Visual Feedback Control of a Micro Lathe Hirotaka OJIMA1, Katsuhiro SAITO1, Libo ZHOU1, Jun SHIMIZU1, Hiroshi EDA11Ibaraki UniversityKeywords: Micro lath, Visual feedback, Position controlAbstractMicromachining progresses rapidly in recent years. In this research, a micro lathe which is installable and operationalinside SEM vacuum chamber has been designed anddeveloped. As a first step, visually guided micro lathesystem is developed with image of CCD camera deviceinstead of SEM image. Unlike the conventional feedbackcontrol which positions the X- Y table only, this scheme offers a direct control of the position, path and speed of thetool tip. Using proposed method, cutting experiment wasachieved, and it is confirmed that developed micro lathesystem is effective to do cutting.1 Introduction Recently, the system capable of producing the micro partsare requested along with the miniaturization1.Micromachining progresses rapidly in recent years. Theexploratory research has approached to a level of accessing asingle molecule or atom. As a driving force, MEMS (microelectronic- mechanical system) has been playing a major rolein making micro components and devices. However, MEMS is based on the photolithography technology and therebyapplicable into limited materials such as siliconmonocrystalline. In orderto meet the demands of miniaturization in electronic and optical applications, alternative micromachining technology which is able toaccess a variety of materials in a 3 dimensional way isrequired2.Micro- Meso Mechanical Manufacturing (M4) offersaccessibility to different kinds of material according to eachobjective, and attains high repeatability and accuracy withthe latest ultraprecision means. There are, however, manyscientific and technological barriers encountered inpragmatic implementation of M4. One of them is the surfacechemistry effects. When machining parts are at micro scale,it is recognized that the surface- area- to- volume ratio will beincreased in both chips and the resulting part as compared to conventional (macro) machining process. Another problem is the direct motion and position control. Sensors that arecapable of directly measuring the relative displacementbetween the tool and workpiece are not yet available.In this research, a micro lathe which is installable andoperational inside SEM vacuum chamber has been designedand developed3. Fig. 1 shows the concepts of the developedmicro lathe. At such oxygen- free condition, cutting tests areconducted to understand surface chemistry effects on micromachining. However, since developed micro lathe issmall in size, rigidity of the lathe is low. Thus the position ofthe tool of the lathe is not able to be controlled accuratelywith a conventional method which controls X- Y table only.Therefore, the vision guided control method is proposed.Theimage from the SEM (scanning electronmicroscope) is digitized by CCD into pixels with 8- bit grayscale. Since each pixel contains 2D positional information,the vision system thus offers an orthogonal coordinate(hereafter referred as the pixel coordinate) for objects inview to refer to. The pixel coordinate is free from themechanical inaccuracy and offers a direct measurement ofSEMCCDMicro latheSEMCCDMicro latheFig. 1. Concept of the developed micro lathe ?H. Ojima, K. Saito, L. Zhou, J. Shimizu, H. Edathe relative position of tool and workpiece. The resolutionincreases together with the magnification of the microscopeand the number of CCD pixels. In this research, a visioncontrol scheme has been proposed and implemented forfeedbackcontrol of the tool movements. Unliketheconventional feedback control which positions the X- Y tableonly, this scheme offers a direct control of the position, pathand speed of the tool tip. As a first step, visually guidedmicro lathe system is developed with image of CCD cameradevice instead of SEM image. 2 Overview of system Actuating moduleSensing moduleProcessingmoduleImage information Actuator signalCapture boardMicro latheXZ stageAMPCPUDiamondtoolCCDWorkpieceD/A boardActuating moduleSensing moduleProcessingmoduleImage information Actuator signalCapture boardMicro latheXZ stageAMPAMPCPUCPUDiamondtoolCCDWorkpieceD/A boardFig. 2. Block diagram of systemTable 1. Specification of systemSize of micro lathe(WDH)909042 (mm)Spindle rotational speed0?8000 (rpm) Depth of cut Traverse feed1010 (mm) Center high adjustment30 (?m)ToolDiamondNose angle / nose radius40() / 2 (?m)Scanning rate20 (frame/s) Total pixels0.3 mega pixel Shown in Fig. 2 is the block diagram of developed microlathe system, which consists of three main modules; the actuating module that drives micro lathe, the sensing module that imports images and the processing module that implements feedback control. Each module is responsible fordifferent function. The actuating module is the core element where the cutting operation is carried out. The sensingmodule imports images from CCD image device, and obtainsthe position of the tool and the workpiece. The other tasks including the image processing and feedback control areexecuted by the processing module. Upper picture of Fig. 1shows the overall appearance of the system. Table .1 showsthe specifications of the system.The actuating module further incorporates a diamondtool with a XZ linear stage, and the sensing module includesa high resolution CCD image device. Through sensingmodule, the appearance of the working area is not only displayed on the monitor to the give the operator the visual information, but also converted into digital signal forsubsequent processing. As the control diagram show in Fig. 2, the movements ofthe diamond tool are governed with the visual feedbackcontrol. The sensing module first abstracts the positions ofthe tool and workpiece by comparing the pre- registeredtemplates with the captured visual information.Corresponding to the relative positions of tool and workpiece, the tool path and speed are calculated and converted intoappropriate pulse train.egfhefghacbdabcd? ?egfhefghacbdabcd? ?Fig. 4. Driving principle of XZ- stageXYZCenter highadjustmentDC motorSpindleXZ- stageMicro latheXYZXYZCenter highadjustmentDC motorSpindleXZ- stageMicro latheFig. 