ARTICLE IN PRESS

0890-6955/$ - se

doi:10.1016/j.ijm

�CorrespondE-mail addr

International Journal of Machine Tools & Manufacture 47 (2007) 996–1002

www.elsevier.com/locate/ijmactool

Direct laser fabrication of thin-walled metal parts underopen-loop control

Li Penga,�, Ji Shengqinb, Zeng Xiaoyanb, Hu Qianwub, Xiong Weihaoc

aSchool of Materials Science and Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100083, PR ChinabState Key Lab of Laser Technology, Huazhong University of Science and Technology, Wuhan 430074, PR China

cState Key Lab of Die & Mould Technology, Huazhong University of Science and Technology, Wuhan 430074, PR China

Received 7 November 2005; accepted 21 June 2006

Available online 22 August 2006

Abstract

Direct laser fabrication (DLF) is an advanced manufacturing technology, which can build full density metal parts directly from CAD

files without using any modules or tools. The investigation on the fabrication of thin-walled parts of nickel alloy using open-loop DLF

process is introduced in this paper. The experimental setup consisted of a CO2 laser, a 3-axis CNC table, a coaxial powder nozzle and a

powder recycler. The 3D-CAD file of a thin-walled metal part was converted into the STL file format and imported into software HUST-

RP to generate ‘pseudo-random’ scanning paths of laser beam. The influence of process parameters on the build height of thin-walled

metal parts was studied by 1–10 layered single-bead stacks of nickel alloy. The result shows that the interference factors which affect the

build height of thin-walled metal parts occur randomly during the process. For open-loop DLF process, thin-walled metal parts can

achieve much better shape quality if the process parameters are suitable. Multilayer single-bead walls were built up with different

scanning velocity to obtain the optimal process parameters of thin-walled parts of nickel alloy. It shows that thin walls of nickel alloy

with uniform height can be built up layer by layer in a certain range of specific energy. However, it is difficult to control the build height

of complex thin-walled metal parts in an accurate manner just using optimal parameters. A special coaxial powder nozzle was designed in

this paper. In a certain range, the deposition thickness of the nozzle is nearly linearly increased with increase in the standoff distance

between the powder focusing point of the nozzle and the deposition substrate. By means of the nozzle, a novel method to control the

build height of thin-walled metal parts using open-loop DLF process was introduced. The difference in build height of a thin-walled part

can be compensated automatically in one or several layers during the process. It is proved that the build height of a thin-walled metal

part can be accurately controlled in theory using the nozzle. A complex single-bead part of nickel alloy whose geometry was designed to

be the well-known Chinese ‘FU’ was fabricated and explained in this paper. The result shows that the shape quality of the sample is quite

good, and actual build height of the sample is 53.54mm while the designed value is 54mm.

r 2006 Elsevier Ltd. All rights reserved.

Keywords: Laser; Nickel alloy; Direct laser fabrication (DLF); Specific energy; Thin-walled metal part

1. Introduction

Direct laser fabrication (DLF) is an advanced manu-facturing technology, which developed from laser claddingand rapid prototyping techniques. During the fabricationprocess, metal powder is injected into the molten pool toform laser claddings and layers along designed filling paths.Because the successive claddings and layers are metallurgicfusion bonded to each other, DLF technology can fabricate

e front matter r 2006 Elsevier Ltd. All rights reserved.

achtools.2006.06.017

ing author. Tel.: +8610 82339349.

esses: peng.lee@tom.com, lip@buaa.edu.cn (L. Peng).

near net-shape metal parts with complex geometry. As anemerging direct manufacturing process, hence, DLFtechnology is a very attractive research field in the lastyears [1].DLF technology has broad applications in die and

mould industry, aviation industry, aerospace industry,energy industry, etc. The fabrication of thin-walled metalparts is a very important research topic for DLFtechnology, especially in aviation and aerospace industry.For example, integral rib structures with large pockets orstructures with high aspect ratio are the most commonfeatures which can be found in the aircraft applications

