Design and feasibility tests of a flexible gripper based on inflatable rubber pockets.pdf

International Journal of Machine Tools fax: +6465147590.E-mail address: (M. Koc-).delicate form of material handling. But, for repetitivecycles, heavy loads and under extreme environments,grippers had to be developed to substitute for humanhands. In the 1960s, after the emergence of modern robots,grippers replaced human hands on numerous occasions.Robot-gripper systems are found to be effective forrepetitive material handling functions in spite of theirinitial capital and ongoing maintenance expenses becauseof their reliability, endurance and productivity. However,variety and rapid response. flexible and reconfigurablemanufacturing systems (FMS and RMS) have emerged asa science and industrial practice to bring about solutionsfor unpredictable and frequently changing market condi-tions 2. In order to fully realize the benefits of RMS andFMS, the grippers, being one of the few direct contactswith the product at the very bottom of the manufacturingchain, must also be designed for flexibility.In the early days of robotic technology applications,understanding about the capability and limitations of the gripper. It was found that objects with different shapes (cylindrical, prismaticand complex), weight (50g20kg.), and types (egg, steel hemi-spheres, wax cylinders, etc.) could be picked and placed without any loss ofcontrol of the object. The range of positioning error for two different part shapes (i.e., prismatic or cylindrical) was found to be 2090mm(translational) and 0.030.91 (rotational).r 2005 Elsevier Ltd. All rights reserved.Keywords: Gripper design; Strategies; Flexible; Selection; Robotic; Rubber1. IntroductionA gripper is an end-of-arm tooling used on robots forgrasping, holding, lifting, moving and controlling ofcost of a suitable gripper may even go higher since theyrequire additional controls, sensors and design needs withregards to being able to handle different parts.In the 21st century, under the influences of globalization,pocketed flexible gripper was designed and built. Feasibility experiments were performed to demonstrate and obtain an overallDesign and feasibility tests ofinflatable rubberHo Choi, MuammerNSF ERC on Reconfigurable Manufacturing Systems, College ofReceived 30 January 2005;Available onlineAbstractIn this paper, we present feasibility test results of a flexible grippercontrol strategies of the existing grippers. A flexible gripper basedture 46 (2006) 13501361a flexible gripper based onpocketsKoc-C3g, University of Michigan, Ann Arbor, MI 48109, USA19 October 2005r 2005design following a literature survey on various types, design andon the use of compliant materials (i.e., rubber) with /locate/ijmactoolARTICLE IN PRESSdrawbacks, cost effective flexible gripper designs havebeen always sought as a viable solution for agile materialhandling systems as an important element of the envisionedFMS and RMS. For example, assembly operations inmany industries make extensive use of dedicated grippersand fixtures. These are part-specific, and therefore, must bemodified or replaced when model changes are introduced.The cost of redesigning, manufacturing, and installingthese grippers and fixtures is substantial (on the order of$100 million per plant per year for automotive manufac-turers) and would be significantly reduced if a more flexiblealternative was developed.In this paper, following an extensive review anddiscussion on different gripper types and design issues inthe first section, a flexible gripper design based on the useof compliant materials and internal pressure (i.e., inflatablerubber pockets) approach is introduced in the second part.This type of grippers conforms to the shape of an object bymeans of elastic gripping elements and pressurization withactive degrees of freedom. In the third section, the resultsof a parametric FEA study are presented to characterizethe performance of the selected configurations of theflexible gripper under different loading and part conditionsto determine the proper parameters setting and thematerial. Finally, in the fourth section, following theprototyping, feasibility tests conducted to characterize thelimits and capabilities of the flexible gripper are explained.2. Literature survey on gripper design and types2.1. Design methodology of grippersWright et al. 3 compared the grippers to the humangrasping system, and categorized the design requirementsof grippers into (a) compatibility with the robot arm andcontroller, (b) secure grasping and holding of the objects,and (c) accurate completion of the handling task. Manyindustrial examples of grippers were also described, and theguidelines for gripper design were presented. Pham et al. 1summarized the strategies for design and selection ofgrippers in different application cases. In their study, thevariables affecting the selection of a gripper were listed as:(a) component, (b) task, (c) environment, (d) robot armand control conditions. The component variables includegeometry, shape, size, weight, surface quality and tempera-ture of objects to be handled. For reconfigurable systems,they divided these components into part families accordingto their shape and size. For the task variables, type ofgripper, number of different parts, accuracy, and cyclewere considered in addition to major handling operationssuch as pick, hold, move and place. For the right gripperdesign at the right place, all aspects should be considered,and multiple validation tests should be