Numerical Simulation of Laser Machining of Carbon Fibre Reinforced Composites

http://pib.sagepub.com/Manufacture

Engineers, Part B: Journal of Engineering Proceedings of the Institution of Mechanical

http://pib.sagepub.com/content/224/7/1017The online version of this article can be found at:

 DOI: 10.1243/09544054JEM1662

2010 224: 1017Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering ManufactureR Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng and G C Lim

Numerical simulation of laser machining of carbon-fibre-reinforced composites  

Published by:

http://www.sagepublications.com

On behalf of: 

  Institution of Mechanical Engineers

can be found at:ManufactureProceedings of the Institution of Mechanical Engineers, Part B: Journal of EngineeringAdditional services and information for

  http://pib.sagepub.com/cgi/alertsEmail Alerts:

http://pib.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://pib.sagepub.com/content/224/7/1017.refs.htmlCitations:  

What is This? 

- Jul 1, 2010Version of Record >>

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

Numerical simulation of laser machiningof carbon-fibre-reinforced compositesR Negarestani1*, M Sundar1, M A Sheikh1, P Mativenga1, L Li1, Z L Li2, P L Chu2, C C Khin2, H Y Zheng2,

and G C Lim2

1Manufacturing and Laser Processing Group, The University of Manchester, Manchester, UK2Singapore Institute of Manufacturing Technology, Singapore, Singapore

The manuscript was received on 20 May 2009 and was accepted after revision for publication on 5 November 2009.

DOI: 10.1243/09544054JEM1662

Abstract: The growing use of carbon-fibre-reinforced polymer (CFRP) composites as high-performance lightweight materials in aerospace and automotive industries demands efficientand low-cost machining technologies. The use of laser machining for cutting and drillingcomposites is attractive owing to its high speed, flexibility, and ease of automation. However,the anisotropic material properties of composites, and issues related to the heat-affected zone(HAZ), charring, and potential delamination during laser processing, are major obstacles in itsindustrial applications. In order to improve the quality and dimensional accuracy of CFRP lasermachining, it is important to understand the mechanism of the transient thermal behaviourand its effect on material removal. Based on the ‘element death’ technique of the finite element(FE) method, a three-dimensional model for simulating the transient temperature field andsubsequent material removal has been developed, for the first time, on a heterogeneousfibre–matrix mesh. In addition to the transient temperature field, the model also predicts thedimensions of the HAZ during the laser machining process. Experimental results obtained withsame process variables using a 355nm DPSS Nd:YVO4 laser were used to validate the model.Based on the investigation, the mechanism of material removal in laser composite machining isproposed. The results suggest that the employed FE approach can be used to simulate pulsedlaser cutting of fibre-reinforced polymer composites.

Keywords: finite element modelling, laser cutting, composites, 355 nm DPSS Nd:YVO4 laser

1 INTRODUCTION

It is challenging to develop a process envelop for lasermachining of carbon-fibre-reinforced polymer(CFRP) composites owing to their inhomogeneousproperties and structures. The decomposition/vaporization of the matrix and fibre in CFRP com-posites occurs in different temperature ranges [1].During laser processing, the temperature at themachining front may not reach the vaporizationtemperature of fibres, but it can be significantlyhigher than the degradation or decomposition tem-perature of the polymer. This results in degradationof the polymer matrix around the ablation site. The

large difference in thermal properties between thetwo constituent materials also results in a large heat-affected zone (HAZ) [2, 3]. The high thermal con-ductivity of carbon fibres in particular results insevere thermal damage to these materials duringlaser processing [1]. Simulation of the laser cuttingprocess is therefore required not only to understandthe HAZ formation, but also to improve the lasercut quality with respect to surface quality anddimensional accuracy [4] and to gain a betterunderstanding of the embedded phenomena andmechanisms.

Different analyses have been reported on HAZprediction in one dimension [1] and two dimensions[5, 6] or the removal depth [7, 8] using analytical andnumerical modelling of laser machining of fibre-reinforced polymers (FRPs) in general and CFRPs inparticular. The feasibility of the finite element (FE)method as an alternative modelling approach has

*Corresponding author: Mechanical, Aerospace, and Civil

Engineering, The University of Manchester, D09 Floor D, Pariser

Building, Sackville Street, Manchester M60 1QD, UK.

email: [email protected]

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

1017

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

been reported in two previous publications. Chenand Cheng [9] developed an FE model to determinethe HAZ during cutting of Kevlar, glass, and carbonfibre composites based on the material propertiesand cutting parameters. They applied a prescribedtemperature–time history to the reference node. Thematerial removal was not modelled and heat loss wasnot considered. Newnham and Abrate [10] presenteda general two-dimensional (2D) formulation for FEanalysis (FEA) of heat transfer in CFRP composites.Their model was based on anisotropic (depending onthe fibre orientation relative to the cutting direction)and homogeneous (volume fraction average of thematrix and fibre) material properties. Their modelshowed a complex temperature distribution near thelaser beam but did not include the material removal.The diversity of the factors involved in modelling ofcomposite machining processes, especially thoseassociated with anisotropic material properties andsimultaneous evaporation of the two different mate-rials (fibre and resin), imposes limitations on theanalytical modelling approaches.

