I originally presented this article, "The Design of Hybrid Flexible Assembly Systems", as a guest speaker at the 6th International Conference on Assembly Automation ...

There is a requirement for a special kind of system to assemble products required in modest volumes with a degree of variety.  A system which is as cost effective and efficient as hard automation, whilst providing the flexibility of manual assembly, is called a flexible assembly system. Within such a system, certain product parts may be required at a different rate to other parts. Some operations may require the flexibility and dexterity of a robot, or even manual labour. The resultant system would be a hybrid of many methods of assembly. This article recommends a technique to be used for the design of such a system, with the aid of a case study.

INTRODUCTION

The factory cost of a product is the addition of the manufacturing cost (e.g. casting, moulding, turning) and the assembly cost (e.g. manual, automatic, robotic).  Industrial engineers continually seek new methods to reduce the factory cost of products. The current trend of exploiting cheap labour in developing nations, through “offshoring” creates a challenge for domestic manufacturers in the developed nations.  Between 40 and 60 percent of the factory cost for many products is associated with the labour content. The majority of this cost is incurred during assembly. There are three reasons for this uneven split between labour costs in manufacturing and assembly.

(i) Manufacturing operations are usually done by, or with the aid of, a machine, i.e. turning, milling, drilling, etc. The manufacturing systems designer does not have the wide choice of the assembly systems designer because some degree of mechanisation must be used. It is then a logical extension to further automate the manufacturing process to reduce labour costs.

(ii) New processes have been developed which eliminate many manufacturing operations. Powder metallurgy is an example of such a process.

(iii) Most products are designed to be assembled manually.  This often means that components are of such a design that they cannot be handled by automatic feeders.  Additionally, many assembly insertion operations are too complex to be automated.

The assembly process is one of the last production processes to be successfully automated by the industrial engineer. However, as much of the factory cost of a product is incurred during assembly, it is this area where great productivity improvements can be made. The design of the assembly system should be undertaken with due consideration of the design of the manufacturing system and of the design of the product. The design of the assembly system, manufacturing system and product should be considered integrally. These three components, when combined, should create a product having the lowest factory cost at the desired level of quality.  The design of a product and it’s associated production system is an iterative process, whereby product design features dictate the design of the production system and the capabilities of the production system determine the product design. The extent to which these actions can be carried out is only limited by the commitment of a manufacturer to a particular production system and product design.

A company may be already committed to a certain manufacturing system if there is prvious investment in capital equipment and tooling.  Additionally, the external dimensions, performance or appearance of the product may be unchangeable. If a product is only part of a much larger assembly, the effect of changing a critical dimension may have expensive consequences for the rest of the much larger assembly. The performance of the re-designed product must be as good as, if not better than, the original design. The product may be one where visual appearance plays an important part in it’s acceptability in the market place. All of these factors place limitations on the engineer being able to specify the optimum product design and production system for that design.

It is easier to design the most economic assembly system for a product prior to commercial manufacture. In this case, there won’t be an inherited investment in manufacturing equipment or tooling, and the product design won’t have been finalised. If the product is well established, and has been produced for many years, the assembly systems engineer may be limited to a re-design of the assembly system alone. This is because a re-designed product may require expensive design modifications to the tooling used for the manufacture of the product parts. In these situations, a hybrid assembly system is required to meet the product requirements. A hybrid assembly system uses a mixture of methods during assembly of the product.

THE COMPONENTS OF A HYBRID FLEXIBLE ASSEMBLY SYSTEM

There are six methods of assembly and the simplest form is MANUAL ASSEMBLY.  For high volume production, the operatives usually work on an assembly line. Other forms of manual assembly are a single worker assembling a complete product and groups of workers assembling a portion of the product.

For a more limited product range, a MANUAL ASSISTED method may be used, whereby workers are assisted by mechanical devices, such as automated parts feeders. The feeders present the parts to the worker in an ordered manner and the assembly time is reduced by eliminating the time taken to separate the parts from bulk random orientation. The reduction in assembly time is the basis for the economic justification of these devices.

The third form of assembly uses AUTOMATIC INDEXING assembly machines. These are rotary or in-line systems with a number of workstations.  Automatic feeders supply components to workheads and they assemble the part to the fixture or part-built assembly. The workstations are ‘special-purpose’ and are dedicated to the assembly of only one product. Production volumes need to be high for the economic justification of these machines. Component quality must also be high to avoid excessive downtime caused by components jamming, etc.

The efficiency of an AUTOMATIC FREE-FLOW assembly machine is less dependent upon component quality.  Transfer of work pieces between    workstations is non-synchronous. There are small buffer stocks between each workstation and other workstations may operate whilst one is stopped due to a fault caused by, for example, a defective part.

The AUTOMATIC PROGRAMMABLE assembly machine has a non-synchronous transfer line and programmable workstations to assemble the parts, which are presented to the workheads by automatic feeders or, in the case of difficult components, part magazines may be used. The workheads execute one, or a number of, operation(s). Different computer programs, for each series of assembly processes, give the flexibility to assemble a variety of product styles on one assembly machine.

