AUTOMATIC ASSEMBLY IN BATCH PRODUCTION A. d'Auria OSAI - Olivetti Sistemiper I'Automazione Industriale - Ivrea
ABSTRACT In I ight manufacturing industries, assembl y operations pla y an impo r tant e c onomic role. The development of advanced techn o logies and the a utomation of manufacturing processes has I ed to ever decreas i ng pl' i c es for c omponents wh i le assemb ~ y operations continue to be carried o ut manually. It may thus happen that it is cheaper to manufa c ture a part th a n it is tu as s emble it. The traditional methodes of mechanising ass e mbly operations c o nsist in rotating table or transfer machines of very high cost and suitable fOI high v o lume productions. The main obstac le to the wide diffusion of the t r aditional automation ecuipment is its rigidity; this makes it unsuitable in the presence of varying production rates, produ c t modifi c ati o ns or produ c t diversification for different markets. Further in rea I i ndustr i a I env i ronments, one often finds smal I and medium sized pr oduction volumes which require a job-lot o rganization rather than a continu o us-flow one. To solve th e se problems o ne needs flexible automat ion ecu i pment h,h i c h may be re a d i I y and eas i I y re-too I ed t o pass from one product to an other. Modern second generati o n computerised industrial rob o t te c hnology enables this important problem area to be fa c ed. However, to develop a new te c hnolog y one must not only design and manufa c ture new ecuipment; but one must also stud y the problems connected with the introducti o n o f such equipment in the pr 0durtion process. I n the f o I I ol,t i ng pages h'e wi I I theref o re attempt to ana I i se the pecuI iar characteristics of the assembly process to determine conditions which must be respected during the design o f the produ c ts and the o rganiz a tional impl i c ations which dri v e fr om the introduction of such e q uipment. Finally we wi I I examine the information flows which interest the assembly pr o cess. 2. Assembly from a technical point of Ylew Let us suppose that the input t o the ass e mb I y pr o cess be n pi I es of random bu I k parts and that the output be m a 0 sembled unit s . The transf o rmation pro c ess, pas s ing f r om input t o o utput, require s the or ientation of each component in a suit a ble wa y and plarinp it in a time-ord e red s equence with respect to the others, ac c ording to wel I-defin e d traj e ctories which are a function of the geometr y o f the parts and of the structure of the unit being assembled. In an automatic process it is convenient that th e tra j ectories be as simple as possible; in parti c ular, it is useful if ea c h component can be assembled independently and the assembly direction is uni ~ ue. In this way it is possible to define an assembl y c y cle in terms of successive overlays, reducing to a minimum the number of degrees of freedom of the parts manipulator. A particular case of assembl y is f o und in sele c tive fitting; in this case, we must first measure a geometric dimension of the c omponent and then fit it to a second one picked from amongst se\eral classes of pre-selected parts. It is thus necessary that the output of the measuring instrument be fed to the system sO as to determine the successive steps.
3. Assembly from a orsanizational point of view The organization of an assembl y line impl ies a flow of materials whi c h guarantees the arrival of the right components a t th e right moment and in the right place so as to ensure the re~uired output of products both in type and in ~uan tity. 153
The assembly organization may be either serial or parallel. It is cal led serial when in one station or work-place al I the components required to f~m the unit are assembled. The organization is said to be parallel when in each of the m stations of the I ine one and only one component is added. A serial organization is suitable to small and medium volume productions, whi le parallel systems are suitable for high vol'Jme productions. A significant parameter of an assembly process is the number ~c~nents to be handled. With a para I I e I organ i zat ion, if the number of components and therefore of stat ions is high, the overall rei iabi I ity of the system, which corresponds to the product of the reliabi lity of each station, is 10h' . On the other hand, if a system is capable of handling only a I imited number of components it is necessary to break down the product to be assembled into a certain number of sub-assemblies. Such sub-assembl ies must be connected in such a way ad to avoid them fal ling apart during the successive unloading operations. This also is a consideration to bear in mind during the design phase.
