Grasping devices and methods in automated

The final version of the present paper can be found at http://www.sciencedirect.com/science/article/pii/S0007850614001887

Please cite the paper as: Fantoni, G., Santochi, M., Dini, G., Tracht, K., Scholz-Reiter, B., Fleischer, J., Lien, T.K., Seliger, G., Reinhart, G.,Franke, J., Hansen, H.N., Verl, A.,2014, Grasping devices and methods in automated production processes, CIRP Annals - Manufacturing Technology, Volume 63, Issue 2, 2014, Pages 679-701, ISSN 0007-8506, http://dx.doi.org/10.1016/j.cirp.2014.05.006.

Grasping devices and methods in automated production processes

Gualtiero Fantoni (2)a, Marco Santochi (1) a, Gino Dini (1) a, Kirsten Trachtb, Bernd Scholz-Reiter (1)c, Juergen Fleischer (1)d, Terje Kristoffer Lien (1)e, Guenther Seliger (1)f, Gunther Reinhart (1)g, Joerg Franke (2)h, Hans Nørgaard Hansen (1)i, Alexander Verl (2)l 1

Abstract

In automated production processes grasping devices and methods play a crucial role in the handling of many parts, components and products. This keynote paper starts with a classification of grasping phases, describes how different principles are adopted at different scales in different applications and continues explaining different releasing strategies and principles. Then the paper classifies the numerous sensors used to monitor the effectiveness of grasping (part presence, exchanged force, stick-slip transitions, etc.). Later the grasping and releasing problems in different fields (from mechanical assembly to disassembly, from aerospace to food industry, from textile to logistics) are discussed. Finally, the most recent research is reviewed in order to introduce the new trends in grasping. They provide an outlook on the future of both grippers and robotic hands in automated production processes. Keywords: Assembly, Automation, Grippers 1.

Introduction

In the last ten years several factors such as the increasing cost of human labour, the spread of automation and the decreasing cost of robotic systems have pushed both industry and academia towards the development of new grippers and robotic hands. While in the past robot hands and industrial grippers were oriented to different goals, nowadays it is often difficult to distinguish a simplified robotic human-like hand from a complex industrial gripper. The fast growth in the field and the development of new grasping technologies merits a review of grasping devices and methods in production processes. In addition the world economic crisis is pushing automation towards new frontiers asking for more flexible, versatile, lightweight, and small grippers able to perform more functions (e.g. fixtureless assembly [136]) than simple grasping and holding during handling [148] (Figure 1). This work contributes to complete some interesting surveys obtained in [113] and [138], and continues the CIRP focus on automatic handling of parts at all scales in different industrial environments [212][121].

Department of Civil and Industrial Engineering, University of Pisa, Pisa Italy Institute for Mechanical Engineering (bime), University of Bremen, Badgasteiner Straße 1, 28359 Bremen, Germany c Department of Planning and Control of Production Systems, BIBA, University of Bremen, Germany d wbk Institute of Production Science, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany e Department of Production and Quality Engineering, Norwegian University of Science and Technology, Trondheim, Norway f Department of Assembly Technology and Factory Management, Institute for Machine Tools and Factory Management, Technical University Berlin, Germany g Institute for Machine Tools and Industrial Management (iwb), Technische Universitat Munchen, Munich, Germany h Institute for Factory Automation and Production Systems, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany i Department of Mechanical Engineering, Technical University of Denmark, Lyngby, Denmark l Institute for Control Engineering of Machine Tools and Manufacturing Units, Stuttgart, Germany a

b Bremen

1/47

2.

The grasping process

The complexity of the grasping process is often underestimated since it looks very familiar for human beings. However the automation of this process creates many problems. In fact the design of a gripper does not depend only on the object characteristics but it is also affected by previous phases as feeding and the following phases such as handling, positioning and releasing. In general correctly fed parts require less versatile grippers with respect to a bin picking situation where the gripper has to face problems such as pieces with different orientation, part tangling, etc. Similarly, handling needs such as high acceleration, reorientation, high precision releasing generate constraints in the gripper design or choice [60]. Neglecting the further requirements due to feeding and handling, the grasping process can be generally described as follows (Figure 1):  Approaching the object: the gripper is positioned nearby the object.  Coming into contact: the contact is achieved. In case of contactless handling, the object is in the range of the force field generated by the gripper.  Increasing the force within certain limits.  Securing the object: the force stops increasing when the desired degrees of freedom of the object are removed and the object stops moving independently from the gripper.  Moving the object. In such conditions the gripper and the object are joined and the object can be moved. Sometimes the process can be carried out by the gripper itself.  Releasing the object. Usually at the macroscale it is caused by gravity when the grasping force is deactivated. At the microscale the problem is more complex since surface forces overcome gravity, therefore other releasing strategies are needed. Monitoring the grasping: force and torque sensors, stick slip sensors, contact sensors, etc. can be used to detect and monitor all the process and particularly the effectiveness of grasping. 2.1 The grasping principles

Figure 1. Typical phases of the grasping process [57]. (Figure reproduced with permission of SAGE Publications Ltd)

