854086__913012697

Embed Size (px)

Citation preview

  • 8/13/2019 854086__913012697

    1/15

    PLEASE SCROLL DOWN FOR ARTICLE

    This article was downloaded by:

    On: 5 February 2011

    Access details: Access Details: Free Access

    Publisher Taylor & Francis

    Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-

    41 Mortimer Street, London W1T 3JH, UK

    Virtual and Physical PrototypingPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t716100703

    New model for shrinkage compensation in selective laser sinteringK. Senthilkumarana; Pulak M. Pandeya; P. V. M. RaoaaDepartment of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi, India

    To cite this ArticleSenthilkumaran, K. , Pandey, Pulak M. and Rao, P. V. M.(2009) 'New model for shrinkage compensationin selective laser sintering', Virtual and Physical Prototyping, 4: 2, 49 62

    To link to this Article DOI 10.1080/17452750802393659URL http://dx.doi.org/10.1080/17452750802393659

    Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

    This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.

    The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.

    http://www.informaworld.com/smpp/title~content=t716100703http://dx.doi.org/10.1080/17452750802393659http://www.informaworld.com/terms-and-conditions-of-access.pdfhttp://www.informaworld.com/terms-and-conditions-of-access.pdfhttp://dx.doi.org/10.1080/17452750802393659http://www.informaworld.com/smpp/title~content=t716100703
  • 8/13/2019 854086__913012697

    2/15

    New model for shrinkage compensation in selective laser sintering

    K. Senthilkumaran, Pulak M. Pandey* and P. V. M. Rao

    Department of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi, India

    (Received 9 May 2008; in final form 5 August 2008)

    This paper presents a new model for shrinkage and a new approach for shrinkage

    compensation to enhance the accuracy of parts produced by selective laser sintering (SLS)

    a solid freeform fabrication process. The present prevailing approach as proposed by

    machine manufacturers is simple but not accurate. A new shrinkage model which

    accounts for part geometry as well as beam offset is proposed in this work. A new

    compensation scheme which accounts for nonlinear shrinkage is proposed, implemented

    and validated. The proposed compensation scheme compensates for shrinkage at everylayer and at every hatch length, unlike a uniform compensation scheme applied to entire

    part. A new algorithm which accounts for this is developed and implemented.

    Experiments carried out with the new shrinkage model as well as with the new

    compensation scheme have shown significant improvement in the accuracy of the parts

    produced which establishes the effectiveness of the proposed methodology.

    Keywords: rapid prototyping; selective laser sintering; shrinkage compensation; beam

    offset; scan length

    Notation

    a Beam offset for hatch line (mm)

    DCAD Desired dimension of CAD model (mm)

    DCSM Dimensions obtained using machine manufacturer

    suggested shrinkage compensation factor (mm)

    DLSM Dimensions obtained using shrinkage models (mm)

    Ia Percentage improvement in accuracy (%)

    L Compensated length (mm)

    Lc Original Hatch length (mm)

    Lm Measured length (mm)

    Ls Actual sintered length (mm)

    LX Hatch length along Xdirection

    LY Hatch length along Ydirection

    s Percentage shrinkage (%)SX Shrinkage in Xdirection (%)

    SY Shrinkage in Ydirection (%)

    SZ Shrinkage in Z direction (%)

    Dc Compensation length (mm)

    Dl Deviation from nominal dimension (mm)

    1. IntroductionSelective laser sintering (SLS) is one among many different

    rapid prototyping (RP) processes where the parts are built

    in a layer by layer fashion. Recently, SLS has gained

    importance due to its ability to process a wide variety of

    materials. SLS can produce functional prototypes and rapid

    tooling components, which necessitates the production of

    high-quality parts. The majority of its applications are in

    aerospace and rapid tooling, where high accuracy levels

    have to be met in order to ensure proper functional

    requirement. Due to this new constraint imposed on SLS

    in terms of part quality, there is a need to study the process

    in detail and to improve part accuracies.

