Scanning system development and digital beam ...

Rapid Prototyping Journal Scanning system development and digital beam control method for electron beam selective melting Chao Guo Jing Zhang Jun Zhang Wenjun Ge Bo Yao Feng Lin

Article information: To cite this document: Chao Guo Jing Zhang Jun Zhang Wenjun Ge Bo Yao Feng Lin , (2015),"Scanning system development and digital beam control method for electron beam selective melting", Rapid Prototyping Journal, Vol. 21 Iss 3 pp. 313 - 321 Permanent link to this document: http://dx.doi.org/10.1108/RPJ-11-2012-0106

Downloaded by TSINGHUA UNIVERSITY At 22:56 01 June 2015 (PT)

Downloaded on: 01 June 2015, At: 22:56 (PT) References: this document contains references to 10 other documents. To copy this document: [email protected] The fulltext of this document has been downloaded 55 times since 2015*

Users who downloaded this article also downloaded: Abby Megan Paterson, Richard Bibb, R. Ian Campbell, Guy Bingham, (2015),"Comparing additive manufacturing technologies for customised wrist splints", Rapid Prototyping Journal, Vol. 21 Iss 3 pp. 230-243 http://dx.doi.org/10.1108/RPJ-10-2013-0099 Ratnadeep Paul, Sam Anand, (2015),"A combined energy and error optimization method for metal powder based additive manufacturing processes", Rapid Prototyping Journal, Vol. 21 Iss 3 pp. 301-312 http://dx.doi.org/10.1108/RPJ-07-2013-0065 Brian N. Turner, Scott A Gold, (2015),"A review of melt extrusion additive manufacturing processes: II. Materials, dimensional accuracy, and surface roughness", Rapid Prototyping Journal, Vol. 21 Iss 3 pp. 250-261 http://dx.doi.org/10.1108/ RPJ-02-2013-0017

Access to this document was granted through an Emerald subscription provided by 453758 []

For Authors If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.

About Emerald www.emeraldinsight.com Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online products and additional customer resources and services. Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation. *Related content and download information correct at time of download.

Scanning system development and digital beam control method for electron beam selective melting Chao Guo, Jing Zhang, Jun Zhang, Wenjun Ge, Bo Yao and Feng Lin


Department of Mechanical Engineering, Tsinghua University, Beijing, China Abstract Purpose – The purpose of this paper is to correct the beam deflection errors and beam defocus by using a digital scanning system. Electron beam selective melting (EBSM) is an additive manufacturing technology for metal parts. Beam deflection errors and beam defocus at large deflection angles would greatly influence the accuracy of the built parts. Design/methodology/approach – The 200 ⫻ 200 mm2 scanning area of the electron beam is discretized into 1001 ⫻ 1001 points arranged in array, based on which a digital scanning system is developed. To correct the deflection errors, the electron beam scans a 41 ⫻ 41 testing grid, and the corrective algorithm is based on the bilinear transformation from the grid points’ nominal coordinates to their measured coordinates. The beam defocus is corrected by a dynamic focusing method. A three-dimensional testing part is built with and without using the corrective algorithm, and their accuracies are quantitatively compared. Findings – The testing grid scanning result shows that the accuracy of the corrected beam deflection system is better than ⫾ 0.2 mm and beam defocus at large deflection angles is eliminated visibly. The testing part built with using the corrective algorithm is of greater accuracy than the one built without using it. Originality/value – Benefiting from the digital beam control method, the model-to-part accuracy of the system is effectively improved. The digital scanning system is feasible in rapid manufacturing large and complex three-dimensional metal parts. Keywords Additive manufacturing, Accuracy, Digital control, Electron beam, Scanning system Paper type Research paper

1. Introduction

increasingly applied in medical implant and aerospace industry (ARCAM AB, 2012). In the EBSM technology, beam deflection errors and beam defocus at large deflection angles would greatly influence the accuracy of EBSM products: 1 Beam deflection errors: The deflection coil generates a magnetic field to deflect the electron beam. Due to the non-linear deflection and the inhomogeneity of the magnetic field, a certain amount of pincushion distortion as shown in Figure 2 can be often found in large area scanning. 2 Beam defocus at large deflection angles: The system uses a focusing coil to focus the electron beam. As illustrated in Figure 3, if the current in the coil is constant, the focal point of the electron beam follows a spherical path during scanning. When scanning a plane, the electron beam focuses around the center and defocuses at a large deflection angle.