The design of an industrial gripper must ensure a secure, robust and reliable grasping. Several grasping principles (Figure 2) have been proposed in the last decades, some of them mimicking the human fingers or animals’ claws or jaws, or exploiting different physical effects. Some principles can be applied only at the microscale (e.g. acoustic levitation or laser tweezers), while others proposed for microhandling are now expanding beyond that field (e.g. van der Waals forces). The grasping principle can be defined as “the physical principle which causes the force effect necessary to get and maintain the part in a relative position with respect to the gripping device” [201]. Mechanical grippers are the most widespread: they are based on friction or on form closure, but also intrusive grippers belong to that class. Suction based grippers and magnetic grippers dominate the automotive field and in particular metal sheet handling [138]. Bernoulli grippers work on the basis of 2/47

airflow between the gripper and the part that causes a force which brings the gripper and part close together. This is now receiving more attention since it acts as a vacuum system, but without coming into contact with the handled part. Other principles are less used in the macro domain, but in the last ten years they have led to interesting applications in micro-handling so that now research teams are trying to exploit them to grasp standard objects. Electrostatic grippers are based on charge difference (sometimes induced by the gripper itself) between the gripper and the part [18], while van der Waals grippers are based on the low force (electrostatic forces) due to the atomic attraction between the molecules of the gripper and those of the object [212].

Figure 2. Grasping principles (adapted from [201] (Figure a reproduced with permission of IFIP)

While capillary grippers use the surface tension of a liquid meniscus between the gripper and the part [122], cryogenic grippers freeze a small amount of liquid and the resulting ice produces the required force [78]. Other grippers are even more complex: for example the ultrasonic based grippers generate standing pressure waves used to lift up a part [166], while laser sources can produce an optical pressure able to trap and move microparts in a liquid medium (optical tweezers [212]). In friction or jaw grippers (Figure 3) grasping is based on the generation of a normal force between the gripper and the object. Such a force generates a corresponding static friction force when the object is lifted. In general the fingertips have a shape which replicates the object section or profile and can provide a self-centring/aligning capability or, in some other cases, can slightly adapt to different objects through pins or soft or rubber surfaces on the fingers. Sometimes even theoretically it is not trivial to distinguish a force gripper (friction) from a form gripper (jaw). Jaw grippers can have also more than two fingers: three-finger grippers are used in industrial applications as for example for grasping and aligning cylindrical parts while for example six-finger grippers are used for assembling rubber o-rings. The most common jaw or friction grippers have at least one movable and one fixed finger, but more often they have two collaborating fingers [138][201].

3/47

Figure 3. Mechanical grippers: a) two fingers; b) piezoelectrically actuated microgripper [25]. (Figure 3a reproduced with permission of Schunk)

For some of the grasping principles, a variety of actuation principles can be used. In the case of jaw or friction grippers their actuation can be electric, pneumatic, hydraulic, and thermal (e.g. in the case of Shape Memory Alloys) or piezoelectric in the case of microgrippers. Since the working principle is based on the application of a force over an object, the deformation induced by the force can imprint high value leather plies or textile, can damage delicate food or wear sensitive parts as silicon wafers, lenses, etc. Therefore an accurate control of the force is necessary or another grasping principle might be more suitable. 2.2 The releasing principles In general the releasing phase is achieved through gravity when the grasping principle is deactivated (Figure 4h). However in some cases gravity is not sufficient. When residual grasping forces remain after the grasping deactivation e.g. in case of ice or glue grasping at the macro scale and of surface forces at the microscale, both passive and active releasing strategies are necessary to allow a reliable and controlled releasing [59]. As shown in Figure 4 releasing strategies can be divided into two groups: passive strategies, obtained by reducing surface forces, and active strategies, where an additional force allows the gripper to release the object.

Figure 4. Releasing strategies (Figure adapted from [59] and reproduced with with permission of IFIP).

Passive releasing strategies can act at the gripper level or at the environment level. Grippers can be (a) made of or coated with conductive materials or can be grounded to prevent electric charge storage or (b) made of the same material as the object to reduce “contact interaction” forces. Moreover their surface can be coated with (c) hydrophobic coating to prevent the adsorption of moisture. Grippers made of (e) hard materials, spherical fingers (g), and fingertips with (f) high surface roughness reduce the contact area (and moreover sharp edges induce the self-discharge effect), while (d) low Hamaker constant coatings reduce van der Waals forces [6]. Changes at environmental level 4/47

