Index Terms: Placement machines, placement head, pick-and-place, electronics assembly equipment, constant force. I. INTRODUCTION

Transcription

1 CONSTANT FORCE LINEAR PERMANENT MAGNET ACTUATORS Johannes J.H. Paulides, Meber, Laurentiu Encica, Meber, Koen J. Meessen Student Meber, and Elena A. Loonova, Senior Meber Abstract: In applications, such as vibration isolation, gravity copensation, pick-and-place achines, etc., there is a need for (long-stroke) passive constant force actuators cobined with tubular peranent agnet actuators to iniize the power consuption, hence, passively counteract the gravitational forces. For exaple, in pick-and-place achines, the passive devices allow the powerless counteraction of nozzles or tooling bits. In these applications, an increasing deand is arising for high-speed actuation with high precision and high bandwidth capability ainly due to the placeent head being at the foundation of the otion chain, hence, a large ass of this device will result in high force/power requireents for the driving echanis (i.e. an H- bridge three linear peranent agnet otors placed in an H-configuration. This paper investigates the cobined constant-force with tubular actuator topology, where the two actuator topologies are separately introduced and the cobination is verified using coprehensive three diensional (3D) finite eleent analyses. Index Ters: Placeent achines, placeent head, pick-and-place, electronics assebly equipent, constant force. I. INTRODUCTION In anufacturing autoation, the ost iportant perforance requireents of robot anipulators for assebly of electronics parts are positioning accuracy, allowable coponent size and high speed otion. Liitations on these perforance characteristics are iposed by echanical design such as structure, aterials used in the linkage, ass distribution techniques and by the drive syste coponents such as gearing, spindle screws and actuators. To overcoe any of these liitations, direct-drive H-bridge robot anipulators are introduced to place large coponents onto PCBs (printed circuit board), as shown in Fig. 1. Fig.1 Current AX-201 coponent ounter. Fig.2 Current oving head which holds the coponents above caera s for identification (2 Pick-and-Place devices are shown). The placeent rate of these achines can be optiized by considering significant factors such as the placeent head travel, the placeent sequence and the position of coponents in the feeder bank. This achine assebles coponents onto a PCB, where in general, each achine has a feeder carrier (or agazine), a PCB table (or worktable), head(s), nozzle(s) (or gripper) and a tool agazine. The PCB table, the feeder carrier and the head can either be oveable or fixed. The feeder reels (tapes holding electronic coponents) are positioned in feeder slots. Most coonly, larger coponents are supplied by tray feeders. The coponents are transported fro these feeders to the placeent position on the PCB using vacuu nozzle(s) which are placed at the end of the head [1], as shown in Fig. 2. TABLE 1 SPECIFICATIONS AX-201 COMPONENT MOUNTER. Maxiu output per hour 5,300 IPC 9850 output per hour 3,900 Placeent accuracy (3s) 25 icron Coponent range 0.4x0.2/165x23 Maxiu coponent height 50 Maxiu board size (L x W) standard 508 x 460 Tape feeding positions 96 (212 (twin tape)) More specifically, in the AX-201 coponent placeent achine ( with the characteristics suarized in Table 1, the PCB is locked in position by a table which is also used to locate the required points for the placeent operations. This PCB table uses a conveyor syste and the placeent heads are oved by an X-Y otion table, as shown in Fig. 2. This achine offers a cobination of high placeent accuracy and speed with an extreely wide coponent range. Therefore, it can be used for either stand-alone applications or as an end-of-line achine in cobination with other

2 placeent achines. A lot of work has been done on iproving the efficiency of coponent placeent achines such as [2-4]. However, ost previous work involves software iproveents, e.g. an offline scheduler to reduce the nuber of coponent being issing fro the feeder carrier. This paper presents a different approach, since there is an increasing deand for high-speed actuation with high precision and high bandwidth capability. As such, the ost generic device of any coponent ounting achine is the placeent head. This device is at the foundation of the otion chain and a large ass of this device will result in high force/power requireents for the H-bridge (three linear peranent agnet otors placed in an H-configuration, e.g., two for the y-axis oveent and one for x-axis oveent). The coplete Pick-and-Place action requires a four degrees-of-freedo robotic otion, where this paper focuses on the short-stroke (50 ) linear otion in the vertical direction to pick and place the coponents, as shown in Fig. 3. To increase the throughput of the total P&P cycle a high force requireent of 300 N is needed to achieve the acceleration level with an added ass of 120 N. Fig.3 Linear part of the placeent head consisting of a tubular peranent agnet actuator cobined with a passive linear constant force actuator. In general, the linear Z-axis oveent of the placeent heads is augented by a echanical coil spring to ensure that the placeent head cannot crash into the PCB when there is a power disruption or software failure. However, this spring force added to the gravitational force of the translator always needs to be actively counteracted by the tubular actuator, which does result in a significant power deand. A ore suitable solution would be provided by electroagnetic eans, in such a anner that the ass of the translator, nozzle or gripper and average coponent can be copensated, estiated to be 120 N in this