Assembly Sequence Structures in Design For Assembly C J Barnes*

G F Dalgleish**

G E M Jared*

[email protected] [email protected] [email protected]

* **

K G Swift**

S J Tate*

[email protected]

[email protected]

SIMS, Cranfield University, Cranfield, Bedfordshire, MK43 OAL, UK. Dept of EDM, University of Hull, Hull, East Yorkshire, HU6 7RX, UK.

Abstract

assembly and manufacturing problems before detailed designs are produced.

Historically, Design For Assembly (DFA) has always been a reactive tool, most often carried out on products in production and, at best, late on in the product introduction process. Thus, a more proactive approach is required. This paper proposes the use of an assembly sequence model as the focus of such a computer-based DFA tool, which would support the development of CAD models and infer or extract relevant information.

The first implementations of DFA took the form of paperbased methodologies that required much laborious, subjective form-filling to complete the analysis. This was often viewed as unnecessarily time and resource consuming with many opportunities for errors. To overcome these issues, several computer-based versions are now available[1,5,3] These systems are stand-alone computer implementations of the manual methods but again, are only proficient when implemented on detailed design solutions. Embedding DFA/DFM techniques within a CAD environment would support interactive early design evaluation. First indications of success with this approach were reported by Jared et al [6]. This work identified that 72% of the DFA interrogations necessary could be extracted from enhanced solid models. In this way, less emphasis is placed upon user interaction and more data is reasoned from a geometric model.

Introduction It is now generally accepted that product design quality has a significant impact on manufacturing costs and timescales and thus, several product development methodologies have become established under the global heading of Concurrent Engineering. One such methodology is Design for Assembly & Design for Manufacture (DFA/DFM) [1,2,3]. These are formal analysis procedures for the validation and evaluation of design suitability for manufacture and assembly, bringing together multi-disciplinary teams to identify possible solutions. These methodologies can claim success with many industrial applications. However, there is some evidence to support the view that not all of the industrial community has heeded the lessons of the DFA principles. One investigation [4] found that on average, current products have around 50% more parts and assembly content and undergo more complex assembly procedures than is necessary. DFA/DFM introduces more than just assembly and manufacturing information. The Lucas DFA method [2], for example, provides scope for Functional Analysis of parts and the generation of Assembly Sequence Diagrams to reduce the number of components and identify assembly issues. These analyses require detailed knowledge of component geometry and function. However, the approach has no real provision for the proactive analysis of the product as the design progresses. Enabling earlier DFA/DFM investigations would alert designers to possible

The importance of generating an assembly sequence within the DFA analysis has been overlooked by many researchers. Little detailed construction assistance is available to the designer in any of the DFA methodologies. This is despite the implications for the results if an inappropriate assembly sequence is used. A recent industrial study [7] found that, in general, companies who applied DFA tools used an existing assembly sequence for analysis. It found that the assembly plan was only considered in detail when the design was essentially fixed. It is clear therefore, within industry there is a need for an effective CAD-based tool to construct assembly sequences and proactively assess assemblability. This means that existing DFA methodologies may need redefining to support this new approach. Using the assembly sequence as a central theme, this paper outlines the implementation of a more proactive approach to DFA and DFM. The relationship between a suitable assembly sequence and the DFA analysis is considered and the supporting role of geometric reasoning is discussed. This research forms part of an ongoing project.

Assembly Sequence Structures in Design For Assembly As previously discussed, the Lucas DFA Methodology requires the definition of an assembly sequence to complete the analysis. Currently this is constructed manually with no systematic validation or evaluation procedures. In fact, no DFA methodology assists the designer to construct the diagram and ensure correctness either syntactically or semantically. Fig 1 details the assembly sequence diagram for a headlamp trim screw before a subsequent DFA analysis showed it was an inherently bad design. Although this product has only 5 parts, it can be seen that the diagram has already become fairly complex. This difficulty increases with the number of components in the assembly, each of the boxes requiring considerable effort to complete. It is clear that to ensure the validity of any DFA analysis, some guidance is necessary to produce a practical assembly sequence. This could include tips on where to start, what is an appropriate sequence and whether all tasks are valid.

Fig 1 Lucas DFA Methodology Assembly Sequence Diagram for a Headlamp Trim Screw Before Redesign. (See [2] for more information.)