have also been adopted: (i) a dry atmosphere decreases surface tension effects (but increases the risk of triboelectrification and the generation of electrostatic force), (k) assembly in fluid eliminates surface tension effects and reduces electrostatic force, (j) vacuum or oxygen free atmosphere reduces the formation of native oxides which increase adhesion, (l) ionized air neutralizes free charges on the surfaces reduces electrostatic forces [6][62]. Conversely, active strategies use an additional force to release the grasped object. Examples are the use of: (m) air pressure generated by compressed air or by heating a series of channels in the end effector, (n) inertial forces as such acceleration or vibration of the gripper support, (o) micro heating to reduce the moisture-liquid forces or to melt the ice in cryogenic grippers, (p) electrostatic force control by shorting out the gripper or tuning the electrostatic force until inverting the polarity. Forces can be varied also by creating differences in adhesion (r, s) with the substrate, (v) by using an additional tool (with low adhesiveness or little contact area), (r) by using liquid with different surface tension or (t) even gluing the part in its final position [160]. The object can be released through its mechanical engagement on the substrate as in the case of (s) snaps or when parts are pushed or scraped against an edge or by decreasing the contact area by: varying the gripper curvature from a flat shape to a curved one, (u) tilting or rolling the gripper or moving the gripper parallel with respect to the substrate. Changes at the fingertip level have also been tested: in mechanical grippers (w) a roughness change reduces adhesion forces, while in capillary grippers the modification of the liquid drop by an (x) electrostatic field, a chemical confinement or by changing the radius of curvature of the tooltip which reduces the contact area. Two examples at the macroscale where grasping and/or releasing are critical are both related to the magnetic principle. The first one is the problem of de-stacking steel sheets from a pile with a magnetic gripper since the magnetic field propagates through ferromagnetic materials. A patented magnetic array tool [119] uses multiple permanent magnets with different magnetic fields to create a very shallow depth of field to destack sheets up to 0.7 mm. The other one is the case of manufacturing of permanent magnet excited electric motors where strong magnetic forces appear during assembling. This problem has found a solution thanks to a gripper provided with an electromagnet able to counterbalance and cancel the field generated by the permanent magnets [110][209]. 2.3 The monitoring methods The presence of the object and its correct grasping are generally monitored through sensors. These sensors may be integrated into the gripper or might be mounted on an external fixture. Different kinds of sensing principles for the three main parameters (presence, force/torque and position/orientation) have been proposed. Figure 5 shows an overview and classification of sensor principles.

5/47

Figure 5. Sensing principles: a) Mechanical switch; b) electrical sensor; c) photoelectric sensor; d) vision based; e) tactile sensor; f) strain gauges; g) force/torque sensor; h) vision based; i) capacitive or electrostatic; j) ledphotodiode (often IR); k) vision based monitoring.

2.3.1 Presence detection Conventional measuring principles follow the idea that the presence of a grasping object activates a physical mechanism, which results in a Boolean electrical signal. These principles can be classified into contacting and non-contacting principles. Sensors of the first category require a direct contact between the object to be grasped and the sensor. A very basic method is a mechanical switch, which is pushed by the object during grasping (Figure 5 a). Electrical sensors that are based on the conductivity of the grasping object require two independent direct contact points to detect the object presence (Figure 5 b). In such a case measuring systems have to be integrated in the gripper finger. Since contactless sensors work with small distances between the object and the sensor surface, they need not to be placed at the gripper fingers. This allows the usage of bigger components that can be added as auxiliary systems. For example, Hall sensors, proximity switches or photoelectric sensors (Figure 5 c) are cheap, reliable and easily mounted on the gripper. Due to these facts they can often be found in industrial applications. Flat-pack inductive proximity sensor

6/47

Figure 6. Contactless presence sensors: (1) flat-pack inductive proximity sensor [9]. (Figure reproduced with permission of Balluff)

Figure 6 shows a commercial example of integrated contactless sensor. Another method of contactless presence detection is the usage of vision based sensors. The tip of the gripper is monitored by a camera and an image processing algorithm searches for known objects (Figure 5 d). Beside presence detection, these systems can easily distinguish different types of objects (shape, colour, size, etc.) and can therefore help to decide whether the observed object is of the desired type or not. Recently, a novel technology for contact sensing has been developed [194]: a small vibration is provided to the fingers and monitored at palm level. When the finger touches an object the measured vibration diminishes and consequently the contact is perceived.

2.3.2 Force/Torque sensing Force/torque sensors provide information about the performance of the grasping process and can be used for a closed loop control of the handling device so enhancing its capabilities. One of the most common functionalities is the grasping of pressure sensitive objects. Furthermore, these sensors can be used to derive information about the presence, size or type of the object. They are also needed for force controlled corrections of the grasping position or for force adaptive trajectory generation [207]. Since the finger tip of a force controlled gripper may be less object specific, these grippers help to overcome difficulties arising from object variety. The sensors can be classified by their physical measuring technique and by their mounting position on the gripper. Figure 5 illustrates three different levels at which the sensor could be positioned. Sensors at wrist level are combined force/torque sensors with multiple degrees of freedom. They are mounted between the last link of the robot and the gripper (Figure 5 g). If a force acts on the gripper, the wrist is deflected. This deflection can be measured through different physical principles – typical sensors use metal or silicon strain gauges. In [49] a force/torque wrist sensor is used to decrease the assembly uncertainty along the z-axis of a precision assembly robot (Figure 7).

Figure 7. Microgripper with a force sensor attached to the wrist [49]. (Figure reproduced with permission of IWF, TU Braunschweig)

Another example of a wrist sensor can be found in [89]. The commercial force torque compliance sensor (Figure 8a) uses an optoelectronic positioning measuring system. Spring elements connect two stacked plates which carry six diodes and six phototransistors, respectively. If an external force acts on the system, the displacement of the plates is measured by the phototransistors and, knowing the rigidity of the springs, the acting force is calculated within an error around ±5%.

