You are on page 1of 4

A Measurement of the Porosity in Aluminum Cast Alloys Using Fractal Analysis

Y. J. Huang, S. Z. Lu Department of materials science and Engineering Michigan Technological University Houghton, MI 49931

Abstract
Porosity is a major defect in aluminum cast alloys. A measurement of the porosity of aluminum cast alloys using fractal analysis is carried out. This method uses two dimensionless parameters, Roughness, D, and Sphericity, , to describe the shape and the roundness of the pores. A Porosity Distribution Map can be constructed to describe the shapes and their distribution of the pores in the microstructure. This method may complement with the conventional quantitative examination for porosity. It is observed that the gaseous pores are described by low D and high values and the shrinkage pores are described by high D and low values. The tensile strengths of the samples are related to the shapes of the pores in microstructure. It is found that the higher the Sphericity, , is the higher the tensile strength.

including the total volume fraction of the pores, the pore sizes, the pore densities and the shapes of the pores.

Characteristics of Porosity
Porosity is a very common defect in aluminum castings. Fig. 1 shows a typical view of microstructure of aluminum silicon alloy with primary dendrites, eutectics and porosity. It is known that the quantity and the appearance of the porosity are very crucial to the mechanical properties of the casting, especially the fatigue property, because the pores in microscale are primary source of initial cracks for the final failure. [1-3]. There are two types of porosities in general, shrinkage and gaseous. Shrinkage porosity is caused by locally insufficient liquid feeding in the late stage of solidification. The contraction due to the volume change in liquid/solid transformation will leave microcavities as a result. This type of porosity is more likely to be seen in unmodified or not well modified Al-Si alloys. Hydrogen porosity is caused by the evolution of dissolved hydrogen in the melt when the solubility decreases upon solidification. Hydrogen porosity is more predominant in the Sr modified alloys. The pores of this type are usually less irregular with rather smooth outlines because they are generated from gas bubbles [4]. The properties of the aluminum silicon cast alloys are sensitive to the characteristics of the porosity

Fig. 1 A typical view of microstructure of aluminum silicon cast alloy with porosity (50X). Individual pores are numerically labeled. Conventionally, volume fraction of porosity, Vp, and average pore size, L, have been used in describing porosity in aluminum castings and the measurement can be carried out with a modern image analyzer [2]. However, the shapes of the pores are also very important to the mechanical property. A pore produces a stress concentration at its tip and the magnitude of the stress concentration depends on both the shape (rough or smooth) and size. If pores cannot be avoided, they should be small and preferably, round in shape to decrease the stress concentration. On the other hand, because the shrinkage pores appear to be more irregular (rough contours) and the gaseous pores have smooth shapes, the shrinkage

Huang and Lu, 1 of 4

pores should be more harmful to castings in general. Therefore, it is necessary and will be interesting to distinguish the two types of pores by a measurement of their shapes. Present paper is to describe a method of fractal analysis for such a purpose. This method can complement with the conventional image analysis in describing porosity in aluminum silicon castings.

Fractal Analysis
Fractal geometry was developed by Mandelbrot about two decades ago [5]. Its principle is universal in any measurement and it has been previously used to numerically describe complex microstructures including graphite flakes and nodules [6-8]. The basic equation is P = PED-1 (1< D < 2 and m < < M) (1)

corner of the Porosity Distribution Map represents the worst possible shape of a pore --- somewhat like a multiply branched flake. The larger the roughness, the more irregular a pore and thus more stress concentration. The pores located close to the lower right corner of the map are assumed to be less harmful than those located close to the upper left corner of the map. This map can be obtained by measuring perimeters, PE, areas, A, with changing yardsticks.

where PE is the measured perimeter, P is the true perimeter, is the yardstick and D is defined as the fractal dimension (1< D < 2). Plotting PE against in Log-Log scale the D value will be obtained from the slope [6-7]. This expression tells us that the true perimeter is actually a function of the yardstick for the measurement. The smaller the yardstick is the more accurate is the measurement, i.e., the measured values are more approaching the true values. It should be noted that there exists an upper limit, M, and lower limit, m, for any shape [6-7]. The upper limit, M, is equivalent to the size of an individual shape so that any measurement with > M will be meaningless. When < m, the measurement is not sensitive to the yardstick chosen. The fractal dimension, D, therefore describes the complexity of the contour of an object. The higher the D value is the more complex the structure. Therefore, it may be more practically meaningful to call the fractal dimension, D, Roughness. To examine whether a shape is close to a perfect sphere, another dimensionless number, Sphericity, , was introduced [3]. = 1 for a perfect sphere (with D =1). The is expressed as follows in Equation 2.

