[机械模具数控自动化专业毕业设计外文文献及翻译]【期刊】现代工业机器人的精度评估-外文文献

Accuracy assessment ofthe modern industrialrobotKen Young andCraig G. PickinIntroductionSince its first appearance in the early 1980s,interest in offline programming (OLP) hasgrown steadily as its benefits have becomeapparent. OLP offers the robot programmer anumber of key benefits, most notably areduction in the time it takes to design andprogram an automated robotic system. Withtraditional teach pendant methods the robotmust be taken out of production whenmodifications or new programs are beinggenerated. A demonstration by Chrysler hasshown that by utilising OLP the time a robotis out of production can be reduced frombetween 12-18 hours to 6.5 hours for eachrobot. Additional figures suggest thatprogramming using OLP can reduceprogramming time by up to 85 per cent forsmall batch programs (Wittenberg, 1995).However, due to problems associated withaccurately modelling both the robotworkspace and the robot itself, OLP is oftenstill used as part of a two part programmingmethodology utilising traditional teachpendant methods for on the job moves.When considering the robot, the mainproblem with respect to OLP is in creatingan accurate kinematic model. Factors suchas machining tolerances in the robotlinkages, compliance and elasticity in therobot arm, encoder resolution, and the lackof repeatability during calibration serve togive each robot a unique signature (VanBrussel, 1990). The term calibration in thisrespect and throughout this paper, unlessotherwise stated, refers to the method ofcommissioning the robot. Typically robotsare factory calibrated using a fixture todetermine absolute joint positioning. Thesepositions are then marked on the robot armlinkages to aid in future re-calibration onthe shop floor.Re-calibration may be necessary duringmaintenance or when calibration data havebeen lost due to power failure. Whencalibrating in this environment, the marksare aligned, possibly in conjunction with afixture, and the absolute joint position set.The robot is then considered to becalibrated.Whereas it is accepted that industrial robotsexhibit good repeatability, robot accuracy israrely quoted and is generally believed to bequite poor, errors commonly being severalThe authorsKen Young is Principal Research Fellow andCraig G. Pickin is a Research Fellow, both atWarwick University, Coventry, UK.KeywordsRobots, Accuracy, CalibrationAbstractThe main drawback to programming robots offline lies inthe poor accuracy of the robot. Robots are mainlyprogrammed using the traditional teach and repeatmethod of programming which requires only goodrepeatability. As a result robots are manufactured withthis in mind. Little is done to improve or even quotefigures for the accuracy, which is generally regarded asbeing poor. A trial has been conducted on three modernserial linkage robots to assess and compare robotaccuracy. Using a laser interferometry measurementsystem each robot has been measured in a similar area ofits working envelope. The results and conclusions fromthis trial show that compared to older robots the accuracycan be remarkably good though it is dependent on acalibration process which is far from robust.Electronic accessThe research register for this journal is available at/research_registers/aa.aspThe current issue and full text archive of this journal isavailable atFeature427Industrial Robot: An International JournalVolume 27.Number 6.2000.pp. 427436# MCB University Press.ISSN 0143-991Xorders of magnitude worse than the robotrepeatability (Mooring et al., 1991). Inaddition, it will usually be found that robotrepeatability remains within its tolerancethroughout the working envelope whereas theaccuracy of the robot may deterioratesignificantly towards the boundaries of theenvelope.Although this situation is far from ideal, ithas in the past been accepted. By utilising thetraditional teach and repeat method ofprogramming, these accuracy errors have noinfluence on the positioning of the robot,repeatability being the important factor.Unlike OLP, this method does not require aprecise Cartesian co-ordinate system. Movingthe robot to its desired position using theteach pendant and recording this position willgenerally guarantee that the robot will returnto this taught position to within a fractionof a millimetre.By programming this position offline it islikely that the final position, due to thesignature effect and lack of preciseco-ordinate system, may be a number ofmillimetres away from the position that isdesired. Based on these factors, it is clear thatrelying solely on OLP methods utilising anominal kinematic robot model alone canresult in serious positional errors.One recent solution to this problem hasbeen the development of calibration tools(Owens, 1994; Schroer, 1994), which resultin a more advanced method of calibratingthe simulation model. These calibrationtools generally involve mapping theaccuracy errors inherent in the real robotworking envelope and using themeasurement data to produce a moreaccurate kinematic model for use in OLP.While