7/47

a) b) Figure 8. a) Force torque compliance sensor [102]; b) Force measuring system with one active module (encircled) and one passive module [89]. (Figures reproduced with permission of Schunk)

Wrist sensors are very useful for monitoring forces outside the gripper’s system boundary like gravity, acceleration, mating or contact forces: however they cannot detect forces inside the gripper. Therefore sensors at finger level are necessary. Usually, strain gauges are adopted to measure the deflection of the fingers while an object is clamped (Figure 5 f). Commercially available systems are delivered as auxiliary modules that have to be mounted between the actuator and the gripper finger [89]. If only grasping forces have to be measured just one finger can be sensorized while the other fingers can remain passive and are used for surface coupling (Figure 8 b). The usage of more than one sensor or multiple-axis sensors allows the implementation of advanced monitoring methods. A gripper finger design, where six sensors are structurally integrated into both fingers of a gripper is described in [207]. H-shaped cut-outs weaken the structure at designated areas to allow strain gauge based deflection measuring along six axes. Figure 9 shows that each finger has three sensors – one for every basic spatial direction (x, y and z). The proposed design allows the measurement of system inherent forces as well as external forces [208]. Thus it eliminates the need for an additional wrist sensor. Furthermore a sensor network with self-learning capabilities allows to connect the individual sensor nodes within the gripper [21].

Figure 9. Gripper with two sensorized fingers [208]. (Figure reproduced with permission of bime, University of Bremen)

8/47

Although the cost of such solutions is mostly very high some commercial examples exist. Figure 10 shows a commercial gripper equipped with tactile sensors.

Figure 10. SDH servo-electric 3-Finger Gripping Hand during grasping (a-d) and detail of the capacitive sensor mounted on the gripper (e) [102]. (Figure reproduced with permission of Schunk)

Another example of an integrated force sensor can be found in [26]. The author describes piezoresistive elements integrated in the fingers of a silicon microgripper. To amplify the strain, a double beam structure is cut in every finger (Figure 11b). The sputtered gold layers are routed over the film hinges of the gripper. Therefore they are deflected during the gripper closure which results in a resistance change of the measuring line.

Figure 11. a) Monolitic microgripper with integrated silicon force sensor [44]; b) Microgripper with integrated force sensors [26]. (Figure reproduced with permission of a) Elsevier ®; b) imt, TU Braunschweig)

The design, fabrication and characterization of microforce sensors attached to the tip of microgrippers is described in [44] and shown in Figure 11a. The sensors consist of silicon cantilever beams and piezoresistive force elements located at their supports. Depending on the task and object to be grasped, the cantilever, which acts as one jaw of the gripper, can be changed as well as the complementary passive jaw (Figure 11a). A certain degree of flexibility is obtained through a tool changer as described in [35].

9/47

Figure 12. Mechanically amplified micro gripper [61].

While sensors at wrist, finger or fingertip level are directly involved in the force transmission, vision based sensors are placed outside the kinematic chain. They observe the contour/shape of the gripper finger and calculate the displacement of designated spots. Vision-based force measurement can be used to determine the grasping force of a microgripper. In [86] a resolution of ±3 mN was achieved. In the gripper described in [61] the small deformations of the fingertips are mechanically amplified to better measure the forces in an indirect way through a camera and image processing techniques (Figure 12). Alternatively small deformations at gripper level can be optically amplified and measured through a laser and a triangulation system [171]. 2.3.3 Measuring of position or orientation The assembly accuracy can significantly be enhanced by monitoring the position and orientation of the object to be grasped. Especially in micro assembly the exact positioning and orienting of objects on a tray or within a feeder system is a very complex task. Alternatively the position and orientation is measured both before the object is grasped and, after grasping, immediately before the object is assembled. A 3D vision sensor and a microgripper for assembly tasks with relative positioning accuracies below 1 µm is described in [28] and shown in Figure 7. A beam splitter is used to project a two-sided view of the gripper tip onto a single camera chip. 3

Application of grasping devices in automated production processes

Grippers have to work with a wide range of parts with very different characteristics which change tremendously in different production processes. They have to meet different requirements because parts differ in (i) size (large in aerospace industry and very small in electronic industry), (ii) flexibility and deformability (both in metal sheet handling and in textile industry), (iii) sensitiveness to scratches and pressure (in silicon wafers and solar cells handling or in food industry), (iv) sensitiveness to static charges and humidity (in electronic industry). Sometimes the design of a gripper is also affected by the environment which varies from standard industrial shop-floors to clean rooms, from submarine to explosive. Moreover grippers have to deal with objects in different conditions: from frozen fish to soft gels in the food industry, from delicate microlenses in microassembly to end-of-life goods (e.g deformed, crushed, rusty household appliances) to be disassembled. They have to perform single operations such as grasping or have to be used such as in packaging where they contribute to grasp, bend, glue [20]. Part size, weight, physical or process properties affecting the grasp reliability can be mapped with reference to different automated production processes as in Figure 13. Other factors such as the feeding conditions (e.g. stacked vs. single steel sheet) or the handling and positioning requirements (e.g. acceleration) and also the environment of the production process (e.g. submarine, vacuum, or explosive atmospheres) make the mapping shown in Figure 13 even more complex. Owing to the infrequency of such conditions Figure 13 has been kept at the minimum level of complexity.

10/47

Figure 13. Automated production processes vs. parts characteristics, properties and assembly difficulties.