3. XZ- stage and micro lathe3 Actuating moduleThe developed micro lathe is shown rightward in Fig. 3. Thislathe consists of the main spindle with the collet chuck withthe DC motor, the centerhigh adjustment using apiezoelectric actuator and XZ- stage which performs both depth of cut (X- axis) and traverse feed (Z- axis). The XZ-stage is driven by the inertial sliding, and is composed of apiezoelectric actuator and the linear guide.XZ- stage is shown leftward in Fig. 3. An accurate toolpositioning is achieved by driving the XZ- stage precisely.Important points of driving the XZ- stage are the control of the driving direction, distance and velocity. Figure 4 shows the inertial sliding mechanism by the saw- tooth wave. Thedirection of the movement is decided by the rising/trailingedge of the saw- tooth wave as shown in Fig. 4. For example, ?Visual Feedback Control of a Micro Lathethe mechanism in the right direction (+) is explained as follows. The voltage gradually rises, and a piezoelectricactuator stretches most in (1). The actuator shrinks based on the centroid in (2) by falling rapidly of the voltage. Only theside where the frictional force is small moves as the actuatorstretches gradually with the ascent of the voltage in (3). Theactuator is stretches again in (4), and advances toward the right direction. The actuator similarly advances also towardthe left direction (- ) if a reverse pulse train is given.00.10.20.30.40200400600Frequency?HzSpeed?mm/s80V40VFig. 5. Velocity change depending on frequency and voltageNext, the velocity control of this mechanism is described.As shown in Fig. 5, the velocity is proportional to bothfrequency of the pulse train and driving voltage.Finally, driving distance can be controlled according tothe number of pulses, because the driving distance by oneplus is about 500?m at 80V or 250?m at 40V. (500,420)(140,420)(500,60)(140,60)XZ(320,240)4123(500,420)(140,420)(500,60)(140,60)XZ(320,240)4123Fig. 8. Experimental condition of linear path controlX pixelZ pixelcount210121011000200300400X pixelZ pixelcount210121011000200300400Fig. 7. Recognition accuracy of tool tip XZ- stageDiamondtoolCCDWork piece(X,Z)ZXXZ- stageDiamondtoolCCDWork pieceXZ- stageDiamondtoolCCDWork piece(X,Z)ZX(X,Z)ZXFig. 6. Visual sensing system4 Sensing module The diamond tool is mounted on XZ- stage, which uses piezoelectric actuator to drive tool. Those mechanicalinaccuracies,mainly caused by thermal expansion,hysteresis/drift in actuators and misalignment of orthogonal axis, may directly deliver a negative effect to the systemperformance. To solve these problems, a vision controlscheme as shown in Fig. 6 is developed. The left picture inFig. 6 shows the micro lathe and CCD image device locatedin Y- axis. From the right picture in Fig. 6, the incomingvisual information from the CCD is digitized into pixels with8- bit gray scale by the sensing module. As each pixel bears 2D positional information, the vision system thus offers anorthogonal coordinate (referred as the pixel coordinate) forobjects in view to refer to. The pixel coordinate is free fromthe mechanical inaccuracy and its resolution increasestogether with the magnification of the CCD. At a 480640 pixel frame used in the current research,for example, the resolution of the pixel coordinate is about6?m when the view of the CCD is twofold magnified. Whenthe CCD is aligned along Y- axis, the position of the tool tipand workpiece is projected into a 2D pixel coordinate (XZ)which is commonly shared by the XZ- stage and workpiece.Driven and controlled by the pixel coordinate, the tool isable to be positioned and moved at the accuracy of pixelresolution with no effect by the mechanical inaccuracy. Inaddition, if the rigidity between XZ- stage and tool is low,positioning of tool tip is not achieved by driving XZ- stage accurately. Thus, more importantly, this operation is aneffective method of positioning for the micro lathe with alow rigidity.Figure 7 shows the recognition accuracy that is made byuse of shape based pattern matching4 to recognize the actualtool tip repeatedly 500 times. We comprehend from thegraph that 88.5% reliability can be achieved within the limesof 1pixel (6?m).5 Processing module For the system which is consisted of the actuating andsensing module in previous section 3 and 4, the visual? H. Ojima, K. Saito, L. Zhou, J. Shimizu, H. Eda feedback control method is described in this section. The tool tip is driven by visual feedback control method with positions of the tool tip and targets from CCD image device. As a first step, we examined linear path control and circular path control of the tool tip. In these path controls, driving frequency is 300Hz (162?m/s). At first, liner path control of tool tip is described. As shown in Fig. 8, the target position is defined as (320, 240) which is the center of the image from CCD, and four kinds of path control are examined. In the case