    Among all types of process-related errors, the effects

    caused by shrinkage are found to have major influence on

    the accuracy of parts produced. Materials exhibit shrinkage

    during thermal cycles, which varies from material to

    material. The shrinkage of crystalline polymer is found to

    *Corresponding author. Email: [email protected]

    Virtual and Physical Prototyping, Vol. 4, No. 2, June 2009, 4962

    Virtual and Physical PrototypingISSN 1745-2759 print/ISSN 1745-2767 online # 2009 Taylor & Francis

    http://www.tandf.co.uk/journalsDOI: 10.1080/17452750802393659

  • 8/13/2019 854086__913012697

    3/15

    be higher than that of amorphous polymers (Venuvinod

    and Ma 2004). The SLS process is always accompanied by

    shrinkage as a result of thermal and phase change effects. In

    practice, the shrinkage is experimentally calibrated and the

    same is compensated during data preparation stage of rapid

    prototyping. The purpose of compensation seems to be that

    of reducing the inaccuracies due to shrinkage and not to

    attempt to eliminate those (Yang et al. 2002). However,notwithstanding the complexity of shrinkage, the basic

    approach used today in most of the RP processes involves

    some form of simple shrinkage compensation. Before

    making the part, users are allowed to set a constant

    shrinkage compensation factor, which is different for

    different direction to overcome the shrinkage effect (Tong

    et al. 2003). The shrinkage compensation factor is found by

    fabricating standard test specimens and deriving a linear

    relationship between the nominal dimensions and fabri-

    cated part dimensions. Here when dealing with shrinkage

    compensation, a fundamental assumption usually made is

    that shrinkage is orthotropic and the shrinkage compensa-

    tion scaling factors are constant for all X, Y and Zdimensions. However, this constant scaling factor is not

    realistic because of the dynamic nature of the SLS process

    and varying layer geometry in part building. Therefore in

    this paper it is proposed to compensate nonuniform

    shrinkage in sliced layer files rather than compensating

    the STL files.

    Jacobs (2000) discussed the effects of shrinkage variation

    on the accuracy of rapid tooling inserts. He found that a

    noise component is present along with mean percentage

    shrinkage for most of the RP processes, and termed it as

    random noise shrinkage. He mainly concentrated on

    random noise shrinkage and found that standard deviation

    of the random shrinkage is directly proportional to the

    mean process shrinkage. He also found that nonuniformity

    in shrinkage is mainly attributed to geometry of the part.

    He concluded that key to accuracy and repeatability of such

    techniques is the reduction of mean process shrinkage to

    the smallest possible level.

    Zhuet al. (2006) studied the shrinkage behavior in metal

    powders. They quantified the two shrinkages namely

    thermal shrinkage and sintering shrinkage. They found

    that in-plane shrinkage (X and Yshrinkage) is very small

    compared to shrinkage in the build direction. The sintering

    shrinkage was mainly caused by densification and is a kind

    of elastic compressive shortening. They suggested thatthermal shrinkage due to cyclic heating can be reduced by

    controlling process parameters. In their work, the thermal

    shrinkage increases with increase in laser power and

    shrinkage decreases with increase in scan speed and scan

    spacing.

    Ning et al. (2006) considered the effect of geometry on

    the shrinkage of the metallic parts. They introduced the

    speed compensation technique based on the scan length. In

    their method, when building a part, the laser scan speed is

    adjusted dynamically according to the scan length which

    varies with geometric shape of the part. The different scan

    speeds for the scan lengths are chosen based on their

    shrinkage values at different speeds. Wang (1999) discussed

    the issues in calibration of shrinkage and beam offset for

    the SLS process. Expressions for shrinkage and beam offset

    in terms of the nominal diameter and error after sinteringwere developed. They also discussed the effect of part

    weight on the percentage shrinkage.

    Manetsbergeret al. (2003) studied the effect of tempera-

    ture, time and pressure on the shrinkage of polymer parts.

    They used a thermal simulation as a basis for shrinkage

    compensation in SLS process. They expressed shrinkage

    values as a function of temperature and also showed a

    linear dependency to the pressure applied. Yang et al.

