- Email: [email protected]
Homogeneity aspects in selective laser sintering (SLS)
Journal of Materials Processing Technology 177 (2006) 348–351
Homogeneity aspects in selective laser sintering (SLS) S. Kolosov a,∗ , G. Vansteenkiste a,b , N. Boudeau a , J.C. Gelin a , E. Boillat c a
ENSMM-LMARC, 26 ch. de l’Epitaphe, 25030 Besan¸con CEDEX, France b ENSMM-LMS, 24 ch. de l’Epitaphe, 25030 Besan¸ con CEDEX, France c EPFL-IPR-LGPP, 1015 Lausanne, Switzerland
Abstract A measure of powder layer heterogeneity has been proposed. It has been tested on the layers realized by different deposition techniques. Good correspondence of the actual measure to the subjective layer quality estimation has been shown. The influence of the laser scanning strategy on the quality of sintered structure has been examined. It is shown that both the sintering precision as well as the inner sintering quality is strongly affected by the hatch distance whereas this fact has been often neglected in the past. © 2006 Elsevier B.V. All rights reserved. Keywords: Selective laser sintering (SLS); Powder deposition; Accumulated fluence
1. Introduction
2.1. Deposition techniques
SLS is a solid freeform fabrication technique. It consists in building a three-dimensional object layer by layer out of a powder selectively heated by laser radiation. The liquid formed by the partially molten material binds the surrounding powder and solidifies when the temperature decreases, which leads to consolidation. This work is dedicated to a study of homogeneity of the sintering process. Homogeneity of fabricated parts is one of the most important criteria imposed upon the SLS process [2]. Questions related to the powder layer deposition are discussed in Section 2. Section 3 describes a simplified sintering model based on the density of energy absorbed by the powder layer and Section 4 concludes the article.
This paper reports on the results of the stainless steel OSPREY 90% −16 m H13 powder deposition tests realized with four different techniques. Their principles are described in this subsection.
2. Powder layer deposition In selective laser sintering the deposition of the powder is an important sub-process. The quality of the powder layer is a very important issue for the process precision and stability [2]. The main quality demands for the layer are to be of constant thickness and homogeneous. However no universal deposition solution suitable for any kind of powder exists until now. ∗
Corresponding author. Tel.: +33 3 81 66 60 76; fax: +33 3 81 66 67 00. E-mail address: [email protected] (S. Kolosov).
0924-0136/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2006.03.188
2.1.1. Classical deposition Initially, a substrate should be covered with loose powder. The top surface of the powder layer is formed by wiping the powder heap with a blade on a certain height from the substrate. This height determines the layer thickness. This technique has been successfully tested with wide range of powders with typical grain size 20–100 m. It however fails in deposition of fine powders with typical grain size <5 m due to powder grain agglomeration. It also gives challenging results for powders in the range of 5–20 m, the quality of the deposited layer is far from being perfect (see Fig. 1(a)). 2.1.2. Pressure-gradient deposition Some significant improvements of the classical deposition technique (see Section 2.1.1) have been proposed. They consists mainly in optimizing the angle and the shape of the scraping surface of the wiping blade. In the current tests the system equivalent to the one described in article [5] has been used. This technique has been found to have a larger range of suitable powder grain size. The deposition tests also show a notable
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
349
Fig. 1. SEM images of powder layer deposited by different technologies and their heterogeneity coefficient ζ.
improvement of layer quality in comparison with the classical deposition (see Fig. 1(b)). 2.1.3. Ultrasound powder compaction The plate vibrating with an ultrasonic frequency compresses loose powder. After the required layer thickness is achieved, the plate should be removed in a delicate manner to avoid destruction of the previously formed powder layer. The powder layer deposited by ultrasound compaction locally (tens of grains) shows a slightly better quality than in the pressure-gradient case. The overall layer quality however cannot compete with the previously described technique because significant cracks appear on a larger scale (see Fig. 1(c)). 2.1.4. Spread method A powder suspended in a highly volatile liquid (acetone was used here) is deposited on the surface to be covered by a powder layer. Under gravitational force the homogeneous suspension redistributes itself and forms a quasi-horizontal surface. After a while the liquid evaporates and the powder remains in a horizontally aligned layer.