Related Research There is a high level of research activity in the field of assembly sequence generation. Many aspects must be considered to ensure a rigorous approach to the problem. Gottipolu and Ghosh [8] identified 5 major issues in automated assembly planning: • • • • •

Component geometry and topology representation Identification of precedence relationships Generation of feasible assembly sequences Assembly plan representation Assembly plan evaluation

The extensive literature in this field tends to rank individual issues according to a particular view of their relative importance and focuses on these exclusively. Thus, there has been no definitive solution proposed for the automatic generation of assembly sequences. Bourjault [9] was the first to propose a method for generating sequences by liaison graphs. De Fazio and Whitney [10] simplified this method to reduce the number of user questions involved. A decomposition approach using AND/OR graphs was proposed by Homem de Mello

and Sanderson [11]. They also presented an algorithm [12] which takes a description of the assembly and returns the AND/OR graph representation of the assembly sequence. Wolter [13] used a constraint graph to represent the precedence relationships in an assembly sequence. A number of research investigations identified knowledgebased heuristics and used these to develop a feasible assembly sequence. Delchambre and Wafflard [14] generated precedence orders found from a backwards planning approach. Swaminathan and Barber [15] applied a case-based reasoning approach and proposed a plan reuse philosophy to generate assembly plans. Much of the recent research in this area has focused upon links to CAD or solid modelling systems with the aim of fully automating the generation of assembly sequences. Although Santochi and Dini [16] inferred some data from a geometric description of the product, the user was required to input insertion trajectories for each component to reduce computing time. A non-directional blocking graph data structure stores face to face contacts in the Archimedes 2 Mechanical Planning System [17,18], from which feasible disassembly trajectories are computed. Gottipolu [19] divided the solution into three stages: precedence knowledge extraction from a solid model, sequence generation and selection using Boothroyd Dewhurst’s DFA Toolkit [4]. Research activity in the area of concurrent assembly design is small compared to that of assembly sequence generation. Most concurrent engineering systems only focus on component design. Li and Hwang [20] identified this gap and proposed a system which linked an assembly sequence generation module, a DFA analysis tool and a redesign suggestion application. Hsu, Lee and Su [21] examined how data generated at the assembly planning stage could produce redesign suggestions. The integration of an assembly planner and a redesign suggestion module was proposed by Kim, Lee and Bekey [22]. It is evident that automatic assembly sequence generation is an extremely complex problem. This is confirmed by the fact that little commercial software is available to industry to assist with this increasingly important task. In addition, very little consideration has been given to its significance within the DFA analysis, especially in a proactive role.

Motivation There are extensive implications for the successful generation of assembly sequences during the early stages of design. Cycle times can be reduced and rework is decreased. The designer is able to produce a product design in which assemblability has been considered, reducing the need for further DFA analysis. Our proposed environment will support the exploration of trade-offs between DFM and DFA ensuring all options are examined before the design is released for production.

Assembly Sequence Construction As mentioned previously, total automation of assembly sequence generation from a product model representation is an area with many unresolved issues, especially concerning the geometric reasoning interface with the modeller. Constructing the assembly sequence is seen as a beneficial exercise as it supports an assembly perspective upon the design. Our proposal is to devise a suitable user interface for interactive generation of the assembly sequence during the design process. This should be available for use even before all the geometry is defined. It must support nonsequential design to ensure that assembly-oriented design and sequence generation are considered proactively in the design process. Help will be available to the designer to ensure a suitable sequence is generated from this approach. This will take the form of hints and tips ( e.g. which component to start with). The DFA analysis will be carried out concurrently with the construction of the sequence, where possible. This necessitates the need for an adapted methodology to deal with only partially defined component geometry and topology.

materials. The application of surface coatings must also be considered at this point. • compatible joining processes - the joining process, once confirmed as being geometrically feasible, must be appropriate for the types of materials and other part characteristics. • compatible with joint characteristics - there are many types of joint from kinematic to interference contact [23], hence the form of the joint must be suitable for the process and material involved. Of course, identification of potential tolerance stacks could also be included as part of the constraint list. As the intention is to devise a system to deal with incompletely defined models, it was felt outside of the scope of this discussion. To evaluate satisfaction of the hard constraints stated above, certain information is required which must either be explicit within the data structure of the solid model or be ascertainable using geometric reasoning techniques. Obviously some information will only be available through user input.

