1 Introduction

In powder bed additive manufacturing, the formation of thin layers is critical. The powder is melted in a selected pattern to build a component. If the spread layer is not of a consistent thickness or density, the build quality will be poor [1].

There are many established techniques to quantify flowability. Each has one or more metrics associated with it and some are standardised [2,3,4,5,6]. There are no established techniques or metrics to measure the spreading of powders (spreadability). The word “spread” does not even appear in the standardized terminology for additive manufacturing [7, 8]. Flowability is sometimes used to infer spreadability, but flowability “is not necessarily correlated to” spreadability [9].

Very simple, inexpensive equipment operated manually can give quantitative spreadability information [10]. However, in manual testing, the investigator can influence the results. Conversely, a custom-built spreading machine has been used to analyse powder via image analysis, video analysis, and laser profilometry [11, 12]. Other specialist machines are under development that can be retrofitted onto additive manufacturing systems [13]. One system uses a custom-built automated spreader to spread thin layers on a balance. This then allows the density of powder to be found [14].

In the current study, an automated measurement technique is proposed. Image analysis provides several metrics that can be combined to understand spreading behaviour. Further developments could also provide additional mass-based measurements to give information about packing within the layer.

2 Materials and Methods

Automated testing machines (“thin film applicators”) are already used in the paint and food industries. One such machine is the TQC Sheen AB4120 (Fig. 1). This machine has a maximum spread area of 43 cm × 30 cm, which is representative of many metal 3D-printers. The range of velocities of the spreader, known as the recoater speed, is between 0.1 and 500.0 mm s−1 to an accuracy of 1%. In early trials, silicon carbide paper was used to provide a rough surface for powder spreading and provide good colour contrast (Fig. 1a).

Fig. 1
figure 1

TQC Sheen AB4120 film applicator with a additional silicon carbide paper; b doctor blade with adjustable height, as used in the current study

Full size image

A doctor blade of adjustable height was used, which reduced the width of the test area to 25 cm (Fig. 1b). Tests were performed on a machined tool-steel plate, which is similar a metal 3D printer’s build plate (Fig. 2).

Fig. 2
figure 2

Camera and lighting setup for experiments in the current study. The LED arrays were turned on at full intensity with a light temperature of 5000 K; room lights were turned off but there were ambient light sources

Full size image

A tripod-mounted Canon 700D camera (18–55 mm lend, automatic mode) was pointed vertically at the geometric centre of the test plate and kept fixed throughout all experiments. The optimum position for lighting was either side at the height of the test area (Fig. 2). This illuminates the plate without forming reflexions in the camera.

Each image file was analysed using the image analysis program ImageJ [15]. A macro was written to automate the analysis:

  1. 1.

    Crop the image.

  2. 2.

    Split the image into three colour channels (red, green, blue)

  3. 3.

    Calculate the difference between the red and blue intensity of each pixel. The result is different for the powder and the steel plate.

  4. 4.

    Apply an automated threshold to classify pixels as powder or steel plate.

  5. 5.

    Measure the total area covered by powder.

  6. 6.

    Draw a convex hull (Fig. 3) around the powder and measure its area.

  7. 7.

    Draw a rectangular bounding box with one edge parallel to the direction of spreading (Fig. 3) around the powder and measure its area.

  8. 8.

    Divide the area measured in Step 5 by that found in Step 6 and, separately, Step 7.

Fig. 3
figure 3

a Custom-designed powder dispenser and b a powder sample dispensed using the funnel. Powder is poured through the top slot of the dispenser and after filling, the funnel dispenser is lifted off the sample

Full size image

Three powders were subjected to spreading and analysis using the procedure outlined above (Table 1). The recoater speed was varied between 50 and 200 mm s−1. The nominal layer height was 50, 70 or 100 µm. Powder was prepared for testing using a custom dispenser, produced using fused deposition modelling (Fig. 4).

TABLE 1 Powders tested. \(d\) is the particle diameter
Full size table
Fig. 4
figure 4

Example of powder outline (blue), convex hull (yellow) and bounding box (pink)

Full size image

Flowability metrics (Hall and Carney flow time and a stability and variable flow rate test in a Freeman FT4 powder rheometer) were also recorded.

3 Results

The image analysis procedure is capable of identifying powder and assigning a convex hull and bounding box (Fig. 5).

