No Modelling of oxide inclusions in steel Calson et al. Returning waves. All Med/high Yes Models entrainment and transport of defects Defects parameterised upon entrainment conditions No film properties modelled (e. Folding of fluid surface Med/high Yes Can model highly aerated flows Proven in industrial applications Oxide inclusion properties modelled (e. Restricted to modelling of oxides in steel Can be computationally intensive at very high degrees of entrainment. density. All (steel castings) Med/high Yes Modelling of the folding mechanism Lin et al. Difficult to quantify Diffusion of scalar is physically unrealistic No film properties modelled (e.g. size) Limitations Comparison of different running systems is virtually impossible. associated only with the discrete modelling of defects is the correct modelling of advection and adherence properties. The accuracy of many current entrainment models is severely reduced as the entrainment thresholds are not known. Not yet applied in 3D Experimental validation required Can be computationally intensive at very high degrees of entrainment. strength. size. density. density. adhesion) Recent results lead to its validity as a quantitative rather than qualitative abilities becoming questionable..

What is a good gating system? or Quantifying quality. Kokot. Simulation of Aluminum Shape Casting Processing: From Alloy Design to Mechanical Properties. as summarised in Table 1. [5] J. . The topic of modeling entrainment defects in casting has received little attention in recent times despite its obvious commercial significance. do oxide films stick to the mould surfaces and or under what conditions. will aid the assessment of whether thin films tens to hundreds of lm in length and only lm in thickness within a fluid can be modelled as a continuum. The Technical University of Denmark. the modeling of defects has been shown to be achievable and advantageous. Reducing energy consumption for melting in foundries. the support of Flow Science Inc and the EPSRC support of NG’s chair (EP/D505569/1). Campbell. Metals and Materials Society. 7. The use of this data could be highly beneficial in improving the accuracy of discrete modelling techniques. and invisible to the naked eye and still challenging to identify even with more sophisticated techniques such as using a scanning electron microscope. [2] V. pp. Although once again challenging to validate. The ability to assess the deformation of the film during advection within a flow and assess its velocity within a known flow field. a three dimensional model containing two or more films would allow the assessment of the representation of thin films with spherical particles and the interaction between two films when they collide. One of the most difficult of these is not the actual modeling of the defect but instead acquiring the knowledge of what to model. It is suspected that a representative experiment rather than an experiment using liquid metal or indirect qualitative evidence will be all that can be achieved in respect to validation of models using these techniques. Campbell. The ability to model a breaking film would potentially allow the assessment of film dimensions for given flow parameters for a given entrainment phenomena. v 2006. in: Department of Manufacturing Engineering and Management.626 C. [4] N. tens to hundreds of lm in length and only lm in thickness. Conclusions In this article the published methods of modeling entrainment defects in metal castings have been reviewed and the topics requiring further research have been highlighted. 2006. San Antonio. 