ARTICLE IN PRESSL. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–1002 997

because these structures can greatly lighten the aircraftwith no damage to the desired strength. Usually, theproduction sequence of complex thin-walled metal parts islong and using traditional manufacturing processes wastemuch material, whereas the DLF process saves themachining time by up to 80% and the overall cost by20–50% [2]. With fast localized cooling of the molten pool,parts with thin walls and high depth-to-diameter aspectratios can be easily fabricated using DLF process. Incontrast, thin-walled structures are difficult to CNCmachining if the depth-to-diameter aspect is more than10:1 [3].

Control of the deposit thickness or build height of thin-walled metal parts is a critical issue since it impacts thequality of the products. In DLF, however, a large numberof parameters govern the fabrication process; theseparameters are sensitive to the environmental variationsand influence each other. Hence, it is difficult to achieveaccurate controls of shape quality using open-loop DLFprocess. Many researches about the closed-loop controlhave been reported in the recent years to improve the DLFprocess [4–6]. In these reports, the real-time sensing andcontrol of build height of metal thin walls is the mostcommon example to prove the effect and necessity ofclosed-loop control to the DLF process.

Despite improving the height precisions of thin-walledparts, closed-loop control system make the hardware ofDLF process be more complicated and cost intensive. Infact, excellent thin-walled metal parts with high precision

HUST-IIIpowderfeeder

ROFIN TR050CO2 laser

coaxial nozzle

3-axis HNC-1

CNC unit

recycler

G-code

focusing mirrorbox

laser beam

x-y tablepowder

z-axis

Fig. 1. Schematic of experimental setup.

Table 1

Basic process parameters for the fabrication of thin-walled parts

Nickel alloy powder, character

Carrying gas, flow rate

Substrate

could also be achieved under open-loop control if rightprocess parameters and special methods were introducedinto the DLF process. In this paper, an investigation on thefabrication of thin-walled metal parts made at the StateKey Lab of Laser Technology, Huazhong University ofScience and Technology (HUST) is discussed in detail.

2. Experimental procedures

The experimental setup for the present work is shown inFig. 1. A HUST-III powder feeder, a coaxial nozzle and a5 kW ROFIN continuous wave CO2 laser (power output0–5000W) were used to deliver and melt metal powders.A 3-axis HNC-1 CNC unit was used to control themovement of x–y table and the raise of the z-axis. In orderto recycle un-melted powders, a powder recycler wasdesigned. The 3D-CAD file of a part was converted intothe STL file format and imported into software HUST-RPto generate the scanning paths of laser beam. The wholesetup is a typical open-loop DLF system.Single-bead walls and parts is good investigating case for

the fabrication of thin-wall metal parts because it is thebasic geometrical feature of thin-walled parts and experi-ences critical processing conditions. In order to find out therelationship between the build height of a thin-walled partand the layer numbers of deposition, 1–10 layered single-bead stacks of nickel alloy were built with a differentscanning velocity. And then, several multilayer single-beadwalls of nickel alloy were built up with a different scanningvelocity to understand the influence of process parameterson the shape of thin-walled parts. During the process, thetraverses of laser beam were carried out in alternatedirections, and a fixed z-increment was used at the end ofeach traverse. The z-increment value was equal to thedeposition thickness of the first layer. By right of the aboveexperiments, the optimal process parameters with whichsingle-bead walls can be built up stably layer by layer wereobtained.Based on the above experimental results, a single-bead

part of nickel alloy with complex geometry was fabricatedwith the obtained optimal parameters. The geometry of thepart was designed to be the well-known Chinese character‘FU’, which means good fortune in English. In order toobtain accurate thin-walled metal parts with uniform buildheight, a novel method was introduced into the process.Nickel alloy powder was used in the present work, and

the deposition substrate is low-carbon steel plate with10mm thickness. The basic process parameters were shownin Table 1.