    (2002) proposed compensation test pieces for the X, Yand

    Zaxes to compensate for the shape distortions caused by

    phase changes during the sintering process and they

    measured shrinkage rates experimentally. With those

    shrinkage rates, a set of equations is proposed which givesthe scale factors of the X, Yand Z axes. The scale factors

    obtained from the compensation test pieces ofX, Yand Z

    axes satisfy the required dimensional accuracy even if there

    are changes in the build positions and in the size of the SLS

    parts. Their work is mainly focused on the study of the

    location of the part on the part bed and does not aim at

    shrinkage variation with process parameters.

    Ning et al. (2005) conducted a series of experiments for the

    direct metal laser sintering (DMLS) process to find the effect

    of hatch length on the material anisotropy, heterogeneity and

    part strength. They concluded that short hatch lines cause

    serious shrinkage and the part becomes less homogeneous.

    They proposed an algorithm to find out optimal hatch

    direction for a typical layer by considering the shrinkage as a

    function of hatch length. Ragunath and Pandey (2007)

    studied the effect of process parameters on the process and

    material shrinkage. They found that scan length influences

    shrinkage in theXdirection. They also predicted that scaling

    factors can have a linear relationship with scan length

    (Figure 1). They derived empirical relations for percentage

    shrinkage in terms of scan length using Taguchi method.

    However they used scaling factors based on the maximum

    dimensions not on the individual scan lengths.

    Several attempts have been made to improve the accuracy

    of RP parts made by other processes like stereolithographyand fused deposition modeling (FDM) by controlling the

    effect of shrinkage. Dao et al. (1999) calculated shrinkage

    compensation factors for FDM parts with varying lengths.

    They observed that mean error after shrinkage compensa-

    tion follows a linear trend and is increasing for increasing

    nominal dimensions. Also the residual errors are found to

    be scattered due to lack of process stability. They attributed

    this trend to the noise shrinkage which is not compensated

    50 K. Senthilkumaran et al.

  • 8/13/2019 854086__913012697

    4/15

    by the scaling factors used. Wang et al. (1996) investigated

    the relationship between post-cure shrinkage and thevarious process parameters for stereolithography by using

    least-square method. They concluded that, as the curing

    degree of the green-state prototype increases, the shrinkage

    encountered reduces. They also found that the curing

    degree is a function of laser power, layer pitch, scan pitch

    and scanning speed. Wenbin et al. (2005) presented finite

    element simulation for stair stepping effect caused by

    material shrinkage in a new process called step less rapid

    prototyping which combines stereolithography and conven-

    tional milling. They reported that the layers had a small

    initial expansion before a large shrinkage. They concluded

    that the light intensity does not have significant effect on

    staircase control. The decrease in layer thickness affects thestair stepping significantly.

    It is evident from the literature review presented above

    that part accuracy is affected highly by shrinkage. Some

    studies (Manetsberger et al. 2003, Wenbin et al. 2005) use

    simulation as a basis for shrinkage compensation. However

    due to the dynamic nature of the process, a realistic

    simulation of shrinkage occurring in the process using

    FEM is often difficult. In most of these approaches which

    use standard test specimen, the assumption is that shrink-

    age is constant. Two important research issues identified are

    nonuniform shrinkage along every Y direction and its

    dependency on geometry and these two are not explicitly

    studied in any previous works.

    2. Nonuniform shrinkage

    The total shrinkage in the SLS process is due to material

    shrinkage, process shrinkage and thermal shrinkage. Dur-

    ing polymer crystallisation, the molecules arrange them-

    selves and occupy less volume thus leading to material

    shrinkage. During processing the powder particles fuse

    together to produce dense parts, leading to a decrease in

    porosity and volume. During heating, the part expands due

    to the co-efficient of thermal expansion and then shrinks

    during cooling. Shrinkage is strongly influenced by the laser

    parameters, powder bed properties, cooling rate and

    geometry. Therefore the total shrinkage b (Shi et al. 2004)

    is expressed as a sum of three shrinkages.

    bbtbsbc (1)

    where bt is the thermal shrinkage, bs is the sintering

    shrinkage and bc is the crystalline shrinkage. The linear

    thermal shrinkage bt (Zhu et al. 2006) can be written in

    terms of the process parameters.

    bta

    AP

    Vh

    1

    rcpLt

    L (2)

    where a is the thermal expansion co-efficient, A is

    absorbtivity of the powder bed, P is the laser power, V is

    the velocity of the laser beam,his the hatch spacing, r is the

    density of the powder bed,cpis the specific heat capacity of

    the powder material, Lt is the layer thickness and L is the

    original scan length.

