Determination of geometric accuracy and surface roughness of small parts of circular and square sections, obtained depending on the printer location in the working space using selective laser melting technology from steel grade 12KH18N10T

The ability of modern companies to respond to changing customer needs as fast as possible is the key to success. One of possible solutions is large-scale implementation of digital technologies into design and production processes. Computer-aided technologies are widely used in design. The production process also involves state-of-the-art layer-by-layer synthesis
technologies or additive manufacturing, which allow to make a model, a prototype or even a functional part as soon as possible. Nowadays the following technologies are mostly used:

• SLA (StereoLithography) technology, raw material – photopolymer;
• FDM (Fused Deposition Modeling) technology, raw material – thermoplastic polymer;
• SLS (Selective Laser Sintering) technology, raw material – powder-like, in most cases, polymeric compound;
• SLM (Selective Laser Melting) technology, raw material – metal powder.

With large scale implementation of 3D printing technologies into the production process at both design and parts manufacturing stages, we can drastically reduce the time needed for new product development, preproduction and manufacturing, labour intensity, and material consumption. Most of the additive technologies (SLA, SLS, FDM) are used for manufacturing
models and prototypes in order to determine the assemblability and performance characteristics and to accelerate and facilitate auxiliary production processes (master models, casting moulds, press moulds). Additive technologies are now and then used for manufacturing non-critical lightly loaded parts (decorative moulding, protective casings, etc.). The SLM technology allows to manufacture not only prototypes [1], but also finished products made of metals and alloys with a large complex of mechanical properties. According to the problem-oriented studies including those conducted at Snezhinsk Physics and Technology Institute (SPTI NRNU MEPhi) in collaboration with Russian Federal Nuclear Center-Zababakhin All-Russia Institute of Technical Physics (RFNC-VNIITF), the chemical composition of metal powders for printing applications may vary. A transition to printing using other materials, interesting for the given manufacturer, can be accomplished quickly enough. Modes for printing new materials are developed using analytical methods to be followed by an analysis of the structure and physical and mechanical properties of printed samples. So far, there are well developed and tested printing methods using powders made of stainless steel and heat-resistant nickel alloy, the chemical composition of which meets Russian state standards. The properties of finished products correspond to physical and mechanical properties of materials based on traditional technologies, but sometimes they are much better (Table 1).

There are some peculiarities in selection of the raw material for applications in the nuclear weapons complex and aerospace engineering. In particular, performance characteristics (physical and mechanical properties) are brought to the forefront while processing properties (manufacturability) are not a limiting factor. Therefore, production based on traditional technologies often involves heavy material and time expenditures. That is why such production technologies cannot be classified as high-tech solutions. Integrating additive technologies into the production process will help improve the efficiently to a great extent. Generally, implementation of the additive technologies into the production sector in Russia leaves much to be desired. The uncertainty of parts accuracy parameters is one of the factors hampering implementation of the additive technologies. In particular, this paper highlights the issues related to geometric accuracy and surface quality parameters for parts based on the selective laser melting (SLM) technology being the most advanced solution for manufacturing
critical functional parts.

According to the analysis, the key issues to be addressed for large-scale implementation of the SLM technology into the production process are as follows: what is the highest geometric
accuracy at the manufacturing stage and what is the surface microchemistry? Thus, the following questions have to be answered in terms of practical application:

• possibility for manufacturing parts with required accuracy and surface quality without additional treatment;
• available stability/repeatability of dimensions of parts to be manufactured;
• whether the part surface shape affects the deviation from nominal dimensions of a 3D model or not;
• possibility of post-processing of parts produced based on the SLM technology;
• which post-processing methods can be used and which ones will be the most efficient;
• if post-processing is needed, which minimum allowance shall be reserved;
• estimated post-processing cost price;
• is it profitable to use 3D printing if post-processing is planned?

The cornerstone is to determine accuracy and stability/repeatability of dimensions in 3D printing.

To determine geometry accuracy, we have to determine the maximum theoretical geometric accuracy, which depends on the physics of the process.

By its physics, the SLM technology is related to Bed Deposition, i. e. the initial formation of a bed with selective internal hardening (melting) of metal powder. A bed is formed due to movement of the work table (deposition platform) in the vertical direction (z-axis) and subsequent distribution of the deposition material powder over the platform. This process includes two limiting factors: 3D printer unit step value (equipment specifications) and deposition material particle size.

According to the 3D printers performance review, the minimum deposition step is 20 µm.