Different production processes (or industrial fields) present different issues to be solved by properly designed grippers. A synthetic overview of the most used grasping principles for each production process is presented in Figure 14 showing how different grasping principles are adopted in industry or proposed by academia. In the following paragraphs the gripping problems and possible solutions are described for the main industrial applications.

Figure 14. Grasping principles vs. applications. 11/47

3.1 Standard mechanical assembly Mechanical assembly has been the first field where grippers have been adopted to automate repetitive tasks. Nowadays, in industrial automated environment a large amount of parts are delivered to the assembly station in an already pre-assembled state (e.g. a piston with piston rings) and correctly fed. There mechanical jaw grippers, electromagnets and vacuum cups play the main role owing to their proven reliability and low cost. Challenges arise in the automotive assembly owing to three new trends: (i) the growth of electric motors (used in the hybrid cars) that implies handling and assembly of stators-rotors, battery and cables just to name a few, (ii) the widespread miniaturization of product parts, (iii) robotics also pushes the borders of automation further and products, manually assembled in the past, will be automated by robots in the near future. Thanks to its high production volume the automotive field can be the forerunner for the adoption of some new grasping solutions (e.g. grippers based on van der Waals forces).

3.1.1 Assembly in aerospace industry Due to outstanding material properties the use of carbon fiber is expected to grow rapidly within the next years. Thus, contemporary efforts in this field of research focus on novel automation technologies for handling, processing, assembly and quality assurance in order to achieve a broad industrialization of composite manufacturing. Today´s manufacturing process is characterized by manual and complex process steps, small batch size, a high number of variants and difficult material properties. These demanding challenges for automation have been the subject of several research activities during the last years. The recent developments are focused on replacement or assistance of manual steps and increasing the reliability of the process chain. Major limitations for an automated handling consist in the flexible, anisotropic and air-permeable material properties and the complex contours of the textiles. Hence novel gripping systems must be highly flexible and must not damage the textile structure. One example is a low vacuum suction device for an automated clearing of debris from an industrial cutter table (Figure 15).

Figure 15. Vacuum gripper for contour-variant parts: cut parts (black)and the gripper (white).

Flexible adaptation to different contours is possible due to selective activation of every single suction hole [168][5][169]. The subsequent process step is the draping of the textiles onto a 3D mould. The end-effector introduced in [169] [162][4] is able to cope with several C-shape and L-shape profiles and quasi-flat surfaces by unrolling the gripped textile in a multistage preforming process. In order to drape Z-shape profiles and concave angles the geometry of the end-effector is modified to a triangle with configurable draping units at the edges (Figure 16). 12/47

Figure 16. Flexible preforming end-effector for Z-shape profiles (a), draping result at concave and convex geometries (b).

The finished preform build-up has to be separated from the mould. To replace the manual process step mechanical ejectors or overpressure can be used. Another possibility is the use of a sliced mould, whose slices can be moved individually in vertical direction in order to separate the preform from the mould [164]. Not all processes can be fully automated. Especially for the assembly and handling of large scaled, compliant or sensitive parts man machine cooperation and collaboration can be helpful to combine human flexibility with robot precision and quality [163].

3.1.2 Disassembly Waste of electrical and electronic equipment (WEEE) is currently considered one of the fastest growing waste streams growing at 3-5 % per year in Europe. WEEE contains diverse substances that pose considerable environmental and health risks if treated inadequately. On the other hand the recycling of WEEE offers substantial opportunities in terms of making secondary raw materials available on the market [96]. For their treatment, namely the separation of various materials, disassembly is recognised as the most effective approach although the development of new dedicated tools and grippers is necessary [185]. However the wide range of WEEE, the great variety of joints within these products as well as unpredictable damage to components and joints due to time and usage require flexible disassembly tools and grippers [13]. Research is trying to develop fast, adaptable and reconfigurable systems for manual disassembly and modular and flexible tools for automated solutions. Both of them share some characteristics. In fact, disassembly processes can be classified in (i) non-destructive, (ii) partly-destructive, destroying joining elements, and (iii) destructive, partly destroying also assembled parts [65]. While non-destructive disassembly needs interfaces and joining elements to be in good conditions, both the other destructive disassembly processes can act also on products in bad condition as WEEE usually are. Destructive disassembly can be performed by few flexible grippers and tools able to create the grasping surface where they can exert forces (acting surfaces). In order to meet high flexibility demands for the disassembly processes, tools should be applicable to a wide range of geometries [188]. The approach of generating new acting surfaces for transmitting forces and torques has been implemented in several prototypical disassembly tools (Figure 17). The prototype of this tool generates the new acting surfaces by using a pneumatically driven internal impact mass. Afterwards, the screw is unscrewed by a pneumatic drive. Conventional bits are inserted in the high-speed clamping system instead of the sharp-edged end-effectors. It has also a special centring device to easily locate the end-effector on the screw head. The device is partly size and geometry independent therefore it can handle different screws, rivets, etc. [186]. Actually the development of modules and standardized end-effectors that can be easily and quickly reconfigured is necessary for disassembling a wide range of WEEE. By recombining the modules, costs for resources can be decreased and a higher flexibility of the tools is achieved. Disassembly cells have been built to test the tools and increase the productivity of a disassembly plant. A wide product 13/47

spectrum from household equipment to electric motors can be disassembled within the cells with minor set-up changes [210].

