Robot Assembly (1/4)

I originally presented this article, "Design for Robot Assembly", as a guest speaker at the UK's 2nd National Conference on Production Research ...

SUMMARY

The design of products and systems for robot assembly requires a new approach to that used for manual and automatic assembly.  Robot assembly is only effective if the robot’s flexibility is used to best advantage.  Additionally, peripheral devices supporting the robot must also be adaptable to handle a wide variety of products and product parts.  This is achieved by using equipment that is not designed specifically to handle a particular type of part with minor modifications to tooling, or the use of a different software application, the robot assembly system can be quickly adapted to assemble a different product or product style.  By this method, robot assembly can be economically justifiable in many situations where it would otherwise have been precluded.

This article discusses the development of robot assembly systems and describes how product design plays an important role in the design of the equipment.

INTRODUCTION

There are three categories of system used in product assembly.  These are manual assembly, automatic assembly and robot assembly.  Whilst assembly can be classified in this manner, it is not uncommon to find an assembly system consisting of two or all of these groups to form a hybrid system.  Manual assembly systems account for the majority of applications.  Automatic systems are used in situations where the demand is high and there is no, or limited, change in the product styles being assembled.  Robots have yet to make a significant impact in the field of assembly.  It’s difficult to technically justify the use of assembly robots as the operation time of programmable devices is longer than that of dedicated automatic equipment.  The economic justification of assembly by robots is equally difficult due to the characteristically small batch sizes for which these systems are appropriate.  This has the effect of increasing the handling and insertion costs of the product being assembled.

Manual assembly is still used for more than ninety per cent of all assembly tasks.  This is because many products are required in low volumes and with a high degree of variety.  Robot assembly could account for more than fifty per cent of all assembly tasks if it could be made to be economic for much smaller annual production volumes. This could be achieved by assembling more than one family of products on one system.  For this approach to be effective, two major conditions must be met.  The proportion of re-usable, or general-purpose, equipment must be high and the time taken to re-configure the system for the assembly of the next product must be low.

Using existing technology, the industrial applications where robots can readily be used have been filled.  These applications include paint spraying, spot welding and materials handling.  Only a very small proportion of existing robots are used for assembly.

Robot assembly system equipment is either general-purpose or special-purpose.  A robot assembly system should have a high proportion of general-purpose equipment and a low proportion of special-purpose equipment. The cost of a system, with a high proportion of general-purpose equipment, can be amortised by all the products that are being assembled by the robot.  This is important when trying to economically justify the use of robot assembly for products required in low volumes.  Under these conditions, many products or product styles, each with a low annual volume, can be grouped together and assembled on a single robot assembly station to obtain a high system utilisation.

Robot Assembly (2/4)

THE HANDLING OF PARTS FOR ROBOT ASSEMBLY

Handling device selection for a particular part depends on the size and geometry of a part, as well as the rate at which the part is required.  Each handling device has its own performance characteristics.  This means that it is suitable for dealing with a limited range of parts.  Small to medium sized parts, with features that can be seen in silhouette, can be handled by the most common of devices, the vibratory bowl feeder.  Parts with no useful features for orientation purposes, or parts with adverse physical properties, are expensive to feed automatically and require special automatic feeding devices.  These types of parts need to be re-designed to reduce their cost for automatic feeding.  There are many properties of a part that would prevent it from being handled by vibratory bowl feeders, such as flexibility and stickiness.  Parts with adverse properties such as these, and larger parts, must be handled by other feeding devices like magazine systems or pallet transfer systems.

The multi-part linear vibratory linear feeder can deliver different parts to a robot assembly station.  It consists of two straight and parallel vibratory orientating tracks on a common drive unit.  The rejected parts fall into return tracks and are brought to the start of the orientating tracks by a reciprocating elevator.  The tracks can be CNC machined from a database of designs that are identified by an automated handling code for a particular part.  Only the orientating tracks are replaced to changeover this multi-part feeder to handle other part types.  The vibratory drive unit and reciprocating elevator are completely re-usable and the cost of these devices is divided between the different part types.  The orientating track for this multi-part feeder is straight and it is much less expensive to produce than the curved orientating track of a vibratory bowl feeder.  Applications of this feeder are limited to parts which require orientating devices simple enough to be produced in one set-up on a horizontal machining centre.