    The crystalline shrinkage occurring during cooling can be

    highly nonuniform along each direction due to strong

    temperature gradient inside the powder bed. There can be

    expansion-shrinkage behavior during the time history of

    sintering. A particular layer can shrink nonuniformly due to

    its position, i.e., high or low temperature regions.

    In order to understand the effect of geometry on

    shrinkage in SLS process, a brief introduction to the

    compensation procedure in injection molding process is

    essential. In injection molding process, different wall

    thicknesses of the part produces different shrinkage.

    When wall thickness increases, more time is required to

    cool the centre of the thicker wall. As polyamide cools more

    slowly, there is more time for crystallisation and stress

    relaxation leading to higher crystallisation and higher

    shrinkage (Jansen et al. 1998). The shrinkage values are

    specified either as a normalised value (mm/mm) or as

    percentage values to the original part dimensions. Different

    percentage shrinkage values are specified for different wall

    thicknesses for a polyamide material. ASTM D955-00 is an

    American standard for measurement of shrinkage in

    molded plastics which specifies that shrinkage values differ

    in directions along the flow to the direction across the flowand it should be measured separately for semi-crystalline

    materials like polyamide (Fisher 2003).

    It is interesting to note that the shrinkage factor is not a

    constant value but it varies according to the geometry in

    many similar processes. Moreover, there is a directional

    effect considering the direction in which laser scanning

    takes place and the direction perpendicular to it. In this

    work, an attempt is made to study the effect of geometry i.e.

    20 30 40 50 60 70 800

    0.2

    0.4

    0.6

    0.8

    1

    Scan length in mm

    Percentage

    Shrinkage(%)

    cX Ls 009647.00302.1 =

    Figure 1. Scan length and percentage shrinkage relation

    (Ragunath and Pandey 2007).

    51Virtual and Physical Prototyping

  • 8/13/2019 854086__913012697

    5/15

    hatch length on shrinkage in the direction of the laser

    scanning and in a direction orthogonal to scan direction.

    3. Development of shrinkage model

    The details of the experimentation for developing a new

    shrinkage model are discussed in next section. As a

    shrinkage model relating hatch length and percentage

    shrinkage needs to be developed, the shrinkage calibration

    part should contain strips of different length to capture the

    effect of scan length. Figure 2 shows some of the common

    geometries used by researchers and RP machine manufac-

    turers for in-plane shrinkage calibration. In order to

    accommodate longer dimensions than older shrinkage

    models a shrinkage calibration part with large variations

    in scan length as shown in Figure 3 is designed. Moreover,

    two shrinkage models are developed by considering the

    direction of scanning. The first model is developed for the

    shrinkage along the scanning direction (X direction) and

    the other across the scanning direction (Ydirection).

    The length of the strips in shrinkage calibration part

    varies from 20 to 200 mm. In order to develop models for

    shrinkage in the Xand Ydirection, two parts were fabricated

    in each case to check the repeatability in each direction.Specimens used in the study were fabricated using

    PA2200 which is a modified nylon 12 developed for use in

    SLS machines by EOS GmbH, Germany. The material is

    semi-crystalline in nature. The material used was refreshed

    powder and the ratio of mixing is 70% used powder and

    30% virgin powder. The schematic of the equipment used is

    shown in Figure 4. It uses a combination of two bins and a

    slot feeder to dispense powder onto the platform. Extra

    powder is collected through two bins situated on both sides

    of the removal chamber. Moreover the environment main-

    tained in the study includes the processing chamber and the

    removal chamber. Nitrogen purging is done and infra red

    heat radiators maintain the required temperature of both

    environments. All the parts are oriented and placed at the

    centre of the platform for building.

