Experiment shows filters reduce melt turbulence defects.

An experiment with four different gating systems demonstrates the importance of filtration in eliminating defects caused by molten metal turbulence in Al-Si-Mg alloys.

Controlling the velocity of molten metal during the initial period of mold filling while the runner system is being primed eliminates the entrapment of oxide-related defects. In an experiment performed at the Univ. of Birmingham, Edgbaston, England, in 1995, four different gating systems were developed to test the use of filters to control velocity and metal turbulence.

The gating systems - named and described as unfiltered untapered (UFUT), unfiltered tapered runner (UFT), horizontal filter (HF) and vertical filter (VF), as shown in Fig. 1 - were formed in resin-bonded silica sand molds with sand and runner system thicknesses kept to a minimum of 1.15 in. and 0.59 in., respectively. A 98% pure aluminum-silicon-magnesium (Al-7Si-Mg) alloy was poured at 1274-1328F (690-720C) with its fill times and other measurements taken via video images, and the degree of turbulence assessed visually.

For each filling system, 30 test bars were cast, 10 each in three molds. After casting, all specimens were heat-treated to near-peak tensile strength, solution-treated for 6 hr at 1004F (540C) and water-quenched prior to a precipitation treatment for 5 hr at 320F (160C). Some test bars were sectioned and polished for metallographic examination, with the remainder machined to produce round tensile test specimens conforming to British Standard BS 10002. Following testing, fracture surfaces were selected at random and examined by scanning electron microscopy (SEM) and energy dispersed X-ray (EDX) microanalysis.

Gating System Fill

The four runner system designs were examined over a range of sprue heights (2.95, 6.3, 11.02, 16.54, 23.03 in.) to create a range of metal velocities in the runner. For the highest sprue heights, horizontal sections were taken at 0.12 in. below the top surface of the metal plates, and the degree of air entrapment was assessed qualitatively by the degree of subsurface porosity [ILLUSTRATION FOR FIGURE 2 OMITTED]. Close inspection of the subsurface porosity of the UFUT castings showed a considerable number of sand grains. However, sand was not found in the UFT castings, despite the presence of subsurface porosity.

In the UFUT castings, turbulent filling at high velocities produced a shrinkage pattern in the casting - forming as cracks throughout the plate. This effect also occurred, to a lesser degree, in the UFT castings, but was absent from the HF and VF castings. This shrinkage pattern also didn't occur in the castings with lower sprue heights.

The filling of the four runner systems was observed using real-time radiography techniques. The velocities calculated from the video images for these castings were used to predict the velocities in the test bar castings as shown in Fig. 3. The UFUT and UFT designs showed a considerable amount of turbulence in the runner and ingate regions. With the use of a filter in the HF and VF designs, the runners filled smoothly and without surface turbulence. However, the HF design allowed a small amount of surface turbulence to develop as the runner failed to begin filling immediately with the first metal to arrive. The flow through the ingate was nonturbulent and controlled by the complete filling of this region of the runner system.

In terms of filling times, all four designs produced similar results. The tapered runners took slightly longer to fill, perhaps because the filling was affected by the surface tension of the aluminum in the narrow region of the runner. Filters in the runner systems also increased filling times.

Optical metallographic examination of the UFUT test bars revealed a number of long oxide-type defects throughout the test bars [ILLUSTRATION FOR FIGURE 4 OMITTED]. The fracture surfaces contained large areas of inclusions on a considerable number of test bars. Examination using SEM techniques revealed similarly large areas of oxides, some with a distinct "collapsed sack" appearance. The UFT castings had oxide inclusions that were more closely tangled on the micrographs [ILLUSTRATION FOR FIGURE 4 OMITTED] as the fracture surfaces showed film on the surfaces.

The HF and the VF castings were free from long lengths of oxide films, with relatively clean fracture surfaces. However, closer examination revealed that what originally appeared to be finely dispersed shrinkage porosity was actually a fine dispersion of oxides associated with fine porosity [ILLUSTRATION FOR FIGURE 5 OMITTED]. Small amounts of oxides - related to shrinkage porosity - were found on the fracture surfaces under SEM examination [ILLUSTRATION FOR FIGURE 6 OMITTED].

Tapered Runner Designs

In an attempt to reduce the velocity of the metal flow, a diffused (expanding) runner was used. The rate of expansion corresponded to an angle of 5 [degrees]. However, separation of the metal flow from the runner walls occurred, resulting in no significant reduction in the metal flow velocity. The tapered design of the runner system helped reduce the degree of air entrapment within the runner, particularly the back-filling that occurred between the ingate and the sprue exit.

SYSTEM DESIGN ANALYSIS

The back-filling of the runner plays an important role in the casting strengths. Experiments using high velocities created by taller sprues show that this area of a runner system will act as a major source of air entrapment in the metal. The film defects trapped in the metal flow have an important role in the initiation of serious solidification defects. Shrinkage defects were clearly visible in the high-velocity UFUT cast plates, while less prominent in the UFT high-velocity castings. Quantification of the degree of air entrapment was affected by machining away the top 0.12 in. of the cast plates. However, there was no evidence of any trapped air in the filtered castings.