Gravity feed track magazines are simply short lengths of track which are loaded manually on-line or off-line.  During off-line loading, a full magazine is substituted for a magazine when it becomes empty.  These magazines are specifically designed for the particular type of part type and cannot easily be re-used for different types of parts.  Although most of the gravity feed track magazine is special-purpose, the cost of these devices is relatively low.  They are useful far feeding large parts and they provide an economic alternative to palletisation.  Parts that are to be handled by this type of device must be stackable, for vertical magazines, and not susceptible to damage when the part is slid into position by the pusher.

The pallet transfer system consists of a walking beam transfer device to load a paternoster, an unload paternoster, and pallets.  Full pallets are elevated by the load paternoster and transferred to the robot working zone by the walking beam transfer device.  Parts are picked from the pallet and the pallets are then indexed to present a new pallet of parts to the robot.  Empty pallets are offloaded from the walking beam by an unload paternoster that produces a stack of empty pallets.  Virtually all of the pallet transfer system is general-purpose, with only the vacuum-formed part retainers being specific to a particular component.  Pallets are loaded by standard means.  Filling of the pallets at the point of manufacture is a very economic way of loading parts, although the cycle time of most manufacturing operations makes it difficult to use this method of loading.  Parts are positively held in position on the pallet by ensuring that they are sandwiched between the underside of one pallet and the top of the one beneath.

Robot Assembly (3/4)

THE ORIENTATION OF PARTS FOR ROBOT ASSEMBLY

An assembly robot will never have the dexterity or intelligence of its human counterpart.  A manual worker is able to pick a part, in random orientation, from a storage bin and orientate it - ready for insertion.  The human senses of sight and touch are used for this purpose.  Whilst a manual worker can perform these tasks with comparative ease, a robot requires a large amount of computer processing power and many feedback devices to achieve any form of intelligence and, even then, the cycle time of the operation is so long as to make it uneconomic to use robots for bin picking.  Automatic feeders for robot assembly must, therefore, present parts to the workhead in a known orientation, or in a limited number of known orientations.  The attitude of the part on the feed track or pallet influences the number of robot degrees of freedom required.  More degrees of freedom are required for those parts which are inserted in a different attitude to which they are presented.  In the case of parts which cannot be presented in one known orientation, the final orientation must be carried out by a robot with extended capabilities.  This involves sensing and part manipulation, to achieve the required insertion orientation.

The robot work envelope poses limitations on the automatic feeder types that can be used in robot assembly.  Each part is presented to the workhhead at the end of a track or on a pallet.  The space occupied by the material of these devices must also be considered when determining the maximum number of parts that can be fed to any one robot.  Robots that can only access parts in the vertical axis must have parts arranged so that they can all be seen in plan view at the part presentation points.  Another problem arises when turret mounted grippers are used.  The grippers can occupy a large volume in space and this makes the avoidance of collision very important.  This situation can be investigated before the design of the robot assembly system is finalised.  The complete assembly process can be studied using computer simulation and there are many three-dimensional graphic simulation packages available that can identify if a collision is likely to occur.

THE INSERTION OF PARTS FOR ROBOT ASSEMBLY

An insertion operation is defined as being the action whereby a part is added to a work fixture, another part, or part-built assembly.  This may involve a simple vertical downwards motion where the part is added to the part-built assembly, without being immediately secured.  Alternatively, it may be a complex motion, such as that required for the application of an adhesive to a part.  Each insertion process may require a different type of end effector and each process takes a certain amount of time to be executed.  It’s possible to categorise each type of insertion process to define the type of end effector required and to estimate the time it would take to carry out the operation.

The end effectors may be accessed by the robot arm in many ways.  The design of end effector, and the method of mounting it onto the arm, influences both the cycle time of the process and the cost associated with the insertion of a particular part.  The simplest, yet most expensive and time consuming, method of accessing an end effector, is to use an
individual gripper, or tool, for each part or insertion process.  The grippers and screwdrivers are stored in a rack within the work envelope of the robot arm.  The relevant tool is picked from the rack, used for the insertion process, and then returned.  The action of picking up the tool, and returning it, can often take longer than the insertion process itself.

Another method of inserting many different designs of parts is to use a multi-functional gripper.  Only one gripper with a multitude of faces is used, for the internal or external gripping of parts.  The time involved with gripper changing is eliminated, but the design of the gripper is complex and other tools cannot be mounted onto the same unit.  Problems may also occur because only one set of jaws is being used for the insertion of many parts.  The gripper designer has to ensure that the gripping force is sufficient to hold the part and yet not too excessive as to cause damage to the part.  The varying force requirements can be met by additional gripper sensing.  This, of course, increases the cost of this design of end effector.