Assembly Sequence Evaluation Assembly Sequence Validation Relevant criteria are necessary for the validation of the assembly sequence. These criteria can be considered as hard and soft constraints. ‘Hard Constraints’ deal with the geometric feasibility of the assembly. Conversely, ‘Soft constraints’ constitute suggestions for ‘best practice’ and so whilst particular options may be feasible they may not be recommended. By definition, a valid assembly sequence will not violate any hard constraints and will satisfy as many of the soft constraints as the user feels is acceptable after consideration of any potential conflicts. Hard constraints: • consistent geometry - there must be no interference between parts within the assembly other than where intended for a particular type of fit. • feasible trajectory - there must be no collisions as each part is brought into contact with the rest of the assembly, either between parts or between the gripping tools used during the insertion process. • feasible joining processes - each part in the assembly will be held in place either by surrounding parts or by some joining process which must be validated in terms of access for the tools required. • stability - every part must remain gravitationally and elastically stable at all times during assembly. Soft Constraints: • compatible materials - when different materials are placed adjacent to one another there will always be a detrimental reaction of some degree and hence, the assembly must be checked for incompatible adjacent

Some form of evaluation is necessary for the quantitative assessment of the assembly sequence. This would be very much application-based and depend upon individual company standards. The results of this evaluation could be used to select candidate designs or measure improvements of redesigns. This evaluation could include • Assembly time of the product based upon accepted standards • Total number of assembly operations, relative to part count • Total number of non-assembly operations, relative to part count • Design Efficiency (Functional Analysis in DFA) • Handling Ratio (from DFA Analysis) • Fitting Ratio (from DFA Analysis) • Conformability Analysis [24]

Influence of the Assembly Sequence On DFA Analysis The significance of early generation of the assembly sequence in DFA analyses is apparent in most applications. The importance of this can be illustrated with an example. Figure 2 shows part of a screen wiper motor assembly. The original assembly sequence defined the end bracket as the base component. All other parts were stacked above and the rivets inserted and fastened simultaneously. The thermosetting polymer brush plate periodically cracked during the riveting process. This problem was only identified during assembly. The implemented solution supported hollow rivets on pegs in a jig and assembled all

other components using the brush plate as the base. The rivets were then formed to secure the arrangement. Although, this is not a well designed product in terms of assembly, it is not certain that a conventional DFA analysis would have detected all these issues. This is because no consideration is given to validating the proposed assembly sequence. If the designer followed the methodology proposed by this paper, an assembly sequence would be constructed and validated at an early stage in the design process. The inherent constraints built into the sequence generation procedure ensure the material and joining process suitability are considered, thus highlighting the issues well before the actual assembly stage.

To support the requirements of the assembly sequencing process it is necessary to either represent explicitly or infer the spatial and temporal aspects of an assembly plan such as : • • • • • • • • •

component locations and orientations mating faces joint characteristics joining processes degrees of freedom insertion trajectories cross-sectional properties mass properties sequence of events

For the purposes of DFA analysis, however, it is also necessary to understand more complex issues such as degrees of symmetry and shape classification.

State of the Art The four-layer product model [6] is one representation which, together with geometric reasoning techniques, may be used to provide the relevant information at the appropriate time during the analysis. Each layer of the product model depends on, has access to or includes all data in the lower layers. They are labelled as follows: Fig. 2 Partial Assembly Of Screen Wiper Motor, Showing Revised Design With Hollow Rivets It can be seen from this example that a poor design can have a well structured assembly sequence, conversely a good design may have poorly organised assembly plan. Our approach would ensure that every DFA analysis was based upon a valid assembly plan. Thus, a more accurate analysis is carried out. This new technique can also highlight many assembly problems caused by poor designs that existing analyses currently overlook. Completing the analysis earlier in the design process addresses the assembly issues before major costs are incurred.

Computer Support It is taken as axiomatic that DFA will be delivered through computer-based tools and furthermore, that interrogation of and inference from a product model is essential. A proactive approach to DFA, as discussed above, requires that the appropriate analyses be invoked at the earliest possible stage of the design process. This will involve decision making with incomplete information. Current DFA analyses depend, for the most part, on geometric considerations to evaluate an assembly. Therefore, to promote the early use of DFA, either the analyses must be redefined or it will be necessary to devise a means of representation of components which will allow the assembly to be evaluated, at least to a limited extent. It is evident then, that data held within a solid model would be invaluable to decision making processes.