2007. The ability to model 3D films would allow the modelling of bubble trails. AFS Trans. However there are currently a range of techniques available to the modeller. Although the development of the technique of modeling the deformation of an oxide film has the possibility to be extremely powerful. It is hard to see where to further develop this model without new experimental evidence to giving further insight into the properties.80]. / Applied Mathematical Modelling 37 (2013) 611–628 the discrete modelling techniques. J. Summary The modeling and quantification of defect entrainment in the casting scenario is in its infancy and is an extremely difficult proposition due to a number of complex problems which have to be addressed. welding and advanced solidification process XI. 2006. or whether they have to be modelled individually with a drag coefficient which requires determining and may be time dependent if the films morphology changes with time. The re-oxidisation inclusion model [79. Burnbeck. Green. www. life cycles and behaviours of defects in steels. Copenhagen. entrainment location and or final location of casting defects. TX. The development of quantitative defect modeling techniques is difficult and complex. what are the characteristics of oxide films created through different entrainment mechanisms and how do oxide characteristics affect the motion of defects within the melt? These questions require experimentalists to work alongside modellers to make further progress in the modeling of entrainment defects in castings. 6.80] currently allow the agglomeration of particles based upon qualitative evidence. The simulation of many phenomena has not yet being undertaken due to the complexity and lack of physical understanding. The Modeling of entrainment defects during casting. The ability to model these tubes would allow insight into under what conditions the bubble trails are torn so as to try and understand not only their formation but their behaviour under common casting conditions. P. but of great industrial significance.but how? Modelling of casting. These tube like structures have thus far never been modelled but are widely believed to be highly detrimental to casting integrity [71–73]. Minerals. it must be considered that validation of these techniques will be extremely challenging. [69. Reilly et al.php.magmasoft. Oxide films are small. Acknowledgments The authors would like to acknowledge the help of The School of Mechanical 114 (1994) 341–347. United States. The University of Birmingham for sponsoring the PhD of CR. References [1] S. For example: do oxide films agglomerate if they collide. this also eases the computational load by reducing the number of particles present within the model.79. However. which providing their limitations are recognised may shed light on the quantity. in: TMS Annual Meeting. Skov-Hansen. Influence in oxide film filling defects on the strength of Al–7Si–Mg alloy castings. 119–126. One such example is the modeling of oxides in steels has been elegantly undertaken and validated by Carlson et al. and therefore further research is urgently required. [3] MAGMASOFT.

Mechanics of Fluids. C. J. J. in: 1st International Aluminium Casting Technology Symposium. [10] C. 99 (1991) 83–90. in: World Foundry Congress 2006. Oxford. H. Math. M. J. J. in: Ductile iron society 2003 Millis. Laty. Cleary. Lee. Samuel. Modelling of the potential for oxide film entrainment in light metal alloy castings. 190 (3–4) (2000) 403–410. Cuesta et al. Lai. Comput. Campbell. Lewy. A new approach to numerical convection. Flow Science Technical Note. Three-dimensional free surface modelling in an unstructured mesh environment for metal processing applications. Birmingham. Prakash et al. Hydraul. Burley. [14] M. Jolly. J. A numerical technique for low-speed homogeneous two-phase flow with sharp interfaces. pp. 193 (39–41) (2004) 4355–4381. Trapp. Courant. [24] N.W. Oguz. Cross. World Environmental and Water Resources Congress. Hydraulic jumps: Bubbles and bores. Sun et al. J. 100 (1) (1928) 32–74. 2010. p. Eng.D. The critical ingate velocity in Al and Mg alloys. Reilly. in Fifth international conference on CFD in the minerals and process industries. [61] H. [47] J. S. in: School of Mechanical Engineering. Gold Coast. Über die partiellen Differenzengleichungen der mathematischen Physik. See. Further developments of running systems for aluminium castings. J. Numerically modelling oxide entrainment in the filling of castings: the effect of the Webber Number. H. Appl.H. Hirt. [39] H. Mech. [16] B. Model. 