Fig. 5
figure 5

Example of image analysis: a photograph of spread layer; b result of subtracting the green colour channel from the red; c automatically identified outline of the powder; d outline overlaid with automatic convex hull; e automatic bounding box

Full size image

Powder C covers the greatest fraction of the area in which the powder could spread. Powder B covers the lowest area (Fig. 6). Similarly, Powder C fills the greatest fraction of a bounding box and convex hull and Powder B the least. For Powder C, the fraction of the available test area and the fraction of the bounding box filled by the powder increase with recoater speed for all layer thicknesses tested. There is a very small effect on the fraction of the bounding box that is covered by powder (Fig. 7a–c).

Fig. 6
figure 6

Coverage of powders indicated during spreadability tests for 50 µm thick layers

Full size image
Fig. 7
figure 7

Spreadability metrics as a function of recoater speed and layer thickness for a, b, c Powder C and d Powder B

Full size image

All three metrics decreased with increased recoater speed for powder B. Powder A had no strong relationship between recoater speed and the metrics. Powder C covered less area at increased recoated speed, like Powder B. However, Powder B filled less of the convex hull and bounding box with increased recoater speed (e.g. Fig. 7d).

Both Powders A and B fail to flow through either a Hall or Carney flow funnel, while Powder C flows very readily (Table 2). Powder C also exhibited the lowest “specific energy” during powder rheometry and Powder B the highest (Table 2). This is the most representative metric for the physical scenario of and agrees with the flowmeter results. In rotating particle analysis, that Powder C exhibited the lowest avalanche and rest angles, while Powder B had the highest.

TABLE 2 Hall, Carney and powder rheometry results for the powders tested
Full size table

4 Discussion

If it is accepted that a “spreadable” powder spreads evenly across its entire width, maintains a constant thickness that is the nominal height of the blade, and that the layer cannot be higher than the height of the blade, one can say that a “spreadable” powder will fill the lowest area as but the highest fraction of the convex hull and the bounding box. These final two results are true for Powder C, but it fills the greatest area. The high filling of the bounding box means that the length of the spread layer is even across the width of the layer and the high convex hull fill fraction implies that there are no protrusions in the spread layer. This is analogous to the use of circularity to characterise the shape of powder particles and other three-dimensional objects. The spread area can be thought of as analogous to the particle size distribution of a powder ensemble. Just as in characterisation of powders themselves, in which size and shape must be considered for a complete description, both the area of the layer and its form must be quantified to describe spreadability. This may be achieved with the proposed measurement and analysis procedure. Powder C was volatile and easily became airborne. It is conjectured that higher recoater speeds led to more powder becoming airborne and settling ahead of the spreading front, covering a larger area (Fig. 8).

Fig. 8
figure 8

50 µm layer spread at 50 mm s−1

Full size image

Powder B was more cohesive and became airborne less than Powder C. Only the height and density of the layer will affect results. A faster recoater speed is more likely to cause defects in the spread layer, which leads to the powder filling a greater area.

Flowability and spreadability may be correlated as the same physical processes occur in both flow and spreading [9]. The rheometry results correlate with spreadability, since the specific energy of Powder C is lowest, implying best flow. Powder B has the highest specific energy, implying worst flow (Table 2 cf. Fig. 6). However, more tests are needed as the specific energies of Powders A and C match within standard deviation. Similarly, the rotating powder analysis showed that Powder C had the lowest avalanche angle and the lowest rest angle, implying best flow, and Powder B had the highest angles.

5 Conclusions

A technique to quantify the spreading behaviour of powders using an image analysis has been developed. Three powders were compared under various conditions that are relevant to metal additive manufacturing. It was found that the powder with the highest flowability gave the highest fraction of bounding box and convex hull filled, but not the smallest total area coverage. This was attributed to the powder becoming airborne and redepositing ahead of the spread layer.

6 Future Work

A wider range of powders and conditions should be tested to assess the suitability of the method. The investigation should also be repeated with a non-magnetic base plate to test the effect of magnetic fields on the behaviour of the powder. The technique can be developed further to include mass-based measurements, such as calculating the mass of powder included in a spread layer of fixed area, which is more representative of powder bed additive manufacturing. Further image analysis could be added, such as the analysis of dark regions to reveal porosity or unevenness [10].