The spread method definitely shows the best quality results in deposition (see Fig. 1(d)). It is however the most difficult technique to implement, it is almost incapable of being automated and the time necessary to form a powder layer is incomparably longer than in dry powder deposition. Moreover, the layer thickness becomes extremely difficult to control. 2.2. Powder layer quality control In Fig. 1 and Section 2.1 the quality of four samples deposited with different techniques was compared. The quality of the layer is however estimated visually which is not always an evident and stable method. For computing purposes the luminosity of a pixel was associated to the height of the powder grain surface. 2.2.1. Measuring procedure In this section a measure of the powder layer quality in term of its homogeneity is proposed. The coefficient of relative heterogeneity (in the context of this article) can be estimated according to the following procedure.
350
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
If f = f(x, y) is the discrete function defined on [0, N] × [0, M] and representing the height of the powder grain surface, then ωg
ω=0 ζ = N
F (ω)
ω=0 F (ω)
(1)
where ωg is the frequency corresponding to the biggest grain diameter D, ωg = N/2D, and F is the y-average of the discrete spectra of function f M N 1 F (ω) = f (x) e−2πiωx/N M
(2)
y=0 x=0
2.2.2. Interpretation The height term hab of the grains surface averaged over a domain with sides a and b is introduced. It is clear that hab will depend on the coordinate of averaging domain and also on the values a and b. Variation (or oscillation) of the average height function hab will be indicated in the Fourier image of the initial height function as a non-zero value at the frequency which corresponds to the sizes a and b (see (2)). One cannot expect the powder layer to be homogeneous on the scale smaller than the grain size. Thus the value ζ is determined as the measure of heterogeneity on a scale larger than the typical grain size, which corresponds to the frequencies [0, ωg ]. It is of course better if all the heterogeneities are concentrated in the high frequency area which leads to small values of heterogeneity ζ → 0. The normalizing in the formula (1) allows avoidance of the influence of the contrast and the lower limit ω = 1 (not ω = 0) eliminates the influence of the image brightness on the final result. 2.2.3. Limitations The procedure described above cannot pretend to be a firm theory of heterogeneity detection for any kind of surface images. The actual formulation does not take a possible anisotropy of the surface structure into account, which is why the averaging in formula (2) is performed only for variable y. To complete this model it would be necessary to implement averaging in any possible direction. Another problem is related to the image quality. Even though the influence of the brightness and the contrast is depressed, the procedure cannot foresee oversaturated or overdarken images, so the image must have reasonable contrast and brightness. It also remains as an open question, ‘what is the heterogeneity of absolutely flat surface f = const’, it formally leads to the uncertainty of ζ in (1) and (2). This case however is similar to the case of certain brightness and zero contrast which is restricted for the procedure according to the previous paragraph. 2.2.4. Results The results of the heterogeneity tests computed according to formulae (1) and (2) are presented below:
Deposition technique
Heterogeneity ζ
Classical deposition Pressure-gradient Ultrasound compaction Suspended deposition
0.16713 0.13037 0.1402 0.0945
As can be seen, the table corresponds to reasoning of Section 2.1. The heterogeneity ζ is in good agreement with subjective estimation of the layer quality. 3. Energy deposition Another SLS sub-process is the laser-matter interaction itself. The model of SLS process based on the approach of continuous locally homogeneous and isotropic medium exists [3,4]. It predicts the temperature evolution as well as the sintering phenomenon within the part. The computation however is very complicated due to non-linearities and memory effects of the material during sintering. A simplified approach to describe the sintering phenomenon has been proposed. It is based on the assumption that the sintering level [3] depends on the integral energy density absorbed by the powder only. It neglects the effect of the energy deposition, of the thermal diffusion [1] and of the sintering state evolution. The energy density absorbed by the material, called accumulated fluence, can be considered as one of the parameters characterizing the sintering. The homogeneity of energy density is a necessary but not sufficient factor describing the homogeneity of sintering. The quality of sintering process can be quantified by the level of heterogeneity of the absorbed energy density. 3.1. Sintering precision Usually the scan strategy is determined by the contour of the desired shape [2] regardless of the laser and the powder param-
Fig. 2. Out-area of sintering depending on hatch distance.