the claims behind these systems areimpressive, it can be argued that they arebypassing the fundamental problemsassociated with real robot accuracy, that is,those associated with the calibration of therobot during commissioning andmaintenance.A recent study at Warwick ManufacturingGroup (WMG) based at the University ofWarwick has measured and compared theaccuracy of a number of industrial robots.This has been conducted with a view toassessing their suitability for use in OLP.MethodA total of three six-axis serial linkage robotshave been measured, namely the CloosRomat 310, The ABB IRB 6400S, and theKUKA KR125 (Figure 1). All three robotsare measured in their unloaded state, and arecommissioned using the manufacturerscalibration methods.With reference to the relevant UK and ISOstandards (BS 4656, 1986; ISO 9283, 1998) atesting framework has been established.Using a Renishaw laser interferometrysystem, accuracy and repeatabilitymeasurements are made within a tolerance of+ 0.001mm. Each measurement is taken aspart of a bi-directional run. These areconducted a total of five times with aminimum of five points. An overrun point isprogrammed at the end of each measurementto remove the effect of backlash whenchanging direction. However, due to thelimitations of this measurement system onlystatic measurements in a particular area of theenvelope are taken. The measurements takenare as follows:.Linear positional accuracy in both the Xand Y axes. These are represented byraw data plots showing relativepositional measurements. A systemdatum is taken at the axis zero point forthe robot co-ordinate system.Figure 1 The measured points for a sample run on the Kuka robot428Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436.Straightness measurements showingdeviation in the X-axis and deviation inthe Z-axis when travelling in the Ydirection. These are represented using theleast squares fit approach. Due todifficulties in aligning the laser systemwith the robot co-ordinate system acorrect datum line is not possible. Theleast squares fit approach involvesdefining a datum such that the sum of thesquares of the distances from the datapoints to the line (graphically measuredparallel to the Y-axis) is at a minimum.Graphically this approach is representedshowing the mean straightness error forboth positive and negative runs, inaddition to the standard deviation forboth sets of run.ResultsCloos Romat 310The linear positional measurements for theCloos Romat 310 are shown in Figures 2and 3. Both measurements have similarcharacteristics for the X and Y-axis. While therepeatability is within the tolerance of 0.1mmspecified by the manufacturer, positionalaccuracy deteriorates quite significantly inone half of the measured envelope for bothsets of measurements. Errors for both axes arewithin an error band of 1.7mm although forone side of the measured envelope this error iswithin a 0.5mm band.The straightness measurements for theCloos Romat 310 are shown in Figures 4 and5. Deviation in the X-axis when travelling inthe Y direction is within an error band ofapproximately 0.7mm across the measuredenvelope. In a similar way to the linearpositional measurements, however, theaccuracy deteriorates in one half of theworking envelope. This error band isapproximately twice that for the opposingenvelope. In contrast, deviation in the Z-axiswhen travelling in the Y direction is similar forboth sides of the envelope, this error bandbeing in the region of 0.4mm.ABB IRB 6400SThe linear positional measurements for theABB IRB 6400 robot can be seen in Figures 6and 7. Accuracy in the Y direction lies withinan error band of approximately 0.5mm, thisbeing roughly constant in both sides of theenvelope. The error band is almost entirelydue to backlash with an indication that thismay be improving towards the boundaries.Due to limitations in the workspace for thisrobot, caused by its proximity to a wall, areduced envelope has been measured in theFigure 2 Linear positional accuracy in the Y direction for the Cloos Romat 310429Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436X direction. Although direct comparisoncannot be made with the measurements takenin the Y direction because of this, similarcharacteristics are evident. Measurements inthis direction show an error band within0.5mm although most of the envelopeexhibits a constant error within a 0.3mm bandwith deterioration towards one side of theenvelope.The straightness measurements for travelin the Y direction for the ABB IRB 6400Sare shown in Figures 8 and 9. Both exhibitsimilar characteristics but deviation in theX-axis has a larger error band than deviationin the Z-axis. This is roughly four timesthat of the Z-axis, maximum error beingwithin bands of 0.4mm and 0.1mmrespectively.Figure 3 Linear positional accuracy in the X direction for the Cloos Romat 310Figure 4 Straightness showing deviation in the X-axis when travelling in the Y direction for the Cloos Romat 310430Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436KUKA KR125Linear positional accuracy for the KUKAKR125 