Direct Flow Simulations in
Pressed Foundry Filters
(Due to the size of graphics in this article, the index has been omitted.  Return to the cover page of this issue)

^a Saint-Gobain Advanced Ceramics Hamilton Ltd -
45 Curtis Avenue, Paris, Ontario, N3L 3T6, CANADA
^b Department of Mechanical Engineering -
Component Technology, Jönköping University Box 1026, S-551 11 Jönköping, Sweden

Abstract
Fluid flow simulation has been used as a tool to understand and optimise gating systems in foundry practice. Gating systems frequently contain filters because of their favourable effect on the flow properties of molten metal. Consequently, filters can have a positive effect on the final quality of cast metals. The present study uses direct flow simulation in the areas around real shape modelled filters. This direct method, rather than the more traditional pressure drop method, gives new insight into the micro flow phenomena of molten metal passing through a filter. The simulations of flow patterns through the filter channel support the theory of cake filtration and thereby promote the fundamental idea of metal filtration.

Keywords: Flow simulation, cake filtration, pressed filters.

1. Introduction
Solid liquid separation technology is widely used in different industrial processes. The phenomena involved are very complex and significant efforts have been made to establish fundamental filtration theories. [1] Generally the principal separation modes are filtration, sedimentation and flotation. Filtering of liquid metals has become standard practice in many foundries as the quality requirements of cast components have increased. A large variety of filters are used and the users generally report the beneficial effect of improved quality. Casting defects related to inclusions are generally reduced and the machinability of casting is normally improved.

However the actual filtration mechanisms are not clearly understood. Various work has which been undertaken, to understand the effect of mould filling on the final properties of a casting, has shown the importance of low fluid velocities contributing to low surface turbulence and less entrapment of surface reaction products into the metallic bulk. [2] Since these effects have been known, the reported quality improvement of filtration was largely attributed to the low fluid flow rate through the filter and the filters role in reducing the turbulence. Reports on microstructure investigation in connection to filters have shown separation and segregation of non-metallic inclusions at the entrance side of filter. [3] This observation indicates the existence of separation phenomena at the filter interface, which are probably a result of complex filtration mechanisms.

Development of fluid flow computation has given us new opportunities to study fluid flow in foundry processes and also the influence of filters on the flow. To fully incorporate a filter into a computer flow simulation, a large amount of extra computing power is necessary. This is because of the intricate structure of a filter. To minimize the amount of extra computer power needed, a special method has been developed to incorporate filters in flow simulation [4]. The filter is considered as a permeable material and a pressure drop is allocated for the fluid through the filter. Both parallel and transverse flow is considered to take into account the anisotropy of the filter media. Filter producers and independent researchers have done considerable work to develop filter data necessary to consider filters in simulations. [5] Validation work reports good accuracy when using a pressure drop to simulate flow through filters [6], however the flow phenomena in close connection to filters is not discussed.

Recent improvements in computational power give us the opportunity to consider the real shape of filters in simulation. Different ceramic filters have been modeled and simulated in their real shape [7] to study metal solidification in a filter and its influence on the flow through the filter. The purpose of this paper is to study the metal flow by simulation in the areas around real shape modeled filters and to compare this to real separation phenomenon observed in filters.

2. Cake filtration
Filtration mechanisms, as they apply to foundry practice, are described by Rushton [1] and a schematic is shown in Figure 1. This mechanism is known as cake filtration. The filtration process starts with solid particles floating in the fluid being trapped on the entrance surface of the filter (Figure 1A). This entrapment is assisted by the presence of low pressure and low velocity areas. At the beginning of the pour, the suspended particles do not affect the flow through the filter (Figure 1B). The separated particles progressively build up a layer known as the "cake" (Figure 1C).

Figure 1. Cake filtration principles

A further increase of the fraction of particles that are suspended will eventually lead to a threshold when the retained particles will start to have an impact on the flow through the filter. As more particles are separated a further, critical threshold will be passed as the cake start to influence the permeability of the filter. The growth of the cake finally leads to the blockage of the flow. A mathematical description of the velocity in a filter involving cake filtration is:

V = DP/m(R_m + R_C)                                                                  (1)
V   –    velocity
DP –    pressure drop
m    –    viscosity
R_m  –    resistance of the filter
R_C  –    resistance of the filter cake

Investigation of the metallic-ceramic interface at the entrance face of a filter is presented in Figure 2. The picture is a micrograph from a part of a gating system including a filter in ductile iron. The filter shown is a pressed ceramic, with a hole of 2.3 mm diameter. On this micrograph, we can observe the entrance part of one filter cell.

In the metallic matrix in contact with the entrance surface of the filter, it is possible to observe a bridge of microscopic inclusions. These are mainly composed of MgO and MgS. The distribution of the inclusions reveals a complex variety of separation processes. The inclusion cake gradually covers the filter channel, which eventually becomes completely blocked. The flow in contact with the entrance surface of the filter will be investigated in the next chapter to determine whether there are suitable conditions for cake formation.

Figure 2. Micrograph of an inclusion bridge

3. Simulation procedure
The simulation has been carried out using the MAGMAsoft simulation software on a gating system, commonly used in foundry practice, which incorporates a filter. Figure 3 shows the gating system considering a horizontally parted casting and a filter in a vertical position. The mould material is considered to be permeable green sand and the metal is considered to be a nodular cast iron. The filling condition was set as a time dependent volumetric function. The size of the gating system components is also commonly used in systems including filters, with an expanding channel area before the filter and a tapering area after the filter.