application. This can be achieved by cobining the tubular peranent agnet (PM) actuator with a constant force actuator as shown in Fig. 3, which results in a significant iproveent of the dynaic behavior. As such, Section II discusses the used optiization routine, Section III the design of the tubular peranent agnet actuator, Section IV the constant force actuator design and Section V the cobined actuator. Finally, Section VI provides the conclusions. II. OPTIMIZATION ROUTINE Analytical deterination of actuator perforance provides an elegant way to design PM achines, e.g. luped paraeter odel or analytical equations [5-7]. However, in order to ipleent these odels certain assuptions need to be considered, hence, odel inaccuracies occur, e.g., flux fringing, coplex shapes, leakage, agnetic saturation, etc. are significant. Therefore, nuerical techniques are coonly used to deterine the field distribution and equivalent electric circuit coponents. However, using solely finite eleent (FE) techniques is tie deanding and therefore rather inefficient for design optiization. An alternative ethod is given by the space apping (SM) technique introduced by Bandler et al. [8]. This optiization technique is surrogate-based where a siple physics based (coarse) odel is exploited and aligned with the coputationally intensive accurate (fine) odel. This technique has successfully been applied in the field of icrowaves for coponent and syste odeling [9]. In this paper, the tubular and constant force actuators have been optiized using the SM technique [10]. The optiization routines together with the corresponding odels are ipleented in a Matlab (The MathWorks Inc.) and Maxwell 3D (Ansoft Co.) environent. Tubular peranent agnet actuator optiization includes the selection of nuerous diensional and perforance variables, e.g. pole pair, tooth thickness, tooth tip diensions, winding configuration, slot nuber, pole-pitch, slot opening, coil-pitch, etc. In general, optiization routines priary design for high efficiency, however, this can be iproved by increasing achine ass within the volue constraints specified. In this optiization, therefore also the translator ass has been included. Further, in slotted actuators, the winding and tooth area are contending for the sae volue, hence, the agnetic and electrical loadings have been optiized by not only considering the agnetic paraeters but also the theral considerations. The proposed design ethodology has the following structure: an optiization proble is forulated for the proposed tubular actuator and an SM variant is ipleented for deterining the

3 corresponding solution; the objective is the iniization of the actuator s ass while providing a specified (static) force response (100 N for a single pole pair) and liiting the levels of iron core flux densities and of generated heat through copper loss; corresponding agnetic and theral odels have been defined, where the tubular actuator optiization is suarized in the next Section, and, consequently, the constant force actuator is optiized in Section IV. The coupled proble for the cobined actuator will be presented in Section V. cobined electroagnetic with theral analysis is used as a basis for this paper. Therefore, the TPMA radial cross-section is shown in Fig. 5 and axial in Fig. 6, respectively. The relative diensions and design specifications (i.e. linear and nonlinear constraints), i.e. force output, the axiu adissible average flux densities in three regions of the iron core and the axiu adissible teperature, are suarized in Table 2. III. TUBULAR PM ACTUATOR A slotted tubular PM actuator, as shown in Fig. 4, is particularly interesting as it has a high force density, no end-windings and zero net radial attraction force between the translator and arature. The tubular actuator consists of a stator and a translator, where the oving agnet actuator is preferred since it does not require connections to the oving part. Fig.5 TPMA radial cross-section: topology and design variables. Fig.4 TPMA (axial cross-section of one pole-pair): topology and design variables. The stator of the TPMA contains coils and is ostly either slotted or slotless, where the highest force density can be achieved when a slotted structure is used [5]. However, the slotted structure also has soe disadvantages. e.g. the reluctance in the airgap is not unifor resulting in an extra force coponent called cogging force. This can be iniized by introducing typical pole-slot cobination, albeit this reduces the winding factor, and hence, the force capability. In this actuator skewing will be applied, which results in the actuator being suitable for this precision positioning application, hence, a sooth force versus position characteristic. As already entioned in Section II, the design objective was to iniize the actuator ass for a specified force output. In addition, upper liits are iposed on the average flux density levels in the iron core and on the peak teperature. This optiization routine has been previously reported in [10], where the Fig.6 TPMA axial cross-section: topology and design variables. TABLE 2 DESIGN SPECIFICATIONS. [x 1, x 2, x 3 ] () [6.0, 6.4, 2.0] [x 4, x 5, x 6 ] () [5.0, 15.2, 23.7] [x 7, x 8, x 9, x 10 ] () [1.8, 5.5, 3.6, 29.6] Airgap, g 1 () 1.0 Force (N) 100 Mean flux density back-iron (T) 1.3 Inner coil teperature ( C) 130 The TPMA optial design proble was solved considering a heat convection boundary condition with a value of 20 W -2 K -1 is specified on the outer lateral surface of the actuator, with an abient teperature of 25 C. The agnetic flux density distribution is shown in Fig. 7. It ust be noted that for this design the shaft aterial was not considered, and thus the shaft was not included in the total calculated ass.