Electron beam selective melting (EBSM) is an additive manufacturing technology building parts directly from a CAD model by melting metal powder layer by layer with an electron beam in a high vacuum chamber. As shown in Figure 1, parts are manufactured by repeating the following steps: the piston moves down a layer thickness, and a layer of metal powder is spread onto the previous layer, and then the electron beam scans and melts the selected areas of the layer. Profiting from its high vacuum environment, EBSM technology is suitable for manufacturing parts with reactive materials with high affinity for oxygen, e.g. titanium. In addition, with the merits of high energy density and high scanning speed of the electron beam, EBSM technology is of higher efficiency and lower cost compared to the laser additive manufacturing process (Yan et al., 2007). Therefore, EBSM technology has become a new concept for metal fabrication from pre-alloyed, atomized precursor powders (Murr et al., 2012). Until 2012, ARCAM AB (2012) has developed two EBSM machines, which are

There already exist a variety of electron beam scanning systems for other applications. Typically in a cathode ray tube (CRT) monitor, the analogue circuits provide saw-toothed wave signals to the deflection coil, driving the electron beam to

The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/1355-2546.htm The authors would like to acknowledge the funding of 2013 Beijing Science and Technology Development Project. Received 20 November 2012 Revised 30 June 2013 30 October 2013 Accepted 23 January 2014

Rapid Prototyping Journal 21/3 (2015) 313–321 © Emerald Group Publishing Limited [ISSN 1355-2546] [DOI 10.1108/RPJ-11-2012-0106]

313

Electron beam selective melting

Rapid Prototyping Journal

Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

Figure 3 Concept of the beam defocus

Figure 1 Schematic of the EBSM technology

Electron beam

Filament Cathode

– 70kV Focusing coil

Gate (0~–1,500V to the cathode) Electron beam gun

Anode Deflection coil

Focusing coil

Deflection coil


f

Vacuum chamber

Powder storage

Δf

Electron beam P (x, y)

(0, 0)

Powder spreader

discretized into pixels and the control data for all the pixels are advance recorded in a matrix. Then, the digital-to-analog (D/A) convertors read in data from the matrix, converting them into analogue signals. Because the beam diameter for lithography is very small (usually several nm), resolution of the scanning system can be very high. Rudolph et al. (2003) designed a control unit for electron beam lithography with a maximum resolution of 217 pixels in x and y direction each. However, scanning area of these systems is very small (⬍ 2 ⫻ 2 mm2), hence the beam deflection errors and beam defocus are neglected. Besides, beam energy for lithography is much lower than that for EBSM. Some scanning systems for high-power electron beam also have been developed. Bahr et al. (1998) proposed an electron beam scanning system for evaporation. Wang and Yao (2004) developed a scanning system for welding, brazing, surface hardening and heat treatment. In these systems, two D/A convertors are used to generate deflection signals. By manually changing the control data sent to the D/A convertors, the beam scanning path can be adjusted and programmed. However, the beam deflection errors cannot be eliminated automatically and the beam defocus at large deflection angles still exists. Companies like TWI (2007) have developed some high frequency digital scanning systems for creation of various heat patterns. By using the scanning system, the welding beam can be split into numerous beamlets and move accurately according to the programmed path. This work is valuable as reference for developing scanning system for EBSM. Qi et al. (2010) introduced a scanning system for EBSM. Two integration circuits are used to generate triangle wave signals, and the integral upper and lower limit voltages are given by the D/A convertors. By changing the data sent to the convertors, endpoints of the scanning line can be adjusted to their given positions. However, as shown in Figure 4, the scanning line still deviates from its given path. Besides, the beam defocus at large deflection angles is still not corrected. In this paper, a digital scanning system for EBSM is proposed. The 200 ⫻ 200 mm2 scanning area is discretized