Each manufacturer involved in production of 3D printers gives their own recommendations on the size range of powders for printing: Phenix (3D Systems) – powder with particle size d₅₀ = 10 µm [1], Conzept Laser – powder fineness d₅₀ = 26.9 µm (25…52 µm), SLM Solutions – d₅₀ = 10...30 µm, Arcam – particle size of 45…100 µm. Numeric value of parameter d₅₀
means that 50 % of powder particles shall not exceed the specified numeric value (if d₅₀ = 40 µm, this means 50 % of powder particles have the particle size less or equal to 40 µm), while other particles may exceed the specified numeric value d50. According to the specified data on various 3D printers, particles of powder used in production mostly exceed the size of 20 µm; therefore, it is technically impossible to obtain a bed of this thickness with considerably close packing of consumed powder’s particles. Based on the above and on 3D printer manufacturers’ recommendations, the bed thickness is usually selected starting from 40 µm with an optional increment in thickness (step by step) by 20 µm. Based on the above information, all curved surfaces can follow the shape of a 3D model only if it’s a stepped shape (Fig. 1). The step height will not be less than the unit step thickness.

Another limiting factor is determined by the melting process itself (Fig. 2).

In the course of product moulding the bed metal is melted under the effect of a laser beam. The size of the unit molten pool will depend on the laser beam focus, metal thermal conductivity, particle packing density in the bed, surface tension formed on the molten metal surface, etc. This is a multi-factor process, but we may note that the unit pool always exceeds the bed area exposed to laser. The size of the area exposed to laser (degree of focus) is determined at the control software development stage and usually varies in the range of 40…80 µm. As the unit molten pool has a specific (semicirclelike) shape (see Fig. 2), irregularities will appear on the surface. The size of irregularities depends on the degree of beam focus and bed thermal conductivity. Moreover, partially molten (“fused”) particles may remain in extreme unit pools, thus distorting the surface microgeometry and changing dimensions of the finished products. The degree of particle fusion in extreme layers may vary, and, therefore, their surface bonding strength will be different. Particles with a minimum fusion value (the particle size usually lies within 18…80 µm) may have a considerable effect on the surface roughness and dimensions, but such particles can be easily removed. If the degree of particle fusion is greater or about 50 %, it takes forces, the value of which is close to the strength of the material itself, to remove a particle, but the effect on changes in the product dimensions will be considerably reduced as compared with the previous case. Based on the above, the surface microgeometry of the products based on the SLM technology will be determined by the size of a unit pool (i.e. by the size of the area exposed to laser) and by the size of particles of raw metal powder. If we rely on the assumption that the
size of raw material particle is 25…45 µm and the unit pool is 40 µm, the surface roughness shall be about R_z = 20.

Hence, the analysis of the selective laser melting process physics has shown that the smaller the deposition step and the smaller the raw material particle size, the higher accuracy and the smaller surface roughness can be obtained. The lower size of particles used in the SLM technology is limited based on technological features of the process [3]. In particular, the
particle size shall ensure the required flowability for bed formation in order to prevent the bed from blowing off by gas flows from the working space. Therefore, the minimum particle size will depend on the specific density of the parent alloy itself.

To obtain specific data on accuracy, we conducted a number of experiments. Based on the previous experience in manufacturing samples and parts using the SLM technology we
found out that in addition to the particle size distribution of the raw powder, the accuracy of resulting dimensions and the surface microgeometry may be affected by the position of samples during printing in the 3D printer’s working space, as well as by their shape and position relative to the vertical axis in the working space during manufacturing. To prove (or disprove) this finding, we selected the following samples of simple shape: 

• rectangular parallelepiped, sizes 10 × 10 × 30 mm;
• cylinder, diameter 10 mm, length 30 mm.

To reveal possible influence of part position within the printing area when compiling control software, 3D models were put in the vertical position and at 45° relative to the horizontal axis.

For SLM-based printing, we used Realizer SLM 100 3D printer. The following parameters were selected for printing: bed layer thickness – 50 µm, laser spot – 20 µm. Raw powder particle size d₅₀ = 20 µm.

We measured the obtained samples using the non-contact method and multi-function measuring means such as 3D scanning (accuracy of 0.005 mm) (Fig. 3) and a micrometer with the accuracy of 0.001 mm (Fig. 4). Measurement data obtained by means of a 3D scanner were processed using the problem-oriented software Magic. Measurement data allows to estimate deviations from the initial model on the entire surface of the sample by a huge number of points. A unit scan contains up to 1 million points. We used 14 unit scans for sample scanning. A disadvantage of the method is higher measurement error (0.005 mm) as compared to micrometer measurement; therefore, this experiment involves scanning as an auxiliary (estimating) check mode, which gives a more comprehensive estimation of deviations of the actual surface, but numeric values have considerable deviations caused, among other things, by surface glittering.

In order to obtain more valid micrometer measurement data, we carried out 8 series of measurement experiments for each sample. The measurement results were processed using statistic analysis methods [4][5]. The results are shown in Fig. 4. With the cylinders put in the vertical position, the spread of actual values was 100 µm, in tilted position – 70 µm, which corresponds to standard tolerance grades 10–11.

