Characterization of microshrinkage casting defects of Al–Si alloys by X-ray computed tomography and metallography

Abstract

The statistical pore size characterization by metallography in the framework of Extreme Value Statistics (EVS) is presented and applied to different sets of cast AlSi7Mg specimens. Specimen production by separate casting or by extraction from automotive cast parts is found to result in different SDAS and porosity (i.e. pore morphology and size) but did not influence the fatigue strength. The application of two equivalent pore size definitions (i.e. maximum Feret diameter and (Area)^1/2) combined with the EVS approach is discussed in terms of predicted critical pore sizes and observed fatigue strengths. The role of casting pore morphology on stress concentration is investigated using the X-ray computed tomography and the finite element method.

Introduction

Cast Al–Si alloys are widely used in fatigue critical structural applications, such as engine blocks, cylinder heads, and chassis and suspension components, for their excellent combination of mechanical and technological properties and to improve automotive fuel economy [1].

Fatigue properties of aluminum castings are strongly dependent on the casting defects and little affected by chemical composition, heat treatment, or solidification time, as reflected by dendrite arm spacing and the shape and size of eutectic silicon and intermetallic phases [2], [3], [4], [5], [6], [7], [8], [9], [10]. Typical defects of casting are macro pores and micro pores and bifilms [5]. While macropores (i.e. larger than a few mm) can be identified by X-ray inspection during quality check, micropores and bifilms are invisible to this kind of inspection. Since the presence of casting defects and discontinuities is almost inevitable in cast aluminum alloys and aluminum alloys have no apparent fatigue endurance limit, a defect-tolerant approach to fatigue design should be used, as proposed in [7], [8], [9], [10], [11]. Such an approach is based on the crack propagation life estimated from the crack growth rate law of the material and the initial crack size estimate. Typically, empirical models based on traditional linear elastic fracture mechanics (i.e. long crack behavior) are used although the short crack behavior has been also considered in [11].

The accuracy of the life prediction strongly depends on the prior knowledge of the defect population for a given material. The size of the largest defect has been recognized as the most important parameter in determining the fatigue properties of aluminum castings. The larger the maximum defect size, the lower the fatigue strength. Therefore, any defect tolerant design approach for materials containing defects should be based on a method to estimate the largest defect size distribution. In this context the approach and procedure developed by Murakami and coworkers [11], [12], represents a fundamental starting point. In short, metallographic inspection of a selected material cross-section and determination of the largest pore size in many fields of view allows constructing a statistical description of the largest pore size using Gumbel’s extreme value distribution. Such a distribution is then used to estimate the largest pore size in realistic part cross-sections by extrapolation [12], [13]. The procedure has been applied by many researchers to a wide range of materials with appreciable success [11], [12], [13], [14]. The Extreme Value Statistics (EVS) approach is based on two key assumptions: (i) the distribution of defects is uniform in the material, (ii) the pore size is well described by the parameter (Area)^1/2 where Area is the area of the largest pore measured on a metallographic section.

However, the combination of complex part geometry and solidification process typical of Al–Si alloys is known to give different pore types and sizes in different sections of the casting [11]. The morphology of casting pores in Al–Si alloys is typically classified either as gas pore or microshrinkage pore [1]. While the former is typically rounded and spheroidal, the latter is branched and elongated. In practice, the distinction is not always straightforward.

The application of metallography (i.e. essentially a 2-D analysis) to pore sizing is expected to give a biased response. In recent years, Buffiére and coworkers pioneered the application of high resolution synchrotron X-ray computed tomography (x-CT) to pore characterization in cast Al–Si alloys and graphite cast irons [6], [15], [16]. The application of x-CT allows the accurate non-destructive 3-D reconstruction of pores within a volume of aluminum alloy (i.e. distribution, size and morphology). In [15] x-CT was used to study the early stages of fatigue crack nucleation and growth from pores. While most of the studied cracks stopped after their length reached the initiating defect size, microcracks initiated on the largest defects or defect clusters continued to grow with short crack behavior. In this last case, crack shape became semielliptical. A substantial part of the lifetime (25% of the total estimated fatigue life) was necessary for the crack to surround the pore and assume a semielliptical shape. The usefulness of x-CT when applied to cast aluminum parts was also discussed in [17].

In this paper the statistical pore size characterization by metallography is initially examined adopting the EVS. The prediction methodology of maximum defect size based on two-dimensional (2-D) metallographic data is applied to individual fatigue specimens produced by direct casting of AlSi7Mg (equivalent to A356) alloy while other specimens were extracted from a complex automotive cast part. Questions emerging from this first approach motivate the investigation of an alternative pore characterization technique. Therefore, x-CT is used to reconstruct the 3-D distribution of casting pores in AlSi7Mg. Shrinkage pores, which are highly irregular in shape, are constructed and assessed in terms of stress concentration using finite element modeling.