Parts Vertical piston

Figure 2 The pincushion distortion in large area scanning

scan frame by frame. A correction circuit is designed to modify the signals so as to eliminate the beam deflection errors. To correct the beam defocus, another circuit provides parabolic wave signal synchronized with the deflection signals to the focusing coil. However, the electron beam in a CRT is of low energy level, and the beam scanning path and frequency are constant. Several high-resolution scanning systems for lithography have been developed. In these systems, the scanning area is 314



Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

Figure 6 Two cases of the beam spot overlapping

Figure 4 Distortion of the scanning line y

FLb

R

Before adjustment After adjustment

R R R

R

x

FLa

Given path

(a)

Actual path

(b)


Notes: (a) Horizontal or vertical overlap; (b) Diagonal overlap into points arranged in array, and the electron beam scans point by point. As long as the points are dense enough, the scanning line would be continuous. Moreover, a digital beam control method is put forward to correct the beam deflection errors and beam defocus.

The transformation from a scanning line to a collection of points can be expressed as the following mathematical formula: y ⫽ f(x), xA ⱕ x ⱕ xB ⇒ 兵Pk(ik, jk), k ⫽ 1, 2, 3. . .n其 (3)

2. Overview of the scanning system

As shown in Figure 7, P1 and Pn are the nearest points to the endpoints A and B, and the other points from P1 to Pn are calculated by using the Bresenham line drawing algorithm (Bresenham, 1996). Here, i and j are the column and row of the point Pk. For the point P(i, j), here are four parameters that need to be controlled: x and y coordinates, beam focal length f and beam current ib. Their corresponding control data are X, Y, F and I. A vector V is introduced to record the control data:

The 200 ⫻ 200 mm2 scanning area is discretized into 1001 ⫻ 1001 points arranged in array, as shown in Figure 5, and the beam diameter of the system is about 0.4 mm. Figure 6 shows the two cases that the adjacent beam spots overlap with each other, and the fluctuations at the edge of the scanning line can be calculated (R is the beam radius):

冉

FLa ⫽ 1 ⫺

冉

FLb ⫽ 1 ⫺

冊

兹3 R ⬇ 0.0268 mm 2

(1)

Vi,j ⫽ [XYFI]i,j

冊

兹2 R ⬇ 0.0586 mm 2

(2)

For all the 1001 ⫻ 1001 points, a matrix M is introduced to record the control data:

The fluctuations in both the cases are very small. Moreover, because of the flow of molten pools, the fluctuations will be much smaller and can be ignored. Therefore, the discretization ensures the scanning accuracy of the system.

M ⫽ V1001 ⫻ 1001 ⫽ [XYFI]1001 ⫻ 1001

(5)

Figure 7 Transformation from a line to points

Figure 5 Discretization of the scanning area j

(4)

y j

y

1,000

1,000

B Pn Pn-1 x x

P2 A

P3

P1

0 0

1,000

i

315

1,000

i



Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

The matrix M is set in advance. When scanning, the control data of the points to be scanned are selected out from the matrix and sent to the memory of four D/A convertors. Then the D/A convertors synchronously convert those control data into analogue signals.

sleeve surrounding the solenoid can prevent the magnetic flux leakage, so that the magnetic field is more concentrated. A flared coil is employed to deflect the electron beam. The flared coil includes two windings, producing nearly uniform magnetic fields in x and y direction separately. A driving circuit shown in Figure 9 is designed to provide current to the coil. A sampling resistor r is connected with the coil, and the voltage across the sampling resistor is fed back to the input, so that the output current iout is linear with the input voltage uin. The ratio of iout (uint: A) to uin (uint: V) is determined by:

3. Hardware configuration


Figure 8 is a block diagram of the hardware configuration of the scanning system. The matrix M is stored in the computer. An arbitrary waveform generator (AWG) with four channels of D/A convertors is used to convert the selected control data into analogue signals. Signals from Channel 0, 1 and 2 are sent to the driving circuits to control the beam deflections and the beam focal length, respectively, and signals from Channel 3 are amplified to control the beam current.

iout Rf1 ⫹ Rf2 ⫽ uin r ⴱ Rf2

4. Digital beam control method

3.1 The arbitrary waveform generator The whole scanning area is discretized into 1001 ⫻ 1001 points. For:

The electron beam is digitally controlled by setting the data (X, Y, F, I) in the matrix M. 4.1 Control data X and Y Control data X and Y in the matrix M are initialized as:

210 ⫽ 1024 ⬎ 1001,

X ⫽ KX ⴱ x⬘i, i ⫽ 0, 1, 2, . . .1000 Y ⫽ KY ⴱ y⬘i, j ⫽ 0, 1, 2, . . .1000

it requires the AWG with a resolution of more than 10 bits and, therefore, the 14-bit resolution AWG meets the requirement. The minimal memory size of the AWG can be figured out:

x⬘ ⫽ ⫺200 ⫹ 5 ⴱ (p ⫺ 1), p ⫽ 1, 2, . . .41 y⬘ ⫽ ⫺200 ⫹ 5 ⴱ (q ⫺ 1), q ⫽ 1, 2, . . .41

Thus, the 128-MB memory in the AWG is adequate to store all the control data in the matrix M. The system is designed to scan the whole area at a maximum speed of five frames per second, and consequently the sample rate of the AWG cannot be less than:

Figure 9 Schematic of the driving circuit for coil +Us

5 ⫻ 10012 ⫽ 5,010,005 ⫽ 5M samples per second

uin

Coil uo

The AWG with 20Ms/s sample rate meets the requirement. –Us

3.2 Coils and the driving circuit Coils are important parts of the scanning system. The focusing coil consists of a solenoid, which generates an axisymmetric magnetic field under the current excitation, and electrons passing through the solenoid converge at one point. Shielding

Rf2

Figure 10 The testing grid

Figure 8 Block diagram of the hardware configuration

Memory Ch 3

Amplifier

Gate of the electron gun

Y(ik, jk)

Ch 2

Driving Circuit

Focusing coil

F(ik, jk)

Ch 1

Driving Circuit

I(ik, jk)

Ch 0

Driving Circuit

M = [ X Y F I ]1001 × 1001

Computer

(7)

Then the electron beam scans a 41 ⫻ 41 testing grid, as shown in Figure 10. The given coordinates of the grid points are:

14 ⫻ 4 ⫻ 10012bits ⫽ 56,112,056 bits ⬇ 7MB

X(ik, jk)

(6)

Deflection coil

k = 1,2,3...N Programmable Arbitrary Waveform Generator (AWG)

316

Rf1

iout

r

(8)




Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

Because of the deflection errors, the grid points would deviate from their given coordinates, exhibiting pincushion distortion. To correct the errors, it needs to measure the actual coordinates of the beam spot, and then update the control data X and Y in the matrix.

Take the point P to be measured as an example. The image coordinates of P, O and A are (uP, vP), (uO, vO) and (uA, vA), respectively. Then, the actual coordinates of P (xP, yP) can be obtained by the rotation-scale-translation coordinate system transformation:

4.1.1 Measurement of the beam spot coordinates It is time-consuming and difficult to measure all the grid points’ coordinates by a caliper. An image detection method is developed to measure the beam spot coordinates. A sheet of smooth paper is attached on the building table, and the electron beam penetrates the paper and leaves holes on it. Then the paper and two high precision rulers are scanned by a linear charge-coupled device (CCD) with a resolution of 1,200 ppi. One ruler is strictly parallel to the scanning direction of the linear CCD, and the other is strictly perpendicular to the scanning direction. As a result, a binary image as shown in Figure 11 is obtained, point O was penetrated by the electron beam when Channel 0 and 1 of the AWG output zero signals, and point A was penetrated when Channel 0 and 1 output 2 V and 0 V, respectively. The positions relative to the building table of O and A are constant. Thus, O and A are used to calibrate the actual coordinate system xOy: O is the origin of the coordinate system, OA represents the x-axis and the direction perpendicular to OA is designated as the y-axis. Now, there are two coordinate systems on the paper: the calibrated actual coordinate system xOy (unit: mm) and the image coordinate system uO’v (unit: pixel). On the image, the beam spots are represented as shadowed regions. Because the beam energy follows a Gaussian distribution, centroid of the shadowed region is defined as the beam spot position. With the image detection tool of software MATLAB, centroid pixels of the shadowed regions are extracted, so that their image coordinates (u, v) can be obtained.

