investigation of deviations caused by powder compaction during 3d printing

Available online at www.sciencedirect.com ScienceDirect Procedia CIRP 57 (2016) 698 – 703 49th CIRP Conference on Manufacturing Systems (CIRP-CMS 2016) Investigation of deviations caused by powder compaction during 3D printing Christoph Schmutzlera,*, Clarissa Boekera, Michael F Zaeha a Institute for Machine Tools and Industrial Management, Beim Glaspalast 5, 86153 Augsburg, Germany * Corresponding author Tel.: +49-821-56883-47; fax: +49-821-56883-50 E-mail address: Christoph.Schmutzler@iwb.tum.de Abstract Powder bed based 3D printing technologies are of great interest for the production of prototypes and low quantity production lots because they allow the realization of highly customized and complex geometries Moreover, significant benefits of 3D printing include an expedited manufacturing and a decreased number of necessary tools in the process chain In contrast to traditional production methods, where material is subtracted or molded, thin layers of powder are successively added and solidified by locally applying binder based on a digital model By continuously depositing material, the powder bed experiences an additional compaction through the increasing weight of the material on top The consequences are dimensional deviations of the component in build-up direction and occasionally even avalanche-like collapses of material The objective of this study is the investigation of these effects and their influencing factors For this reason, occurring deviations and their underlying causes are examined Test objects were manufactured and examined with the objective of identifying the main influencing factors by varying the length, position and surrounding conditions of the objects within the building chamber Significant parameters were analyzed in greater detail in order to generate rules, regarding their effect on the occurrence and magnitude of the identified deviations Finally, an approach for the compensation of these deviations prior to the production by a pre-deformation of the digital model was implemented © Published by Elsevier B.V This ©2016 2015The TheAuthors Authors Published by Elsevier B.V.is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of Scientific committee of the 49th CIRP Conference on Manufacturing Systems (CIRP-CMS 2016) Peer-review under responsibility of the scientific committee of the 49th CIRP Conference on Manufacturing Systems Keywords: 3D Printing, Powder compaction, Deviations, Compensation Introduction The ongoing development of additive manufacturing (AM) technologies enables new applications beyond rapid prototyping [1] AM technologies increase the time flexibility regarding the production of individualized geometries and raise the design freedom [2] This leads to a continuously expanding market and thus to higher requirements in reference to size and shape accuracy [3] In powder bed based 3D printing (binder jetting) thin layers of powder are selectively solidified by the deposition of a binder using an ink-jet printhead The binder reacts with the powder and creates locally limited solidifications [4] By connecting the steps of applying thin powder layers and selectively depositing binder by lowering the powder bed, the parts are manufactured [5] For the investigations described in this paper a polymethylmethacrylate material system called PolyPor distributed by the company voxeljet was used Effect of powder compaction The powder used for 3D printing represents the raw material for the produced components, but loose, non-printed particles below the parts also stabilize them in the building chamber [6] A characteristic of the powder bed is a high percentage of pores in the interstice between the particles Especially for small grains (e g average particle size 50 µm), adhesive and electrostatic forces have a greater impact than gravitation, resulting in a lower bulk density of the powder bed than theoretically possible [7] Each new layer steadily increases the weight affecting the previously generated layers [8] The consequences are a continuous compaction of the powder 2212-8271 © 2016 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the 49th CIRP Conference on Manufacturing Systems doi:10.1016/j.procir.2016.11.121 699 Christoph Schmutzler et al / Procedia CIRP 57 (2016) 698 – 703 measured deviation during the process and randomly occurring collapses of the powder bed (cf Fig 2)[9] These effects are additionally supported by vibrations of the doctor blade and by the applied shear stress during the layer application [10] The compaction of the powder also affects the components, resulting in deviations in vertical direction (z-axis)[9] In a pre-study, z-rods with marks at regular intervals were created on a voxeljet VX800 system After production the rods were digitalized using a laser scanner type ScanControl Compact 2700 Subsequently, the edge points of the marks were separated by means of a case distinction and the total difference to the target value related to the start point of the rod was calculated and plotted in a diagram, cf Fig mm -1 -2 -3 -4 -5 -6 -7 z-position 100 200 300 mm 500 assumed shrinkage defect caused by collapse of the powder bed Fig Determined deviations caused by powder compaction Collapses of the powder bed occur randomly distributed over the entire build volume, whereby no compensation is possible In order to avoid this defect and to analyze only the deviation caused by continuous powder compaction, a printed stabilization grid fixed on a printed base plate was used In contrast to typical support structures in AM, the grid supports the entire powder bed and not only the components Additionally, the following investigations were