Robotic assembly is used for the assembly of products with large product variety, required in low volumes.  Assembly operations are carried out by a robot which, itself, transfers the completed product onto the next operation.

THE DESIGN OF HYBRID FLEXIBLE ASSEMBLY SYSTEMS

The assembly process has two constituent parts and these are; the handling of components and the insertion of components. The design features of a part must be examined to decide if it can be automatically handled automatically or if it must be handled manually or placed in magazines.  Similarly, the insertion process must be analysed to decide what type of workhead is required.

Various organisations have developed procedures that help the designer to estimate how easy it is to handle and orientate components by assigning a handling code to each part. The maximum feed rate and relative cost of the feeding method can then be estimated from this code. The parts which would require expensive automatic feeders or which could not be fed at the required feed rate can be identified.  These parts must then be handled manually or in magazines/pallets.  Additionally, certain parts cannot be handled automatically because they have other bad feeding qualities, e.g. they may be flexible or too light. The previously mentioned estimation systems also help the system designer to forecast the relative cost of the workhead required to insert a part into a part-built assembly. Those operations which require a complex path of insertion, or a large thrust, require more expensive workheads than for simpler operations. A list of parts (with their associated automated handling codes) and a list of operations (with their allocated automatic insertion codes) can be constructed from the preceding information.

If the product parts are listed in order of increasing handling difficulty levels then the most economical method of feeding a part to the workhead can be determined. Parts with low handling difficulty levels are fed by conventional vibratory feeders and, as the difficulty level increases, specially designed feeders/magazines/pallets/manual handling are used. The relationship between the handling difficulty level and the type of feeder to be used depends upon the required return on investment for the equipment.

The insertion operations can also be listed in order of insertion difficulty levels to determine the most economical method of insertion of a part into a part-built assembly. Greater difficulty levels can mean that the equipment is more expensive and, for assembly robots, more degrees of freedom are required for an insertion operation. If the difficulty level is too high then it’s necessary to employ manual workers for some operations.

When an assembly system is designed for a new product, the cost of parts handling and insertion can be reduced through re-design of the product. It’s usually not viable for an existing product to be re-designed, because of the tooling modification cost in the manufacture of the parts. Inevitably, therefore, the most economical method of assembly is limited to the existing product design, without design efficiency improvements.

The assembly handling and insertion codes determine which feeding method and insertion device are most appropriate for each part and operation. The part-built assembly has to be transported to each workstation between operations. This will either be synchronous or non-synchronous motion.  Synchronous machines are generally less expensive than non-synchronous types, but they are limited by how many parts can be assembled on one machine. This is due to downtime and the space available.

It is desirable to construct a product from as many sub-assemblies as possible to achieve a high overall efficiency of the assembly system. These sub-assemblies should be common to all product styles, within the family of products. The variety can then be created in the final assembly of the product. If this approach is adopted then sub-assemblies will be required at a rate which is enough to justify the use of automatic indexing machines having dedicated workheads. The output from these machines can then be sent to the final assembly line via free transfer lines, to create a buffer stock of sub-assemblies. The buffer stock is necessary to minimise the effect of any indexing machine downtime.

CASE STUDY - THE DESIGN OF A HYBRID FLEXIBLE ASSEMBLY SYSTEM FOR SPEEDOMETERS

The case study describes how a hybrid flexible assembly system was designed for the assembly of a mechanical drag cup speedometer. This type of speedometer is the most widely used today and its design has not changed over the last 50 years. If there is already a heavy investment in capital equipment for the manufacture of the individual parts then it is not economical to re-design the product for automatic assembly.

The input shaft of the speedometer carries a permanent magnet. The flexible drive shaft from the engine drives the input shaft, thus setting up a rotating magnetic field. A metallic cup is situated in this field and is continuously connected to the pointer. As the input shaft rotates, a torque is produced at the spindle, which is proportional to the speed of the input shaft. The spindle is free to rotate and yet is restrained by a delicate hairspring. The spring rate is chosen to be linear over the range of the spindle angular deflection, thus providing a pointer movement that is proportional to the input shaft speed. The hairspring returns the pointer to zero when the vehicle is at rest. A series of gears from the input shaft convert the rotation of the flexible drive shaft to a rotation of the odometer wheels. Gear ratios typically vary from 600:1 to 2000:1.

There are 25 parts used in the assembly of the speedometer and more than 50 product styles can be obtained by a variation in the design of six parts. These are the dial, second worm gear, third worm gear, odometer sub-assembly, hairspring and pointer sub-assembly. The total annual production volume for all the styles is in excess of one million units. An individual style may be required in volumes between 200 and 200,000 per year. Clearly, these volumes require an assembly system which has flexibility to handle such large demand fluctuations.