is shown in Figures 10 and 11.Accuracy in the Y direction is within an errorband of 1.8mm with much of this error beingevident in one side of the envelope. The errorfor the X direction is slightly less, being withina band of 0.8mm although like themeasurements for the Y direction, the errorband is greater in one half of the envelope.Repeatability is within the manufacturersstated tolerance of 0.2mm.Figure 5 Straightness showing deviation in the Z-axis when travelling in the Y direction for the Cloos Romat 310Figure 6 Linear positional accuracy in the Y direction for the ABB IRB 6400431Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436Straightness measurements when travellingin the Y direction for the KUKA KR125 areshown in Figures 12 and 13. Deviation inthe X-axis is within an error band of 0.7mmand is almost symmetrical about the centreof the measured envelope. In a similar way,deviation in the Z-axis is almost symmetricalabout the centre of the envelope althoughwith a smaller error band, this being within0.2mm.Discussion of resultsThe results from this trial show that of thethree robots measured, none exceeded aFigure 7 Linear positional accuracy in the X direction for the ABB IRB 6400Figure 8 Straightness showing deviation in the X-axis when travelling in the Y direction for the ABB IRB 6400432Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.4274361.8mm error band in any of the threedimensions measured, with repeatabilitybeing within the specified tolerance. Whilethe Cloos Romat 310 and the KUKA KR125 exhibited similar error bands, the ABBIRB 6400 appears to be the most accurate,never exceeding a 0.5mm error band in anyof the three dimensions measured. It isnotable that for the linear positionalmeasurements for the Cloos Romat 310 andthe KUKA KR 125, robot accuracydeteriorates significantly in one half of themeasured envelope, while the samemeasurements for the ABB IRB 6400 havean almost constant error band throughoutthe envelope. The likely reasons for theseerrors are calibration errors on one ormore joints.Figure 10 Linear positional accuracy in the Y direction for the KUKA KR125Figure 9 Straightness showing deviation in the Z-axis when travelling in the Y direction for the ABB IRB 6400433Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436When considering the calibration errorsresponsible for the accuracy errors evidentin this trial it is important to analyse theeffect of each joint in turn. Analysis of theeffect of a small angular error in any of thejoints shows that errors in joints 1-3 makethe major difference for positional errors,with joints 4-6 mainly affecting theorientation. Any error introduced into joint1 has the effect of rotating the robot co-ordinate system without affecting the overallaccuracy of the robot as measured in thismanner. It is therefore apparent that theerrors found in these results are due toerrors being introduced by poor calibrationon joint 2 and joint 3. Errors in these twojoints tend also to introduce errors whichshow up as straightness errors centred onthe centre of the axis system. These errorsare evident in Figures 6, 12 and 13.Conclusions and recommendationsThis trial has established the accuracy andrepeatability errors inherent in the modernindustrial robot. While the trial is somewhatlimited in that only static measurements aretaken in a particular area of the envelope, itdoes serve to give an indication of theseerrors. The errors are not large and are lessthan those recorded for older robots(Mooring et al., 1991) and significantlylower than expected when the trials began.The results show that OLP can in fact be aviable method of programming for certaintasks where this level of accuracy isacceptable. Any improvement in robotaccuracy, however, would be welcomedwhen utilising OLP. Whereas certainsignature errors like backlash can onlyrealistically be rectified by increasing thecost of manufacture, much of the accuracyerror, it is felt, can be reduced by a moreaccurate method of calibration whencommissioning the robot.The serial link robots used in this trial useone of two methods of calibration. This iseither by defining the axis absolute positiondatum using a calibration fixture, or bylining up scribe lines on each axis to achievethe same aim. The KUKA KR125 measuredas part of this study for example utilises acalibration fixture. The accuracy in spite ofthis is comparable to the two robotscalibrated using the scribe line method.This would indicate that both the toleranceinherent in the fixture and its method offixing to the robot are introducing an errorinto each joint axis. In a similar way, it isdoubtful that the scribe line method used onthe Cloos and ABB robots is repeatable.Figure 11 Linear positional accuracy in the X direction for the KUKA KR125434Accuracy assessment of the modern industrial robotKen Young and Craig G. PickinIndustrial Robot: An International JournalVolume 27.Number 6.2000.427436Accurate and repeatable positioning of theselines by the manufacturer is unlikely, as is arepeatable method of aligning th