4 pereability. Starting fro the general for of Apere s law: H dl = J d A = Ienc (2) C S and considering the agnetic field strength to be (piece-wise) constant on the integration path, the siplified expression is obtained: H l H l = 0 (3) g g + Fig.7 TPMA flux density distribution for the electroagnetictheral design solution. IV. OPTIMAL CONSTANT FORCE ACTUATOR DESIGN Passive constant force actuators (CFA) and their respectively force-displaceent characteristics are discussed in [11-12]. In general, this actuator is placed externally in parallel to the linear actuator. Further, no position sensor is required and an enhanceent could be achieved if the given force characteristic could be varied by eans of a predeterined constant current force-displaceent actuator [12]. For shorter strokes, couple of c s or less, solenoids producing a force depending on current are used. These solenoids have a very siple structure and are therefore aenable to ass production. However they, in order to produce force, require a constant current excitation and usually still have to use a spring to return to their initial position. This paper, therefore, presents the optial design of a long stroke constant force-displaceent actuator topology, as shown in Fig. 3, that fits inside the tubular actuator with a constant passive (without energy consuption) force independent on the position for 90% of the stroke for a passive force level of 120 N, respectively. Initially, the force capability of the constant force actuator can be derived fro well established analytical expressions. This siplified analytical odel is derived based on the following assuptions: no leakage fluxes or fringing effects are considered, and the agnet and iron relative pereabilities are taken to be equal to 1 and, respectively. For a agnet aterial having a linear deagnetization characteristic, with a working point that lies on the linear region, the flux density is given by: B + µ µ = Br o r H, (1) where, B is the working flux density, H is the corresponding agnetic field strength, B r is the reanent flux density and µ r is the relative recoil Further, it is assued that the airgap and agnet flux are equal. Hence, for the surface-ounted agnet rotor structure the average agnet flux density, fro (1), is: B Br = lg S 1+ l S g, (4) where l g and S g are the airgap length and surface area, and l and S are the agnet thickness and the agnet area, respectively. The airgap flux density can be derived fro (5) as: S B B S g =. (5) g The force is then derived fro the rate of change of agnetic co-energy with respect to the translator displaceent, where the force level, for the direction of travel along the z-axis, is then calculated by: W ' F =, (6) z z I = const then, by substituting (4) and (5) in (6), the following expression, which is independent of the z-axis displaceent due to the exclusion of axial leakage in the analytical odel, can be obtained for the force aplitude: 2 Br 1 Fz = cl 2µ 0 lg1c 1+ lcg1 1.(10) lg2c 1+ lcg2 In this, c g and c are the width of the airgap and agnet flux path, respectively, and the expressions for calculating the various paraeters considered are suarized in Table 3. In this, l g1 and c g1 are, correspondently, the length and area of the airgap of the overlapping part, l g2 and c g2 are the length and airgap of the non-overlapping part and l and c are

5 the length and area of the peranent agnets. The various diensions of x 11 to x 14 and g 2 are shown in Fig. 8 and suarized in Table 4. Further, n p represents the nuber of poles, where the use of a low nuber of poles in CFAs (typically 2-4) is ipleented. TABLE 3 PARAMETERS FOR THE PASSIVE CFA l g1 2g c g1 2π (x 1+x 2+g/2)/n p/2 l g2 2π (x 1+x 2+g)/n p/2 c g2 2π (x 1+x 2+g)/n p/2 l 2 x 3 c 2π (x 1+x 2+g+x 3/2)/n p/2 agnet thickness could provide the sae force, also an upper constraint could be ipleented on the agnet thickness. Further, the CFA force output could also be varied by adding a stator f using a eandered wound coil in between the agnets, which allows for deviations of the constant force characteristic aplitude [12]. To increase the force level and density, the polepair nuber selection is significantly influenced by considerations regarding the size and leakage, since increasing the nuber of poles, reduces the stationary and translating back-iron thicknesses. However, it ay also lead to a higher leakage flux, and