The speedometer consists of four sub-assemblies and twelve parts. The dial sub-assembly has three parts, the first worm sub-assembly has six parts, the speed cup sub-assembly has two parts and the frame sub-assembly has two parts. Each sub-assembly is a self-contained unit and does not require any holding of the parts for stability between workstations.

Synchronous assembly machines are most economical for the high volume assembly of a small number of parts. Each sub-assembly contains six or less parts, making them most suitable for this method of assembly.

A rotary indexing machine for the FRAME SUB-ASSEMBLY is used for the assembly of two components. There are eight workstations on this machine to allow for non-value adding operations in addition to the direct insertion process. The handling difficulty level of the bearing means that it is presented by a specially designed feeder. It is impregnated with oil and this doesn’t allow the part to be handled by a conventional vibratory feeder. The frame cannot be handled by an automatic feeder because it is large and has no symmetry about any axis. The complex shape of the frame means that it cannot be magazined and it is, therefore, palletised. A robot places the frames onto the machine because they are picked from several hundred pallet locations.

The rotary indexing machine for the SPEED CUP SUB-ASSEMBLY uses a simple pressing operation to secure the speed cup to the spindle. There are four workstations for; the assembly of the spindle to the fixture, the speed cup to the fixture, the pressing of the speed cup onto the spindle and an output station. Both parts are fed by vibratory bowl feeders and inserted by dedicated workheads.

The FIRST WORM SUB-ASSEMBLY consists of six components, all of which are fed by vibratory bowl feeders. The indexing machine uses ten dedicated workstations to complete the sub-assembly. The first worm shaft is burnished before final assembly.  This operation is executed after the rotary indexing machine, on a free-transfer line. Two burnishing stations are used, in parallel, to achieve the cycle time. The free transfer line also provides a buffer stock of completed sub-assemblies before the final assembly line.

The rotary indexing machine for the DIAL SUB-ASSEMBLY assembles three parts. Only the pointer stop can be automatically fed and so the dial and label use special feeding methods. Different designs of dials are used to create product variety. However, only the print face and diameter of the dial are variable and the dial is picked from a magazine, on the reverse face, by a dedicated workhead. The label is applied by a conventional labelling device.

All sub-assembly indexing machines are linked to the final assembly machine by free-transfer lines, for overall system efficiency. This also creates space for auxiliary operations to be carried out on the sub-assemblies before final assembly. The speed cup sub-assembly is dynamically balanced before final assembly, and this is done with the aid of two robots. The programmability of a robot is required for the 'decision making' operations of this process. Feedback from the balancing machine determines whether the sub-assembly has to be balanced more than once or, in the case of it being excessively out of balance, it is rejected.

There are twenty six workstations used for the FINAL ASSEMBLY of the speedometer, making it necessary to use a free-transfer linear machine to allow buffer stocks to be created between each workstation, to maintain high system efficiency. Of the twelve parts used during final assembly; seven parts are handled by conventional vibratory bowl feeders, two parts by multiple vibratory feeders, one part by pallet, one part by manual handling and the remaining part by actual manufacture on the assembly line.

The parts which are fed by vibratory feeders are small components with either useable symmetry or definite asymmetry. These are inserted into the part-built assembly by dedicated workheads.

The two parts to be handled by multiple vibratory feeders are the second worm gear and the third worm gear. These parts are changed to produce the various gear ratios used to create different product styles. The disruption to production, during product changeover, is minimised by using a group of vibratory feeders which deliver one particular second or third worm. The pick up point of the workhead is thus quickly changed to the output of a particular feeder for the assembly of a different style.

The jewel--plate sub-assembly is a large and delicate part which cannot be fed by an automatic feeder.  It can, however, be palletised.  A robot picks up the jewel-plate sub-assembly from the pallet and inserts it into the part-built assembly.  The operation is relatively complex and an operator has been retained at this station to assist the robot when difficulties arise.

The hairspring is a delicate part that can’t be handled by an automatic feeder. The insertion process is also difficult because the end of the spring is welded to a stub on the jewel plate. This part is assembled manually by two workers in parallel, because of these difficulties.

The second worm gear retaining pin is manufactured from wire and it is most cost effective to manufacture this part on the final assembly line by a guillotining operation. The bending of the pin is carried out simultaneously to the part being inserted and secured.

CONCLUSIONS

1) Product re-design for ease of assembly creates worthwhile savings in assembly costs.  However, particularly for large products, these cost savings must be offset against the additional tooling modification costs for the manufacture of re-designed components.

2 ) When assembling a product which has :
a) Many parts
b) Many variants in the product family
c) A large annual production volume
d) Many common sub-assemblies

a hybrid flexible assembly system is required and it will combine manual, automatic and robotic assembly methods.

3) Sub-assemblies, having a fixed content, are always best assembled on dedicated automatic assembly machines.

4 ) Variable content sub-assemblies are most economically assembled using either
a) Assembly robots
b) Flexible free-transfer machines

5) Transfer between sub-assembly production units and final assembly need large buffers to de-couple these two activities and reduce downtime.