investigation of deviations caused by powder compaction during 3d printing

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

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

Procedia CIRP 57 ( 2016 ) 698 – 703

ScienceDirect

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

© 2015 The Authors Published by Elsevier B.V

Peer-review under responsibility of Scientific committee of the 49th CIRP Conference on Manufacturing Systems (CIRP-CMS 2016)

Keywords: 3D Printing, Powder compaction, Deviations, Compensation

1 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

2 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

© 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 scientifi c committee of the 49th CIRP Conference on Manufacturing Systems

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 1

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 1 (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 1 - 2 mm The defect also

affects commercial components, cf Fig 2

3 Description of the deviation

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 Addition-ally, 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 z-position 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 5 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 3 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

Fig 2 Defect caused by a collapse of the powder bed Fig 3 Description of the deviation as a combination of a linear and a

-2,5 -2,0 -1,5 -1,0 mm

0

z-position

-7

-6

-5

-4

-3

-2

-1

0

0 100 200 300 mm 500

z-position

defect caused by collapse

of the powder bed

Fig 1 Determined deviations caused by powder compaction

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

4 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 z-rods 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 z-coordinate In Fig 4 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

-3,0 -2,5 -2,0 -1,5 -1,0 mm

0

z-position

series placed at the botton series placed at the top approximated function Fig 4 Comparison between the two staircase-shaped series

5 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 1 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

signifi-cant 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 5

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:

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

The distribution of the maximal deviation over the x-y-plane

of the building chamber is shown in Fig 6 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 1

Fig 6 Distribution of the maximal deviation of the quadratic proportion Fig 5 Comparison of several measurements

-4,5

-4,0

-3,5

-3,0

-2,5

-2,0

-1,5

-1,0

-0,5

mm

0,5

z-position

1 trail

2 trial (1 measurement)

2 trial (2 measurement)

measured values quadratic portion linear portion

6 Compensation

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

mid-point 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 1 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:

Table 1 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]

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 1 and 5 a higher

statistical spread of the measured values was observed

Therefore, the results within the zones 2-4 are analyzed first

Fig 7 compares the measurement data of the z-rods

compensated according to table 1 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

pre-deformed 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 1 (middle), an extrapolation

of the trend from the zones 2-4 was used due to the high

statistical spread, cf section 5 The measured deviations of the

produced rods, pre-deformed and uncompensated, show a

similar trend compared to the rods in zone 2-4 and the pre-deformed 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 5 (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 8 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

Fig 8 Evaluation of compensating the quadratic portion only -2,5

-2,0 -1,5 -1,0 -0,5 mm

0,5

z-position

compensated not compensated

-2,5 -2,0 -1,5 -1,0 -0,5 0 mm

1,0

z-position

(not compensated) (not compensated) (not compensated) Fig 7 Evaluation of the compensation in zone 2, 3 and 4

Parallel to this investigation a case study with two

benchmark geometries was conducted, using the parameters

according to table 1 The objective was to verify the behavior

of the continuous powder compaction and the developed

compensation procedure Thus, a not compensated and a

pre-deformed benchmark part were produced, cf fig 9 The

benchmark object was positioned 10 mm above the bottom and

in two different zones (zones 4 and 5) due to a size of

approximately 186 x 177 x 155 mm³

Therefore, a transition of the compensation parameters from

zone 4 to zone 5 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 8 As observed at the test rods, an overcompensation was

also determined Nevertheless, a reduction of the deviation of

up to 50 % could be achieved

7 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

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

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: Anwender-Leitfaden 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

Fig 9 Evaluation of the compensation by refence to the benchmark objet