1 (2001) 15–30. 132 (5) (2010) 051011. Hoque. Materials Science and Engineering. Eur. Birmingham. . Barkdudarov. Methods Appl. [52] J. R. Methods Appl. Comput. R. Towards the ultimate conservative difference scheme. Fluid Mech. Aliabadi.W. freezing and defect creation in low pressure die casting. Chem. Phys.C. 58 (11) (2006) 62–65. Chanson. [28] Flow3D. J. Chanson. Eng. [30] M. Chanson.D. Aliabadi. in Third international conference on CFD in the minerals and process industries . Volume of fluid (VOF) method for the dynamics of free boundaries. in Materials and Metallurgy. Methods Appl. [9] L. 46 (1) (2008) 35–44. Extension of SPH to predict feeding. [26] W. [35] P. Math. Welding and Advanced Solidification Process X.W. Eng. scale effects and self-similarity of stepped spillway flows. Phys. Int. W. Technol. 2004. J. Friedrichs. Cross.-Y. Australia. Bubbly flow structure in hydraulic jump. Ramshaw. Sci. Yonggang. sixth ed.R. J. [56] F. Air entrainment by a viscous jet plunging into a bath. Sci. Isawa. 2010. J. Chanson. Reilly et al. [45] E. V. [17] R. 122 (2002) 82–93. 2006. Pekguleryuz. Pericleous et al. Similitude and scale effects of air entrainment in hydraulic jumps. Crown Plaza. 1995. Cross et al. Reilly et al. [37] R. [60] C. [32] P. J. 2003. F. Comput.4 User Manual. 38 (1–2) (1993) 291–302. M. London. [29] M. C. [58] W. C. Green. Lewis. Campbell. [21] T. 1998. Mater.S. 21 (4) (1976) 438–453. M. J. A. Process.flow3d. Tian et al. Struct. 2004. Meth. Manuf. Mater. [48] B. 144 (3–4) (1997) 227–233... A statistical distribution function of wide applicability. Chanson. 2008. AFS Trans.W. Mech. Oxide entrainment structures in horizontal running systems. [49] R. 20/10/2009. Process.N. [12] H. Model. [34] P. 404 (2000) 151–177. and defect formation in ductile iron castings. p. [23] K. 22 (11) (1998) 895–906. Melt oxidation behavior and inclusion content in unmodified and Sr-modified A356 alloy-their role in pore nucleation.D. 38 (2003) 1255–1267.E. Pita. Mech. Modeling turbulent entrainment of air at a free surface. 2003. Cleary et al. Chirazi.W. ALE method for solidification modelling. [15] C. Prosperetti. The University of Birmingham. Lai. 59 (2004) 255–262. H. Phys. Ohl. Physical Review Letters. 104 (1996) 763–768. Hirt.. P. www. 93 (week ending 17/12/04). Reilly et al.magmasoft. Barkhudarov. Dross formation during metal transfer operations. Using criterion functions to quantify entrainment in castings. Mater. Tezduyar. / Applied Mathematical Modelling 37 (2013) 611–628 627 [6] P. Kirkaldie. Fluids 45 (3) (2008) 513– 521.E. 560. The University of Birmingham.W. [22] J. R. Preliminary SPH modeling of oxide formation during the mould filling phase in DC casting of extrusion billets.M. Numerical study of large amplitude free surface motions. Melbourne. Hernandez-Ortega et al.. Teodosiu. Hsu. Campbell. Massey.M. Modelling of Casting. Effect of melt cleanliness on the formation of porosity defects in automotive aluminium high pressure die castings. in: TMS2009 Shape Casting Symposium. [42] H. IL. [13] K.A.D. J. 1998. UK. [55] C. EDICT for 3D computation of two-fluid interfaces. Quere. Enhanced-discretization interface-capturing technique (EDICT) for computation of unsteady flows with interfaces. J. Abbes. Eggers. Butterworth Heinemann. R. Eng. Melbourne. [53] C. A. Modeling microporosity in aluminium–silicon alloys: a review. Felicelli.A. [31] MAGMASOFT.R. Gualtieri. Experimental assessment of scale effects affecting two-phase flow properties in hydraulic jumps.. [18] A. Chanson.P. Influence of oxides on porosity formation in Sr-treated Al–Si casting alloys. Appl. Sci. 1994. Ann. J. Green. J. Prosperetti. N. 88 (3–4) (2010). Modelling of Casting. Reilly. H. in: Department of Metallurgy and Materials. Mech. The University of Birmingham. Math. symposium 11. 