    Maximum power available with CO2 laser in the SLS

    machine used for the study is 58 W. Only 62% (36 W) of the

    maximum power is used in the present experiment. This is

    mainly due to the fact that curling is observed at higher

    laser powers. A maximum scan speed of 4500 mm/s is

    chosen for minimum build time. The laser spot size is kept

    around 0.6 mm. The energy density used in the experiment

    is 26.66 KW/m2. The process parameters used for contour

    are low laser power and scan speed compared to hatching in

    order to achieve a good surface finish. If the part is not

    allowed to cool in a controlled environment for long time,

    the part tend to warp due to faster cooling in the outside

    environment. During faster cooling significant stresses

    develop causing post-build warpage. So the part is allowed

    to cool inside the platform for five hours. There are four

    types of exposure strategy available with the current SLS

    equipment used for the study. The scanning lines aregenerated and exposed (1) along Xdirection only, (2) along

    Ydirection only, (3) exposure along both Xand Ydirection

    and (4) along Xdirection in one layer and along Yin next

    layer (parallel toXand Yon alternate layers). The exposure

    strategy used in the present study has scan lines only along

    Xdirection (type 1).

    There are considerations in fabricating shrinkage calibra-

    tion parts using different process parameters for contour

    Figure 2. Some shrinkage calibration parts found in the literature: (a) part used by Wang (1999); (b) part used by Yanget al.

    (2002); (c) part used in EOS method (EOS 2003); (d) part used in DTM method (Pham and Dimov 2001).

    52 K. Senthilkumaran et al.

  • 8/13/2019 854086__913012697

    6/15

    lines and hatch lines (Figure 5). Mostly the contour lines

    are scanned with a low laser power and high beam speed as

    compared to the hatch lines. For the present experiments,

    the contouring option is switched off to accurately capture

    the shrinkage behavior of the hatch lines. The diameter of

    the sintered zone is usually larger than the laser diameter

    and is called as spot diameter. In order to compensate the

    dimensional error due to spot diameter, the laser beam

    should be offset from the boundaries of the cross section of

    the object and is called beam offset. The estimation of these

    beam offset values for the SLS process is described in Wang

    (1999). In the present work, in order to estimate the beam

    offset for the hatch lines, a part is designed with dimensions

    250.66 mm. The dimensions of this part are chosen

    such that when fabricated, this part will have single line

    exposure of laser beam. The design, arrangement and

    fabricated specimen are shown in Figure 6. The process

    parameters used for this experiment is listed in Table 1. The

    thickness of the single hatch line part is measured and its

    value is found to be 0.555 mm. No beam compensation is

    applied while fabricating specimens. It is incorporated while

    calculating percentage shrinkage.

    Figure 3. Shrinkage calibration specimen: (a) design (b) length of strips (c) arrangement in platform and (d) fabricated

    specimen.

    Figure 4. Schematic of the SLS process.

    53Virtual and Physical Prototyping

  • 8/13/2019 854086__913012697

    7/15

    The percentage shrinkage is calculated from the following

    relation:

    sLc a Lm

    Lc a100 (3)

    where Lm is the measured dimension of the part after

    sintering and cooling and a is the beam offset.

    The exposure strategy used for shrinkage calibration

    experiments is type 1. The process parameters used for thisexperiment are given in Table 1. The dimensions of the

    strips in calibration specimen are measured and the

    percentage shrinkage is calculated using equation (3). The

    percentage shrinkage is plotted against nominal dimensions

    of the strips. The results showed good repeatability of the

    shrinkage pattern between the two specimens fabricated in

    each direction. The orientation of the specimen and the

    results of these experiments are presented in Figure 7.

    Percentage shrinkage increases with nominal dimensions for

    strip lengths between 20 to 160 mm and becomes constant

    after 170 mm. Also significant amount of expansion rather

    shrinkage is found between strip lengths 20 to 110 mm. As

    discussed earlier, there is a shrinkage-expansion behavior

    during time-history of sintering. Between the strip lengths

    20 to 110 mm, the expansion dominates the shrinkage. Thisexpansion or increase in dimension is attributed to the fact

    that sintering is unconstrained. In most of the SLS systems,

    the laser scans the top surface of a heated powder bed to

    form the area enclosed by contours of the layered object in

    raster scan mode (hatching) in combination with the

    outlining of the cross sections of the part in vector scan

    Figure 5. Exposure strategy showing contouring and hatching (type 4).