Ni–24Cr–4Si–2.5B–3Fe–2.5Mo–0.6C (wt%), �140/+320 mesh

Ar, 8L/min

Low carbon steel plate, 10mm thick

ARTICLE IN PRESS

Fig. 2. Cross-section of single-bead stacks with different scanning velocity and layer number. From left to right, the layer numbers are 1, 2,y,10.

(a) Scanning velocity of 4mm/s. (b) Scanning velocity of 8mm/s.

0 4-1

0

1

2

3

4

5

Sta

ck h

eigh

t (m

m)

Build height Ideal height Height difference

(a)

0.0

0.5

1.0

1.5

2.0

Sta

ck h

eigh

t (m

m)

Build height Ideal height Height difference

(b)

2 6Layer number

8 10

0 42 6 8 10Layer number

Fig. 3. Relationship between build height of single-bead stacks and layer

number. (a) Scanning velocity of 4mm/s. (b) Scanning velocity of 8mm/s.

L. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–1002998

3. Results and discussion

3.1. Influence of process parameters on build height of

single-bead stacks

For a given laser power (2000W), powder feeding rate(5.8 g/min) and diameter of laser spot (2mm), 1–10 layeredsingle-bead stacks of nickel alloy were built with scanningvelocity of 4 and 8mm/s. Fig. 2 shows the cross-section ofthe stacks, and Fig. 3 shows the relationship between thebuild height of the single-bead stacks and the layer number.As shown in Fig. 3, the build height of the stacks increasedwith increase in layer number when the z-increment value isequal to the deposition thickness of the first layer. It alsoshows that the difference between the build height and theideal height of the stacks is random. The ideal height of asingle-bead stack can be determined by the product of thez-increment value and its layer number. It indicates that theinterference factors which affect the quality of thin-walledmetal parts occur randomly during the open-loop DLFprocess. Therefore, it is difficult to precisely predicate thefinal height of a thin-walled metal part under open-loopcontrol without using any special control method. How-ever, we can see that the control of build height withscanning velocity of 8mm/s is much better than that withscanning velocity of 4mm/s. Consequently, thin-walledmetal parts can achieve much better shape quality if theprocess parameters are suitable.

3.2. Influence of process parameters on shape quality of

single-bead wall

Fig. 4 shows the multilayer single-bead walls of nickelalloy fabricated with different scanning velocity valuesfor a given laser power (700W), powder feeding rate(5.8 g/min) and diameter of laser spot (1mm). The walls are20 layers when scanning velocity were 2 and 8mm/s,whereas 30 layers when scanning velocity were 4 and6mm/s. As shown in Fig. 4, build height at both ends of thewall is much lower than that in the middle of it whenscanning velocity is 2mm/s (Fig. 4(a)). As there is anincrease in scanning velocity, the difference in build height

between the ends and the middle of the wall decreasesgradually (Fig. 4(b)). When scanning velocity is 6mm/s, thesingle-bead wall of nickel alloy can obtain uniform heightalong whole scanning track (Fig. 4(c)). As the single-beadwall shown in Fig. 4(d), however, too faster scanning

ARTICLE IN PRESS

Fig. 4. Single-bead walls fabricated with different scanning velocities: (a) 2mm/s, (b) 4mm/s, (c) 6mm/s, (d) 8mm/s.

Velocity of CNC TableSpecific Energy

Start point End point

Sca

nnin

g ve

loci

ty

Spe

cific

Ene

rgy

Fig. 5. Schematic of relationship between specific energy and scanning

velocity along a scanning path.

L. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–1002 999

velocity results in higher build height at both ends of thewall when scanning velocity is 8mm/s.