The most efficient method of accessing a multitude of end effectors is to mount them onto an indexing turret.  Between eight and twelve tools can be housed on one unit, depending on their size.  Grippers, screwdrivers and othertools are mounted in a circle.  This may be about a vertical, horizontal or inclined axis.  The use of universal mounting plates, between the turret and the end effectors, allows interchange-ability of grippers and tools for product changeover.  The time lost, due to gripper changing, is minimised because indexing of the turret occurs between movements to, and from, the parts feeders.

Most products, or sub-assemblies, have many possible sequences of assembly and it is important to recognise the most appropriate sequence, particularly in robot assembly.  In all forms of line assembly, where moving work carriers are employed, it is good practice to secure parts as soon as possible because subsequent work carrier movements may cause a part to be displaced. This suggests certain precedences.  If no movement of the part-built product occurs during assembly then the securing of parts is not important and a sequence of assembly can be chosen which involves a minimum number of gripper changes.

Consideration also has to be given to the appropriate action needed when a malfunction occurs.  The decision to scrap, rectify or dismantle depends on the; value of the part-built assembly, frequency of the malfunction, labour cost and sequence of assembly.  In single station robot assembly, an overriding consideration is the cost of gripper changing.  The
optimal sequencing, linked with appropriate product design, can significantly reduce this cost.  Computer software applications are available which, given the precedence constraints, identify the optimal sequence to minimise gripper changes.  The cost of error recovery is important.  The alternative actions need to be examined at each stage in the assembly build and the cost of these actions should be determined for all possible sequences.  This activity is influenced by the chosen criteria of; minimum cost, maximum production or maximum profit.

Robot Assembly (4/4)

PRODUCT DESIGN FOR ROBOT ASSEMBLY

Three factors determine how easy it is to use an assembly robot for a product.  Each product part should be examined with respect to these three important qualities. In order of priority, they are the; necessity of the part to be separate from those which have already been assembled; ease with which the part can be handled, and the ease with which the part can be inserted.  By considering these factors in turn, the most economical design of product can be chosen for robot assembly.  A measure of the assemble-ability of the product is the 'design efficiency', and this is related to the above factors.

A part is considered to be necessarily separate from those previously assembled if one of four conditions apply to the part.  Otherwise, it can be eliminated.  Firstly, if the part or sub-assembly moves relative to its mating part during the normal function of the final assembly then it must be a separate part.  Secondly, if the part or sub-assembly must be of a different material than its mating part (eg. for insulation, vibration damping) then it must be a separate part.  Thirdly, if disassembly of the part or sub-assembly must be allowed for (e.g. servicing requirements, recycling) then it must be a separate part.  Finally, if the part or sub-assembly, when combined with it’s mating part, would prevent the assembly of other separate parts (except where the part's only function is to fasten) then it must be a separate part.

The majority of insertion processes take place along, or about, the vertical axis.  If the action of insertion for a part is not in the vertical axis then the process should be analysed to see if the more complex insertion path is really necessary.  If possible, it should be re-designed to take place in only one axis.  The vertical axis is always the preferred axis because the weight the part acts in this direction and assists, not hinders, the operation.  The robot cost is lower if insertion processes are kept simple.  This is because complex operations need more robot degrees of freedom and each degree of freedom requires an individual pneumatic, hydraulic or DC servo motor which increases the cost of the equipment.  Additionally, the potential profitability of the equipment is reduced because the cycle time of the operation will also be increased.

CONCLUSIONS

The use ofassembly robots will increase in the future if the ancillary equipment, i.e. end effectors and parts feeders, are as flexible as the robot.  The feeding devices should present the parts in a known orientation so that the dexterity required from the robot is low.  The cycle time of the operation would be lowered and, consequently, the assembly rate increased.  The flexibility of the feeders is ensured by using devices with a low special-purpose content.  An indexing turret, used for gripper mounting, minimizes the time lost due to gripper changing.  For any form of gripper mounting, the cycle time can be minimised by using a sequence of assembly which needs the least number of gripper changes.  Operator involvement can be minimised by developing strategies which allow the robot to recover from error situations, without the assistance of manual labour.  The cost of robot assembly can be minimised by designing the product for robot assembly.  This involves using the minimum number of parts and ensuring that the parts can be easily handled and inserted.