    Figure 6. Beam offset calibration parts (a) single line exposure (b) design (c) arrangement and (d) fabricated specimens.

    54 K. Senthilkumaran et al.

  • 8/13/2019 854086__913012697

    8/15

    mode (contouring). In hatching, the laser scans across the

    powder surface in one dimension, turning the laser on and

    off at the boundaries of the contour. The complete length of

    laser scan in one dimension is dependent upon of the size

    and shape of contours dictated by part geometry. As the

    exposure strategy used for experimentation had only

    hatching lines and no contour lines, the expansion during

    sintering is not constrained by the contour lines as shown in

    Figure 8. A more detailed hypothetical explanation for this

    in-plane expansion is dealt by Zhu et al. (2006).

    Moreover, the pattern of shrinkage obtained shows that

    the percentage shrinkage is lower for lower scan lengths and

    higher for higher scan lengths unlike the pattern obtained in

    literature (Ning et al. 2005, 2006, Jacobs 2000). This

    reversal of trend is mainly due to the fact that lower strip

    length receives less energy in the present experiments

    compared to works reported in literature. The length of

    scan during which acceleration and deceleration of galvano

    mirrors takes place, while the laser takes turn at the

    boundaries, is compensated in the present work and the

    laser is switched off while scanning these lengths and hence

    no exposure. This length of unexposed compensated region

    is independent of part size, shape and location. This method

    produces less energy in smaller strip lengths unlike the

    earlier approaches and produce less densification and

    shrinkage.

    In addition, the scans are incremented in a manner suchthat the present scan overlaps the previous scan to obtain

    structural integrity of the part. Accordingly, the time

    between adjacent and overlapping scans of the two hatch

    lengths will widely vary according to the different lengths of

    scan. Such variation in this time is the prime cause for

    greater temperature deviation, inconsistent sintering rates

    leading to nonuniform shrinkage. Also larger length of scan

    will have enough time for sintering to happen and produce

    more shrinkage. In order to compensate this nonuniform

    shrinkage, a shrinkage model is developed relating percen-

    tage shrinkage and length of scan along the Xdirection.

    Figure 9 shows the shrinkage pattern for calibration

    specimens oriented along the Ydirection of the platform. It

    is to be remembered that this pattern is obtained with parts

    where the scanning direction is still parallel to the X

    direction and only the variation in length of strips is in Y

    direction. The percentage shrinkage increases with increase

    in strip length until the strip length is 110 mm and then

    starts to decrease with increase in strip length. Unlike the

    trend in the X direction, there is no expansion which is

    Table 1. Process parameters set for the experiments.

    Parameter Value

    Laser power (W) 36

    Hatch spacing (mm) 0.3

    Scan speed (mm/s) 4500

    Part bed temperature (oC) 176

    Hatching Parallel to X

    Figure 7. (a) Orientation of specimen; (b) percentage shrinkage versus strip length alongXdirection.

    55Virtual and Physical Prototyping

  • 8/13/2019 854086__913012697

    9/15

    found in the specimens. As the first scan line itself acts as a

    barrier to expansion during sintering along the positive Y

    direction, the sintering is usually partially constrained in theY direction of the platform while scanning direction is

    parallel to the X direction. More crystalline structures are

    formed when the material is kept longer in the temperature

    range of 140 to 160oC. The shrinkage of the material will be

    higher if the material is undergoing a slow cooling in this

    temperature range. Moreover, as the width of the specimen

    is small, the time interval between two consecutive scan

    lines is not sufficiently long for the surface to cool down;

    the temperature in the region will gradually build up,

    resulting in higher temperature and longer liquid-phase

    time. This will cause greater gradients in temperature to the

    surrounding and a relatively faster cooling than the longer

    length strips (Ning et al. 2006). Therefore the material

    shrinkage is lesser for smaller strip lengths as the part cools

    rapidly allowing minimum time to form the crystallinestructures.