conducted. Toreduce this exhaustive effort, Pham et al. 4 developed anexpert system for selecting robot grippers. They built aH. Choi, M. Koc- / International Journal of Machinehybrid expert system that employs both rule-based andobject-oriented programming approaches.2.2. Gripper types and classification by driving forceGrippers could be also classified with respect to theirpurpose, size, load, and driving force. Typically, grippermechanisms and major features are defined by their drivingforces. The driving forces for robot grippers are usuallyelectric, pneumatic, hydraulic; or in some cases, vacuum,magneto-rheological fluid and shape memory, etc.Grippers with electric motors have been used since 1960,abreast with robot technology. Many other grippersadopted motor driven mechanisms. Basically, this type ofsystems included step motors, ball screws, encoders,sensors and controllers. As the arms approach the object,distance, force, weight and slip are detected by sensors. Atthe same time, a controller regulates the force, speed,position and motion. Friedrich et al. 5 developed sensorygripping system for variable products. They used multiplesensors to measure the grasping force, weight and slip.Mason et al. 6 and Kerr et al. 7 presented thefundamentals of grasping with multi-fingered hands. Leeet al. 8 comprehensively reviewed the field of tactilesensing. For contact and slip, Tremblay et al. 9 consideredslip detection, and Howleg et al. 10 divided slip into fourstages; pre-slip tension, slip-start, post-movement, and stopto better analyze grasping of parts.Another way of actuating the robot gripper is throughpneumatic (or hydraulic) systems. Pneumatic systems havebeen developed because of their simplicity, cleanliness andcost-effectiveness. Warnecke et al. 11 and Wright et al. 3developed a soft pneumatic gripper which can handle softmaterials such as eggs. Ottaviano et al. 12 developedgrasp-force control in two-finger grippers with pneumaticactuation. They proposed force control in a two-fingergripper with a sensing system using commercial forcesensors. A suitable model of the control scheme has beendesigned to control the grasping force. Experimentsshowed the practical feasibility of two-finger grippers withforce controlled pneumatic actuation 12. Lane et al. 13used hydraulic force for a sub-sea robot hand. They offerednatural passive compliance to correct for inevitablepositioning inaccuracy with simple design and minimummoving parts. The gripper finger relied on the elasticdeformation of cylindrical metal bellows with thin con-voluted walls. The convolution ensured that the assemblywas significantly stiffer in the radial direction than thelongitudinal one. Therefore longitudinal extension wasmuch greater than radial expansion when subjected tointernal hydraulic pressure. The modular finger tipcontained a variety of sensors and interfaces. The fingertip contact zone contains both a strain gage and apiezoelectric vibration sensor. Closed-loop position controlwas used. It was driven by hydraulic pressure measuredfrom sensors within each tube 13.Grippers based on vacuum forces are designed and usedmainly for deformable and lightweight part handling.Tools (a) three rows ofpocket design. In all cases, note the multiple holes and pins on the uppersensor array. For an integrated robotic gripper system,they also suggested hierarchical control architecture andfuzzy logic formulation. For manipulating payloads withmultiple robots, Sun et al. 27 described an approach tonon-model based controls of multi-robot systems.2.4. Flexible gripping strategiesIn terms of accomplishing flexible gripping tasks, fivedifferent strategies were suggested by Pham and Yeo 1 toachieve the flexibility in a cost-effective manner. The firstkind of gripper gains its flexibility from a number ofnotches on the gripping surfaces so that objects withvarious shapes can be handled. Obviously, notching ofgripper fingers is only suitable for the parts of similar sizeand weight. Another flexible gripper concept is based oninterchangeable gripper fingers. This method is moreflexible and reliable than notching method when thegripper is equipped with a finger changing apparatus anda standard set of fingers. Another strategy is to change thegripper itself. This method can be used when a singlegripper cannot handle a whole set of parts with differentsizes, geometries and weights. These grippers need variouspayload applications such as electronics and small precisemachining.Universal grippers are also suggested to be a viableway of handling a wide range of objects with differentshapes and weight 1. The universal grippers are groupedinto two categories: active and passive grippers. Passivegrippers automatically conform to the shape of the objectsby means of gripping elements which are elastic or havepassive degrees of freedom. It was reported that, withpassive gripping, it is difficult to ensure the precise positionof the gripped object with respect to the robots coordinateARTICLE IN PRESS00.511.522.533.544.550 100 200 300 400 500 600 700Strain (%)Stress (MPa)100% modulus1.7 MPa300% modulus2.9 MPaFig. 