In the current paper, a three-dimensional (3D) FEmodel is developed for predicting the transient tem-perature field together with the subsequent materialremoval during laser machining of CFRP composites,for the first time. A commercial FE program (ANSYS)is used for this purpose utilizing its ParametricDesign Language. For the first time, the modelrepresents a heterogeneous mesh (fibre and matrix)for the composite with anisotropic material proper-ties. The material properties are input according tofindings from thermogravimetric analysis (TGA) andspectrometry analysis on beam absorption, and thetemperature dependency of the thermal conductivity[11] is applied. The numerical approach used enablesefficient prediction of material removal during theprocess in a pulsed moving laser beam. The ablationdepth is hence predicted by the FEA simulation andnot pre-defined. The model also simulates the chipremoval mechanism in CFRPs (using consequentialspaced scanning of the material by the laser beam[12]) for the first time. The experimental resultsobtained with a 355nm diode pumped solid-state(DPSS) Nd:YVO4 laser, being reported for clean cut-ting of these materials [12], was used to validate theFE results under similar processing parameters.

2 EXPERIMENTAL SETUP

An Avia-X high-power Q-switched third-harmonicNd:YVO4 DPSS system with a wavelength of 355 nmwas used for the experiments. The output beamprofile was Gaussian in shape. All experiments wereperformed at a frequency of 40 kHz, pulse duration of25ns, and a maximum average output power of 10W.

The laser beam was delivered to the workpiece usinga galvanometric scanner with an f-theta lens achiev-ing a spot size of 25mm and 0.3mm depth of focus.The focal plane of the laser beam was set at the sur-face of the workpiece that was mounted on a com-puter numeric control X–Y–Z table.

Fully cured [0�/90�]2 (i.e. 0.3mm thick) and [0�/90�]6(i.e. 1mm thick) CFRP composite laminates wereused for the experiments. The volume fraction ofcarbon fibres (7mm in diameter) was 0.60 and theresin was Nelcote� E-765 epoxy. The experimentalinvestigation showed that the quality of the processdeteriorated for the 1mm thick samples if a singletrack was used. This was due to the small spot size ofthe beam (25mm) that resulted in a high-aspect-ratiocut which required a high number of passes to cutthrough the material. This not only affected thematerial removal rate (MRR) but also increased theaccumulated heat effect, which increased the qualitydefects. The multiple tracks strategy [12] was there-fore adopted for processing the 1mm thick samplesto improve the MRR and the quality.

The 0.3mm samples were machined with multiplepasses on a single track of 30mm long cut (referred toherein as single-line cutting). Figure 1(a) presents asimplified sketch of the material removal procedureduring single-line cutting. In each pass some pene-tration occurs until eventually the material is cutthrough. For the 1mm thick samples, on the otherhand, multiple passes on two linear tracks of 30mm

(a)

(b)

spacing

1st laser path

2nd laser path

chip removal region

Parent material

Through Cut

1st Pass 2nd Pass 3rd Pass nth Pass

Fig. 1 Strategies used for laser machining: (a) sketch oflaser beam scanning on a single track with multiplepasses, i.e. single-line cutting; (b) sketches of laserbeam scanning on two tracks with multiple passes,i.e. double-line cutting

Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture JEM1662

1018 R Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng, and G C Lim

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

long cut (two-line cutting; referred to herein asdouble-line cutting), were performed as depictedschematically in Fig. 1(b). The second track wasscanned with different spacing from the first track toinvestigate the scanning spacing effect on chip for-mation. As depicted in Fig. 1(b), in some of theregions between the tracks that gain enough energyto disintegrate the matrix (but not enough to vaporizethe fibres), fibre chips are formed (shown as thedash lined region in Fig. 1(b)). This chip removalmechanism improves the MRR as well as reducingthe thermal defects [12].