transferred to a more recent and stable VX1000 system in order to reduce the probability of the occurrence of this defect Since the first and the last layer are almost in correct zposition at the end of the printing process, cf section 2, the difference between the start and the end point of the scatter-plot can be ascribed to the shrinkage of the material during solidification For the subsequent investigations it is assumed, that this shrinkage is linearly distributed over the component, as discussed in section in more detail In order to set the start and end point of the digitalized data to 0, the measurement values are adapted by subtracting a linear function This adjustment of the data was made to isolate the assumed effect of the continuous powder compaction effect For the description of the deviation, these adapted values are approximated by a quadratic function, cf Fig Preliminary work has already shown that a graphical derivation of the measured data by comparing the differences of neighboring marks shows a linear correlation [9] This also suggests a quadratic relationship The curve can be explained as follows: The lower layers are compacted by the increasing weight load and are therefore thinner than intended As a consequence, the real thickness of the subsequently applied powder layers increases In the middle of the building chamber (related to the z-axis) both effects, the compaction and the resulting increased layer thickness, almost balance Hence, component areas in the middle of the building chamber are shifted downwards, but the distances between the marks are nearly constant In the upper area of the building chamber the thickness of the applied powder layers continues z-position 0 100 200 300 mm 500 mm measured deviation For this pre-study all specific compensation operations of the 3D printer had been disabled and standard process parameters were used The manufactured z-rods use the height of the building chamber to full capacity Due to the accurate axial control of the build platform, the first and the last layer are almost in the correct z-position This is due to the lack of compactable powder below the first layer and the missing load on top of the last layer, except for some finishing layers [8] Therefore, the shown deviation measured at position 500 mm is primarily caused by shrinkage of the material (induced by the ongoing solidification reaction of the binder [11]) after producing the last layer At a z-position of about 380 mm a step in the measured data was determined The data of the digitalized z-rod shows an analog distortion in exactly this position, cf Fig (right) This defect was determined for all z-rods produced during the same manufacturing job In this case the collapse of the powder bed caused a deviation of approximately - mm The defect also affects commercial components, cf Fig Description of the deviation -1,0 -1,5 -2,0 -2,5 measured values adapted values Fig Defect caused by a collapse of the powder bed linear function approximated function Fig Description of the deviation as a combination of a linear and a quadratic function Christoph Schmutzler et al / Procedia CIRP 57 (2016) 698 – 703 to rise due to the increased number of compacting layers below and they are less compacted due to a lower weight load Thereby, the distance between the marks is greater than desired, but the total deviation declines Influencing factors In this paper potential influencing factors concerning the continuous powder compaction are analyzed in order to identify regularities The objective was to determine significant factors regarding the occurrence of the effect and to develop a compensation technique Therefore, three preliminary tests were executed following a two-stage fractional design of experiments The following factors were identified for the investigation, using a failure mode and effect analysis (FMEA): x the height of the test object, x the position in reference to the x-, y- und z-axis, x the distance to the edge or to the middle of the building chamber and x the distance between two neighboring test objects Powder wetted with binder shows a different behavior compared to loose powder Thus, a variation of the component’s height in build-up direction may also affect the resulting deviations This is important as the advanced design freedom in additive manufacturing often leads to components with significantly differing sizes within one build-up process In addition, it is known from experience and prior studies that the position of components within the building chamber also affects the intensity of powder compaction [9] Hence, the influence of changing the component’s position in z- as well as in x-y-direction was also investigated For industrial printing it is economical to achieve an optimal exploitation of the building chamber Therefore, components are usually printed in close proximity to one another The components may stabilize the powder bed but the evaporating binder may also influence nearby Thus, the distance between individual objects was analysed regarding the mutual influence of components being placed closely together For investigating the influence of the height, z-rods with a length of 250 mm and 450 mm were compared to each other The short rods were either placed 10 or 210 mm above the bottom of the building chamber in order to identify effects related to the z-position Short and long rods were positioned in each edge of the building chamber to simultaneously evaluate the influence of the position in x-y-direction Moreover, testing parts were placed in the middle of the building space in order to compare them with those in the border area According to consulted manufacturing experts, the diffusion area of evaporating binder