Mold Erosion & Sand Inclusions

Close examination of the porosity in the UFUT casting revealed sand grains. These grains are related to the oxide films trapped during the filling process. No sand grains were found in any of the other castings. These observations are consistent with the theory that mold erosion occurs as a natural consequence of turbulent filling systems. In calmly filled molds, the metal rolls over loose sand grains, and the surface film on the metal pins the sand in place. The mold walls are supported by the pressure of the metal and mechanically separated from any bulk turbulence by the surface film on the liquid. In turbulently-filled molds, the hammering of the surface by collapsing bubbles oxidizes and burns away the binder, mechanically dislodging the grains of sand with lost or weakened bonds.

Despite this theory, the transport of sand from a turbulent runner system into the mold cavity is unlikely because the minute grain area presents such a small cross section that any drag force on the sand is minimal. However, if the grains were attached to the surface films, then the force on the grains is magnified by a factor of 100 or 1000. Thus, the sand grains have been removed from the mold walls, possibly plucked off by the surface film, then wrapped up within the tangled network of films and transported through the casting with the oxide films acting as sails. Mold erosion has, therefore, less to do with mold strength, but it is strongly dependent on the degree of surface turbulence during filling.

Casting Defects

The source of failure for each test bar could not be identified. Inclusions discovered on the surface of the UFUT and the UFT castings regularly covered large areas of the fracture surfaces. These were identified as oxide film defects, using SEM and microanalysis techniques. Most of these films appeared as new films (1 micron thick) probably created during the filling of the casting. A few films were identified as older films that had probably come through with the metal from the crucible.

Oxides could be found throughout most of the polished metallographic sections. In the case of the UFUT design of the runner system, longer films were found relating to the trapped porosity. This confirmed the dry-side-to-dry-side mechanism of film entrainment. The films discovered in the UFT castings were tangled films. Close examination of the filling process revealed a number of tight rolling vortices that were trapping air, that may have created such defects. Furthermore, the high-flow velocity and resultant high Reynolds number (105) produced greater bulk turbulence to further tangle the film after entrapment.

Action of Filters

To investigate whether the filter was working effectively to hold back oxides, filters from castings HF and VF were sectioned longitudinally and examined metallographically. No expected buildup of films was found at the front or in the interior of the filter. Only isolated films were detected in the metal, inside the filter. This contrasts with sections through filters published by filter manufacturers. It seems that these published results were probably obtained using aluminum melts that were far dirtier than would be found in normal foundry practice.

Since oxides were not being held back by the filter, the question arose: Where were the oxides that would have been formed during the filling process?

For those castings with a filter in the runner, the polished sections exhibited finely dispersed microporosity. On closer examination, this porosity was seen as small lengths of oxide [ILLUSTRATION FOR FIGURE 5 OMITTED]. SEM techniques revealed smaller areas of films associated with pores exhibiting dendrites [ILLUSTRATION FOR FIGURE 6 OMITTED]. These observations highly suggest that all these pores were originally air bubbles, trapped as a result of the dry-side-to-dry-side entrapment mechanism. In addition, in some instances they have grown a little by subsequent precipitation of hydrogen or by a small contribution of shrinkage. Probably, both hydrogen gas and shrinkage contributions to growth resulted in the observed dendritic morphology of the interior of some pores [ILLUSTRATION FOR FIGURE 6 OMITTED].

The filter allowed a significant amount of small lengths of oxide into the casting. These may have been torn from films that had partially adhered to the filter surface or may have been films that were simply broken into smaller fragments as the metal was forced through the filter. Such a film-chopping mechanism suggests that a runner system design that has significant turbulent filling prior to the filter is more likely to have a fine dispersion of film fragments to aid the formation of fine microporosity. Alternatively, the filter may have created such films in its filtering action, which naturally involves the repeated separation and recombining of flowing streams on a microscale. More research is needed to elucidate the filtering mechanism.

Despite the fact that the filters did not filter out and retain all the oxide, they remained highly efficient at reducing the velocity of the metal flow and, therefore, controlling the degree of surface turbulence during the filling process. The importance of the flow downstream of the filter is emphasized by the HF runner, which did not fill completely, allowing a free surface to remain, which later backfilled. This may have impaired its action. Although the VF design filled turbulently before the filter, the runner after the filter filled immediately and completely, with a low metal front velocity.

Tapered Runner Design

As the runner size was reduced, the metal velocity through the ingate became high, leading to turbulently filled castings and defect entrapment. In an attempt to overcome this problem, the later areas of the runner were tapered at an angle of 5 [degrees] in an attempt to expand the metal stream while maintaining fluid attachment, thus reducing the metal velocity both in the runner and through the ingate.

The flow of metal through the expanding section of the runner did not follow the shape of the runner and had little effect on the metal velocity. (The design of the runner was based on fluid principles developed on the theory of a full system, not one that had to be primed; thus, separation was likely to occur.) However, the design of the runner reduced the degree of film and air entrapment created during the back-filling at the front end of the runner.

With velocities in the runner greater than 1.5 m/sec, the flow through the ingate became turbulent. This occurred irrespective of the ingate size because of the dominant momentum effect that forced the flow up the side of the ingate farthest from the sprue. The degree of air entrapment increased as the velocity increased. Further investigations still must be made to develop an improved approach to the design of runner systems.