3 FE MODELLING

3.1 Procedure and assumptions

The two explained processing cases, namely single-line cutting and double-line cutting, were studiedusing FEA. The mesh geometry was restricted to thecomputation time and space requirements. In thecurrent study it was aimed to configure a realisticmodel of the material with actual prediction of theablation depth. This necessitated modelling of ashallow depth of the experimental sample, enoughfor analysing a single pass of the beam. The length ofthe mesh was also restricted by the number of ele-ments adaptable to the modelling time and its sta-bility. Therefore, in the following initial trials, thedimensions of the mesh geometry used in the modelwere adopted as 300mm · 100mm· 50mm. A lengthof 300mm was selected which allowed analysis of thechip removal mechanism in double-line cutting andof the HAZ configuration in single-line cutting. Single-line cutting was modelled for analysis of the effect ofscanning speed while double-line cutting was mod-elled for analysis of the effect of scanning spacing onthe chip removal mechanism. For symmetry, single-line cutting was modelled in the middle of the meshlength (i.e. distance of 150mm from the mesh edge)and the scanning speed varied. The double-line cut-ting, on the other hand, was modelled by varying thebeam path spacing (i.e. 75, 100, 150, and 200mm)from the edge of the mesh at a constant scanningspeed (i.e. 100mm/s) to investigate the character-istics of chip formation. The edge of the mesh wasconsidered as the cut in the first laser beam path (seeFig. 1(b)) due to the limitation of the computationdomain. The depth of 50mm, on the other hand, wasthe optimum to analyse the actual prediction of theablation depth after a single pass both for the single-and the double-line cutting. The assumptions andstrategies in the model included:

(a) the laser beam scan direction is linear and per-pendicular to the fibre orientation;

(b) the fibre distribution is considered to be uniformbased on the volume fraction;

(c) a single pass on a single track of laser machiningis considered;

(d) beam scattering inside the groove and its inter-action with the vapour are ignored.

3.2 Geometric model and FE mesh

From microscopy of the cross-section of the com-posite lamina, the fibre arrangement was simplifiedto be uniform with a diameter of 7mm and a spacing(filled with the resin) of 1mm for the modelled 50mmdepth of material. The voids were neglected owing totheir low content, i.e. 2 per cent [13]. Figure 2 showsthe 3D mesh generated for the laminate with regionsrepresenting fibres and resin.

3.3 Material properties

An important problem in numerical simulation ofcomposites is the lack of available property data.Every attempt has been made to determine and userealistic material properties for the model. The tem-perature of decomposition for the material wasdetermined experimentally from TGA. As shown inFig. 3, the material decomposition starts at 593K,with carbon loss starting at 773K. At 1155K,96 per cent weight loss is observed. This was assumedas the total weight loss for the simulation. The heatreleased or absorbed from the epoxy decomposition,nitridation, and oxidation was ignored. The coeffi-cient of beam absorption of the material at the inci-dent wavelength was determined by reflectivityspectrum analysis on the samples using a Jena Ana-lysis Specord 250UV spectrophotometer. For 355 nm(in line with the experimental system) the reflectivitywas found to be 7 per cent. High ultraviolet absorp-tion of the polymer resin [14] contributes to the low

Fig. 2 FE mesh used for analysis

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

Numerical simulation of laser machining of carbon-fibre-reinforced composites 1019

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

value of reflectivity. Other properties of the fibres andepoxy used for modelling purpose are reported inTable 1 [5,15].Moreover, in order to generate a realisticmodel, temperature-dependent anisotropic thermalconductivity of the material was used (Table 2). Gen-erally, the thermal conductivity of CFRP decreases astemperature increases [11]. This can be attributed tothe carbon fibres, where the phonon–phonon scatter-ing path, as the dominant conduction mechanism, isinversely proportional to the temperature at mediumto high temperatures [16]. On the other hand, beingartificial graphite and hence having hexagonal crystal-line structure, carbon fibres exhibit a 2D layer structurewith anisotropic thermal conductivity [17]. The valuesapplied for the carbon fibres followed the heat flowtrends recommended in reference [17], while for theepoxy the trend in reference [18] was used.

3.4 Governing equations and solution strategy

A transient thermal problem was solved with thermalloading applied according to the laser pulse shapeand number of pulses. The number of pulses incidentat each beam spot was calculated according to thescanning speed and pulse length, i.e. 25 ns. The‘element death’ methodology (available in ANSYS)was used for simulating the material removal by

ablation. An element with a temperature higher thanthe decomposition temperature of the resin or fibrewas considered to represent material that had beenablated. Such an element was considered to be dead,with insignificant effect in subsequent analysis. Thegoverning equations for the problem have beenestablished [19] as follows.

3.4.1 Time-dependent heat conduction equationunder the irradiating surface

riCi@T n; tð Þ

@t¼ @

@nKi

@T n; tð Þ@n

� �þQ tð Þ 1� Rð Þ; ði ¼ f;mÞ

ð1Þwhere r, C, and K are density (kg/m3), specific heat(J/(kgK)), and thermal conductivity (W/(mK)),respectively, and indices f and m refer to fibre andmatrix. n is the normal vector, R is the beam reflec-tion, Q is the constantly distributed heat flux (W/m2),and t is time in seconds.