Fig. 2 Schematic representation of the Porosity Distribution Map

Experimental Procedure
The analyses were conducted on a Clemex image analyzer with an automatic stage and built-in software for fractal measurement. On the average, the pores are widely spaced, so magnification selection is important. If the magnification is too low, say, 25X, it is very hard to obtain fine details of pores, and the sphericity measurement would be inaccurate. On the other hand, if the magnification is too high, 200X for instance, where the resolution is satisfactory, one may observe one or two pores in a field of view, and some microstructural fields may not contain any pores. In this research, a compromise is made by using 50X magnification, where both the sampling problem and accuracy of measurement are satisfied to some extent. In the measurement, guard frames are used to eliminate the pores that are incomplete in a field of view. The guard frame defines the portion of an image that is included in the analysis. It is adjusted to include only the features that are completely within the image. This is done by adjusting the guard frame so that the object count points (OCP) of the features of interest are within the frame. The OCP is the top left pixel of each object. By defining the size of the guard frame, incomplete pores can be removed from being considered in an analysis. Samples from A319 and A356 aluminum-silicon castings, provided by GM Powertrain Group, were used to test the method. The samples are from fractured tensile bars, from which the mechanical properties have been measured. The area where the measurements are conducted on each sample included almost all the

= 4 A/P2
where P = PED-1, thus

(0 < <1 and 1< D < 2)

(2)

= (4 A/PE2)2(1-D)

(0 < <1 and 1< D < 2)

(3)

Where A and PE are measured area and perimeter of each individual pore. Similar to the Nodularity Map for ductile irons, a Porosity Distribution Map can be constructed to describe various shapes of pores using these two dimensionless numbers of Roughness, D, and Sphericity, . In the present case, as schematically presented in Figure 2, a location on the porosity map represents a certain pore. The location of = 1 and D = 1 represents a perfect circle. The D = 1 axis represents regular (Euclidean) shapes (eg. smooth ellips etc.). As is decreases along this axis the shapes are more elongated showing a departure from a perfect sphere. The locations of 1 < D < 2 represent less regular shapes. The higher is the D value the more complex or rough are the outlines of the shapes. The upper left

Huang and Lu, 2 of 4

polished surface and the measured areas are composed of at least 30 fields of view at 50x magnification. A software for fractal analysis was also developed at Michigan Tech [9]. This software allows the analysis to be performed on any PC (Windows) without the need for modern image analysis system.

for the two groups of samples from A356 and A319 castings. Each data point in this diagram represents a sample (not an individual pore in this case). It is very obvious that of the two groups of samples, the A356 group has more gaseous pores that are relatively small and regular, while the A319 group has more shrinkage pores that are more irregular.
1.35

Result and Analysis


A typical result of such a fractal analysis for the labeled pores in the view field shown in Fig. 1 is listed in Table 1. As can be seen, the pores #7 and # 8 have the worst shapes and they are represented by low values of and relatively high values of D.
Roughness
IIndividual pores
1.3

Weighted Average

1.25

D = 1.19 = 0.11
1.2

Table 1 Measured Result for the view Field in Fig. 1 No. 1 2 3 4 5 6 7 8 9 Fractal Dimension, D 1.11 1.01 1.16 1.13 1.03 1.11 1.11 1.12 1.08 Sphericity, 0.62 0.27 0.57 0.59 0.99 0.64 0.20 0.20 0.79

D = 1.03 = 0.97

1.15

1.1

1.05

1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Sphericity

Fig. 3 Porosity Distribution Map for sample 2-24 of A319 alloy. Each data point represents an individual pore in the sample.