Figure 17. Grippers used in the disassembly cell. (Figure adapted with permission of © Emerald Group Publishing).

The system described in [187] for the disassembly of washing machines consists of three co-operating robots (Figure 17). Together with some disassembly tools some special grippers are here used as follows. A scissor gripper can cut cables and tires capturing the cut-off parts for later disposal. A screwnail gripper for plastic parts consists of a rotary drive with a screwnail endeffector and linear moveable needles to fix the object in position while generating the surfaces for handling or loosening the object. A screwnail gripper for heavy tumble systems consists of two “gripper modules” with robust and powerful pneumatic drives which drive two screwnail endeffectors.

Figure 18. a) 3 degrees of freedom manipulator with an expanding gripper (b). Movements of the gripper during insertion (c1), rotation of the three arms (c2) and their radial expansion (c3-4). 14/47

Similarly a 3 d.o.f. (degrees of freedom) manipulator equipped with a gripper can grasp discarded washing machines by acting on the internal surfaces of the rotating drum[178]. It consists of a hydraulic actuator able to move three expanding radial arms; these arms act on the internal surface of the drum deforming it and realizing a stable connection with the gripper (Figure 18). This feature makes five external faces of the appliance entirely free, thus obtaining an optimal accessibility for the worker to the parts to be disassembled (motor, electronic boards, etc.) and allowing a stable and reliable grasping.

3.1.3 Handling of non-rigid parts in textile and leather industry Handling non-rigid parts is a very important issue in textile and leather industries, but also in other sectors such as food processing, aerospace industry, biomedical materials, etc. Defining the concept of non-rigid part (or flexible part) is not easy. Taking into account the problems occurring in handling, a part can be considered “flexible” or “non-rigid” if, under the action of forces usually exerted during manipulation (i.e.: weight, inertia forces, grasping forces), its deformation is greater than at least one of its dimensions According to this definition, a metal sheet is a non-rigid part since its weight can deform the part at the corners of a quantity greater than its thickness if it is grasped in the center of the upper area. Therefore, the most evident behavior of a non-rigid part during handling is the change of shape which creates several difficulties. Other problems derive from their surface characteristics (porosity or delicateness) that, in some cases, demand specific solutions. These aspects were deeply investigated one decade ago in [189]. These problems depend both on the shape of the object and on the material.

Figure 19. Problems in handling non-rigid parts. 15/47

Some examples of non-rigid shapes and materials typically handled in industry are: 2D shapes (flat) as fabrics, leather plies, paper sheets, metal sheets etc. and 3D shapes as plastic parts, rubber parts, food, bags of liquid or granular material. Force closure, form closure and material bond are basic principles that can be applied to grasp workpieces. Force closure connections are realized by friction created by springs, magnets, suction cups, or other principles. Force closure grippers ensure sufficient holding forces but might damage the surface. Suction grippers are restricted to airtight materials [198]. Examples of grippers based on form fit are needle gripper and carden grippers [184]. Needle grippers puncture the workpiece and can be adapted to different material properties. Carden grippers consists of a flexible strip covered with a multitude of thin needles that intrude the first layer of the object. Both grippers damage the surface of the material and are therefore limited to non critical operations [189]. The problems occurring in handling of non-rigid parts are peculiar and different for each step of the process (Figure 19):  Problems in grasping. When exerting the grasping force the object significantly changes its shape, producing unexpected behavior in its position with respect to the gripper (Figure 19a). In addition, in de-stacking operations (ply separation), non-rigid parts such as textiles or leather plies are stacked a ply over the other: the problem is to correctly grasp only one ply without involving the others (Figure 19b).  Problems in moving. During the robot movement, inertia forces can change the shape of the object causing different troubles such as unexpected releasing of the object (Figure 19c) or collisions with obstacles along the programmed trajectories (Figure 19d).  Problems in releasing. Releasing could be very difficult since the part can assume unexpected positions that prevent a correct placement; for instance, a corner of a leather ply could fold during releasing, not allowing a correct spreading of the ply on a surface (Figure 19e); the deformation of a rubber tube during grasping could interfere with the assembly operation at the end of the trajectory (Figure 19f).

Figure 20. Grasping principles of non-rigid parts (XXX: very good; XX: good; X: fair).

A non-rigid object can be grasped in different ways according to its shape and material. In order to avoid or minimize the problems listed in Figure 19, the generally used grasping principles can be classified as: mechanical grasping, ingressive grasping, adhesion grasping (divided in electromagnetic, electrostatic, suction, air jet and cryogenic grasping). Figure 20 summarizes the capability of each principle in grasping different non-rigid parts. 16/47

Mechanical grasping is based on the use of fingers. It often causes the deformation of the object due to the forces exerted by the fingers during closure. This effect can be minimized by a proper selection of the surfaces to grasp. In grasping 2D objects, two different approaches are possible:  grasping in the central part of the upper surface. This approach is easy to perform but obviously produces an evident deformation of the part, introducing difficulties in ply separation or in releasing on a flat surface.  Grasping from the edges. It often requests special fingers, like those ones shown in Figure 21a and also in this case ply separation could be very difficult [53]. The advantage of this method consists in moving the part maintaining a vertical position: the part falls down due to gravity, without unexpected deformation of the part, obtaining an easy releasing on a flat surface.