Figure 3. Simulated gating system	Figure 4. Pressed ceramic filter used in the simulation

Three different simulations were performed. One simulation considers that there is no filter in the geometry, a second simulation assumes that the filter as a permeable material causing a pressure drop through it, and the third simulation considers a real, pressed ceramic filter (Figure 4.) The simulation parameters are given in the appendix. The calculated filling time is equal in all the cases ( t_filling = 18.7 sec.), which is obviously due to the commonly imposed filling conditions. The fact that there are no differences in filling time between the cases with and without filters reveals the neutral contribution of the filter to the total filing time.

Figure 5 shows sequences from the simulated flow for the different filtering conditions.

Fraction filled	No filter	Filter as a permeable solid	Real shaped filter
2.39%
3%
5%
30%

Figure 5. Simulated flow sequences for different filtering conditions.

Comparing the case with the filter as a permeable solid and the case of a real shaped filter, it is evident that the flow at a macroscopic level is very similar; the flow velocities at different filled fractions are comparable. This observation underlines the correctness of modeling the filter as a permeable solid when only the thermal interaction between the metal and gating system is of interest.

Observing the flow at a higher magnitude, it is possible to see clear differences, especially in areas that are in contact with the filter element. In the case of the solid permeable filter (Figure 6), there is a homogenous flow distribution in the filter and the velocity decreases from the entrance surface of the filter to the exit surface, proportionally to the imposed pressure drop characteristics for this calculation method.

Figure 6. Velocity distribution in a permeable solid filter at 30 % fraction filled	Figure 7. Velocity distribution in real shaped filter at 30% fraction filled

In the case of a real shaped filter (Figure 7), the simulation predicts a distribution of the flow velocity according to the shape and size of the filter channels.

Figure 8. Velocity distribution in real shaped filter at 30% fraction filled	Figure 9. Velocity distribution in real shaped filter at 30% fraction filled

A perpendicular view of the same area (Figure 8) shows the specific velocity distribution in connection with the vertically placed filter. Light colours represent high velocities while dark colours represent low velocities. A low velocity area exists at the entrance side of the filter between the streams of higher velocity, indicated by arrows.

Figure 9 shows the distribution of low velocity areas up to a half millimeter from the entrance face of the filter, where the low velocity areas are interconnected in a complex way. Considering the fundamental theories of cake filtration, the presence of these low velocity areas helps to promote particle suspension.
According to this simulation, pressed filters exhibit conditions that promote particle suspension and which can then initiate cake filtration.

To underline the complexity of the phenomena exhibited at the entrance side of the filter, the distribution of low velocity areas, up to a half millimeter from the entrance surface of the filter, are presented at different fractions filled in Figure 10.

Figure 10a. Fraction filled 8%	Figure 10b. Fraction filled 25%
Figure 10c. Fraction filled 50%	Figure 10d. Fraction filled 75%

The results seen at different fractions filled close to the filter surface show a transition between interacting high velocity and low velocity areas. The low velocity area is expanding but also the high velocity areas are also changing. At this time there is no known method for calculating metal flow in combination with solid particles in suspension. However, the frames of Figure 10 indicate that a consideration of the particles in suspension, and thereby the influence of the cake on the flow in connection with the filters, would give a different flow pattern and promote the understanding of filtration mechanisms.

The flow with maximum velocity at the exit face of the filter is seen in an area where the streams are interacting and sudden changes between high and low velocity areas coexist. Suitable conditions for separation and cake formation exist at this filter exit and further studies will be performed on this specific point.

4. Conclusions
A real shape simulation on pressed foundry filters has been provided. While the traditional pressure drop method is adequate for investigating macroscopic flow phenomena in the molten metal within a runner system, this new method is able to show information on the fluid flow close to the filter itself. A gating system commonly used in foundry practice was simulated. Complex flow patterns were observed at the entrance and exit surfaces of the filter. High velocity areas adjacent to low velocity areas create sharp velocity gradients. The results are discussed in relation to general filtration mechanisms, where a cake filtration mechanism is the most credible mechanism appearing in foundry filtration. The simulation results clearly show the possibilities for particle separation due to the presence of low velocity areas in the vicinity of the entrance surface of the filter. The physical properties of real inclusions existing in foundry practice are not included in the calculation, although it is expected that these would change the flow characteristics of the metal adjacent to the filter.

5. Acknowledgements

The simulation work has been carried out in collaboration with Foundrysoft AB in Sweden.

6. References

1. A. Rushton, A. S. Ward, R. G. Holdich: Solid-Liquid Filtration and Separation Technology, 2nd Completely Revised Edition, 2001.

2. J.Campbell : Casting, 1991

3. Saint-Gobain Advanced Ceramics Hamilton, Molten Metal Filtration - An Engineered Balance - Internal report.

4. MAGMAsoftÒ 4.2, User manual, Part 2.

5. J.Bäckman, I.L.Svensson: Evaluation of Filter Parameters from Direct Observation of Metal Flow in Aluminium Casting, 1st Intl. Conf. Gating, Filling and Feeding of Aluminium Castings, Memphis, USA, 1999.

6. J.M.Dumaillet, G.M.Wilson: A Comparison of Flow Modification through Cellular Foundry Filters using both Water Modelling and Simulation Software, Part 1, 2002, Transactions of the American Foundry Society V 110 Paper No 02-106 P 187-198, 2002

7. J-C.Gebelin and M.Jolly: Numerical Modelling of Metal Flow Through Filters, Modelling of Casting, Welding and Advanced Solidification Processes X, 2003, Destin, USA, pp. 431-438.

Appendix I

Colour scale for the simulation results presented in Figures 5 to 10.

• The real shaped filter introduced in the gating system is a pressed ceramic filter with nominal dimension of 50x50x10 mm and cell diameter of f2.3 mm.

Back to top