Considerations on the qual ity of the components used In the assembly process
Modern assembly processes are bases on Fordism,that is, on the principle, appi ied for the first time by Henry Ford in the production of automobiles,of the interchangeabi I ity of component parts by means of the appropriate use of tolerances. Further, mass production processes guarantee that the parts be on average of sufficient qual ity but the statistical control methods which are normally applied result in the acceptance of lots which contain a percentage, be it a small one, of faulty parts. Whi le this fact is perfectly acceptable in manual assembly, since it is the assembly line worker which carries out a final selection when the part is not suitable, it causes a series of problems and of stops in processes based on automatic systems of the deterministi c type. This fact is particularly serious in parallel systems; in these, each reject in one of the m stations may prejudice the correct functioning of al I the succeeding ones. Attempts have been made to el iminate this problem by placing,after each assembly station, a station which controls the assembly operations; this station blocks the c ompletion of the sub-assembly in the successive stations of the assembly line. Even with this solution, there is the risk of drastically reducing the overal I productivity of the system. Modern computerized systems for serial assembly work tend to resolve this problem by applying to the grippers suitable general-purpose force and position sensors; these enable the system to check that each phase of the assembly process has been carried out successfully. Further, should this not happen,it is possible to activate suitable emergency procedures (eventually of an interact i ve nature) to enab I e the system to term i nate the operat ion successfu I I y. Naturally these senSOI"S and the relative pr oce dures I"or k o nly I,hen the fault in the part makes it impossible to assemble it. Whenever the defect is of a nature which has no bearing on the esse mbly process (eg esthetic type of defects) it wi II be necessary to carry out controls on the output of the system.
5. Parts feeding Normally, in a produ ct ion process, the parts are produced in lots and stored in a random bu I k manner. On I y in a few cases are the components supp lied ordered, in special containers, e.g. integrated circuits or reeled components. Normally however, the storing of parts in an or dered manner may be discarded for the following reasons: a) the final operations of the production cycle randomize the components (heat treatments, surface treatments etc) b) the high number and cost of special purpose containers c) the difficulty of handl ing the circulation of a high number of specific containers. r n manua I assemb I y processes the h'ork-stat ions are therefore fed from pi I es of random bulk parts; the worker, coordinating his movements under the contl',, 1 of his sight and touch organs, pre-orients each part before assembl ing it. The elementary steps of the process consist of separating one and only one piece from the pi le and successively orienting it in space during the closing-in phase so as to have it in the most suitable position to faci I itate the final "placing" trajectory. If one attempts to design an anthropomorphic type of system, one is faced with technological and, more importantly, econcmic difficulties. Some interesting attempts have been made in important research institutes, but these were aimed
Automatic assembly in batch production
at relatively simple cases of manipulation of prism~tic objects. On the other hand however, the real industrial environment recuires equipment capable of repeating thousands of times the same cycle and with adeguate performance. We feel therefore that the solution must entai I the design of general-purpose orientation devices capable of operating off line at high speed and feeding to the assembly system proper, the parts In an ordered manner and in specially designed containers. The possibi lity of rapidly converting the orientation device from the handl ing of one component to another, would enable this work to be carried out at the input to the assembly machine; this could reduce to a minimum the number of specific containers required. 6.
Information flow in assembly processes
From what stated in the preceed i ng paragraphs, it may be seen that in each of the steps In which we have subdivided the automatic assembly process, it is necessary to add to the components being assembled a certain amount of information. This concept may be summarized in the following way: - Random parts + information on required orientation = ordered parts - Ordered parts + information on the trajectories = set of elementary assembly operations - Set of elementary operations + information on their se~uence = assembly cycle - Assembly cycle + operating conditions information fed back from the system = hi stor ica I date. For a serial system fed from a general purpose orientation device,the various steps of the process may be represented by the following block diagrams.
Random bulk parts
Measure shape and position of part
Yes C91~ulate the p£ Sltlon ot the part relative to the correct one
Move part into the correct position
Set of oriented parts Fig. 1
A. d' Auria
Set of the elementary assembly operations
A, B •• , N
Parts feeding (preoriented)
Elementary assembly routines for each part (control led by the sensors).
•. , Rn
We may have several
Reject part A and try again
The system goes in stop
Unload partly assembled unit and start new cycle from the beginning
Shift a certain distane In a certain direction and try again.
Automatic assembly in batch production
Fig. 4 Assembly cycle The set of elementary assembly operations integrated with informations about the secuence of operation yeld the right assembly cycle. Such sequence IS not univocable determinated because the branch of each recovery procedures (fig.2) can produce different path and then different cycle. The artificial intel I igen~ techniques may help in solving the problem of finding the optimal strategy.
Finally, at the running time it system behav i our
possible to log historical data about
tl umber of assemb I ed un i ts
Number of rejected units Number and type of faulty parts Number of recovery calls Stops of the system ( No feeded parts)