The variation in build height between both ends andmiddle of a single-bead wall is caused by the accelerationand deceleration of CNC table. Because of acceleration atstart portion and deceleration at end portion in acontinuous path, the specific energy at both ends of a pathis different from that in the middle of it. The specific energyE (J/mm2) can be determined by a simplified equationwhich is as follows:

E ¼P

DS, (1)

where P is laser power (W), D is diameter of laser spot(mm) and S is scanning velocity (mm/s). Fig. 5 shows thevariation in the specific energy along a scanning path, inwhich we can see that the specific energy at both ends of thepath is greater than that in the middle of it because of lowerscanning velocity. For a given laser power, diameter oflaser spot and powder feeding rate, collapse will occur atboth ends of a single-bead wall if scanning velocity is toolow, which will cause the build height at both ends of thewall to be lower than that in the middle of it. On thecontrary, too fast scanning velocity will not cause collapsebut higher deposition thickness at both ends of the wall.Therefore, there is a suitable range of the specific energy inwhich single-bead walls can be built up stably layer by layerunder open-loop control. According to our previous workreported in the literature [7], the optimal specific energy forparts of nickel alloy is 100–200 J/mm2. The above result

shows that the range of specific energy is also acceptablefor the fabrication of thin-walled parts or nickel alloy.

3.3. Fabrication of complex single-bead metal parts

A complex single-bead part of nickel alloy wasfabricated based on the above experimental results. Theparameters for the process were: laser power, 700W;scanning velocity, 6mm/s; diameter of laser spot, 1mm;powder feeding rate, 5.8 g/min, with which single-beadwalls of nickel alloy can be built up stably layer by layer.The geometry of the part was designed to be the well-known Chinese character ‘FU’, which means good fortunein English and the size is 70mm wide and 70mm long. In

ARTICLE IN PRESS

Fig. 6. Thin-walled part of Chinese character ‘FU’ fabricated by open-loop DLF process without using any special control method. (a) Sample and

(b) simulation of scanning velocity.

-1 0 20.0

0.1

0.2

0.3

0.4

0.5

Dep

ositi

o th

ickn

ess

(mm

)

Ar, 4L /min Ar, 8L /min Ar, 14L/min

Standoff distance between powder focusingpoint and substrate (mm)

31

(a)

(b)

Powders PowdersLaser beam

Powder focusing point

Substrate

d

Fig. 7. Influence of standoff distance between powder focusing point of

coaxial powder nozzle and deposition substrate on the deposition

thickness of the nozzle. (a) Schematic of the standoff distance of the

nozzle. (b) Relationship between the standoff distance and the deposition

thickness.

L. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–10021000

order to improve the shape quality of the part, the scanningpaths of laser beam was designed as ‘pseudo-random’ pathsin each layer using the control software HUST-RP.

Fig. 6(a) shows a sample of the part when thez-increment is equal to the deposition thickness of the firstlayer. We can see that the build height on the top layer ofthe sample is not uniform, and some places are abnormallyhigher than other places in the scanning paths. As shown inFig. 6(a), such places are almost at the corners wherescanning direction is suddenly changed. It is also caused bythe change of traveling speed of CNC table. At the sharpcorners in the tool paths, the movement controller of CNCunit has to decrease its vector velocity to change themovement direction. Hence, scanning velocity at thesecorners is lower than normal places, and the thin-walledparts maybe collapse at these places or maybe higher thanother normal places because of the increase of specificenergy. Fig. 6(b) shows the simulation of the variations inmovement velocity along the whole scanning paths, and thebig points in the paths are the corners where scanningvelocity is less than 4mm/s. We can see that these pointsare almost same as those at higher places shown in Fig. 6(a)excepting at one or two points.