reaches up to 30 mm surrounding the component Thus, the z-rods were placed either in a distance of 10 or 50 mm from each other The desired distance was achieved by printing additional testing objects, which were not evaluated Moreover, subjacent objects can affect the compaction of the powder This was implemented by creating additional plates placed under special z-rods As a result, the objects at the edge of the building chamber displayed less deviation than the ones in the middle This confirms the dependence on the position in x- and y-direction Moreover, the deviation is nearly constant along certain axes parallel to the x- and the y- direction and almost point symmetrical to the center of the building chamber Regarding the influence of surrounding components, no meaningful effect was detected Neither the varying distance between the individual objects nor the additional plates below the objects affected the deviation in a significant way The variation of the component length and the position in vertical direction had the greatest influence on the defect The short testing rods positioned in the upper section of the building chamber showed smaller deviations compared to the long zrods and to the short testing objects placed at the bottom of the building chamber In order to study the influence of the component height and its z-position more closely, two series each with six z-rods and a variation in height from 150 to 450 mm were placed staggered along the y-direction of the plant In the chosen area for these parts, only a minor influence of the position was determined before In the first series, all z-rods with different lengths were placed at the bottom of the building space while in the second series all test objects were aligned at the top of the building chamber Based on this investigation, a dependence of the identified deviation on the z-position was determined Therefore the measured values were matched to the corresponding zcoordinate In Fig the deviation to the target value of all scatter-plots of both series with different length and start positions are shown It is apparent, that all scatter-plots show the same trend Consequently, the effect is almost independent from the component’s length, but correlates to the respective position in z-direction Thus, a common description for the deviation is possible and the deviation can be described by one function throughout the entire height of the building chamber z-position 0 100 200 300 mm 500 mm measured deviation 700 -1,0 -1,5 -2,0 -2,5 -3,0 series placed at the botton series placed at the top approximated function Fig Comparison between the two staircase-shaped series 701 Christoph Schmutzler et al / Procedia CIRP 57 (2016) 698 – 703 Rules for powder compaction After having identified essential influences in a first trial, the following set of experiments was focused on examining the most significant factors in more detail For this trial an experimental design, the central composite design (CCD), was used With a CCD, factors can be examined precisely, as every factor can be investigated on five different stages As mentioned above, the deviation can be described by a quadratic function, wherein the maximum of the function depends on the respective area in x- and y-direction Accordingly, the positions in the x-y-plane were grouped into five different areas for the subsequent experiments These areas are rectangular zones with increasing size, located concentric around the middle of the building chamber The distance between zones is defined by the CCD and the dimensions of the building chamber, cf table This distribution allowed a reduction of the examined factors and a more efficient analysis When comparing rods which have been produced depending on the respective zone, the trends of the measured values in the zones 2-4 are comparable and show the same tendency In the zones in the middle (zone 1) and close to the edge (zone 5) fewer measuring points were gathered resulting in a higher statistical spread Thus, for subsequent compensation, the determined trend was extrapolated for these zones The functions of each zone represent the mathematical description of the effect for the pre-deformation In preparation for this compensation the reproducibility of the results was checked by comparing different build-up processes Parallel to the studies, three z-rods were produced at the same position in the building chamber during two different build-up processes The rods produced in the second build-up process were measured at two different moments in time in order to examine the influence of time Comparing the measured data, a significant difference in the values between the functions was found, but they share the same trend Thus the scatter-plots were separated into a linear and a quadratic portion, cf Fig z-position 0,5 mm 100 200 300 mm 500 By inspecting the graphs in more detail the difference between the measurements only occurs within the linear portion The quadratic functions of the measurements are almost identical The differences between the linear functions can be interpreted as a continuous shrinkage of the material after the finish of the build-up process This assumption is supported by a consideration of the time, which passed between the printing and the measurement The first measurement of the second trial was conducted within four days after the printing process As a result the linear function of this experiment shows the smallest deviation In the first trial the time period between production and measurement amounted to approximately two weeks Thus, the time interval as well as the linear deviation was greater during the first experiment The major period of time, about four weeks, passed between the printing