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
351
starts is about 0.25 J/mm2 . This value is of course valid only for continuous exposition during 1 ms and for the powder used in these tests. We thus can find a compromise to choose the hatch distance in a way to diminish out-sintering zone and to achieve good quality inside the area to be sintered. 4. Conclusions
Fig. 3. Energy density profile in transversal direction, Hd = 50 m.
eters. However this simple method sometimes leads to fallout in sintering precision and quality. The sintering takes place in a certain area around the center of the laser beam due to the thermal diffusion and the variation of laser intensity within the beam. The zone of sintering thus exceeds the desired one for a certain distance called “out-area of sintering”. The experimentally examined dependence of this distance on the hatch distance Hd is shown in Fig. 2. The experimental study has been performed by sintering H13 powder (see Section 2.1) under the Nd:YAG laser of 10 W power in continuous wave mode. The beam diameter was 30 m and the scan speed was 30 mm/s. Fig. 2 indicates that the hatch distance should be increased in a way to make the out-sintering area smaller. 3.2. Homogeneity of energy density On the other hand, if the hatch distance exceeds 30 m then the zones covered by sintered powder inside the area to be consolidated no longer overlap and as a result, no solid object can be produced. The same effects can be observed by analyzing the absorbed energy density (see Fig. 3). According to Fig. 3 and the experimental results, the minimal energy density at which the sintering
A procedure to estimate the powder layer quality by means of the so called heterogeneity coefficient ζ has been proposed. This coefficient shows good accordance with subjective quality by visual estimation. The advantages of this procedure are its independence with respect to personal perception, brightness and contrast. A well known model based on the global energy density absorbed by the powder has been implemented. It has been used to study the influence of the hatch distance on the precision and also the quality of the sintered layer. A way to determine the optimal hatch distance has also been shown. Acknowledgements The authors acknowledge industrial partners: Laser Cheval, Groupe S3i, Tenax SA and Pierre Bercher SA for supportion INTERREG project. They are also grateful to Mr. M. Matthey and Mr. J. Jhabvala from Swiss Federal Institute of Technology, Lausanne, Switzerland and Mr. C. Roques Carmes from ENSMM, Besancon, France for their help and consiltations. References [1] P. Fischer. Non-ablative laser beam interaction with materials. PhD thesis, UNIBE, IAP, Bern, Switzerland, 2003. [2] N.P. Karapatis. A Sub-Process Approach of Selective Laser Sintering. PhD thesis, EPFL, LGPP, Lausanne, Switzerland, 2001. [3] S. Kolossov. Non-linear Model and Finite Element Simulation of the Selective Laser Sintering Process. PhD thesis, EPFL, LGPP, Lausanne, Switzerland, 2005. [4] S. Kolossov, E. Boillat, R. Glardon, P. Fischer, M. Locher, 3D FE simulation for temperature evolution in the selective laser sintering process, Int. J. Machine Tools Manuf. 44 (2004) 117–123. [5] G. Vansteenkiste, N. Boudeau, H. Leclerc, T. Barrire, J.C. Gelin, C. Roques-Carmes, C. Millot, C. Benoit, E. Boillat, Investigations in direct tooling for micro-technology with SLS, in: Proceedings of Fourth International Conference on Laser Assisted Net Shape Engineering, Erlangen, Germany, 2004, pp. 425–434.