3.4.2 Boundary conditions

At the top surface where the laser heat flux is appliedand heat losses are considered

Ki@T

@n¼ Q tð Þ 1� Rð Þ � h T � T1ð Þ; ði ¼ f;mÞ

ð2ÞAll surfaces except the groove side and the top sur-face are considered adiabatic. Hence, convection

0

20

40

60

80

100

120

273 473 673 873 1073 1273Temperature (K)

)%(thgie

W

593K

773K1155K

Fig. 3 TGA results for the used CFRP in air

Table 1 Properties of the CFRP used for the analysis[5, 15]

Property Fibre Epoxy

Volume fraction 0.60 0.40Density (kg/m3) 1800 1200Thermal conductivity at ambienttemperature (W/(mK))

50 0.1

Specific heat (J/(kgK)) 710 1884Decomposition temperature (K)* 1153 698

*Decomposition temperature was applied according to the TGAresults.

Table 2 Anisotropic temperature-dependent thermalconductivity, K (W/(mK)), of carbon fibre andepoxy resin as used in the model

Temperature(K)

Carbon fibre [17] Epoxy matrix [18]

Kx Ky¼Kz Kx¼Ky¼Kz

298 50.00 36.06 0.100323 48.79 35.76 0.155350 46.97 34.55 0.148373 45.45 33.64 0.140400 43.94 32.42 0.135473 39.70 29.18 0.130500 38.48 28.09573 35.15 25.64600 33.94 24.73673 30.91 22.76700 29.94 22.09773 27.45 20.42800 26.52 19.82873 24.33 18.39900 23.58 17.85973 21.73 16.48

1000 21.06 16.001073 19.61 14.931100 18.35 14.011173 17.09 13.081200 15.83 12.16

Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture JEM1662

1020 R Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng, and G C Lim

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

heat loss on the groove side depending on the inter-face substrate follows

Ki@T

@n¼ �h T � T1ð Þ; ði ¼ f;mÞ ð3Þ

while the adiabatic surfaces follow

Ki@T

@n¼ 0; ði ¼ f;mÞ ð4Þ

In the above, T, T1, and h denote the cell tempera-ture, ambient temperature, and heat transfer coeffi-cient, respectively. The air convection coefficient andambient temperature are taken as 50W/(m2K) and293K, respectively.

4 RESULTS AND VALIDATION

Composites 0.3mm and 1mm thick were selected tostudy the effects of scanning speed and scanningspace, respectively. Therefore, two FEA simulationcases were studied to understand the effect of speedfor single-line cutting (Fig. 1(a)) and the effect ofspacing for double-line cutting (Fig. 1(b)), as shownin Table 3. Case I was to study the effect of speed onHAZ, ablation depth, and total number of passesrequired for a through cut. Case II was to find theeffect of spacing in removing the material. In thelater case, the speed was kept constant at 100mm/s,which was selected by experimental trials that pro-vided acceptable thermal damage. Other processparameters (frequency of 40 kHz, pulse width of25nm, and laser power of 10W) were kept constantthroughout the analysis.

4.1 Effect of laser scanning speed

The effect of laser scanning speed was studied duringmachining of the 0.3mm thick composite with mul-tiple scans on a single track (Fig. 1(a)). Figure 4 showsthe temperature distribution and the correspondingablation depth for various speeds of 50, 100, 200, and800mm/s. It can be seen from Fig. 4 that a severeHAZ was present in the direction of the fibres. This is

due to the higher thermal conductivity of the fibrescompared with the resin. However, it can be observedfrom Figs 4(c) and (d) that for higher speeds heatdissipation through the fibres was sufficient to causedecomposition/vaporization of only the adjacentpolymer matrix over a short distance beyond thelaser-machined region.

As illustrated in Figs 4(a) and (b), at scanningspeeds of less than 200mm/s severe thermal damagewas predicted (exceeding the mesh geometry).Figure 5(a) illustrates the experimental result at thescanning speed of 50mm/s, showing high thermaldamage to the cut edge. As the scanning speedincreased the model predicted less thermal damage(Figs 4(c) and (d)), in line with the experiments(Figs 5(b) and (c)). At 800mm/s speed although therewas a HAZ, it was not sufficient to cause much of thepolymer matrix to disintegrate, leading to a shorterfibre pull-out (Fig. 4(d)). This predicted a clean cut-ting process; i.e. short fibre-pull out at 800mm/sscanning speed agreed well with the experimentalresults where for the same process parameters onlyshort fibre pull-out (< 30mm) was observed (seeFig. 5(c)). The decrease in fibre pull-out withincreasing scanning speed is due to a reduction in theinteraction time at higher speeds, which results inshorter heating phases thereby reducing the thermalinput at each point along the cut path to cause fibrepull-out through heat conduction. Moreover, fromFig. 4, the ablation depth was predicted to decreasewith an increase in the scanning speed. This wouldlead to an increase in the number of passes requiredfor a through cut, which again agreed with theexperimental findings [12]. Also seen from Fig. 4 isthe irregular length of fibres and irregular tempera-ture profile along the sides of the wall. This is due tothe difference in distance that the beam is trans-mitted between subsequent pulses at each of thescanning speeds, which exhibit individual heatingand cooling cycles per beam position. As can bejudged, the increase of scanning speed reduced theirregularities. This is a consequence of a reduction inaccumulated heat from subsequent pulses as thespeed increases and agreed with a similar behaviourobserved in the experiments [12].