and the second measurement of the second trial Accordingly, the linear function shows the largest deviation This also indicates that the linear part of the overall function can mostly be ascribed to the shrinkage effect Thus, the quadratic portion is caused by powder compaction and therefore mainly used in the following investigations Nevertheless, in the edge regions of the building chamber a smaller value for the linear shrinkage portion was determined In the remaining regions, this value was almost constant The remaining values after subtracting the linear functions are approximated by the following quadratic function: ݂ሺ‫ݖ‬ሻ ൌ ܽሺ‫ ݖ‬െ ݀ሻଶ ൅ ݁ (1) The function is described by the variable e that is defined as the deviation at the vertex The remaining parameters are derived from the boundary conditions: The start and end point of the function is 0, the function is symmetric and the vertex is located in the middle of the function Thus, the variable d is defined as 225 mm and the variable a is calculated by ܽൌ ି௘ ௗ; (2) The distribution of the maximal deviation over the x-y-plane of the building chamber is shown in Fig for the quadratic portion This value corresponds to the variable e in order to describe the deviation in each area of the building space by the function presented in equation measured deviation -0,5 -1,0 -1,5 -2,0 -2,5 -3,0 -3,5 -4,0 -4,5 trail trial (1 measurement) trial (2 measurement) measured values quadratic portion linear portion Fig Comparison of several measurements Fig Distribution of the maximal deviation of the quadratic proportion across the building chamber 702 Christoph Schmutzler et al / Procedia CIRP 57 (2016) 698 – 703 Compensation z-position Table Parameters for compensation zone distance to the middle in x-direction [mm] distance to the middle in y-direction [mm] linear compensation [mm] quadric compensation [mm] 30 2,60 0,35 128.5 59.5 1,92 0,51 272.5 147.5 2,03 0,70 416.5 235.5 1,90 1,03 515 295 1,16 For each compensated object a not pre-deformed counterpart was produced in order to compare and evaluate the results As described in section 5, in zone and a higher statistical spread of the measured values was observed Therefore, the results within the zones 2-4 are analyzed first Fig compares the measurement data of the z-rods compensated according to table to the not pre-deformed ones The result shows an improvement of the dimensional accuracy The deviation was reduced by an average of more than 80 % Subtracting the linear portion (shrinkage) from the not compensated rods, the defect caused by powder compaction decreased by 35-50 % on average Although the deviation was reduced by the compensation, a parabolic defect in the opposite direction emerged Comparing the results of the not predeformed rods with the first and the second trial, a smaller quadratic deviation was determined Possible reasons for these variations could be differing environmental conditions (like temperature and air humidity) or changing batches of powder, with different grain size distributions or altered flowability The consequence was an exaggerated compensation resulting in a smaller improvement than theoretically possible For the compensation of zone (middle), an extrapolation of the trend from the zones 2-4 was used due to the high statistical spread, cf section The measured deviations of the produced rods, pre-deformed and uncompensated, show a 100 200 300 mm 500 measured deviation mm -0,5 -1,0 -1,5 -2,0 -2,5 zone (compensated) zone (compensated) zone (compensated) zone (not compensated) zone (not compensated) zone (not compensated) Fig Evaluation of the compensation in zone 2, and similar trend compared to the rods in zone 2-4 and the predeformed ones were also affected by the overcompensation This suggests, that the applied compensation method can also be used in the middle of the building space In zone (edge area) a smaller linear portion was mentioned in the pre-studies Thus, only the quadratic portion was used for compensation Analyzing the results, in the lower half of the building chamber a significant improvement was observed Towards the end, both scatter-plots have almost the same deviation, cf fig Isolating the quadratic portion by subtraction the identified shrinkage, the result was a greater overcompensation than in the other zones Thus, it could be shown, that a segmentation into a linear and a quadratic portion is also promising in this case z-position 0,5 100 200 300 mm mm measured deviation In this study, the mathematical method of Free Form Deformation (FFD) was used in order to adapt the virtual geometry data of the test rods and a benchmark object During the FFD, not the object itself is deformed, but rather a cuboid surrounding the part [12] All changes of the cuboid are scaled to the interior object Thus, even a pre-deformation of complex geometries is possible In order to use the FFD, a mathematical description of the deviation is required [13] The implementation of the compensation proceeded in two separate steps First the quadratic and afterwards the linear compensation was performed For the quadratic compensation the overall height of the rods was kept constant and the midpoint was shifted to the corresponding value Using a 3rd order spline function for the connection of midpoint and edges, the pre-deformation was scaled throughout the rod according the quadratic function of equation For the linear compensation, a unidirectional scaling in z-direction was performed Thus, the overall length of the rods was increased by the measured value of the shrinkage The parameters for compensation depending on the related zone are presented in table 1: 1,0 -0,5 -1,0 -1,5 -2,0 -2,5 compensated not compensated Fig Evaluation of compensating