cos␪ ⫽ sin␪ ⫽

uA ⫺ uO (uA ⫺ uO)2 ⫹ (vO ⫺ vA)2 vO ⫺ vA

(uA ⫺ uO)2 ⫹ (vO ⫺ vA)2 uP ⫺ uO vO ⫺ vP xP ⫽ cos␪ ⫹ sin␪ ku kv uP ⫺ uO vO ⫺ vP yP ⫽ ⫺ sin␪ ⫹ cos␪ ku kv

(9)

ku and kv are the ratios of length(mm) to pixels in the two directions, which can be obtained by counting the pixels numbers along the rulers on the image. By using this image detection method, it costs only several seconds to measure all the grid points’ coordinates on a single piece of paper. 4.1.2 Deflection error correction Principle of the error correction can be expressed in mathematics: (x, y) ⫽ f2关f1(x⬘, y⬘)兴

(10)

共x⬘, y⬘兲 are the given coordinates, and (x, y) are the actual coordinates. f1 is the correction function and f2 is the distortion function. As long as f1 ⫽ f2⫺1, the actual coordinates coincide with the given coordinates. The function f2⫺1 can be determined by the given coordinates and the actual coordinates under the condition of f1⬅I. As shown in Figure 10, the testing grid divides the area into 40 ⫻ 40 unit fields, and the size of each unit field is 5 ⫻ 5 mm2. Within each unit field, the function f2⫺1 can be approximated as a bilinear transformation:

Figure 11 Schematic of the image detection method

x⬘ ⫽ Ax ⫹ By ⫹ Cxy ⫹ D y⬘ ⫽ Ex ⫹ Fy ⫹ Gxy ⫹ H

(11)

To obtain the coefficients A ⬃ H, the actual coordinates and the given coordinates of the unit field’s four corners are substituted into equation (11):

冤冥冤冤冥冤

x⬘1 x1 x2 x⬘2 ⫽ x3 x⬘3 x4 x⬘4 y⬘1 x1 x2 y⬘2 ⫽ x3 y⬘3 x4 y⬘4

y1 y2 y3 y4 y1 y2 y3 y4

x1y1 x2y2 x3y3 x4y4 x1y1 x2y2 x3y3 x4y4

冥冤冥冥冤冥

1 A B 1 ⴱ C 1 D 1 1 E F 1 ⴱ G 1 H 1

(12)

The equation (12) has eight equations and eight unknowns, so there is a unique solution for the coefficients A-H. Then within the unit field, the control data X and Y in the matrix M are modified as: Xnew ⫽ KX ⴱ 共Ax⬘ ⫹ By⬘ ⫹ Cx⬘y⬘ ⫹ D兲 Ynew ⫽ KY ⴱ 共Ex⬘ ⫹ Fy⬘ ⫹ Gx⬘y⬘ ⫹ H兲 317

(13)



Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

By using the above algorithm, the control data X and Y within all the unit fields are modified to improve the beam deflection accuracy. The above steps need to be repeated until all the grid points’ deviations are below ⫾0.2 mm.

large parts. Therefore, a testing part is designed to evaluate the manufacturing accuracy. The testing part comprises five cylinders arranged in a right angle. Cross-section of the part is showed in Figure 12. The diameter of each circle is 20 mm, and the distance between adjacent circles is 75 mm. The circles are labeled with C1-C5, and the coordinates of their centers are (xCi, yCi), (i ⫽ 1, 2, 3, 4, 5). Then, four vectors are defined by the following equations:

4.2 Control data F Assuming that the control data F in the matrix M are constant: F ⫽ Fconst

(14)

X1 ⫽ ((xC3 ⫺ xC1), (yC3 ⫺ yC1)) X2 ⫽ ((xC2 ⫺ xC1), (yC2 ⫺ yC1)) Y1 ⫽ ((xC5 ⫺ xC1), (yC5 ⫺ yC1)) Y2 ⫽ ((xC4 ⫺ xC1), (yC4 ⫺ yC1))

the electron beam would defocus at large deflection angles. The solution to correct the beam defocus is to introduce the compensation component: (15)

The dimensional accuracy is evaluated by comparing the measured distance with the nominal distance. Dimensional deviation in x direction:

As illustrated in Figure 2, when the electron beam is deflected to the point P (x, y), the traveling distance of the electrons L is larger than focal length f. The ratio of the compensating amount ⌬f to f is calculated:

兹x ⫹ y ⫹ f ⫺ f ⌬f ⫽ f f x2 ⫹ y2 (x2 ⫹ y2) ⬇1 ⫹ ⫺1⫽ 2f 2 2f 2 2

2

ⱍX1ⱍ ⫺ 150 ⫽ 兹(xC3 ⫺ xC1)2 ⫹ (yC3 ⫺ yC1)2 ⫺ 150(mm) (22)

2

Dimensional deviation in y direction:

(16)

ⱍY1ⱍ ⫺ 150 ⫽ 兹(xC5 ⫺ xC1)2 ⫹ (yC5 ⫺ yC1)2 ⫺ 150(mm) (23)

And the relationship between f and its control data F is: f⬀

1 F2

The straightness accuracy is evaluated by calculating the angles between the two vectors in the same direction. Straightness deviation in x direction:

(17)

具X1, X2典

Then, it can be deduced that: ⫽ cos⫺1

⌬f ⫺2⌬F ⴱ F ⫹ ⌬F 2 ⫺2⌬F F2 ⫺1⫽ ⬇ ⫽ 2 f F (F ⫹ ⌬F) (F ⫹ ⌬F)2 (18)

(xC3 ⫺ xC1) ⴱ (xC2 ⫺ xC1) ⫹ (yC3 ⫺ yC1) ⴱ (yC2 ⫺ yC1)

(°)

兹(xC3 ⫺ xC1)2 ⫹ (yC3 ⫺ yC1)2 ⴱ 兹(xC2 ⫺ xC1)2 ⫹ (yC2 ⫺ yC1)2 (24)

Combining equations (16) and (18): ⌬F ⫽

⫺F 2 (x ⫹ y2) ⫽ K⌬F ⴱ (x2 ⫹ y2) 4f 2

Figure 12 Cross-section of the testing part (19)

y

Equation (19) demonstrates that the compensating amount is proportional to the square of the beam deflection. The four corners of the scanning area are tested to obtain the proportional coefficient, and K⌬F is the average of the four results.

C5 20

75 mm

m m

4.3 Control data I Control data I in the matrix M is proportional to the given beam current ib: I ⫽ KI ⴱ ib

Y1 C4

x

(20)

The beam current can be uniform or non-uniform when the electron beam scans the powder layer. Once the given beam current is changed, the control data I in the matrix will be renewed.