A comparison of the extent of fibre pull-out pre-dicted by the FE method and the averaged experi-mental results is shown in Fig. 6. Fibre pull-out isconsidered as the region in which the temperatureexceeds the decomposition temperature of the poly-mer matrix but remains less than the vaporizationtemperature of the fibre. As shown in Fig. 6, at the lowrange of speed, i.e. 50–200mm/s, the model predic-tion for fibre pull-out exceeded the mesh boundary(shown as the dashed line in the FEA results) andhence the exact prediction of temperature distribu-tion was not achieved. The FEA results showed less

Table 3 Process parameters used for the study

Simulation no. Case Speed (mm/s) Scan spacing (mm)

1 Case I 50 N/A2 100 N/A3 200 N/A4 800 N/A5 Case II 100 756 100 1007 100 1508 100 200

N/A, not applicable.

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

Numerical simulation of laser machining of carbon-fibre-reinforced composites 1021

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

deviation from the experimental results as the scan-ning speed increased. This highlights the difficultiesof considering in the model various phenomena thatare included in the real-life situation particularly atlower scanning speeds (which show higher heat

accumulation and e.g. enhance the effect of tem-perature dependency of the material properties).

Figure 7 compares the ablation depth predicted byFEA with the experimental results. In the lower speedrange, longer interaction time led to accumulation of

(a) (b) (c)

Fig. 5 Scanning electron micrographs showing experimental results at different scanning speeds:(a) 50mm/s; (b) 200mm/s; (c) 800mm/s

b()a( )

d()c( )

htapmaeB

htapmaeB

htapmaeB

htapmaeB

Fig. 4 HAZ profile after one track for various scanning speeds: (a) 50mm/s; (b) 100mm/s; (c) 200mm/s;(d) 800mm/s

Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture JEM1662

1022 R Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng, and G C Lim

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

heat, which resulted in rapid material removal. Onthe other hand, in the high speed range the heatinput decreased (i.e. number of pulses per beamposition) and consequently less thermal distortionand less material removal were found. As can beobserved from Fig. 7, in contrast to fibre pull-outanalysis (see Fig. 6), the FEA results agreed well withthe experimental results. For low scanning speeds themodel provided closer predictions to the experi-mental results, while at higher speeds, i.e.200–800mm/s, the simulated result diverged slightlyfrom the experiments.

4.2 Effect of spacing between laser scans

As discussed in section 2, for cutting thick samples amultiple tracks strategy with appropriate spacingbetween consecutive passes was considered toenable effective material removal and minimize the

thermal effect on the samples. As shown in Fig. 1(b),during the first laser track the composite wasmachined to some depth, and the next track wasperformed with some spacing from the first track.During the second track the composite wasmachined to some depth; at the same time the resinbetween the two tracks was fully decomposed andultimately the chip produced between the two tracks(although the fibre was not fully decomposed) wasremoved. This spacing between the two laser trackshas a significant effect on the MRR and the numberof passes required for machining. It is important tounderstand the effect of spacing on chip formationand also to predict the optimal spacing which willdecompose the resin between the two laser trackswith less thermal damage. In this study, the scanningspeed was considered as 100mm/s, as in the experi-ments, and four different spacing parameters (75,100, 150, and 200mm) were considered. The distancewas considered from the groove side, which wasassumed to have been formed in the previous lasertrack.

Figure 8 shows the FEA results of laser machiningfor various values of spacing distance from the grooveside. It can be seen from Fig. 8 that a higher tem-perature was predicted along the groove side than atthe block side for distances less than 150mm. This isdue to better heat dissipation through the block sideto the parent material by conduction and compara-tively less heat transfer from the groove side to theatmosphere by convection. As depicted in Fig. 8, thelength of the fibre chips on the groove side increasedwith spacing and reached an optimum at 150mmspacing. Decomposition of resin on the groove sidedid not occur when the spacing distance increased to200mm (Fig. 8(d)) and the formation of fibre chipswas not predicted between consecutive laser tracks.A comparison of FEA and experimental results on theablation depth through chip formation at differentscanning spaces is given in Fig. 9. Here, although theFEA results underestimated the ablation depth, itshowed the same trend as the experiments. Thisproves the capability of FE modelling to incorporatesimilar predictions of the experimental situation indouble-line processing.