Homogeneity aspects in selective laser sintering (SLS) S. Kolosov a,∗ , G. Vansteenkiste a,b , N. Boudeau a , J.C. Gelin a , E. Boillat c a
ENSMM-LMARC, 26 ch. de l’Epitaphe, 25030 Besan¸con CEDEX, France b ENSMM-LMS, 24 ch. de l’Epitaphe, 25030 Besan¸ con CEDEX, France c EPFL-IPR-LGPP, 1015 Lausanne, Switzerland
Abstract A measure of powder layer heterogeneity has been proposed. It has been tested on the layers realized by different deposition techniques. Good correspondence of the actual measure to the subjective layer quality estimation has been shown. The influence of the laser scanning strategy on the quality of sintered structure has been examined. It is shown that both the sintering precision as well as the inner sintering quality is strongly affected by the hatch distance whereas this fact has been often neglected in the past. © 2006 Elsevier B.V. All rights reserved. Keywords: Selective laser sintering (SLS); Powder deposition; Accumulated fluence
1. Introduction
2.1. Deposition techniques
SLS is a solid freeform fabrication technique. It consists in building a three-dimensional object layer by layer out of a powder selectively heated by laser radiation. The liquid formed by the partially molten material binds the surrounding powder and solidifies when the temperature decreases, which leads to consolidation. This work is dedicated to a study of homogeneity of the sintering process. Homogeneity of fabricated parts is one of the most important criteria imposed upon the SLS process [2]. Questions related to the powder layer deposition are discussed in Section 2. Section 3 describes a simplified sintering model based on the density of energy absorbed by the powder layer and Section 4 concludes the article.
This paper reports on the results of the stainless steel OSPREY 90% −16 m H13 powder deposition tests realized with four different techniques. Their principles are described in this subsection.
2. Powder layer deposition In selective laser sintering the deposition of the powder is an important sub-process. The quality of the powder layer is a very important issue for the process precision and stability [2]. The main quality demands for the layer are to be of constant thickness and homogeneous. However no universal deposition solution suitable for any kind of powder exists until now. ∗
Corresponding author. Tel.: +33 3 81 66 60 76; fax: +33 3 81 66 67 00. E-mail address: [email protected] (S. Kolosov).
0924-0136/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2006.03.188
2.1.1. Classical deposition Initially, a substrate should be covered with loose powder. The top surface of the powder layer is formed by wiping the powder heap with a blade on a certain height from the substrate. This height determines the layer thickness. This technique has been successfully tested with wide range of powders with typical grain size 20–100 m. It however fails in deposition of fine powders with typical grain size <5 m due to powder grain agglomeration. It also gives challenging results for powders in the range of 5–20 m, the quality of the deposited layer is far from being perfect (see Fig. 1(a)). 2.1.2. Pressure-gradient deposition Some significant improvements of the classical deposition technique (see Section 2.1.1) have been proposed. They consists mainly in optimizing the angle and the shape of the scraping surface of the wiping blade. In the current tests the system equivalent to the one described in article [5] has been used. This technique has been found to have a larger range of suitable powder grain size. The deposition tests also show a notable
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
349
Fig. 1. SEM images of powder layer deposited by different technologies and their heterogeneity coefficient ζ.
improvement of layer quality in comparison with the classical deposition (see Fig. 1(b)). 2.1.3. Ultrasound powder compaction The plate vibrating with an ultrasonic frequency compresses loose powder. After the required layer thickness is achieved, the plate should be removed in a delicate manner to avoid destruction of the previously formed powder layer. The powder layer deposited by ultrasound compaction locally (tens of grains) shows a slightly better quality than in the pressure-gradient case. The overall layer quality however cannot compete with the previously described technique because significant cracks appear on a larger scale (see Fig. 1(c)). 2.1.4. Spread method A powder suspended in a highly volatile liquid (acetone was used here) is deposited on the surface to be covered by a powder layer. Under gravitational force the homogeneous suspension redistributes itself and forms a quasi-horizontal surface. After a while the liquid evaporates and the powder remains in a horizontally aligned layer.