1.15 1.14 1.13

As another example, a Porosity Distribution Map for the sample 2-24 is presented in Figure 3. Each point in the map corresponds to an individual pore that is described by two numbers, Sphericity, , and Roughness, D. The points towards lower-right corner represent smoother and more spherical shapes and locations towards the left-upper corner represent the undesired, more irregular pores. In this map, pore A is one of the worst shapes with the measured values of = 0.11 and D = 1.19, while pore B is almost a perfect sphere with minimal roughness ( = 0.97 and D = 1.03). Since shrinkage pores are usually larger and of more irregular than gaseous pores, they usually occupy the left-upper position. Our study indicates that there exists a critical value of the sphericity, by which the two types of pores can be separated and this value, from our limited measurements, seems to be = 0.3. The pores with < 0.3 are normally shrinkage pores and the pores with > 0.3 are normally gaseous pores. With this criterion, it will be very easy to calculate the percentage of such different porosities. For example, we may find that 70% of pores in a sample are gaseous type and the rest are the shrinkage pores. In the example above, the weighted average values are = 0.29 and D = 1.11. These values are used as a representative for this particular sample. It can be seen that the shrinkage pores are probably more dominated ones in this case. Figure 4 is a summary of all the measurements containing all of the weighted average values

1.12 Roughness 1.11 1.1 1.09 1.08 1.07 1.06 1.05 0.2 0.3 0.4 0.5 Sphericity 0.6

356

319

0.7

0.8

Fig. 4 Porosity Distribution Map for all the measurements of weighted average values for the two groups of samples from A356 and A319 castings. Each data point represents a sample. The weighted average values are used to relate with the test results of mechanical properties for all the samples. Fig. 5 shows the relationship between the weighted values of sphericity and the tensile strength (TS) for all the samples measured. Again, each data point here represents a sample. It appears that the samples from A319 castings have lower TS and low sphericity in general and the

Huang and Lu, 3 of 4

samples from A356 castings have higher TS and higher sphericity. There is a trend that the higher the sphericities are the higher the TS.
3.50E+08

3.00E+08 TS (Pa)

2.50E+08

2.00E+08

356
1.50E+08 0.2 0.3 0.4 0.5 0.6

319

4. B. T. Lee, S. Z. Lu and A. Hellawell, Solidification Processing97, Sheffield, UK (1997) pp. 243-247 5. B. B. Mandelbrot, The Fractal Geometry of Nature, W. H. Freeman and Co. New York (1982) 6. S. Z. Lu and A. Hellawell, An Application of Fractal Geometry to Complex Microstructures: Numerical Characterization of Graphite in Cast Irons, Acta Metall., 42 (1994) pp 4035-4047 7. S. Z. Lu and A. Hellawell, Using Fractal Analysis to Describe Irregular Microstructures, pp. 14-17, JOM, December 1995 8. S. Z. Lu and A. Hellawell, A Fractal Method of Nodularity Measurement in Ductile Iron, AFS Transaction, Vol. 107 (1999) No. 99-195, pp757-762 9. Y. F. Zhang, S. Z. Lu, MicroFractal Verion 1.0, Windows Software (Win98/95) Free to download from http://www.mm.mtu.edu/~szlu/fractal/Fractal1.html (1999)

0.7

Weighted Sphericity

Fig. 5 Relationship between weighted average sphericities and the tensile strengths for all the samples from the two groups of A319 and A356 castings.

Conclusion
1. Fractal analysis can be applied to the porosity measurement to describe the shapes of the pores in aluminum silicon cast alloys using two dimensionless parameters, Roughness, D and Sphericity, . This method may complement with the conventional quantitative examination for porosity. 2. A Porosity Distribution Map can be constructed to describe the shapes and their distribution of the pores in the microstructure. 3. It is observed that the gaseous pores are described by low D and high values and the shrinkage pores are described by high D and low values. 4. The tensile strengths of the samples are related to the shapes of the pores in microstructure. The higher the Sphericity, , is the higher the tensile strength.

Acknowledgment
This project is supported by the Metallurgy Program of the Division of Materials Research (DMR9520173) at the National Science Foundation in conjunction with GM Powertrain Group, Ford Motor Company. The authors would like also to express their many thanks to Norman Carter and at GM Powertrain and James Boileau at Ford Motor Company for their valuable suggestions and discussions, and for providing the test samples.

References
1. N. Roy, A.M. Samuel and F.H. Samuel, Met. and Mater. Trans. 27A (1996) pp. 415-429 2. N.Roy, L. Zhang, P.R.Louchez and F.H. Samuel, J. Materials Science 31 (1996) pp. 1243-1254 3. S. Viswanathan, A. J. Duncan, AFS Transaction (1998) pp. 411417

Huang and Lu, 4 of 4