thus, a decrease of the average airgap flux, hence, decreased forces. This is illustrated by Fig. 8, which shows the influence of varying the nuber of poles within the CFA. Clearly it can be seen that a reduced nuber of poles results in a saller outer radius, hence, leakage flux is significant when the outer radius is relatively sall. This also results in a saller stator ass of the CFA, however a slight increased translator ass. Fig.9 CFA Force versus stroke characteristic. V. COMBINED ACTUATOR The cobined tubular and constant force actuator topology requires a 3D representation, as shown in Fig. 10a, where the actuator is odeled in agnetostatic 3D finite eleent (FE) by eans of Maxwell 3D fro Ansoft Co. However, this does require approxiately 50,000 tetrahedra eleents to achieve an accurate force response (for an active length of 198 Fig. 10b). By utilizing the syetry inside the odel this finite eleent odel can be reduced to a quarter for the CFA of Fig. 8, as shown in Fig. 10a. Fig.8 CFA radial cross-section: topology and design variables. TABLE 4 CONSTANT FORCE ACTUATOR PARAMETERS n p = 4 n p = 6 n p = 8 n p = 10 x 11 () g 2 () x 12 () x 13 () x 14 () Outer radius () Translator ass (kg) Stator ass (kg) (a) The agnetic loading, B r, deterines the specific force capability, as it is clear fro (10). Further, it sees that scaling the surface ount agnets produces an increase in airgap flux density. However, if the airgap could be reduced a significantly saller (b) Fig.10 Cobined tubular peranent agnet with constant force actuator (a) 3D finite eleent analyses (b) cross-section with overall paraeters.

6 The relative evaluated diensions and geoetry are given by cobining Tables 2 and 4. These diensions are separately optiized designs cobined into a single actuator. The peranent agnets are assued to be a sintered NdFeB with a reanence of 1.23T and can be either radially or parallel agnetized. Further, the standard non-linear BH-curve for ild steel, AI 1010, is used for both stationary as translating back-iron. The force-displaceent characteristic of the cobined actuator is shown in Fig. 10. In this figure a stroke of 0 (initial position) corresponds to the stator with agnets and translator of the CFA being fully aligned and Fig. 11 shows the resulting force characteristics. Using the 3D finite eleent analyses, siulations at various positions of the CFA inner part (stator) have been undertaken, which results in the force acting on the cobined translator is the su of the TPMA (approxiately 300 N), CFA (120 N) force and gravitational force responses, as shown in Fig. 11. This gives that this cobined actuator allows for the copensation of a ass of soe 120 N when oving downwards fro the equilibriu position, hence, in this working area has an increased acceleration. Fig.11 Cobined tubular peranent agnet with constant force actuator: Force versus z-axis position characteristic. VI. CONCLUSIONS Several aspects regarding the design of a long stroke actuator that can copensate for translator and nozzle ass in a tubular peranent agnet actuator have been discussed. First the application has been discussed, where clearly the highest dynaical capability for a predeterined voluetric envelope is needed. As such, a slotted tubular PM achine and constant force actuator for 300 N and 120 N have been designed considering electroagnetic and theral aspects, for which an optiization routine based on the space apping idea is considered. The uniqueness of the optiization approach allows the passive constant force actuator to be optiized using just a single analytical equation as a coarse odel and extended 3D finite eleent analyses as a fine odel. This actuator exhibits a constant force characteristic (120 N) of approxiately 90% of the stroke (z-axis position). Further, it has been shown that the constant force decreased with increasing nuber of poles, where the outer radius has to increase by 25% when increasing fro 4 to 10 poles. Finally, this paper showed that the cobined actuator produces the required force for the stroke of 50 in a single (downwards) z-axis direction. REFERENCES [1]. J.L. Horijon, Patent US2007/ , Coponent placeent unit as well as a coponent device coprising such a coponent placeent unit, [2]. M. Ayob and G. Kendall, A triple objective function with a Chebychev dynaic pick-and-place point specification approach to optiise the surface ount