The spread method definitely shows the best quality results in deposition (see Fig. 1(d)). It is however the most difficult technique to implement, it is almost incapable of being automated and the time necessary to form a powder layer is incomparably longer than in dry powder deposition. Moreover, the layer thickness becomes extremely difficult to control. 2.2. Powder layer quality control In Fig. 1 and Section 2.1 the quality of four samples deposited with different techniques was compared. The quality of the layer is however estimated visually which is not always an evident and stable method. For computing purposes the luminosity of a pixel was associated to the height of the powder grain surface. 2.2.1. Measuring procedure In this section a measure of the powder layer quality in term of its homogeneity is proposed. The coefficient of relative heterogeneity (in the context of this article) can be estimated according to the following procedure.
350
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
If f = f(x, y) is the discrete function defined on [0, N] × [0, M] and representing the height of the powder grain surface, then ωg
ω=0 ζ = N
F (ω)
ω=0 F (ω)
(1)
where ωg is the frequency corresponding to the biggest grain diameter D, ωg = N/2D, and F is the y-average of the discrete spectra of function f M N 1 F (ω) = f (x) e−2πiωx/N M
(2)
y=0 x=0
2.2.2. Interpretation The height term hab of the grains surface averaged over a domain with sides a and b is introduced. It is clear that hab will depend on the coordinate of averaging domain and also on the values a and b. Variation (or oscillation) of the average height function hab will be indicated in the Fourier image of the initial height function as a non-zero value at the frequency which corresponds to the sizes a and b (see (2)). One cannot expect the powder layer to be homogeneous on the scale smaller than the grain size. Thus the value ζ is determined as the measure of heterogeneity on a scale larger than the typical grain size, which corresponds to the frequencies [0, ωg ]. It is of course better if all the heterogeneities are concentrated in the high frequency area which leads to small values of heterogeneity ζ → 0. The normalizing in the formula (1) allows avoidance of the influence of the contrast and the lower limit ω = 1 (not ω = 0) eliminates the influence of the image brightness on the final result. 2.2.3. Limitations The procedure described above cannot pretend to be a firm theory of heterogeneity detection for any kind of surface images. The actual formulation does not take a possible anisotropy of the surface structure into account, which is why the averaging in formula (2) is performed only for variable y. To complete this model it would be necessary to implement averaging in any possible direction. Another problem is related to the image quality. Even though the influence of the brightness and the contrast is depressed, the procedure cannot foresee oversaturated or overdarken images, so the image must have reasonable contrast and brightness. It also remains as an open question, ‘what is the heterogeneity of absolutely flat surface f = const’, it formally leads to the uncertainty of ζ in (1) and (2). This case however is similar to the case of certain brightness and zero contrast which is restricted for the procedure according to the previous paragraph. 2.2.4. Results The results of the heterogeneity tests computed according to formulae (1) and (2) are presented below:
Deposition technique
Heterogeneity ζ
Classical deposition Pressure-gradient Ultrasound compaction Suspended deposition
0.16713 0.13037 0.1402 0.0945
As can be seen, the table corresponds to reasoning of Section 2.1. The heterogeneity ζ is in good agreement with subjective estimation of the layer quality. 3. Energy deposition Another SLS sub-process is the laser-matter interaction itself. The model of SLS process based on the approach of continuous locally homogeneous and isotropic medium exists [3,4]. It predicts the temperature evolution as well as the sintering phenomenon within the part. The computation however is very complicated due to non-linearities and memory effects of the material during sintering. A simplified approach to describe the sintering phenomenon has been proposed. It is based on the assumption that the sintering level [3] depends on the integral energy density absorbed by the powder only. It neglects the effect of the energy deposition, of the thermal diffusion [1] and of the sintering state evolution. The energy density absorbed by the material, called accumulated fluence, can be considered as one of the parameters characterizing the sintering. The homogeneity of energy density is a necessary but not sufficient factor describing the homogeneity of sintering. The quality of sintering process can be quantified by the level of heterogeneity of the absorbed energy density. 3.1. Sintering precision Usually the scan strategy is determined by the contour of the desired shape [2] regardless of the laser and the powder param-
Fig. 2. Out-area of sintering depending on hatch distance.