    The magnitude of shrinkage in the Xand Ydirections is

    less compared to shrinkage along the build direction (Z

    direction). Apart from that, the process parameters that

    affect the in-plane shrinkage does not affect the Zdirection

    shrinkage. For example, the effect of scan length on the Z

    shrinkage is negligible (Ragunathet al.2006). In addition to

    that nonuniformities in shrinkage in the Z direction

    depends upon the placement of the part in the build

    chamber, i.e. a part placed at the bottom of the platform

    will have more shrinkage than the same part placed at the

    top of the build chamber due to the weight of the powderacting upon it. It is observed from the authors previous

    studies that part bed temperature, and scan spacing mostly

    influence the Z shrinkage (Ragunath and Pandey 2007). So

    in this study shrinkage models are developed for Xand Y

    directions only and for compensating Zdirection shrinkage

    a constant shrinkage factor is calculated using the method

    suggested by the machine manufacturer (SZ3.5%).

    Unconstrained sintering without contour exposure

    Constrained sintering with contouring and hatching

    Figure 8. Expansion in X direction owing to exposure

    strategy

    Figure 9. (a) Orientation of specimen; (b) percentage shrinkage versus strip length alongYdirection

    56 K. Senthilkumaran et al.

  • 8/13/2019 854086__913012697

    10/15

    For fitting a model, to the percentage shrinkage as a

    function of scan length, the data from the experiment

    conducted using a specimen shown in Figure 3 with

    parameters listed in Table 1 is used. A model is fit between

    percentage shrinkage and scan length in both the Xand Y

    directions using curve fitting toolbox in MATLAB. The

    fitted shrinkage model for X and Ydirections (SX and SY

    in%) are given below

    SX0:292428:5

    LX(4)

    SY3:12105L2Y0:007271LY0:09478 (5)

    Figure 10 shows the fitted curves for the shrinkage in X and

    Ydirection. The coefficient of determination (Montgomery

    20 40 60 80 100 120 140 160 180 200

    -1.2

    -1

    -0.8

    -0.6

    -0.4

    -0.2

    0

    0.2

    0.4

    0.6

    Strip length (mm)

    Percentageshrinkage(%)

    X direction

    Y direction

    Negative shrinkage

    Figure 10. Comparison of shrinkage model in X and Y

    direction.

    Figure 11. Illustration of new compensation method: (a) STL file; (b) slicing; (c) contour showing hatch lines; (d) offsetting

    hatch lines based on shrinkage model; (e) comparison of original and final compensated geometry.

    57Virtual and Physical Prototyping

  • 8/13/2019 854086__913012697

    11/15

    and Runger 2007) for the fitted shrinkage model in the X

    andY direction are 98.67% and 81.67% respectively.

    4. Proposed compensation procedure

    When the shrinkage scaling factors are constant, shrinkage

    compensation is relatively easy. The scaling transformation

    can be used to offset the vertices of the triangles of the STL

    file to the single scaling factor value in each X, Yand Z

    direction. However if the shrinkage scaling factors vary

    with hatch length, shrinkage compensation requires off-

    setting each scan line of the part. Therefore a sliced layer

    file is preferred rather than a STL file for the ease of

    compensation. Compensating on sliced data is proved

    successful by Tong et al. (2008) for compensating machine

    errors for SLA and FDM processes.

    In order to achieve this, the STL file of the part is sliced

    into layers and is stored in a layered file format called CLI

    file. In order to generate the scan lengths, entire scan path

    of the laser has to be calculated. The contour information is

    extracted and all the co-ordinate points of the contour are

    sorted based on whether the contour is internal or external

    contour. After reading the contour information hatch lines

    and its lengths are found using intersection of hatch lines

    with the contours. Then compensation value for every hatch

    length is calculated using the fitted shrinkage models. Hatch

    lengths are compensated at its two ends. After offsettingeach contour at its end points, the contour is rebuilt with its

    new vertices and a new compensated CLI file is written.

    Then this file is used for part building.

    Figure 11 illustrates this procedure with an example of a

    sphere. The STL file of the sphere (Figure 11(a)) is sliced

    with uniform slice thickness of 150 mm (Figure 11 (b)).