5 DISCUSSION

The present modelling simulates laser machining ofcomposites by incorporating realistic mesh geo-metry, an advanced numerical procedure for estab-lishing the material removal process, and anisotropicthermal conductivity. During the numerical simula-tion the material removal process is updated at theend of every time step according to the TGA diagramdiscussed in section 3.3. Compared with previous

0

10

20

30

40

0 200 400 600 800

Scanning speed (mm/s)

)mµ(

ssaprephtped

noitalbA

ExperimentsFEA simulation

Fig. 7 FEA prediction of ablation depth after a single passcompared with experimental results

0

30

60

90

120

0 200 400 600 800

Scanning speed (mm/s)

)mµ(tuollup

erbiF

ExperimentsFEA simulation

FEA mesh length limit

Fig. 6 Predicted HAZ after a single pass compared withexperimental results (the dashed line for FEA refersto fibre pull-out exceeding the mesh geometry)

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

Numerical simulation of laser machining of carbon-fibre-reinforced composites 1023

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

numerical models this simulation addresses theactual situation in the machining process. In pre-vious FE models the material removal was usuallysimplified or not included. To the authors’ knowledgethis study is the first to show the realistic ablationprofile in composite machining by a laser.

The mesh used in the present model consists ofseparate fibres and resin. It is more realistic com-pared with previous approaches and it reflects theshape and size of actual composites, while previousmodels usually considered a simple uniform mesh

with anisotropic material properties which do notrepresent the reality. This complex mesh con-secutively increased the number of elements to392 475, which forced the model to be limited to asmall portion of the actual machining condition.Although the model does not reflect the wholedomain, the FEA results provide a fairly good insightinto the phenomena of composite machining, givingresults close to the experimental ones.

The thermal conduction of CFRP is anisotropic anddepends on various factors such as the fibre direction,its geometry, and volume fraction [20] (compositesbeing heterogeneous), while it exhibits high depen-dency on temperature and decreases significantly atelevated temperatures (typically above 773K).Although anisotropic thermal conductivity has beenconsidered previously [10], experimental verificationof such consideration, incorporating its temperaturedependency in FEA simulations, has not been repor-ted for laser machining of composites. In order tosimulate the real-life situation, the current analysisconsiders anisotropic thermal conductivity by incor-porating its variation with temperature. Such a con-sideration enables the model to consider thefibre–matrix interface role during the thermal process.

Using the FE method, provision of reliable thermalpredictions during laser processing of compositematerials depends on various factors. Among these isthe accurate modelling of heat propagation into the

Fig. 8 Predicted material removal for various values of spacing distance: (a) 75mm; (b) 100mm;(c) 150mm; (d) 200mm

0

10

20

30

40

50

70 90 110 130 150

Scanning spacing distance (µm)

)mµ(

htpednoitalb

A

ExperimentsFEA simulation

Fig. 9 FEA predicted and experimental ablation depth atdifferent scanning spaces in double-line cutting

Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture JEM1662

1024 R Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng, and G C Lim

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

material, which requires the adoption of an accuratemodelling strategy that recognizes the complexitiesrelating to thermal damage to these materials. Ther-mal diffusivity and thermal conductivity play impor-tant roles in the propagation of heat within amaterial. As shown in Fig. 4, the FE model predictsthat fibres rapidly conduct heat away, leaving a largerHAZ as compared with the extent of fibre pull-out.This enables sufficient heat for decomposition of thesurrounding matrix to some extent, which is propor-tional to the interaction time. It also shows that for agiven laser power the model predicts the morphologyof the laser-machined sample to be strongly influ-enced by the scanning speed. In particular it has beenfound that, at low velocities, the samples are char-acterized by fibres leaning out of the matrix sur-rounded by a large HAZ with loss of material. As thecutting velocity increases, this structure tends todisappear. The increase in speed also reduces the kerfwidth and thermal damage to the composite due tothe aforementioned reason.

The complex structure and composition of com-posites makes it difficult to find or calculate theirmaterial properties. This is one of the reasons for thedeviation in experimental and FEA modelling results.As shown in Fig. 6, the model predicts a reduction inthe fibre pull-out as the speed increases. The com-putational time and stability restricted the meshgeometry input which would provide the real fibrepull-out predictions in the low speed range, i.e.50–200mm/s, of the model. The overestimation sug-gests that a more complicated formulation of theprocess as well as the material properties is requiredfor more accurate modelling. The ablation depth isalso overestimated in the FEA model as comparedwith the experiments (Fig. 7). However, it shows closeagreement with the trend of the experimental results,which implies the capabilities of FEA modelling inthe field. The ablation depth shows high sensitivity tothe scanning speed over the lower ranges of speed(50–200mm/s) while smoother results are obtainedfor the higher speed ranges (200–800mm/s).

The process of chip formation, which is the mainfactor involved in double-line laser machining ofthick composites, is predicted in the FE model, asillustrated in Fig. 8. As can be seen from Fig. 1(b), thedouble-line processing would in an ideal situationlead to control of the HAZ and effectively use the HAZto dissociate the matrix molecules and remove theintact fibre chips in between the processing tracks.The model predicts fibre chip formation for all spa-cings less than or equal to 150mm. In other words, amaximum of spacing of 150mm is required betweentwo consecutive laser tracks to decompose the resinbetween them and consecutively remove the fibrechip. This coincides with the experimental finding ofa spacing of three to five times the beam spot dia-

meter (in the range of 150mm) being required foreffective chip formation in multiple track processing[12].