75 mm


F ⫽ Fconst ⫹ ⌬F

(21)

Y2

C1

C2 X1 X2

5. Accuracy evaluation method The developed scanning system and the digital beam control method are aimed to improve the manufacturing accuracy of

75 mm

318

75 mm

C3



Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

Figure 14 Result of the testing grid scanning

Straightness deviation in y direction: 具Y1, Y2典 ⫽ cos

⫺1


(°)

兹(xC5 ⫺ xC1)2 ⫹ (yC5 ⫺ yC1)2 ⴱ 兹(xC4 ⫺ xC1)2 ⫹ (yC4 ⫺ yC1)2 (25)

Additionally, the angular accuracy is evaluated by calculating the angles between vectors X1 and Y1: Angular deviation: 具X1, Y1典 ⫺1

⫽ cos


兹(xC3 ⫺ xC1)2 ⫹ (yC3 ⫺ yC1)2 ⴱ 兹(xC5 ⫺ xC1)2 ⫹ (yC5 ⫺ yC1)2

(°) ⫺ 90(°)


(26) This method of accuracy evaluation is used for the following reasons. First, the influence of the surface roughness can be reduced as much as possible (Brajlih et al., 2011). Second, by defining the vectors, the position and the orientation of the measurement coordinate system have no impact on the measuring results.

6. Experiments and results Figure 13 is the picture of the machine EBSM-250 which adopts this developed digital scanning system. Accelerating voltage of the electron beam gun is about 70 kV, and the beam current is 0-50 mA. The 41 ⫻ 41 testing grid was scanned on paper before and after error correction. The beam current was 0.5 mA and the beam residence time on each grid point was 20 ms. Figure 14(a) shows the testing grid scanned by the electron beam before correction. Pincushion distortion can be seen in Figure 14(a), for example, at the right edge of the grid, the point gradually deviates from the reference line as it moves down from the middle of the edge. While in Figure 14(b) which shows the testing grid after correction, almost all the points along the right edge of the grid are exactly on the reference line. It demonstrates that the beam deflection errors are largely reduced. Actual coordinates of all the grid points are Figure 13 Prototype of the EBSM-250 machine measured by the image detection method, and it is found that the maximum error occurs at the lower right corner of the grid and it reduces from 2.84 to 0.18 mm. It proves that the accuracy of the corrected beam deflection system is better than ⫾0.2 mm. Experimental observation shows that the beam spot has nearly the same brightness in the center as in the corners. In Figure 14(b), all the holes penetrated by the electron beam are nearly of the same size. It indicates that the beam defocus at large deflection angles is eliminated visibly. To further verify the effectiveness of the corrective algorithm, a three-dimensional testing part was built with and without using the corrective algorithm. A 316L stainless steel starting plate with dimensions of 230 ⫻ 230 ⫻ 10 mm was placed in the vacuum chamber and then was pre-heated for about 40 minutes. The beam current for preheating was 25 mA and the scanning velocity was 20 m/s. The 316L stainless 319




Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

Table I Accuracy evaluation of the testing parts

steel powder with an average particle size of 30 um was used to build the testing part. The layer thickness was 0.1 mm. For each layer, a 180 ⫻ 180 mm2 square on the layer was pre-heated for 70 seconds by the electron beam, with a scanning velocity of 10 m/s. In this step, the beam current was increased every 10 seconds and was, in order, 1, 3, 5, 7, 10, 15 and 20 mA. Then, the electron beam scanned the five circular areas for three times at a velocity of 1 m/s while the beam current was increased from 1 to 5 mA (increased by 2 mA per scan). Lastly, the electron beam scanned the five circular areas twice at a velocity of 0.25 m/s while the beam current was 5 and 10 mA in the two scans, respectively. A total of 100 layers were built. The beam path scanning the circular area is illustrated in Figure 15. Space between the scanning lines was 0.2 mm, and the directions of the two adjacent scanning lines were opposite. After the previous layer was finished, the orientation of the scanning lines rotated 90°. This would help to reduce the probability of defects. Figure 16 is the photo of the testing parts built with and without using the corrective algorithm. The parts were measured using a ZEISS Contura G2 coordinate measuring machine. For each cylinder of the testing part, the center of its circular cross-section was measured at the same height. The model-to-part deviations were calculated and listed in Table I, and it indicates that the testing part built with using the

Measuring results

Without correction

With correction

Dimensional deviation inx direction Dimensional deviation iny direction Straightness deviation inx direction Straightness deviation iny direction Angular deviation

⫺4.7 mm ⫺2.7 mm 0.51° 0.51° 0.94°

0.26 mm 0.65 mm 0.16° 0.14° 0.080°

corrective algorithm is of much greater accuracy than the one built without using the corrective algorithm.