If the z-increment value is equal to the depositionthickness of a certain layer, then, it is very difficult toachieve a good thin-walled metal part with uniform buildheight because differences in deposition thickness along thewhole scanning paths will accumulate layer by layer underopen-loop control. In order to eliminate the accumulationof height difference, a coaxial powder nozzle with specialdeposition character was designed in the present work.Fig. 7 shows the variation in deposition thickness of thenozzle along the standoff distance between powderfocusing point of the nozzle and deposition substrate fordifferent flow rate of carrying gas, Ar. The processparameters for the curve were: laser power, 700W;scanning velocity, 6mm/s; diameter of laser spot, 1mm;and powder feeding rate, 5.8 g/min. We can see that thedeposition thickness of the nozzle is sensitive to thestandoff distance if flow rate of carrying gas is suitable(8 L/min). As shown in Fig. 7(b), the deposition thickness

of the nozzle has nearly linear growth with the standoffdistance when the distance is less than 1.0mm. The slope ofthe linear function is equal to 1.0 approximately, and the

ARTICLE IN PRESSL. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–1002 1001

maximum thickness of a deposition is about 0.48mm in therange.

By means of the nozzle with such a special character, wefind a novel approach using which the build height of athin-walled metal part can be well controlled in an accuratemanner under open-loop control. For DLF process, aslight difference in build height usually exists in the toplayer of thin-walled metal parts. If the difference can becompensated in one or successive several layers, the partswill get a uniform height finally. It requires that the rate ofdeposition at higher places should be slower, and that atlower places should be faster than normal places. For anarbitrary position in the top layer of a thin-walled part, asshown in Fig. 8, the standoff distance to the level offocusing point of the nozzle is dn after n layers ofdeposition. Assume that the deposition thickness at theposition in the (n+1)th layer can be stated as

hnþ1 ¼ k dn þ b, (2)

where hn+1 is the deposition thickness, k the slope and b

the intercept. Then, the standoff distance dn+1 in the

-1 0 20.0

0.1

0.2

0.3

0.4

0.5

Dep

ositi

on th

ickn

ess

(mm

)

Standoff distance between powder focusing pointand substrate (mm)

Ar, 8L/min

1 3

Level of powder focusing point

Convexposition

Normalposition

Concaveposition

Thin-walled metal part

Substrate

d1d2

d3

Fig. 8. The difference in build height of a thin-walled metal part can be

compensated automatically in one or several layers by the nozzle.

(n+1)th layer will be

dnþ1 ¼ dn þ Dz� hnþ1 ¼ ðDz� bÞ þ ð1� kÞdn, (3)

where Dz is the z-increment value of DLF process. Ifthe standoff distance at the position in the first layer is d0,Eq. (3) will be

dn ¼ ð1� kÞnd0 þXn�1i¼0

ð1� kÞiðDz� bÞ, (4)

where n ¼ 0, 1, 2, 3,y,. If the standoff distance is d0 andd 00, respectively, for two arbitrary positions with differentbuild height in the first layer, the difference in build heightbetween them in the nth layer will be

d 0n � dn ¼ ð1� kÞnðd 00 � d0Þ. (5)

Eq. (5) explains the transmission formula for thedifference in build height of a thin-walled metal part. Ifthe difference in build height can be eliminated in one orseveral layers, k must satisfy

j1� kjo1; i:e: 0oko2. (6)

At this time, d 0n � dn ! 0 as n-N. Specially if k ¼ 1,the difference in build height of a thin-walled metal part ina layer will be eliminated in the next layer at once.According to Eq. (4), the build height at a position in thenth layer

Hn ¼ nDzþ d0 � dn ¼ nDzþH0 � ð1� kÞnd0

�Xn�1i¼0

ð1� kÞiðDz� bÞ, ð7Þ

where H0 is the build height of the first layer. Specially ifk ¼ 1, Eq. (7) becomes

Hn ¼H0; n ¼ 0

ðn� 1ÞDzþ ðH0 þ bÞ ¼ ðn� 1ÞDzþ C; n ¼ 1; 2; . . .