3. Flow stress curve of neoprene rubber material used in the FEanalyses 27.Table 1Coefficient of friction of neoprene rubber with different part materialsMaterial Coefficient of static friction on rubberWax 0.5970.01Aluminum 0.6670.01Steel 0.6970.01H. Choi, M. Koc- / International Journal of Machine Tools (a) selected conceptual model, (b) assembled gripper, and (c) gripper attached to a robot.H. Choi, M. Koc- / International Journal of Machine Tools & Manufacture 46 (2006) 135013611356mm, c 70mm, D 45264mm, DD is the initial distance between jawflexible griper and (b) side view of a jaw.two values and the geometry, average friction coefficientswere calculated. Similarly, the material model of rubberwas constructed based on the data obtained from MatWebfor neoprene 29 as depicted in Fig. 3. A hybrid elementtype (C3D8H, fully incompressible hyper-elastic material)was used to model the rubber material. The thickness of therubber was considered to be 1.6mm whereas the side platethickness was 18mm. Fig. 4 illustrates the FE model,definition of the initial conditions (i.e., initial distancebetween jaws), and the measurement of vertical partdisplacement during gripping (Dy), whereas Table 2presents the factors considered in the FE analyses (valuesof initial distance between jaws, pressure and part weight),their ranges and levels as well as the response.ARTICLE IN PRESSFig. 10. Illustration of pick and place operation with an egg without damaging or breaking it.Fig. 9. Tested part types (shape and weight).H. Choi, M. Koc- / International Journal of Machine Tools & Manufacture 46 (2006) 13501361 1357Table 5Specifications of the parts used in the feasibility testsPart type MaterialHemispherical part, Fig. 9a SteelStepped cylindrical part, Fig. 9b AluminumPrismatic wax part, Fig. 9c WaxCylindrical wooden part, Fig. 9d WoodEgg, Fig. 10Weight Size5kg 60C290C270mm (w, h, d)1kg 60C290mm (OD, h)67gr 30C230C260mm (w, h, d)300gr 70C290mm (OD, h)50gr 30C250mm (OD, h)(as(DDlarlysufficidisplwhenThe designed flexible gripper operates in two phases. First,it roughly encloses the part with parallel-jaws (size plates)approaching to the part without any contact. Secondly, therubber pockets attached to parallel-jaws are inflated tograsp and hold the part securely. By controlling the initialjaw displacement (DD) and pressure (P), various partshapes and weight can be handled accurately and softly.4. Feasibility tests on the flexible gripperVarious tests at different levels were performed to verifythe feasibility of the flexible gripper. First, in order todemonstrate that it could handle a variety of part shapes,parts with different shapes were picked, moved and placedwithout any handling problems such as slippage, dropping,or breakage, as shown in Fig. 9. For instance, around twodozens of eggs were picked and placed repeatedly withoutdropping and damaging as depicted in Fig. 10. In thesecond level, we tested the sensitivity of the gripperparameters on different part types. We changed thepressure between 30 and 80KPa, initial jaw distanceARTICLE IN PRESSDD (mm) 7.5 4 4Tools & Manufacture 46 (2006) 13501361In the second set of FE analyses, elasticity of the jaws (sideplates) was also taken into account in addition to the effectsof double-rubberpocketand non-flatparts. Thus, besides thevertical displacement (Dy) of the part, horizontal displace-ment (Dx) of the jaw was also measured and reported in theFE analyses. Three different cases of (DD, P and W)wereconsidered as presented in Table 3 and Fig. 6.Asitcanbeobserved in this figure, elasticity of jaws affect the parthandling accuracy considerably when Fig. 6a and b arecompared for a single pocket and flat part type case.Regardlessof the processconditions (DD, P, W), verticalandhorizontal displacements increase 48 times (to 2040mmfrom 5mm and below). It should be noted, in the case of rigidjaws (Fig. 6a), there is no horizontal displacement. A similartrend in the increase of part displacements is also observedfor non-flat part as depicted in Fig. 6c. However, upon use ofdouble-rubber pockets, as shown in Fig. 6d, displacementerrors particularly in the vertical direction are reduced downto 10mm levels. Horizontal displacement is slightly decreasedby the use of double pockets since the rubber materialconform to the part shape. In summary, as a result of the FEanalyses, we can conclude that:(1) The concept of inflatable rubber-pockets for flexiblegripping is feasible.(2) Range of placement errors was predicted to be between40 and 75mm under different pressure and part weightcircumstances.(3) Multiple rubber-pocket approach is potentially moredesirable to reduce part handling errors.(4) However, manufacturing (i.e., molding) of small mush-room-like pockets would be too costly and lengthy.Therefore, for prototyping and feasibility experiments,a flexible gripper with single-rubber pockets wasselected due to its simplicity, relatively low cost andeasy interchangeability and maintainability.Detailed specifications of the designed gripper are listedin Table 4. A three-dimensional rendering of the flexiblemorebetweengripperarethat optimal conditions may change for heavier andcomplex-shaped parts since the friction conditionsthe part and rubber would change.jawnotedg), optimal handling conditions are found to bethe pressure is at low levels (50KPa) while the initialdisplacement is around 67mm. However, it should bepart(C24100500pressure (P) is at its medium levels (6070KPa) for aweight of around 1.5kg (W). When the part is lightinitialandshown in Fig. 5) excessive initial distance between jawsm) leads to poor handling (Dym) of the part particu-when the part is heavy (Wm) and pressure is notent (Pk). The optimal handling (i.e., minimal partacement, Dy) conditions can be achieved when thedistance between the jaws (DD) is around 78mmAs a result of the first set of FE analyses conductedaccording to the matrix in Table 2, we can conclude thatH. Choi,