7. Summary The beam deflection accuracy this paper focuses on is the first important factor which influences the model-to-part accuracy of the system. In this paper, a digital scanning system with a resolution of 1001 ⫻ 1001 for EBSM has been developed. Benefiting from the digital beam control method, the beam deflection errors and the beam defocus at large deflection angles were largely reduced by the corrective algorithm. The beam deflection accuracy is better than ⫾0.2 mm over the 200 ⫻ 200 mm2 scanning area and the model-to-part accuracy of the system is effectively improved by using the corrective algorithm. In addition to the beam deflection accuracy, there are some other factors influencing the model-to-part accuracy, such as materials properties, beam diameter, beam processing paramete Therefore, the manufacturing accuracy needs to be further studied and improved in future. Furthermore, with the benefit of the programmable scanning system, a variety of scanning strategies (e.g. the scanning strategy in Figure 15) can be realized with the scanning system. Hence, this developed scanning system is suitable for studying the effect of different scanning strategies on the temperature fields and the parts’ microstructures.

Figure 15 Electron beam path

References

(a) (b) Notes: (a) For layer 1, 3, 5 [...] ; (b) for layer 2, 4, 6 [...]

ARCAM AB (2012), “EBM systems, software and auxiliaries”, available at: www.arcam.com/products/index. aspx (accessed 10 August 2012). Bahr, M., Hoffmann, G., Ludwig, R. and Steiniger, G. (1998), “New scan and control system (ESCOSYS™) for high power electron beam techniques”, Surface and Coatings Technology, Vol. 98 Nos 1/3, pp. 1211-1220. Brajlih, T., Valentan, B., Balic, J. and Drstvensek, I. (2011), “Speed and accuracy evaluation of additive manufacturing”, Rapid Prototyping Journal, Vol. 17 No. 1, pp. 64-75. Bresenham, J. (1996), “Pixel-processing fundamentals”, IEEE Computer Graphics and Applications, Vol. 16 No. 1, pp. 74-82. Murr, L.E., Gaytan, S.M., Ramirez, D.A., Martinez, E., Hernandez, J., Amato, K.N., Shindo, P.W., Medina, F.R. and Wicker, R.B. (2012), “Metal fabrication by addictive

Figure 16 Testing parts built by EBSM

320



Chao Guo et al.

Volume 21 · Number 3 · 2015 · 313–321

manufacturing using laser and electron beam melting technologies”, Journal of Materials Science and Technology, Vol. 28 No. 1, pp. 1-14. Qi, H.B., Wang, J.Q. and Yang, L.N. (2010), “Development of high speed scanning system for the fabrication of 3D metal parts using electron beam selective melting technology”, 2009 International Conference on Manufacturing Science and Engineering, Zhuhai, 26-28 December 2009, Trans Tech Publications, pp. 4001-4004. Rudolph, S., Heim, S., Rudolph, D. and Schmahl, G. (2003), “Fast control unit for electron beam lithography systems especially for X-ray and EUV optics”, 7th International Conference on X-Ray Microscopy, Grenoble, 28 July 20022 August 2002, EDP Sciences, pp. 203-205.

Sanderson, A. (2007), “Four decades of electron beam development at TWI”, Welding in the World, Vol. 51 Nos 1/2, pp. 37-49. Wang, X. and Yao, S. (2004), “Control on electron beam scanning track in brazing”, Transactions of the China Welding Institution, Vol. 25 No. 6, pp. 31-34. Yan, Y.N., Qi, H.B., Lin, F., He, W., Zhang, H.R. and Zhang, R.J. (2007), “Produced three-dimensional metal parts by electron beam selective melting”, Chinese Journal of Mechanical Engineering, Vol. 6 No. 43, pp. 87-92.

Corresponding author


Feng Lin can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website: www.emeraldgrouppublishing.com/licensing/reprints.htm Or contact us for further details: [email protected]

321