(,

(8)

where constant C ¼ H0 þ b. Eq. (8) shows that the buildheight of a thin-walled metal part can be accuratelydetermined by the z-increment value and its layer numberfor open-loop DLF process if the coaxial powder nozzlehas the deposition character shown in Fig. 7.Obviously, the z-increment value of the process should

not exceed 0.48mm so that the condition shown in Eq. (8)can be met. In order to improve the shape quality of thepart, the z-increment value is set to 0.25mm in the presentwork. Fig. 9(a) shows the final sample of ‘FU’ which wasfabricated for 8 h, and we can see that the shape quality ofthe sample is quite good. The designed height of the part is54mm, and the actual build height is 53.54mm. Fig. 9(b)shows another sample fabricated in the present work.

4. Conclusions

The fabrication of thin-walled parts of nickel alloy usingopen-loop DLF process is introduced in the paper. Theprocess parameters and special control method for

ARTICLE IN PRESS

Fig. 9. Samples of thin-walled metal parts. (a) Chinese ‘FU’ and (b) logo of HUST.

L. Peng et al. / International Journal of Machine Tools & Manufacture 47 (2007) 996–10021002

improving the shape quality of single-bead walls andcomplex thin-walled parts was discussed in detail. Some ofconclusions can now be summarized:

(1)

The interference factors which affect the build height ofthin-walled metal parts occur randomly during DLFprocess under open-loop control. In open-loop DLF,thin-walled metal parts can achieve much better shapequality if the process parameters are suitable.

(2)

In open-loop DLF, thin walls of nickel alloy withuniform height can be built up layer by layer in acertain range of specific energy. If the specific energy ofthe process is too high, collapse will occur at both endsof a single-bead wall. Otherwise, the both ends of thewall will be higher than the middle of it if the specificenergy is too low.

(3)

In open-loop DLF, it is difficult to control the buildheight of a complex thin-walled metal part in an accuratemanner just using optimal parameters. If the depositionthickness of coaxial powder nozzle is linearly increasedwith increase in the standoff distance between thedeposition substrate and the powder focusing point ofthe nozzle, and the slope of the linear function k meets0oko1, the difference in build height of a thin-walledmetal part will be compensated automatically in one orseveral layers. It is proved that the build height of a thin-walled metal part can be accurately determined by the z-increment value and its layer number in theory underopen-loop control if the slope of the linear function k ¼ 1.

(4)

Accurate thin-walled metal parts with uniform buildheight can be fabricated by DLF process without usingany closed-loop control methods.

References

[1] G.K. Lewis, E. Schlienger, Practical considerations and capabilities for

laser assisted direct metal deposition, Materials and Design 21 (4)

(2000) 417–423.

[2] F.G. Arcella, Titanium alloy structures for aircraft applications by the

AeroMet laser additive manufacturing (LAMSM) process,

in: Proceedings of the 22nd ICALEO, 13–16 October 2003, Adam,

Special Lecture.

[3] C. Atwood, M. Griffith, L. Harwell, E. Schlienger, M. Ensz,

J. Smugeresky, T. Romero, D. Greene, D. Reckaway, Laser

Engineered Net Shaping (LENSTM): a tool for direct fabrication of

metal parts, in: Proceedings of the 17th ICALEO, 16–19 November

1998, Orlando, pp. Section E 1–9.

[4] J. Mazumder, D. Dutta, N. Kikuchi, A. Ghosh, Closed loop direct

metal deposition: art to part, Optics and Lasers in Engineering 34 (4–6)

(2000) 397–414.

[5] H. Dongming, R. Kovacevic, Sensing, modeling and control for laser-

based additive manufacturing, International Journal of Machine Tools

& Manufacture 43 (1) (2003) 51–60.

[6] L. Jichang, L. Lijun, In-time motion adjustment in laser cladding

manufacturing process for improving dimensional accuracy and

surface finish of the formed part, Optics & Laser Technology 36 (6)

(2004) 477–483.

[7] L. Peng, Y. Taiping, L. Sheng, L. Dongsheng, H. Qianwu, X. Weihao,

Z. Xiaoyan, Direct laser fabrication of nickel alloy samples,

International Journal of Machine Tools & Manufacture 45 (11)

(2005) 1288–1294.