DESIGN

THE NAMING OF

PARTSDesign is an essentialoverhead to any companyor project. The problem

facing design managers ishow to reduce the design

cost without loweringstandards.

Brian Hodgson describeswork at Southampton

University which has thepotential to reduce design

costs by up to 50% insome areas.

The expression 'to reinvent the wheel' iswell known to all good designers, and itis an activity which is clearly counterpro-ductive. However, the administrative

systems within which most designers work actuallyencourages repetition of design with the resulting,and yet unnecessary, duplication of design effort. Afurther problem in industry is that many good de-signs are produced to solve the wrong problem. Ifthe design problem is badly defined, the resultingsolutions are often far from the optimum, resultingin over design with higher costs, poor reliability, lostmarkets etc. Our work at Southampton Universityaddresses these two problems with the ultimateobjective of enabling designers to find design solu-tions which satisfy all the problem needs, but at thesame time avoid design duplication.

The approach used is one of identifying the part,system or assembly by its functional requirementsand the essential characteristics which it must have.In effect, this means identifying the part by its designspecification. Once the functional capability of anitem is used as its identifier within the system, it isthen possible to retrieve the item to meet a similarfuture need, without further expenditure of designeffort. This has many advantages, because not onlydoes it reduce design time, it also reduces the rangeof items being produced, the need for spares, andleads to a progression towards standardisationwhich is driven by product need, not formulated bycommittee.

This approach is known as: "specification basedparts-numbering". In this case the specification is thedesign specification defining the part or system need.

Specification based parts-numberingA good designer will set about a design by first

ensuring that the definition of the problem is correct.Any solution which satisfies this definition, or spe-cification, will fully satisfy the design need. A de-sign specification used to define the design problemcan be divided into two main categories: the func-tions which the design must perform and the char-

acteristics which the design must have in order tofully satisfy the requirement.

To illustrate this, consider the simple exampleof an electrical socket. The functions it must performare: to retain the plug, transfer electrical current tothe plug from the incoming wires, to insulate thesupply, and possibly provide location relative to awall or bulkhead. The characteristics would includesuch aspects as: current rating, voltage rating, pinconfiguration, impact resistance, thermal limits, sizeenvelope, appearance etc. If a system can be devisedwhereby the part number of the component, or theassembly, can be made to contain information de-scribing the specification which that item satisfies,then the drawing of the item can be filed accordingto its functional capabilities.

This means that a designer intending to designan item would first define the specification to besatisfied by the item, ie the function which it mustperform and the characteristics which it must have,and then would be able to define the part numberwhich would ultimately be used to describe the itemwhich they intend designing. However, before start-ing a new design they could use the part number tosee if a suitable item already existed, and if so, theymay then be able to use it, possibly with modifica-tion, and so avoid wasted design effort. Alternative-ly, they may find that the existing item is not totallysuitable, but it may be possible to introduce a newdesign capable of satisfying the previous need, aswell as the new need, with a consequent reductionin stockholding.

There are many benefits which result from spe-cification based parts-numbering and some willnow be outlined. The first is that of identifyingstandard items regularly used by a company orindustry. At the present time, standardisation nor-mally follows a route of a company realising that ituses a number of very similar components with onlyminor variations between them. A good examplemight be electric motors. A decision is then takento try and rationalise the variety of motors beingused, and this would normally be done by having a

MANUFACTURING ENGINEER NOVEMBER 1991

DESIGN

It's not enough to usethe latest technology

to design your products -by focusing on the function

of each part as it isdesigned, savings and

efficiency are possible

working group review the motors needed and thenspecify certain types which would be preferred.

The problem with this exercise is that stan-dardisation is now "frozen" to developments inthose particular motors. If, over a period of years,the selected motors lag behind technical develop-ments, then the designs incorporating them aretechnically lagging. The specification based parts-numbering approach would be to define all themotor's functions, which may include more thanproviding mechanical power, eg speed or torquecontrol, and then define the essential characteristicssuch as: running torque, starting torque, pull-uptorque, pull-out torque, running speed, maximumor minimum inertia, reversibility etc.

Normally, standardisation can become ob-sessed with geometrical form and the specifying ofmaterials from which the item must be made. Takingthe motor as the example, the essential geometricalcharacteristics are quite limited and would includesuch aspects as the drive shaft to mountingplate/flange configuration and dimensions, shaftdiameter and the maximum volume envelope. Thematerials would be specified by defining the char-acteristics representing the environment, eg tem-perature, humidity, supply voltage.