2005. S. The control of the initial fall of liquid metal in gravity filled casting systems. [38] H. B. Mater. TMS. [50] R. Physical modeling and similitude of air bubble entrainment at vertical circular plunging jets. Jolly.T.S. Tracking defects. Gas porosity and metal cleanliness in aluminum casting alloys. Eng.D. S. Chem. [11] M. [59] N. AFS Trans. A. FSI-03-TN61. Statistical distributions of fracture strengths of cast Al–7Si–Mg Alloy. [27] MAGMASOFT. [7] D. Usmani. [54] C. [46] C. Liu. Eng. Appl. Lewis. C. Eng. S. 2007. 1992. Comput. Emadi. Tracking Defects. 2006. model verification. Numer. Weibull. D. K. Tezduyar..pdf. Nicholls. Comput. J. Hirt. A.T. B. [33] M. J. [41] F. Harrogate. Murzyn. Res. Samuel. 3D SPH simulations of the aluminium ingot casting process. Van Leer. San Francisco.S. Laslaz. The John Campbell Symposia. London. Australia. 35 (4) (2005) 787–806. [8] G. Harlow. 155 (3–4) (1998) 235–248. Aoki. IV. Mechanism of air entrainment by a disturbed liquid jet. in: Shape Casting. V4. Eng. Lorenceau. Chanson. [19] A. [44] Z. USA. Flow 3D Version 9. A fluid–structure interaction method for highly deformable solids. second ed. Gruzleski. Academic press. 39 (9) (1984) 1357–1372. Fluids 12 (2000) 1710.N. UK. Behr. [40] H. Phys. A finite element model for the simulations of mould filling in metal casting and the associated heat transfer. B/Fluids 26 (3) (2007) 367–384. D. Entrainment of air into liquids by a high speed continuous solid surface. [20] T. Technol. Exp. 2003. Bellet. Rosemont. [51] T. Usmani. Comput. A.R.W.R. Fachinotti. Mech. Zaidi. 1996. Fluids 9 (1966) 842. [57] C. Air Bubble Entrainment in Free-surface Turbulent Shear Flows. 34 (11) (2010) 3189– 3201. Hirt. Backer. 415–422. J.R.W. Welding and Advanced Solidification Process XII 12 (2009) 419–426.4 Manual. Sci. Griffiths.D. Light Met. www. K.W. Cuesta et al. J. Australia. Finite element simulation of mold filling using marker particles and the k-e model of turbulence. [25] N. [36] C. Computer simulation study on the influence of geometry on the critical gate velocity for molten aluminium. Welch. 2003. 18 (1951) 293–297. Mech. Development of quantitative casting quality assessment criteria using process modelling. Castings. The University of Birmingham: Birmingham. 2003. Computational modeling of mould-filling and related free surface flows in shape casting: an overview of the challenges involved. Methods Appl. Comput. in: 16th Australasian Fluid Mechanics Conference. On the mechanism of air entrainment by liquid jets at a free surface. An experimental and numerical study of flow patterns and air entrapment phenomena during the filling of a vertical die cavity. Chapman & Hall.J. in: Aluminium Cast House Technology 4th Australasian Asian Pacific Conference.D. Comput. Oguz. The analysis of mould filling in castings using the finite element method. [43] H. 39 (1981) 201. Nguyen. Physical modelling. 23 (3) (1977) 276– 299. 2003. Using the calculated Fr number for quality assessment of casting filling methods. Phys. Modeling.

J.C. M. Mater. in: The 65th World Foundry Congress. 2009. 2009. J. Beckermann. Estimation of porosity defects with consideration of oxide entrapment. J. Modelling of Casting. McBride et al. 1999.. Welding and Advanced Solidification Process XII. Numerical simulation of the movement. in: Shape Casting. in: The Department of Engineering. Bounds et al. Benchmark testing the flow and solidification modeling of Al castings. Estimation of porosity defects in Al-alloy and spherical-graphite iron castings. in: Modeling of casting.R.T. / Applied Mathematical Modelling 37 (2013) 611–628 [62] J. [64] S. Filling and Feeding of Aluminium Castings.. Texas. J. pp. 2000. Divandari. Modelling of Casting. [85] B. [80] M. Jolly. Mater.R. Modelling of the tilt casting process for the tranquil filling of titanium alloy turbine blades. in: Department of Metallurgy and Materials. heat transfer. [67] W. 1 (1) (1999) 71–78. Blair et al. 2005. 