by William Hahne, graduated master student from Uppsala University

For 20 weeks, I have been working with AMEXCI for my master thesis and as the project is wrapping up, I can safely say that it has been an exciting period, full of new experiences. The purpose of the master thesis was essentially to develop process parameters that could yield higher build rates and low amounts of pores. In the end, three samples were chosen as potential candidates for this purpose, corresponding to an increase in build rate between 30% and 60% and relative densities between 99,99% and 99.91%.

Project Background – why? -how?

Laser powder bed fusion (L-PBF) is an additive manufacturing system of interest because of its many degrees of freedom and is a good compromise with regard to speed, accuracy, and flexibility, compared to other additive manufacturing technologies. In order to improve the competitiveness of L-PBF versus traditional manufacturing, however, L-PBF must become faster and also cheaper. One way to increase the productivity and improve the cost and time competitiveness is to increase the layer thickness, which was the aim for this master thesis.

In the L-PBF process, a metal build plate is covered by a thin layer of powder, which is selectively melted by a laser and recoated by a subsequent powder layer. The process is repeated, and the component is manufactured layer by layer from the ground up. Once finished, the component is cut from the build plate. A typical laser powder bed fusion set up can be seen below.

Figure 1: A simplified schematic of the laser powder bed fusion set up.

The L-PBF process is dependent on a large number of process parameters. To achieve results within the project timeline, only some of the most critical parameters were included in the scope of the project, namely laser power, scanning speed, hatch spacing, and powder layer thickness. Laser power is the energy input from the laser, scanning speed is the velocity of the laser during printing and hatch spacing is the distance between the laser scan tracks.

As the powder layer thickness is increased, so is the build rate and thus also the productivity. In this project, powder layer thicknesses of 80 µm and 100 µm were investigated for the sake of increasing the productivity, compared to the 40 µm or 60 µm layers used traditionally. To keep the loss of the sample’s mechanical properties to a minimum, the relative density was studied along with productivity. This was of particular relevance since the risk for porosity typically increases as the build rate is increased. Therefore, the process parameters need to be optimized in order to achieve a balance between build rate and porosity. Unfortunately, in laser powder bed fusion, there are multiple pore-producing mechanisms. Like for example, gas pores, lack of fusion pores, and spatter related pores:  

• Gas pores can come from either gas entrapment or from preexisting gas in the powder feedstock. Gas entrapment occur since the cooling rate in laser powder bed fusion is very high and preexisting gas can come from the powder manufacturing process itself. Because of this, gas pores are often inevitable. Fortunately, they do not contribute much to the overall porosity value due to their small size.

• Lack of fusion pores can be defined as powder particles that have not been fully melted or incorporated into the melt pool. These types of pores can generally be recognized as large and irregular in shape.

• Spatter is a typical mechanism occurring during laser-powder interactions. It can be described as the ejection of particles close to or directly form the laser spot and can either come from unstable melt pool formations or vapor jets. To transport spatter away from the build plate, an inert gas flow is implemented in the build chamber. If the spatter is too excessive however, spatter might land back on the powder bed and causing the laser to interact with them instead of the powder.

The first two sample batches were printed with a powder layer thickness of 80 µm, and the other three with a powder layer thickness of 100 µm. Each sample batch contained around 24 samples, resulting in 119 samples in total. Every DoE was constructed based on results obtained in previous iterations, based on both the samples build rates and their level of porosity.

Since a powder layer thickness of 80 µm has been investigated at AMEXCI before, the relative densities achieved was quite high. Most pores seen were characterized as gas pores, but spatter-related pores also played a dominant role. Considering the comparatively large amount of spatter generated in the higher layer thicknesses, this was an expected result. Less expected was the relatively low levels of spatter-related porosity found in the 100 µm batches where no correlation like this could be drawn. The sample that achieved the best balance between porosity count and build rate for these prints, was sample 48, with increase in productivity of 30% with a relative density of 99,99%.

In the first 100 µm print, more pores were present in the samples printed with the highest scanning speed and hatch spacing, indicating an under-melting effect. This was confirmed by excessive lack of fusion pores observed in multiple samples. This resulted in a much lower relative density value in comparison to the previous sample batches. Despite this, some samples obtained a relatively high relative density. The printed batch 1 samples can be seen in Figure 2 where the volume of the spheres correspond to the level of porosity.

Figure 2: 100-micron batch 1 samples (blue), positioned according to their process parameters. The respective porosity can be correlated to the sphere volume

For the second 100-micron print, the scanning speed was decreased in order to minimize the under-melting effect. At the same time, the hatch spacing was increased in order to increase the build rate. Since no correlation could be drawn between laser power and porosity at these high laser power levels previously, the power values were kept the same. In this batch, pores seemed more common for the samples printed with the lowest scanning speed, indicating an over-melting effect. Some samples obtained pores due to lack of melt pool overlap because of their high hatch spacing and low energy input. The melt pools were also electrolytically etched in order to reveal their melt pools. The observed melt pools extended up to 400 µm, or the equivalent of 4 powder layers. Spatter that landed on the samples could therefore have gotten remelted up to 4 times. This is a likely reason why spatter generally didn’t pose a dominant problem for the 100 µm prints. The printed samples can be seen in Figure 3.

Figure 3: 100 µm batch 2 samples (orange), positioned according to their process parameters. The respective porosity can be correlated to the sphere volume.

For the third and last sample batch, the scanning speed was increased slightly in order to minimize over-melting and the hatch spacing was decreased slightly in order to minimize lack of melt pool overlap. Even though no correlation could be drawn between porosity and laser power, the laser power was also increased slightly. For this sample batch, pores were slightly more present for the highest hatch spacing because of lack of melt pool overlap, and some spatter influence was seen. The batch 3 samples can be seen in Figure 4.

Figure 4: 100 µm batch 3 samples (green), positioned according to their process parameters. The respective porosity can be correlated to the sphere volume.

In the end, three samples were singled out as promising candidates. These samples can be seen in the Table 1 below.

Sample #Powder layer thickness [μm]Increase in build rate (%)Mean porosity (%)
48803099,99
601005099,94
1121006099,91
Table 1: The three samples considered to have the best build rate – mean porosity balance of all investigated samples.

The master thesis can be found in its entirety on the DiVa portal, titled “Optimization of laser powder bed fusion process parameters for 316L stainless steel” .