By describing a standard motor in this way anymotor which satisfies these essential criteria is ac-cepted as meeting the standard, but at the same timenewly developed motors meeting the same criteriaalso satisfy the standard. Consequently, the stand-ard does not become technologically frozen as canhappen with the usual standardisation procedure.

The motor illustrates another benefit of specifi-cation based parts-numbering: that of integratingthe numbering of bought-in components into the

same system for those manufactured in-house. Twobad practices encountered in many present daydesign offices are those of assuming that a need canonly be met by designing and manufacturing itemsin-house, and the second is the failure of designersto realise that bought-in components are required tosatisfy a design need, and therefore the designspecification for that need must be formulated beforethe component is sought.

By integrating the numbering of bought-in com-ponents through specification based parts-numbe-ring, the designer is alerted to these shortcomings.Essentially, because the part number is based on thedesign specification defining that need, then the partnumber will be the same whether the item is ofin-house manufacture or bought-in.

This has the advantage that when the designerhas formulated the design specification to satisfy theneed, then the code representing that specificationis generated meaning that the designer is encour-aged to consider items bearing the same or similarpart numbers whether made in-house or bought-in.Furthermore, the need to define the specificationfirst ensures that any bought-in parts are selectedon the basis of a properly defined need; many de-signers, at the present time, are inclined to browsethrough manufacturers' catalogues, with the sup-plier imposing the definition on the design.

Another problem encountered in many indus-tries is the need to trace components within assem-blies. Sometimes this is because items or sub-sys-tems have to undergo modifications or because alater generation of the item becomes available andreplacement is needed. To undertake this task onpresent day numbering systems requires extensivecross-referencing facilities to be established early in

the design management to ensure that all like itemsare traceable regardless of where they appear in thetotal system. With specification based parts-num-bering all items or sub-systems performing similarfunctions will possess a similar part number.

An unusual benefit was suggested by one de-signer. In his case the company was generatingmore part numbers than were required for the pro-duct. This came about because, being a large or-ganisation, when the designers were undertakingdesign schemes, or layouts, the company needed toallocate numbers to the parts being generated.

In a smaller organisation, the procedure wouldbe for the designer to lay out ideas as a designscheme, which would comprise a general arrange-ment drawing of the proposed idea, showing thecomponents and sub-systems which make up thedesign. Normally, at that stage the design would bediscussed and some items would be deleted oradded, and some would be merged.

When the scheme had been finalised parts num-bers would be added. In the case of a large organi-sation, the scheme may have to be sent to severaldifferent sites or locations, even in its tentative form,and this would require items to be numbered forpurposes of identification in subsequent discus-sions. When parts were then deleted or merged theirnumbers still remain registered and held within thesystem. With specification based parts numberingthis does not matter, because when a similar func-tional requirement occurs the part number will bereactivated and used.

In a sense, a system of specification based partsnumbering establishes a universal numbering sys-tem based on functional requirements and, there-fore, a set of numbers or codes exists to describe all

MANUFACTURING ENGINEER NOVEMBER 1991

DESIGN

functions, and whether a particular part number isused within a company depends on whether thatcompany is currently producing, or using, itemsperforming that function.

Parts coding of systems, subsystemsand components

The approach described so far has been limitedto the coded description of individual items. How-ever, in industry, the need is for numbering systemswhich describe products completely at all sub-sys-tem levels. It is within the capabilities of specifica-tion based parts numbering to give consistencybetween levels.

As an example, consider the design of a hydraulicactuator. The primary function of the actuator is toconvert hydraulic energy into mechanical energy interms of a force being applied throughout the stroke.The actuator will also possess various charac-teristics, eg the values of the force with stroke posi-tion, stroke length, environmental resistance. Thecombinations of the codes for the functions and thecharacteristics will uniquely describe the actuator.

Within the actuator, there will be various othercomponents, such as the piston and its seals. Thesein turn will have their own functions and charac-teristics which will enable them to be describeduniquely by their own parts numbers. Therefore, thefinal product, ie the actuator, will have a code de-scribing its design specification, as will the otherindividual items of which it is comprised. The ac-tuator might, in turn, be used in a larger systemwhere, again, its part number will be consistent withits function.

To link an item to an assembly it is still necessaryto have a parts schedule, as with other numberingsystems. Numbering systems which are based on ahierarchial division of a project into its componentsstill require a schedule to identify the quantitiesrequired for an assembly.

The research at SouthamptonThe concept described so far appears simple.