• component model - solid model together with attribute information such as surface finish, etc. • assembly - position and orientation information for each component within the assembled product • component interaction model - mating faces information incorporating joining processes, method of assembly • assembly plan - time sequence information i.e. sequence of assembly operations including nonassembly processes. An essential part of the four layers is the representation of features. Unfortunately, the concept of a feature is rather imprecisely defined. This is because the interpretation of a feature often depends upon the field of application or the viewpoint of the individual. For instance, a slot may be regarded as the union of three cuboids or the intersection of two depending on whether there is a material removal or material joining process in mind. Historically, there have been two approaches to generation of features information: ‘Design-by-Features’ and ‘Feature Recognition’ [25]. The first method, also commonly known as ‘feature representation’ and ‘form-feature modelling’, requires the user to select fragments of a component from a standard library to compile the product description. The second method often involves post-processing an existing solid model to find occurrences of types of feature using a computer program. Clearly a combination of these two systems will be necessary to cater for the effects of different viewpoints.

As already stated, DFA has several requirements relating to symmetry and these remain to be properly specified in geometric terms. Obviously, there may be approaches to symmetry detection which take account of the features of a component. Many researchers have developed successful feature recognition algorithms for boundary representations of solid models. A few have tackled the issue of detection of symmetry but these approaches have little bearing on this application. Interrogation methods for features within a constructive solid geometry (CSG) solid model have been somewhat neglected whereas a promising algorithm for detection of symmetry has been developed, which also has applications in nesting and docking problems. Woodwark’s Algorithm [26] involves a multidimensional search to match instances of a template with the object of interest. Again, this work is inappropriate for this application. The field of representation of incomplete models seems to be relatively unexplored to date.

Further Development of Geometric Reasoning and Product Modelling

Implementation For the purposes of this project it has been established that the interface of the proposed system must be an environment with which the designer is familiar and, logically, this must be the CAD tool. There are many design systems available. However, for the purposes of geometric reasoning which must take place in the course of the DFM/DFA analyses, it is essential that the design tool includes a solid modeller with an open and accessible interface, and furthermore, that it can represent nonmanifold objects. The ACIS solid modeller is a ‘kernel’ modeller which fulfils such requirements. It is anticipated that the system will incorporate a windows based graphical user interface (GUI) and other software tools which would ideally be written in a compatible language. It is foreseen that such useful tools may include a database package for materials, processes and DFM/DFA score tables, a data management system to handle the links between all parts of the system and a knowledge-based/rule-based system to hold the decision support information. Figure 3 shows the potential structure of such a system.

For the purposes of this research, additional methods of shape classification are required. It is proposed to implement present methods of feature representation and recognition and pursue the development of: • the creation of an environment to represent and evaluate an incomplete product model • the development of an algorithm for identification of major axes and/or degrees of symmetry within a boundary representation of a solid model. Several means have been identified to deal with an incomplete product model: • use of a solid modeller with the capability to handle non-manifold objects • propagation of assemblability constraints through the product model • implementation of knowledge-based system methodologies. It is proposed to devise methods for symmetry detection which exploit the presence of features information and the explicit nature of a B-rep solid model. The three main obstacles to these geometric reasoning problems are: • the lack of a standard API for solid modellers [27] • controlling computational complexity • accurate, unambiguous articulation of relevant features. The first may be addressed by using a kernal modeller with an open, accessible interface. Experimentation with the implementation of feature recognition algorithms and their effect on DFA will be required to tackle the other two problems.

Fig 3 Proposed System Structure Diagram

Conclusion We have proposed that an assembly sequence could form the basis of a more proactive approach to DFA in a solid modelling environment, invoking functions and accessing information as required from a ‘product model’. It is intended that this new methodology should be used concurrently within the design process. We have shown that for earlier implementation of the DFA analysis, the underlying product model is required to deal with instances

of incomplete data. It is necessary to redefine the existing Lucas DFA Methodology to ensure suitability for analysis with partially defined component geometry. We have shown evidence that this new technique would identify assembly issues that conventional DFA analyses generally overlook.

Acknowledgements The authors would like to express their thanks to Peter Robinson and Hong Mei, University Of Hull, John Todd and Mark Limage, Rover Group Ltd., Graham Hird , CSC and Henry Merryweather, RADAN Computational Ltd for their assistance with this work. This research has been carried out under EPSRC Grant Numbers GR/K 78485 and GR/K 74401.