S. Kolosov et al. / Journal of Materials Processing Technology 177 (2006) 348–351
351
starts is about 0.25 J/mm2 . This value is of course valid only for continuous exposition during 1 ms and for the powder used in these tests. We thus can find a compromise to choose the hatch distance in a way to diminish out-sintering zone and to achieve good quality inside the area to be sintered. 4. Conclusions
Fig. 3. Energy density profile in transversal direction, Hd = 50 m.
eters. However this simple method sometimes leads to fallout in sintering precision and quality. The sintering takes place in a certain area around the center of the laser beam due to the thermal diffusion and the variation of laser intensity within the beam. The zone of sintering thus exceeds the desired one for a certain distance called “out-area of sintering”. The experimentally examined dependence of this distance on the hatch distance Hd is shown in Fig. 2. The experimental study has been performed by sintering H13 powder (see Section 2.1) under the Nd:YAG laser of 10 W power in continuous wave mode. The beam diameter was 30 m and the scan speed was 30 mm/s. Fig. 2 indicates that the hatch distance should be increased in a way to make the out-sintering area smaller. 3.2. Homogeneity of energy density On the other hand, if the hatch distance exceeds 30 m then the zones covered by sintered powder inside the area to be consolidated no longer overlap and as a result, no solid object can be produced. The same effects can be observed by analyzing the absorbed energy density (see Fig. 3). According to Fig. 3 and the experimental results, the minimal energy density at which the sintering
A procedure to estimate the powder layer quality by means of the so called heterogeneity coefficient ζ has been proposed. This coefficient shows good accordance with subjective quality by visual estimation. The advantages of this procedure are its independence with respect to personal perception, brightness and contrast. A well known model based on the global energy density absorbed by the powder has been implemented. It has been used to study the influence of the hatch distance on the precision and also the quality of the sintered layer. A way to determine the optimal hatch distance has also been shown. Acknowledgements The authors acknowledge industrial partners: Laser Cheval, Groupe S3i, Tenax SA and Pierre Bercher SA for supportion INTERREG project. They are also grateful to Mr. M. Matthey and Mr. J. Jhabvala from Swiss Federal Institute of Technology, Lausanne, Switzerland and Mr. C. Roques Carmes from ENSMM, Besancon, France for their help and consiltations. References [1] P. Fischer. Non-ablative laser beam interaction with materials. PhD thesis, UNIBE, IAP, Bern, Switzerland, 2003. [2] N.P. Karapatis. A Sub-Process Approach of Selective Laser Sintering. PhD thesis, EPFL, LGPP, Lausanne, Switzerland, 2001. [3] S. Kolossov. Non-linear Model and Finite Element Simulation of the Selective Laser Sintering Process. PhD thesis, EPFL, LGPP, Lausanne, Switzerland, 2005. [4] S. Kolossov, E. Boillat, R. Glardon, P. Fischer, M. Locher, 3D FE simulation for temperature evolution in the selective laser sintering process, Int. J. Machine Tools Manuf. 44 (2004) 117–123. [5] G. Vansteenkiste, N. Boudeau, H. Leclerc, T. Barrire, J.C. Gelin, C. Roques-Carmes, C. Millot, C. Benoit, E. Boillat, Investigations in direct tooling for micro-technology with SLS, in: Proceedings of Fourth International Conference on Laser Assisted Net Shape Engineering, Erlangen, Germany, 2004, pp. 425–434.