the quadratic portion only 500 Christoph Schmutzler et al / Procedia CIRP 57 (2016) 698 – 703 future work the reproducibility must be examined in more detail For this purpose, appropriate parameters have to be determined in order to predict or avoid fluctuations, such as these due to flowability of the powder or environmental conditions The influence of this effect may also vary with the usage of different plants According to Eschey [3], in additive manufacturing each plant might offer its own shrinkage characteristic For both, the dependence on the powder and the environmental conditions as well as the influence of the machine, a solution has to be found in order to examine the reproducibility and to increase the significance of the results Finally, the compensation method can be transferred to industrial use cases, with the objective of producing more accurate components using 3D printing Fig Evaluation of the compensation by refence to the benchmark objet Parallel to this investigation a case study with two benchmark geometries was conducted, using the parameters according to table The objective was to verify the behavior of the continuous powder compaction and the developed compensation procedure Thus, a not compensated and a predeformed benchmark part were produced, cf fig The benchmark object was positioned 10 mm above the bottom and in two different zones (zones and 5) due to a size of approximately 186 x 177 x 155 mm³ Therefore, a transition of the compensation parameters from zone to zone was used for the compensation Although only a compensation of the quadratic portion was applied in zone 5, an improvement could be observed, because the benchmark part was placed in the lower half of the building space, cf Fig As observed at the test rods, an overcompensation was also determined Nevertheless, a reduction of the deviation of up to 50 % could be achieved Conclusion and Outlook The investigation has shown that an increase of the dimensional accuracy is possible by adapting the digital geometry data In this paper, a distinction was made between deviations caused by shrinkage of the material during solidification and variances induced by a compaction of the powder during the 3D printing process For the shrinkage a classical compensation by scaling the component according to the experience is recommended Excluding the effect of shrinkage, the first and the last layer in the building chamber are on the correct z-position Since the defect caused by powder compaction just affects the layers in between, it was analyzed separately Statistically distributed collapses in the powder bed were avoided by adapting the process conditions in order to implement a compensation As a result, the deviation caused by continuous compaction of the powder can be described via a quadratic function This function can be applied to the entire building chamber and is described by the maximum value depending on the x- and y-coordinate Nevertheless, differences in the magnitude of the deviations were observed between the different build-up processes and the experiments were performed on only one single machine In Acknowledgements The results were developed within the research project “Intelligent Deformation Compensation for 3D-Printers” subsidized by the Bavarian Research Foundation The authors sincerely thank the foundation and the partners for their support and the good cooperation References [1] Wohlers TT Wohlers report 2014: 3D printing and additive manufacturing state of the industry annual worldwide progress report Fort Collins, Colorado: Wohlers Associates; 2014 [2] Günther D, Heymel B, Günther JF, Ederer I Continuous 3D-printing for additive manufacturing In: Rapid Prototyping Journal, vol 20 Iss 4; 2014 pp 320–327 [3] Eschey C Maschinenspezifische Erhöhung der Prozessfähigkeit in der additiven Fertigung München: Utz; 2013 [4] Kellner I Materialsysteme für das pulverbettbasierte 3D-Drucken München: Utz; 2013 [5] Zäh MF Wirtschaftliche Fertigung mit Rapid-Technologien: AnwenderLeitfaden zur Auswahl geeigneter Verfahren München: Carl Hanser; 2006 [6] Gibson I, Rosen DW, B Stucker B Additive manufacturing technologies: Rapid prototyping to direct digital manufacturing New York: Springer; 2010 [7] Schulze D Pulver und Schüttgüter: Fließeigenschaften und Handhabung, 2nd ed Berlin, Heidelberg: Springer Berlin Heidelberg; 2009 [8] Lee S, Sachs E, Cima M Layer position accuracy in powderǦbased rapid prototyping, Rapid Prototyping Journal, vol 1, no 4; 1995 pp 24-37 [9] Schmutzler C, Günther D, and Zäh MF, Minimierung der Deformationen von 3-D-gedruckten Bauteilen München: Utz; 2014 [10] Budding A, Vaneker T New Strategies for Powder Compaction in Powder-based Rapid Prototyping Techniques Procedia CIRP, vol 6; 2013 pp 527–532 [11] Kottlorz C 3D-Druckverfahren für kompakte und mechanisch stabile Formkörper; 2014 [12] Sacharow A Kompensation von Formabweichungen durch adaptive Freiformdeformation der CAD/CAM-Daten Essen: Vulkan; 2013 [13] Schmutzler C, Zimmermann A, Zaeh MF Compensating warpage of 3D printed parts using free-form deformation 48th CIRP Conference on Manufacturing Systems; 2015 703 ... defect caused by collapse of the powder bed Fig Determined deviations caused by powder compaction Collapses of the powder bed occur randomly distributed over the entire build volume, whereby no... by shrinkage of the material during solidification and variances induced by a compaction of the powder during the 3D printing process For the shrinkage a classical compensation by scaling the... improvement of the dimensional accuracy The deviation was reduced by an average of more than 80 % Subtracting the linear portion (shrinkage) from the not compensated rods, the defect caused by powder compaction