Moreover, as the spacing decreases below 150mm,the model predicts more severe thermal damage tothe material (Figs 8(a) and (b)). This refers to theundesired heat accumulation between the processinglines due to insufficient space for heat dissipationinto the material and causing severe damage, parti-cularly in 75mm spacing (Fig. 8(a)), where no chipformation is predicted. All the material between thetwo tracks in this case, i.e. 75mm spacing, is pre-dicted to be vaporized in agreement with the experi-ments. As a notable feature of the model, the lengthof the fibre chips is predicted to increase in propor-tion to the scan spacing up to 150mm, above whichthe materials (fibre and resin) between consecutivelaser passes are not removed and subsequently lasermachining by the chip removal mechanism fails(Fig. 8(d)).

An advantageous feature of the model is successfulprediction of fibre pull-out, i.e. intact fibres withdecomposed matrix. Figure 10 compares typical fibrepull-out in the experiments (for the purpose ofclarity, the surface after a through cut is depicted)and the FEA modelling result at similar processingparameters. This capability enables the model toeffectively predict the chip removal mechanism indouble-line processing. For 150mm as the optimumdistance, the typical length of fibre chips in theexperiments is about 100mm. This represents thetrend of the modelling results, where the predictedlength of fibre chips exceeds the mesh geometry asshown in Fig. 8(c). A comparison of a typical chipformed during the experiments at 150mm spacingdistance and the FEA modelling result of the similarcase is given in Fig. 11.

The model is capable of predicting more effectivematerial removal in the double-line processing case.As is shown in Fig. 9, the ablation depth variation ispredicted with similar trend and in reasonably closeagreement to the experiments. The uncertainty of theresults can be attributed to factors such as vapourformation, vapour–beam interaction, and beamscattering inside the groove that were not consideredin the FEA model.

Generally the FEA is capable of predicting similartrends to the experimental result for both single-lineand double-line cutting. Despite all attempts tosimulate the material as close to the real-life situationas possible, the simulation overestimates the fibrepull-out while underestimating the ablation depth.Although the difference between the model and theexperiments for ablation depth in single-line cuttingis small (Fig. 7), for the double-line cutting the dif-ference is larger (Fig. 9). The wide differences both inthe length of fibre pull-out and the ablation depth

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

Numerical simulation of laser machining of carbon-fibre-reinforced composites 1025

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

reflect the complexities involved in modelling a meshthat can fully satisfy the specifications of a compositestructure. Ignoring other phenomena in the model,such as laser beam interaction with the plume andbeam absorption by the groove wall, also contributesto the wide difference between the experimentalresults and the FE simulations.

6 CONCLUSIONS

A 3D FE model of heat flow and material removalduring ultraviolet laser machining of CFRP compo-sites has been developed using more a realistic for-mulation of the process conditions and materialproperties. Temperature distribution and ablationdepth at different laser beam scanning speeds andspacing distance have been predicted. The results ofFEA simulations showed similar qualitative andphysical agreement with the experiments. HAZ and

ablation depth were predicted to be more sensitive tospeeds in the lower range of 50–200mm/s as com-pared with the higher range of 200–800mm/s. Parti-cular phenomena involved in the experiments suchas burning at low scanning speeds (50mm/s), fibrepull-out, chip formation, and clean cuts at the highscanning speeds (800mm/s) were successfully pre-dicted in the FEA model. The study revealed thefeasibility of applying FEA modelling for simple(single-line cutting) and more complicated (double-line cutting) machining of CFRP composites usingheterogeneous meshing, in particular at scanningspeeds higher than 200mm/s. For the double-lineprocessing, the FEA model predicted an optimalspacing of 150mm between the two tracks whichclosely matches the experimental results. The FEAfindings showed better quality machining at the highscanning speeds and for the double-line processingcase, where fibre pull-out was less important com-pared with that in the single-line processing.

(a) (b)

Fig. 11 (a) FEA predicted chip formation compared with (b) typical chip formed in the experiments at150mm scan spacing in double-line processing

(a) (b)

Fig. 10 (a) Effective modelling of fibre pull-out in the FEA compared with (b) the surface morphology ofthe experimental result at a scanning speed of 50mm/s

Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture JEM1662

1026 R Negarestani, M Sundar, M A Sheikh, P Mativenga, L Li, Z L Li, P L Chu, C C Khin, H Y Zheng, and G C Lim

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from

Discussion of the results of the fibre pull-out andmatrix recession show the importance of includingthe fibre–matrix interface features in similar models.The other factors that should be considered for theaccuracy of the model include vapour formation,material–vapour interaction (its role on heat inputcalibration), beam absorption at the groove wall, aswell as polymer matrix decomposition and fibrevaporization temperatures, mechanisms that are notfully understood at extremely high heating rate oflaser machining.

