Oxides, porosity and fatigue performance of AlSi10Mg parts produced by selective laser melting

Highlights

• Fatigue properties of additively manufactured (AM) AlSi10Mg are similar to castings.

• Oxide-driven porosity is the most common internal defect in the AM parts.

• Pores and oxide particle size at the part surface control fatigue life.

Abstract

It is well-known that the fatigue behavior of cast aluminum alloy parts is largely determined by internal defects, particularly pores and inclusions. In this work, it is shown that such imperfections are also present in AlSi10Mg parts produced by selective laser melting, and serve as sites to initiate fatigue cracks. The effect of hatch spacing and building orientation on tensile and fatigue properties was tested. Oxide-driven pores dominate the fatigue resistance of the samples in this work. The larger oxide particles which are associated with crack initiation likely form by oxidation of metal vapor during part manufacture.

Introduction

Metal additive manufacturing has the advantages of flexible geometric design and less material lost [1], [2]. Full utilization of these advantages requires predictable mechanical properties. In the work presented here, the focus is on selective laser melting (SLM) of gas-atomized powder of the aluminum alloy AlSi10Mg (composition in Table 1) to produce parts with close to 100% density. AlSi10Mg is similar in composition to A360 aluminum casting alloy (9.5%Si, 0.5%Mg and Fe less than 1.5%) [3]. In conventional casting processing, A360 has advantages of high fluidity (a result of its near-eutectic composition), low shrinkage, good weldability and high corrosion resistance [3].

Previous researchers studied selective laser melting of various alloys, including TiAl6V4, Inconel 718, Type 316 Stainless steel and AlSi10Mg, fabricating metal parts and testing these for mechanical properties [4], [5]. Previous work on the microstructure and mechanical properties of parts built from AlSi10Mg focused on the effect of powder properties [6], [7], internal porosity of parts [8], [9], [10], crystallographic texture [11], tensile properties [12], [13], surface roughness [14], hardness [15] and creep behavior [13]. Mechanical properties of AlSi10Mg alloys reported in literature are summarized in Table 2, including AlSi10Mg parts produced by SLM and die-cast A360 parts.

There has been limited work on fatigue performance [17], [19]. Brandl et al. [17] reported that platform temperature, building direction and heat treatment (peak-hardening) all affected the fatigue life of AlSi10Mg parts produced by SLM, with heat treatment having the largest effect. Maskery et al. [19] also found that heat treatment (solution treatment, 1 h at 520 °C, followed by artificial aging for 6 h at 160 °C) significantly improved the ductility and fatigue performance of AlSi10Mg parts built by SLM (without subsequent machining), although heat treatment caused some loss of strength. Siddique et al. [25] reported that increasing the base temperature (to 200 °C) during selective laser melting (of a slightly different alloy, AlSi12) decreased residual stress, eliminated large pores and decreased scatter in fatigue results, but causing lower static strength and lower average fatigue strength.

The extensive work on the fatigue behavior of cast aluminum alloys also provides useful insights into the factors which can be expected to control fatigue of SLM parts. Most reported results are for the casting alloy A356, which contains about 7%Si [3]. The fatigue resistance of cast aluminum alloys was reported to be controlled by microstructural defects, typically the largest porosity [26], [27], [28], and sometimes oxide films [29]. Other microstructural factors which can affect fatigue life include secondary dendrite arm spacing (SDAS) or cellular spacing, grain size (including modification by Sr), second-phase particles, intermetallic inclusions, and oxide inclusions (casting dross) [3], [27], [30].

The work reported here studied the effect of porosity and building orientation on the tensile and fatigue behavior of AlSi10Mg parts produced by selective laser melting, testing the expected correlation between defects and fatigue. Building conditions were based on standard values recommended by the machine and powder supplier (see Table 3). Two experimental variables were tested: hatch spacing and sample orientation during building (see Table 4). The effect of building direction on fatigue life was examined by testing samples built along two directions: one with the long axis of cylinder samples aligned with the build direction (denoted by Z), and the other with the long axis of the samples horizontal (perpendicular to the build direction, denoted by XY) [31].

Overlap between melt pools is expected to affect defect formation; to study this, three hatch spacings were used (where the hatch spacing is the distance between subsequent laser scans while building the same layer): 0.16 mm, 0.19 mm (the default value as listed in Table 3) and 0.22 mm. For comparison, the melt pool width under the building conditions used here is approximately 0.22 mm and the melt pool depth is approximately 0.11 mm [32].

Hatch spacing is expected to affect both porosity due to incomplete melting and the elimination of defects upon repeated remelting of the part during subsequent laser passes.

Sufficient overlap of melt pools to avoid incomplete melting is obtained if the following geometric criterion holds [33]:

$${\left(\frac{H}{W}\right)}^2+{\left(\frac{L}{D}\right)}^2⩽1$$

where H is the hatch spacing, L is the layer thickness, W is the melt pool width, and D is the melt pool depth. Of the three hatch spacings used in this work (0.16, 0.19 and 0.22 mm), the two smaller spacings satisfied the criterion for full melting. While the largest hatch spacing did not satisfy the criterion, a geometric model [33] predicted that unmelted material would be only 0.02% of the part volume, and hence would not be expected to affect part density significantly.

Hatch spacing would also affect the number of times each volume in the part is melted (n_melt). Assuming that the melt pool cross-section can be approximated as two half-ellipses with width W and total depth D (as shown in Fig. 1), for each laser pass with an arbitrary length x, the volume deposited is HLx and the total melted volume is (πWD/4)x. The ratio of the melted volume to the deposited volume gives the average melting count:

$$n_{\text{melt}}=(\pi /4)(W/H)(D/L)$$

Based on this expression, the hatch spacings used in this work gave average melting counts of 4.0 (for H = 0.16 mm), 3.3 (for H = 0.19 mm) and 2.9 (for H = 0.22 mm).