placeent achine, European Journal of Oper. Research, Vol. 164, pp , [3]. J.J.H. Paulides, L. Encica, J.W. Jansen, R.A.J. van der Burg, J.L. Horijon, E. Loonova, Robot Architecture for a Contactless Industrial Pick-and-Place Machine, Proc. 11th Int. Conf. on Electrical Machines and Systes. Wuhan, China, pp. 1-6, [4]. K.J. Meessen, B. Gysen, J.J.H. Paulides, E.A. Loonova, Halbach Peranent Magnet Shape Selection for Slotless Tubular Actuators, IEEE Trans. on Magnetics, Vol. 44, Issue 11, pp , [5]. J. Wang, G. W. Jewell, and D. Howe, A general fraework for the analysis and design of tubular linear peranent agnet achines, IEEE Trans. on Magne., vol. 35, no. 3, pp , [6]. B.L.J. Gysen, E.A. Loonova, J.J.H. Paulides, A.J.A. Vandenput, Analytical and Nuerical Techniques for Solving Laplace and Poisson Equations in a Tubular Peranent-Magnet Actuator: Part I. Sei-Analytical Fraework, IEEE Trans. on Magn., Vol. 44, pp , [7]. B.L.J. Gysen, E.A. Loonova, J.J.H. Paulides, A.J.A. Vandenput, Analytical and Nuerical Techniques for Solving Laplace and Poisson Equations in a Tubular Peranent Magnet Actuator: Part II. Schwarz Christoffel Mapping, IEEE Trans. on Magn., Vol. 44, pp , 2008 [8]. J.W. Bandler, Q.S. Cheng, S.A. Dakroury, A.S. Mohaed, M.H. Bakr, K. Madsen, and J. Søndergaard, Space apping: the state of the art, IEEE Trans.Microwave Theory Tech., Vol. 52, pp , [9]. J. W. Bandler, Q. S. Cheng, D. H. Gebre-Maria, K. Madsen, F. Pedersen, and J. Søndergaard, EM-based surrogate odeling and de-sign exploiting iplicit, frequency and output space appings, IEEE MTT-S Int. Microwave Syp. Dig., pp , [10]. L. Encica, J.J.H. Paulides, E.A. Loonova, A.J.A. Vandenput, Electroagnetic and Theral Design of a Linear Actuator Using Output Polynoial Space Mapping, IEEE Trans. on Ind. Appl., Vol. 44, pp , [11]. J.J.H. Paulides, L. Encica, E.A. Loonova, A.J.A. Vandenput, Passive and active constant forcedisplaceent characteristics of a long-stroke linear actuator, Proc. Int. Conf. on Electrical Machines, pp. 1-6, [12]. L. Encica, Space-apping optiization applied to the design of a novel electroagnetic actuator for active suspension, PhD thesis Eindhoven University of Technology, pp , 2008.

7 1 1 Johannes Paulides (j.j.h.paulides@tue.nl) was born in Waalwijk, The Netherlands in He received the B.Eng. degree fro the Technische Hogeschool s-hertogenbosch in 1998 and the M.Phil. and Ph.D. degrees in electrical and electronical engineering fro the University of Sheffield in 2000 and 2005, respectively. Since 2005, he has been a Research Associate at Eindhoven University of Technology, and siultaneously is a director of Paulides BV and Advanced Electroagnetics BV, sall SMEs based in the Netherlands producing electrical achines and prototype electroagnetic devices. His research activities span all facets of electrical achines, however in particular linear and rotating peranent agnet excited achines for autootive and high precision applications. Laurentiu Encica (l.encica@tue.nl) was born in Bucharest, Roania, in He received the Dipl. Eng. degree fro the University "Politehnica" of Bucharest, Roania, in 2002, and the Ph.D. degree fro the Eindhoven University of Technology, Eindhoven, The Netherlands, in He is currently affiliated as a post-doctoral researcher with the Eindhoven University of Technology, The Netherlands. His research interests are in coputer assisted analysis, design and optiization of electroagnetic actuators. Koen Meessen (k.j.eessen@tue.nl) received his MSc degree fro Eindhoven University of Technology, The Netherlands, in 2008, where he his currently working towards his Ph.D. degree on a high acceleration tubular peranent agnet actuator for pick-and-place achines. Additionally he is a eber of the Robocup football tea, where he is responsible for the electroechanical actuators and their control. His current research interests are in design of high-perforance electroagnetic peranent agnet actuators. Elena A. Loonova (e.loonova@tue.nl) was born in Moscow, Russia. She received the M.Sc. (cu laude) and Ph.D. (cu laude) degrees in electroechanical engineering fro Moscow State Aviation Institute (TU), Moscow, Russia, in 1982 and 1993, respectively. She is currently an Associate Professor at Eindhoven University of Technology, Eindhoven, The Netherlands. She has worked on electroechanical actuators design, optiization, and developent of advanced echatronics systes.