ACKNOWLEDGEMENTS

The authors would like to thank the support offeredby the Agency for Science, Technology and Research(A*STAR) of Singapore and the joint support of theEngineering and Physical Sciences Research Council(EPSRC) and the Technology Strategy Board (TSB),UK under the ELMACT grant DT/E010512/1.

� Authors 2010

REFERENCES

1 Tagliaferri, V., Di Ilio, A., and Visconti, I. C. Laser cut-ting of fibre-reinforced polyesters. Composites, 1985,16(4), 317–325.

2 Dell’Erba, M., Galantucci, L. M., and Miglietta, S. Anexperimental study on laser drilling and cutting ofcomposite materials for the aerospace industry usingexcimer and CO2 sources. Compos. Mfg, 1992, 3(1), 14–19.

3 Aoyama, E., Inoue, H., Hirogaki, T., Nobe, H.,Kitahara, Y., and Katayama, T. Study on small diameterdrilling in GFRP. Compos. Struct., 1995, 32(1–4), 567–573.

4 Cenna, A. A. andMathew, P. Evaluation of cut quality offibre reinforced plastics – a review. Int. J. Machine ToolsMf., 1997, 37(6), 723–736.

5 Pan, C. T. and Hocheng, H. The anisotropic heat-affectzone in the laser grooving of fibre-reinforced compositematerial. J. Mater. Process. Technol., 1996, 62, 54–60.

6 Uhlmann, E., Spur, G., Hocheng, H., Leibelt, S., andPan, C. T. The extent of laser-induced thermal damageof UD and crossply composite laminates. Int. J. MachineTools Mf., 1999, 39, 639–650.

7 Chryssolouris, G., Sheng, P., and Choi, W. C. Investi-gation of laser grooving for composite materials. Ann.CIRP, 1988, 37(1), 161–164.

8 Sheng, P. and Chryssolouris, G. Theoretical model oflaser grooving for composite materials. J. Compos.

Mater., 1995, 29(1), 96–112.9 Chen, C. C. and Cheng, W.Material properties and laser

cutting of composites. In Proceedings of the 23rd

International SAMPE technical conference, Kiamesha

Lake, New York, USA, 21–24 October 1991, pp. 274–287

(Society for the Advancement of Material and Process

Engineering, Covina, California, USA).10 Newnham, P. and Abrate, S. Finite element analysis of

heat transfer in anisotropic solids: application to man-

ufacturing problems. J. Reinf. Plast. Compos., 1993, 12,

854–864.11 Griffis, C. A., Masumura, R. A., and Chang, C. I. Ther-

mal response of graphite epoxy composite subjected to

rapid heating. J. Compos. Mater., 1981, 15, 427–442.12 Li, Z. L., Chu, H. Y., Lim, G. C., Li, L., Marimuthu, S.,

Negarestani, R., Sheikh, M., and Mativenga, P. Process

development of laser machining of carbon fibre rein-

forced plastic composites. In Proceedings of the 27th

International Congress on Applications of lasers and

electro-optics (ICALEO), Temecula, California, USA,

20–23 October 2008, pp. 222–231 (Laser Institute of

America, Orlando, Florida, USA).13 Nelcote Advanced Composite Materials. E-765 epoxy

prepregs: product overview, 2008, available from http://

www.parkelectro.com/parkelectro/images/E-765.pdf

(access date 9 September 2009).14 Allcock, H. R. and Lampe, F. W. Contemporary polymer

chemistry, edition 2, 1990, p. 624 (Prentice-Hall, Engle-

wood Cliffs, New Jersey).15 Fenoughty, K. A., Jawaid, A., and Pashby, I. R.

Machining of advanced engineering materials using

traditional and laser techniques. J. Mater. Process.

Technol., 1994, 42, 391–400.16 Savage, G. Carbon–carbon composites, 1993 (Chapman

& Hall, London).17 Ho, C. Y., Powell, R. W., and Liley, P. E. Thermal con-

ductivity of the elements: a comprehensive review. J.

Phys. Chem. Ref. Data, 1974, 3(1), 1–796.18 Sundqvist, B., Sandberg, O., and B€ackstr€om, G. The

thermal conductivity of an epoxy resin at high pressure

and temperature. J. Phys. D: Appl. Phys., 1977, 10, 1397–

1403.19 Carslaw, H. S. and Jaeger, J. C. Conduction of heat in

solids, edition 2, 1959 (Oxford Science Publications,

Oxford, UK).20 Halpin, J. C. Primer on composite materials: Analysis,

1984 (Technomic Publishing Company, Lancaster,

Pennsylvania, USA).

JEM1662 Proc. IMechE Vol. 224 Part B: J. Engineering Manufacture

Numerical simulation of laser machining of carbon-fibre-reinforced composites 1027

at Loughborough Univ of Tech on May 9, 2012pib.sagepub.comDownloaded from