The major problem is one of establishing a universalsystem of coding functions and characteristics.Normally, the number of functions which an itemmust perform is small compared with the charac-teristics it must possess. Furthermore, a significantproportion of the characteristics required of an itemare directly related to its functional requirements.Therefore, it is likely that the final form of the partsnumber coding will lead with the function code andbe added to by the characteristics element of thecode. Research to date has centred on basic con-cepts related to the grouping of functions. Thefindings suggest that it is possible to uniquely rep-resent all possible functions with a code of less thanthree characters, although there may be presenta-tional, or data handling, advantages in extending thenumber of characters used. One aspect to be con-sidered is the representation of items which are

capable of multiple functions. The relative import-ance of certain functions in particular applicationsmight suggest that the total function code may groupthe codes, for the individual functions, into a priorityorder, but this could cause items not to be retrievedfor new designs if the required functional capabilityhad been given a low ranking. Present thinkingsuggests that items having multiple functional ca-pabilities should not have their codes ranked.

Whilst it is likely that this work will endeavour toachieve a universal system for classifying functionit is only feasible to devise a system for the classi-fication of characteristics in a finite area. This areawill be dictated by the nature of the work from thecompanies participating with the University in thisstudy. There are certain problem areas in the classi-fication of characteristics.

One of these is the problem of defining theinterface between two components. Since the char-acteristicscode is based on the design specification,the interface for the item must be defined in terms ofthe two or more items with which it must effectivelyinterface, but these in turn will have their interfacedefined in terms of the item being considered. Asan example, take two items directly connected toeach other to form a pipe, probably a stop valve anda regulator. If the regulator has a flange connector,then the valve characteristics must require a flangeconnector, but if the regulator has not been defined,how should the valve interface be defined?

Another is the realisation that characteristics arenormally represented by parameters to which areattached values, which are normally dimensional.This means that numerical values included in theparts code have to be such that they do not precludethe item's use in another application.

Alternative parts coding systemsThere are essentially three different types of

parts-numbering systems currently in use. Theleast satisfactory is the "next number out of thebook" system which is used by many companies. Ithas little to commend it other than it provides aunique number to each item. The system works byhaving a reference book, or a computer data base,with a string of available numbers. When a new partis designed a description of the part is placed along-side the number and the number is allocated. Oftena product will comprise parts which are numberedat very different places in the "book".

Some companies try to refine the system byallocating blocks of numbers to products leavingfree numbers for subsequent additions. The "book"is often a very poor route into identifying drawingnumbers, as it depends heavily on the drawingdescription; often the designer will describe thedrawing's appearance rather than the item's func-tion. When the author once queried why a particulardesign office described all its drawings as 'weldedassembly', the response came that this was to com-ply with NATO terminology.

A number of larger companies number drawingson a project hierarchy basis. The first charactersdefine the project, the next the major assembly,progressing along the character set until the individ-ual component is uniquely identified. This systemhas merit in relating the items to the project and inproviding an extendable numbering system with adefinite project identity. However, it does not en-courage the use of common parts between assem-blies unless the parts are given an identifying partnumber as part of a standardisation process. Other-wise, identical parts on two different projects wouldhave different parts numbers.

The most developed parts numbering system todate is that based on group technology, for example,the Brisch or the Opitz systems. These systems areparticularly strong in rationalising componentshape to reduce variety in manufacture and to enableplant to be grouped into manufacturing cells in orderto optimise plant layout and the allocation of itemsto manufacturing facilities. However, this approachstill requires the design to be undertaken to the detailstage before the rationalisation can be made. Byusing the functions approach of specification basedparts numbering much of this design effort can beeliminated. There is merit, from a manufacturingpoint of view, in using a group technology numbe-ring system in parallel to describe the geometry ofcomponents which will need to be manufactured.

ConclusionsSpecification based parts numbering has great

potential for significantly reducing design costs andraising design quality, by focusing the designer'sattention on the design specification at the start ofeach design task. The first stage of the work ondesign functions suggests that this could be usedas a first stage in a parts numbering system which,even though not complete, could significantly re-duce costs and raise standards. The work in develo-ping the coding of characteristics will become moreproduct specific and will be centred around theproducts of collaborating companies to the benefitof these organisations.

At the moment, conventional parts-numberinghas little benefit in the reduction of design costs andthe raising of design standards, in fact, some sys-tems significantly increase design office loading.More recent developments in parts-numbering havemade significant improvements to manufacture andit is the intention that specification based parts-numbering will produce similar improvements inreduced design office costs and higher designstandards. EQFor more information enter ME34

Companies wishing to participate in this workshould write to Dr Geoff Pitts or Mr Angus Tavner,Department of Mechanical Engineering, The Univer-sity, Southampton S09 5NH.

MANUFACTURING ENGINEER NOVEMBER 1991