Based on measurement results obtained by using both methods, we found out the following regularity: most of actual dimensions of cylinders are a little bit smaller than the nominal dimensions. This can be explained by the features related to preparation of control software for printing [6]. Control software for a 3D printer is complied using the 3D model represented in the STL format (triangulated model). When the cylindrical surface is represented by triangles, there is an error associated with this inaccuracy (Fig. 5). 

To reduce deviations of actual dimensions of the curved surfaces from the nominal value, two methods can be used:

• improving the accuracy of the STL model by reducing the separation step;
• recalculating and increasing (decreasing) the nominal dimension of the 3D model used for developing control software for 3D printer.

The first method can be used for products of comparatively simple shape. For products with a complex three-dimensional outline, if the step is reduced, the model may be too large,
and additional errors will appear, caused by 3D model uniformity check. Besides, this leads to stricter requirements to materials and hardware.

Implementation the second method requires test printing, determination of deviations, and integration of correction coefficients for 3D model dimensions. The advantage of the method
is the possibility to take into account the influence of the thermal expansion coefficient and at the same time the influence of the printed part’s position in the working space. In fact, we implemented the second method with the separation step partially reduced. Actual dimensions of printed cylinders after integration of correction coefficients are shown in Fig. 6.

After correction coefficients were integrated into the 2D model (corrected printing) with the cylinders in the vertical position, the spread of actual values was 52 µm; in tilted position –
50 µm. These values do not exceed the tolerance for standard tolerance grade 9.

After integrating correction coefficients, the spread of actual values of cylinders irrespective of their position in the 3D printer’s working space during manufacturing is 58 µm. This value
exceeds the tolerance by standard tolerance grade (IT9 – 52 µm) by 6 µm, but it is considerably smaller than the tolerance by standard tolerance grade 10 (IT10 – 84 µm). According to
experimental data processing, most of the actual dimensions of the printed samples (95 %) are grouped close to the nominal dimension with the spread of 40 µm. Further improvement of accuracy of manufacturing is possible, but it will be difficult due to particle fusion, i. e. due to surface roughness. Dimensional accuracy can be improved by using minimum post-processing, for example, sand blasting.

Figures 7 and 8 show the results obtained by measuring actual dimensions of test print prisms.

The surface shape analysis with the help of 3D scanning revealed powder particles fused to the prism faces, resulting in a dramatic deviation of actual dimensions from nominal values
during measurements with a micrometer. We applied sand blasting and obtained actual dimensions of prism faces close to nominal values (Fig. 9, Table 2). 

According to the experiment results, the dimensional repeatability is rather high. The analysis of actual dimensions of the printed samples revealed the following specific features:

• vertical samples (cylinders and prisms) have a small degree of taper at the height of 30 mm within the limits of up to 6 µm;
• samples tilted at 45° are free of taper during manufacturing; the spread of actual dimensions is smaller than that of vertical samples by 42 %.

During printing based on the SLM technology, we checked surface roughness on 40 samples with different position of surfaces relative to the vertical axis in the 3D printer’s working space. Measurements have revealed that the surface roughness does not depend on the surface position in the working space during printing or on the shape. Vertical surfaces have minimum roughness R_a = 1.6...2.5. The higher the deviation of the surface from the vertical axis, the higher the roughness value. It can reach R_z = 40...50 on horizontal surfaces.
Thus, according the analysis of selective laser melting process physics and the experiment in printing using stainless steel 12KH18N10T, we can make the following conclusions:

• the accuracy of manufacturing complies with the standard tolerance grade 9–10;
• the spread of actual dimensions is smaller if a part is slightly tilted relative to the vertical axis (up to 43°) during printing;
• samples based on the SLM technology have high dimensional repeatability;
• the shape of the surface being formed affects the accuracy of dimensions;
• to improve the accuracy, correction coefficients may be integrated into a 3D model at the control software preparation stage;
• SLM-based surfaces may be exposed to sand blasting, which considerably reduces surface roughness and brings actual dimensions closer to predetermined values;
• the roughness of machined surfaces depends on the surface position in the working space during manufacturing; the closer the surface to the vertical axis, the lower the roughness, and vice versa.

The accuracy of manufacturing (standard tolerance grade 9–10) may fully comply with the requirements for unmated surfaces, i. e. such surfaces may be printed based on the selective laser melting technology (SLM) without further machining. Surfaces with more strict accuracy requirements (standard tolerance grade 6–7) and lower roughness require further machining. That is why, a new phase of development shall be intended for determining the depth of a defective layer when manufacturing parts based on selective laser melting.