Removal of staircase artefacts using the binder jetting process

Almost all three-dimensional printed components have noticeable discrete steps that are easily recognisable on certain surfaces. These steps are the result of the layer-by-layer build-up process during 3D printing. If the surfaces of the component to be produced have their normal direction close to parallel to the build-up direction, this leads to visible steps. Their appearance depends on the selected layer thickness and they are particularly pronounced at inclinations of up to 20° (see Figure 1).

The layer thickness plays a decisive role in the cost-effectiveness of the 3D printing process. A reduction leads to an increase in the layers required for the same build height, which results in longer process times and therefore higher costs. In addition, the commonly used sands with an average grain size of around 140 µm do not allow an unlimited reduction in layer thickness. Eliminating the steps by using thinner layers is therefore both economically and technically impractical. Steps are often neglected in roughness measurements. If these were taken into account, stepped components would have a considerably increased roughness. 

If conventional roughness parameters do not adequately consider the steps, 3D-printed components often cannot fulfil customers' specific requirements. The influence on possible notch effects in moulded components has not yet been extensively researched, but seems plausible. Steps pose a significant challenge, especially for small components, high precision requirements and components for flow applications.

The accuracy can be controlled by specific modifications to the binder content. This relationship is now to be integrated into the print data on a microscopic level in order to increase or decrease the binder content locally and thus prevent the formation of steps. This involves manually adjusting the print data in image processing software, followed by iterative tests of various modelling approaches. The observation shows that by applying overlapping, almost continuous curves, the steps can be equalised regardless of the slope. The result of this improved process is shown in Figure 2.

The results of the data manipulation presented show that the steps can be resolved using suitable gradients. Optical 3D scans of the surfaces revealed that the remaining deviation from the 3D model lies within a range smaller than a medium-sized grain of sand. The change in component colour is remarkable: in conventional printing, the component has a uniform dark green colour, whereas, in smoothed printing, darker areas can be seen in the former step fillet. This is due to the increased concentration of the binder in these areas.

With regard to the casting technique, the question arises as to whether the increased binder content leads to an increase in gas porosity. Micrographs of castings showed no signs of an increased gas tendency. The resulting gas porosity appears to be unaffected by the local change in binder content. One explanation for this could be that a local increase in the binder content, as shown above, also binds additional sand to the mould and therefore only marginally changes the actual binder content.

Typically, real components that can benefit from the improvements shown have a demanding geometric complexity. Manual data manipulation, especially considering the usual free-form surfaces, would not be practicable, not least because of the considerable time required. For this reason, a method was developed based on established slicing algorithms, making it possible to automatically derive the modified print data from a 3D model. The combination of sand and binder, as well as the desired layer thickness and the existing print head, are taken into account as input parameters in the program.

The smoothed surfaces are already clearly visible in the sand moulds (Figure 3). Using alternative data preparation, half of the mould with conventional steps and half of the mould with smoothed surfaces can be produced on the test system while maintaining the same shift time. The resulting cast part »Impeller«, built on an S-Max system from the manufacturer ExOne, illustrates the industrial possibilities of this process (Figure 4). In the case shown, one half of the mould was constructed conventionally, while the other half was created using the processed data. The two-part mould can be used to compare the change in surface quality.

The project presents various methods to improve the added value of 3D-printed moulds and cores. With an optimised surface, demanding customer requirements can be met not only in terms of surface quality, but also in terms of testability to higher standards. It also opens up new perspectives in technology. For example, the improved surface quality in 3D printing enables further development of design freedom in additive processes. Complex, functional surfaces, such as those that occur at the trailing edge of flow surfaces, can be provided with additional vortex generators to allow the flow to break off in a targeted manner and thus increase efficiency, similar to golf balls.

In addition to the elimination of steps, this project also gained insights into the causes of dimensional deviations. Based on this knowledge, dimensional deviations in the entire build volume can be compensated for in the future using data-driven compensation, resulting in increased accuracy of the printed parts. The introduction of new printing strategies has also laid the foundation for improving edge definition.

The tests in this project were carried out with a print head that is grey-scale capable. The 3D printers currently available are almost exclusively equipped with print heads that are black and white-capable and use a different resolution and different print modules. The adaptation of the data and the consideration of the hardware of the available devices on the market were carried out by automating the process, taking into account the black-and-white print modules.