References [1] G. Boothroyd, P. Dewhurst. Product Design For Assembly. Boothroyd Dewhurst, RI, USA 1989. [2] Lucas Engineering Systems Ltd, University Of Hull. Design For Manufacture and Assembly Practitioners Manual. Version 10, 1993. [3] S. Miyakawa, T. Ohashi. The Hitachi Assemblability Evaluation Method. Proceedings International Conference on Product Design For Assembly, 1986. [4] G. Boothroyd. Product Design For Manufacture And Assembly. Computer-Aided Design, 26(9):505-520, 1994. [5] CSC Computer Sciences Ltd. TeamSET Users Manual. Version 2, 1996. [6] G.Jared, M.Limage, I.Sherrin, K.Swift, Geometric Reasoning and Design For Manufacture. Computer Aided Design, 26(7) : 528 - 536, 1994. [7] D. Polfreman, Early Evaluation Of Assembly Sequences - An Investigation. Thesis Cranfield University, 1996. [8] B. Gottipolu, K. Ghosh. An Integrated Approach To The Generation Of Assembly Sequences. International Journal Of Computer Applications In Technology. 8(3/4):125-138, 1995. [9] A.Bourjault. Contributionune approche methodologique de l’assemblage automatise: elaboration automatique des sequences operatiores Thesis d’Etat Universite de Franche-Comte, Besancon, France, 1984. [10] T.De Fazio,D.Whitney. Simplified Generation Of All Mechanical Assembly Sequences. IEEE Journal Of Robotics and Automation 3(6): 640 - 658, 1987. [11] A.Sanderson, L.Homem de Mello. And/Or Graph Representation Of Assembly Plans. IEEE Transactions on Robotics and Automation. 6(2):188 199, 1990. [12] A.Sanderson, L.Homem de Mello. A Basic Algorithm For The Generation Of Mechanical Assembly Sequences. Computer Aided Mechanical Assembly Planning. Kluwer Academic Publishers, 1991.

[13] J.Wolter. On The Automatic Generation Of Assembly Plans. Computer Aided Mechanical Assembly Planning Kluwer Academic Publishers, 1991. [14] A.Delchambre, A.Wafflard. A Pragmatic Approach To Computer Aided Assembly Planning. Proceedings IEEE Conference on Robotics and Automation, 1990. [15] A.Swaminathan, K.Barber. An Experience Based Assembly Sequence Planner For Mechanical Assemblies. IEEE Transactions on Robotics and Automation 12(2): 252-267, 1996. [16] M.Santochi, G.Dini. Computer Aided Planning Of Assembly Operations : The Selection Of Assembly Sequences. Robotics and Computer Integrated Manufacturing, 9(6):439 - 446, 1992. [17] R.Wilson, J.Latcombe. Geometric Reasoning About Mechanical Assembly. Artificial Intelligence 71(2) 1994 [18] S.Kaufman, R.Wilson, R.Jones, T.Calton. Archimedes 2 Mechanical Assembly Planning System. IEEE International Conference on Robotics and Automation, pages 3361-3368, 1996. [19] B.Gottipolu. An Approach For Assembly Sequence Planning Using Solid Models. Thesis University Of Montreal, 1994. [20] R.Li, C.Hwang. A Framework For Automatic DFA System Development. Computers In Industrial Engineering 22(4):403-413, 1992. [21] W.Hsu, C.Lee, S.Su. Feedback Approach to Design for Assembly By Evaluation Of Assembly Plan. Computer Aided Design, 25(7):395 - 410, 1993. [22] G.Kim, S.Lee, G.Bekey. Comparative Assembly Planning During Assembly Design. Proceedings IEEE International Symposium on Assembly and Task Planning, pages 319-332, 1995. [23] A.Lin, T.Chang. Framework For Automated Mechanical Assembly Planning. Journal Of Mechanical Working Tech. 20:237 to 248, 1989. [24] K.Swift, J.Booker. Engineering for Conformance. The TQM Magazine. 8(3):54-61, 1996. [25] J.Shah. Assessment Of Features Technology. Computer Aided Design. 23(5):331 - 343,1991. [26] S.Parry-Barwick, A.Bowyer. Multidimensional Settheoretic Feature Recognition. Computer-Aided Design, 27(10):731-740, 1995. [27] Geometric Modelling Society. Introducing Djinn - A Geometric Interface for Solid Modelling. Information Geometers Ltd., 1995.