of liner path control, the angle formed by the target position and the present position of the tool tip is fed back to achieve the path control. Figure 9 (a) shows the resultant path of the tool tip without feedback control, and (b) shows that with feedback control. In the case of the path without feedback, final errors of four paths are between 5pixels (30?m) and 15pixels (90?m). On the other hand, the path with feedback follows along the target path, and final error is within 2pixels (12?m).Next, the circular path control which is multi-axial interpolation is described. The condition of the circular path control is shown in Fig. 10. The center of the target circular path is defined as (320, 240) which is the center of the image from CCD, and the radius of the target path is 100pixels (600?m), moreover the tool tip is driven from starting point (220, 240) along counterclockwise direction repeated 3 times. In the case of circular path control, we consider to feed back not only the angle formed the center of the target circular path and the present tool position, but also the deviation of the radius which is the error between the radius of the target circular path and the distance from the center of the target path to the present tool position. In the case of the driving the path without feedback control, the tool is driven by the angles prepared in advance. Figure 11 (a) shows the resultant path of the tool tip without feedback control, and (b) shows the path with feedback control of the angle only, and (c) shows the path with feedback control of the angle and radius. Figure 11 (a) shows that the resultant path departed from target path, and the center and the radius of the path are deflected from those of the target path. Figure 11 (b) shows the center of the resultant path matches the center of the target path, but extends the radius of the resultant path as the path goes around. Moreover, Fig. 11 (c) shows that the resultant path 0100200300400500100 200 300 400 500 600X pixelZ pixelgoal12340100200300400500100 200 300 400 500 600X pixelZ pixelgoal1234 (a) without feedback control 0100200300400500100 200 300 400 500 600X pixelZ pixelgoal12340100200300400500100 200 300 400 500 600X pixelZ pixelgoal1234 (b) with feedback control Fig. 9. Experimental results of linear path control (220,240)(320,240)XZCCW(220,240)(320,240)XZCCWFig. 10. Experimental condition of circular path control 0100200300400100200300400500X pixelZ pixel0100200300400100 200 300 400 500X pixelZ pixel0100200300400100200300400500X pixelZ pixeltooltarget tooltarget tooltarget 0100200300400100200300400500X pixelZ pixel0100200300400100 200 300 400 500X pixelZ pixel0100200300400100200300400500X pixelZ pixeltooltarget tooltarget tooltarget tooltarget tooltarget tooltarget (a) without feedback control? (b) with feedback control of the angle (c) with feedback control of the angle and radius Fig. 11. Experimental results of circular path control ?Visual Feedback Control of a Micro Lathefollows the target circular path closely, and errors are5pixels (30?m). From mentioned above, it is confirmedthat proposed feedback control method is effective toposition the tool tip of the micro lathe.Fig. 14. SEM picture of a resultant brass (4)(1)(2)(3)(4)(1)(2)(3)Fig. 13. Experiment of cutting a brass bar (300,180)(240,300)R=60(404,240)(300,300)30?XZ(c)(b)(a)30?(300,180)(240,300)R=60(404,240)(300,300)30?XZ(c)(b)(a)30?Fig. 12. Tool path strategyFinally, using the proposed control method, cutting of a brass bar is experimented. As shown in Fig. 12, the tool tipis driven with circular and linear paths. The target pathmoves 1pixel (6?m) rightward every 1 lap, then the tool tipis achieved to cut. In this experiment, total depth of cut is150?m. Figure 13 shows theappearance of cuttingexperiment, and cutting of a brass bar advances from (1) to (4). Figure 14 shows the resultant brass bar which is taken apicture by SEM. From this picture, developed micro lathesystem can implement cutting well.7 Conclusion This paper described a vision guided micro lathe which isdeveloped for fundamental research. The system consisted ofthe actuating module and the sensing module was made upand the visual feedback control method was proposed.Driving principle of the XZ- stage of the micro lathe wasinvestigated, and control method of the XZ- stage wasproposed. Using the CCD image device and image processing, the accurate position of tool and workpiece onthe micro lathe was obtained. The visual feedback controlmethod for the system cosisted of the actuating module andthe sensing module was proposed. Using the proposed control method, linear path control and circular path controlwas experimented and was able to be controlled that error iswithin the limes of 5 pixels (30 ?m). Moreover, cuttingexcperiment of a brass bar was achieved, it is confirmed thatdeveloped micro lathe system is effective to do cutting.8 References1Okazaki Y., (2001) NC micro lathe, Journal of Societyof Grinding Engineers (in Japanese), Vol.45, No.6:279- 282.2Libo Z., Kuriyagawa T., (2002) M4 trends in America, Journal of the Japan Society for Abrasive Technology(in Japanese), Vol.46, No.12: 594- 597.3Yokoyama I., Nishida K., Zhou L., Eda H., KawakamiT., (2003) Experiment & Design of Ultra- precisionMachine tool, Proceedings of IbarakiDistrictConference (in Japanese) : 165- 1664Qiu Z., Zhou L., Nishida K., Shimizu J., Eda H.,IshikawaT., (2004) Development of Cell Manipulation System under Microscope, Proceedings of the 1st International Conference on Positioning Technology : 457- 461?