Figure 21. a) Grasping fabric from the edges [112]; (b) vacuum gripper grasping a CFRP fabric; (c) Needle gripper grasping a non-rigid object; (d) Electrostatic gripper. Figures reproduced with permission of (a) Elsevier; (b, c) Schmalz; (d) IPT Fraunhofer.

Ingressive grasping is divided in non-intrusive methods (Velcro systems) and intrusive mechanisms (needles Figure 21b). The prehension principle is based on the partial penetration (e.g. of needles) on the upper surface of the object, therefore it is exclusively used in grasping fabrics or foods. By controlling the depth of penetration the gripper is able to correctly separate objects from a stack. A typical needle gripper for textiles is represented in Figure 21b. The coordinated extension of four needles is commanded by two pneumatic cylinders. The holding force in needle grippers depends on the penetration angle and number of needlesas well as on the on the elasticity of the workpiece. As the elasticity is often small, the maximum grasping forces are limited. The same behaviour can be observed in carden or velcro grippers. They are made of staple-holes or hook-loop pairs, respectively. The holding force can only be enhanced by increasing the number of pairs in contact. An upper limit of the force exists owing to geometric constraints to staple/hook size. 17/47

Adhesion grasping is based on different principles such as suction cups (vacuum grippers), electromagnetic or electrostatic, air jet or cryogenic grippers. Handling workpieces with low porosity is a comparatively easy task for vacuum grippers. On the other hand, porous or permeable surfaces, such as textiles made of carbon fiber (CFRP) or composites such as Kevlar or woven fabrics, present a challenge. Producers have therefore designed special vacuum systems with ejectors able to generate high volume flow, ensuring reliable handling for not only composite textiles, but also for non-rigid, unstable components, extremely thin and delicate foils or even thin circuit boards (Figure 21c). A highly flexible electrostatic gripper (Figure 21d) is capable of lifting semi-finished textile products made of carbon fibers and other materials and putting them down again with pin-point accuracy without any damage. Electrostatic phenomena are the key aspects of this solution and they can be applied to both conductive and insulating materials, but their performance decreases in presence of water and dust. In vacuum grippers problems as the delicateness and the porosity of the grasped surface (e.g. leather or food) are addressed by modifying both the material of the cup and the texture of its contact surface. As reported in [53] the material and the cup geometry allow the high mating level with the leather surface, while the small channels of the texture over the contact area reduce and hide the imprints generated in the contact by the vacuum(Figure 22).

(a)

(b)

Figure 22. The deformable vacuum cup for leather ply grasping: a) front view; b) lateral view during a stripping force test.

The holding forces of suction grippers can easily be increased by enhancing the vacuum in case of workpieces with high air permeability. An alternative solution for grasping a non-rigid object by adhesion is based on the use of the cryogenic principle (cryo-gripper illustrated in Figure 23a [195]). A cryogenic gripper using Peltier elements is described in [88] [190]. Recent experiments have led to the hydro adhesive gripper [195]. The principle exploits the liquid solid transition and the criogripper freezes water as active means for material bond. For attaching a non-rigid part, hydro adhesive grippers spray a little amount of water on the part surface. The gripper is equipped with a Peltier element that freezes the water and causes the adhesion. Releasing of the gripped objects can be accomplished by reversing the current of the Peltier element or by air pressure [132]. A further adhesion gripping principle is based on a novel Coanda effect ejector (Figure 23b). This ejector allows the construction of a slim, plate-shaped vacuum gripper with multiple independent suction heads. Each suction head is powered by a recently patented lateral Coanda ejector that ensures gripping power on all non-rigid parts even in case of porous ones [131]. Bernoulli grippers have also been tested with deformable objects and have demonstrated interesting results such as in the case of leather plies [40]. Advanced textiles as Kevlar, Twaron, glass and carbon fibers have been finding applications in many advanced products [72]. Their high cost is mainly due to picking, draping and assembling textiles in a mould. New grippers have been suggested for automating the process which today is still manual.

18/47

Figure 23. a) Cryogenic gripper [195]; b) Coanda effect gripper handling textile [131]. (Figure a reproduced with permission of Elsevier)

Preforming is a critical step in the manufacture of continuous fibre reinforced polymers. The transformation of a 2D textile into a near-net-shape 3D fibre part represents a huge challenge for automation. Currently, preform production is mainly manual, but several research approaches exist for automating this process step. The research project described in [70] investigates preform technologies, which are suitable for mass production. One development is shown in Figure 24. The so called “drape gripper” is able to pick up a 2D textile semi-finished product and drop it in a 3D shape, like a “cap-profile”. By means of this system it has been shown that the draping of decoilable forms can be carried out through a handling system. In the future, the results should be transferred to more complex geometries [70].

Figure 24 a) Drape gripper draping a cap-profile out of 2-dimensionally textile semi-finished products. b) Sensor based Coanda gripper.