    Hatch vectors (dexels) are generated with a hatch gap of 300

    mm for circular slice geometry (Figure 11(c)). The hatch

    vectors along Xdirection corresponds to scan path of the

    laser and the hypothetical hatch vectors along Ydirection

    are perpendicular to these scan lines. It should be remem-

    bered that there is no exposure on these hatch lines along

    the Ydirection and these are virtual hatch lines generated

    for the purpose of the compensation.

    Most of the degenerated cases in calculating intersections

    of hatch vector with contour geometry are solved with thehelp of odd/even parity checking. The generated hatch

    vectors are compensated depending upon their hatch

    length. When a dexel with length Lc is fabricated, due to

    linear shrinkage, the dexel length reduces to Lm. The

    deviation due to shrinkage of original length Lc is Dl.

    From Figure 11(d), it is observed that in order to obtain a

    length Lc a dexel with length L has to be used. Thus the

    compensated value is LLcDc, where DcsLc1 s

    is the

    compensation length to be added to every dexel of the part.

    For a part having scan length Lc, Dc is calculated using the

    shrinkage model and appended at the two ends of the dexel

    by the amountDc/2. First, the dexels in the Xdirection arecompensated using shrinkage model (SX) developed in the

    X direction (equation (4)). Then compensated contour is

    reconstructed from the end points of the hatch vectors.

    Following this dexels in Ydirection for this new compen-

    sated contour are generated and compensated using shrink-

    age model (SY) developed for the Ydirection (equation (5))

    and then compensated contour is reconstructed from the

    end points of the hatch vectors. The comparison of original

    and compensated contour is illustrated in Figure 11(e). All

    the compensated contours are then written into a CLI file

    for fabricating the slices.

    Unlike the conventional compensation techniques, the

    compensation length varies nonlinearly with the originalhatch length. As percentage shrinkage is a function of hatch

    length and during compensation it is again multiplied by the

    hatch length, the amount of compensated length varies

    nonlinearly with hatch length. Figure 12 compares the

    original and compensated geometry for a semi-circular

    section with a magnification in compensation length. The

    implementation of this compensation methodology for X

    directionis illustrated in theformof a flow chart in Figure 13.

    Thesame procedure is repeatedfor thecompensation in theY

    direction. This new method of compensation can be used for

    any four types of exposures mentioned earlier in section 3.

    But for each type of exposure new shrinkage models need to

    be developed following the steps outlined in section 3.

    5. Results and discussion

    The shrinkage compensation algorithm and the developed

    shrinkage models are verified by conducting experiments.

    Improvement in accuracy is quantified by fabricating parts

    following two different compensation methods:

    -20 0 20 40 60 80 100 120 140 160 180 200-20

    0

    20

    40

    60

    80

    100

    120

    140

    160

    180

    200

    Compensated geometry

    Original geometry

    Y

    X

    Figure 12. Comparison of geometry before and after

    compensation using the developed shrinkage model.

    58 K. Senthilkumaran et al.

  • 8/13/2019 854086__913012697

    12/15

    . The first part is compensated by applying a scaling

    transformation to STL file with a constant shrinkage

    scaling factor (calculated using method suggested by

    SLS machine manufacturer (EOS 2003)).

    . and the other part is compensated using the shrinkage

    model and new compensation method

    The improvement in accuracy for the model developed in

    this work for both the Xand Ydirection is studied using a

    standard benchmark part as suggested by Ippolito et al.

    (1995). A comparison experiment is conducted to quantify

    the improvement in accuracy due to the inclusion of the

    nonuniform shrinkage term. In one part a machine

    Start

    Input CLI file

    Read layers, contours, vertices and their attribute

    Find the max and min values of x & y for each contour

    Generate Hatch lines

    Find Intersection points of hatch

    lines with contours

    Sort the intersection points in ascending order of X co-ordinates

    Generate scan lines from the intersection points

    Offset scanline by c

    Write new contour data from compensated scan line

    Output compensated CLI file

    Stop

    IF hatchline> H

    Initialise first layer & contour

    Increment layer_no

    Increment

    contour_no

    If layer_no