915–933. Holliday. and solidification phenomena.A. Lin. 20 (2004) 505– 513. in: AFS 105th Casting Congress. [76] I. Nashville. Ohnaka et al.. 11 (97) (1997) 645. 603–610. Chicago. Des. Carlson. Sako et al. Modelling of Casting. Hill.. [70] B. Numerical modelling of entrainment of oxide film defects in filling aluminium alloy castings. 2002. Porosity formation mechanism in Al and Mg alloy castings and its direct simulation. Welding and Advanced Solidification Process XII 12 (2009) 443–450.. Divandari. Sato et al. Campbell.D. A computational model for defect prediction in shape castings based on the interaction of free surface flow. Aachen. welding and advance solidification processes IX. [81] X. Particle-fluid coupling. TN. Albers. [73] M. M. Korea. [83] X. Modelling of Casting. [79] M. 2001. Investigating surface entrainment events using CFD for the assessment of casting filling methods. Flow Science Technical Note. M. Campbell. J. Mechanisms of bubble trail formation in castings. Tomiyama et al. Welding and Advanced Solidification Process XII 12 (2009) 427–434. AFS. The determination of inclusion movement in steel castings by positron emission particle tracking (PEPT). in: 1st International Conference on the Gating. J. Zhao et al. [66] W. Griffiths et al. Gyeongju. Nucl. Eng. pp. Kimatsuka et al. Shaker Verlag. Modelling of Casting. Mechanisms of bubble damage in castings. breakup and entrapment of oxide films during aluminum casting. Using OFEM to quantify entrainment in castings. Alum. 17 (6) (2004) 321–331. Sci. Mold filling simulation for prediction gas porosity. 2001. Ohnaka et al. Mater. Welding and Advanced Solidification process XII. University of Paisley. Reilly. J. 53–60. Wang et al. Taipei. [63] D. 1995. 2006. Trans. Divandari. Mater. Reduction of surface turbulence during filling of sand castings using a vortex-flow runner.. Blair et al. The John Campbell Symposia. [72] M.D. Welding and Advanced Solidification Process VII. 2004. The Chinese Foundrymens Association. Campbell. Modelling of Casting. in: Melting of Casting and Solidification Processes VI (6th pacific rim conference). The application of positron emission particle tracking (PEPT) to study the movements of shape castings. 1999. in: D7th Asian Foundry Congress. San-Fransisco. Taiwan. AFS Trans. Complex free surface flows for mould filling using centrafugal casting.R. 2002. Influence of gas phase on mould filling for sand casting.. M. [88] C. . Steel Foundry Society of America. [65] H. Green. Metall. The mechanism of bubble damage in castings.. in: 58th Technical and Operational Conference. 77–84. symposium.R. pp. Influence of oxide film defects generated in filling on mechanical strength of aluminium alloy castings. Goodheart. FSI-99-TN50. Modeling of reoxidation inclusion formation during filling of steel castings. Modeling of oxides entrapment during mould filling of Al-alloy castings. Shape casting: 3rd international. 57 (5) (2005) 29–34. Griffiths et al. Reilly et al. [74] A.L. in: Proceedings of the 7th Asian Foundry Congress 2004. [71] M. Yang. [90] X. [69] K. 2005. Int. in: The 65th World Foundry Congress. Campbell. [86] Y. Reilly et al. [89] C. [68] C. Trans. 43 (2008) 6853–6856. Modeling of oxides entrapment during mould filling for Al-alloy castings. B 31 (3) (2000) 515–527.W. The University of Birmingham. [84] X. Designing reliable castings.. C. J. 2009. Effects of liquid metal flow behaviour in the filling on the final mechanical strength of castings. 2004. Dallas. Paisley. AFS. Hirt. Predicting the occurrence and effects of defects in castings. Sirrell. 2010. Sharif. [77] H. A three-dimensional particle tracking method for bubbly flow simulation. J. 231–238. Mechanism of filtration in reduction of casting defects due to surface turbulence during mold filling. Dai et al. [75] I.628 C. Jakumeit. Sirrell. M. Yang et al. Gyeongju. Sci. Tapei. [82] J. 2001. N. Campbell. Korea. Technol. Jolly. Welding and Advanced Solidification Process XI. pp. [87] Y. Birmingham. pp.. Res. Dai. Cast Met. K. 175 (1–2) (1997) 77–86. [78] A.