Another solution in the field of automated handling of semi-finished textiles made of carbon fibre is a low pressure Coanda gripper, shown in Figure 24b. It represents a suitable solution with a high automation level. The grasping force is exerted through a series of holes located on a flat cylindrical surface where two annular electrodes are used to measure the grasping force [73]. The contact resistance measured between the two electrodes and the semi-finished textile depends on the grasping force. Actually the higher the force the lower the contact resistance measured between the two electrodes. A control unit and a regulating valve 19/47

maintain the gripping force at a low but reliable value. Thus air consumption can be reduced to a minimum while still guaranteeing a reliable grasping. 3.1.4 Handling of hazardous parts Hazardous environments sometimes complicate also the design of the gripper. Standard mechanical jaw grippers are usually used to handle parts produced from processes like sand-casting, metal forming or plastic moulding both when the part is at high temperature right after being casted or formed (Figure 25a) and right before product assembly when this part is cooled to room temperature. Special solutions are adopted: grippers need to be completely sealed from the die casting environment, fingers are in stainless steel, air cooling systems prevent the gripper itself from overheating [100]. However mechanical jaw grippers have several limitations in handling such parts: the variability of the object shapes due to flashes and burrs can prevent stable grasping, the gripper cannot completely enclose the part, sharp edges can damage the gripper itself [217], parts are often too heavy to be handled safely by both commercial and specific grippers [28]. Some grippers for explosive atmospheres (gas/air, dust/air) are described in [100]. All the possible sources of ignition from the system, i.e. electric, electrostatic and friction have been removed. Among the technical solutions (e.g. insulation of electric parts) anti-static and non-sparking coating are applied to all the mechanical components of the gripper. A different approach developed for handling big submarine pipes is the ball-and-taper gripper [50] (Figure 25b). It is a cylindrical gripper acting on the inner surface of a pipe. It consists of a connector with a series of balls held securely in tapers that are machined into the connector surface. The balls expand radially when the tapers move axially. The gripper is inserted into a close fitting tube, the connector pushes the balls out of the tapers, thus increasing the diameter of the connector and providing a powerful multi-point grasp.

Figure 25. a) Jaw gripper for hot parts; b) ball-and-taper gripper (Figures reproduced with permission of a) Schunk, b) First Subsea Ltd)

3.2 Electronic assembly In electronic industry almost all the objects to be grasped are flat, the environment is clean (usually assembly takes place in clean rooms) and the production rate is very high. Moreover the component spectrum has a wide range with the component size from a few tens of microns in case of SMDs (surface mounted devices) to 200 mm in case of silicon wafers or solar cells. In general parts are mechanically strong but fragile and their surface is very sensitive, even to contact. In electronic assembly pick and place machines are used for assembling electronic components (like resistors, capacitors, inductors, IC, diodes, etc.) in surface mount technology (SMT) on printed circuit boards (PCB). These machines are able to place electronic components with an immense speed of up to 150,000 parts per hour. Electronic components e. g. 01005 ceramic chips with a size of 0.4 mm by 0.2 mm or flip-chips with bump sizes from 60 µm to 30 µm (pitch 100 µm [64][75][182]) can be placed with an accuracy of down to several micrometers. The placement head of a typical pick and place machine collects the components from the feeders and places the components at the exact position and orientation on the PCB. The accuracy in placing components is guaranteed by an integrated vision system. The camera optically measures the position 20/47

in x and y direction and orientation of each component before placing on the PCB (Figure 26). A second camera registers fiducial marks on the PCB to measure the position of the PCB in the machine. The components are grasped within a couple of milliseconds by vacuum nozzles. The collect and place head grasps up to 20 components from the component feeders and places them sequentially on the PCB. Up to 37,500 components per hour at an accuracy of 40 µm (3σ) can be realised by one placement head in this mode. The head with the highest placement accuracy of up to 10 µm (3σ) is the pick and place head, that picks one component from the feeder and places it onto the PCB. However, the placement performance is reduced to 6,000 components per hour.

Figure 26. Pick and place machine with placement head on portal system. (Figure reproduced with permission of ASM AS).

The flexibility in placing a wide range of component shapes is obtained by automatically changing the nozzles for each job. Additional sensors can be integrated in the machine like laser triangulation sensors to measure the planarity of the contacts of large-area components. A component height sensor measures the height of mechanically sensitive components. The head can also measure electrical characteristics of the components during the placement process. A self-diagnostic analysis detects blocked nozzles and performs a nozzle cleaning cycle. Moreover further tools for diverse processes like glue dispensing and laser soldering can be integrated. Also a process module, that enables pick and place machines to place components on 3D circuit carriers, is applicable. This module has integrated motors to extend the machines’ kinematics [156]. In the future even smaller parts with sizes of 0.3 mm by 0.15 mm must be gripped and handled reliably by SMD pick and place machines. Different sectors such as chemical analysis (e.g. fluid injection analysis, microvalves), medical sector (e.g. drug delivery, heart pacemaker, hearing aid), electronic sector (e.g. Micro Fuel Cells) and telecommunications (e.g High frequency applications, Bluetooth antennas, resonators) need electronic parts based on ceramic substrates due to material properties such as stability at high temperatures and resistance to chemical agents, and/or biocompatibility. To produce such components green ceramic tapes are used. They are made of a mixture of ceramic powder, polymer binder and solvent and are stored in tapes with a thicknesses varying between 100 μm and 400 μm. The production process of a device consists of several phases needing grasping, handling and positioning: firstly a tape is cut in parts, then they are handled and processed (printed, coated, embossed, etc.), later grasped again and positioned with high accuracy (