Collaborative product design and development

Transcription

1 CHAPTER 6 Collaborative product design and development D.T. Pham & N.S. Gourashi Manufacturing Engineering Centre, School of Engineering, Cardiff University, United Kingdom Abstract Applying computer techniques to the processes in the product design and development cycle can improve product quality, shorten lead time and reduce the total development cost. Product design and development is usually a team activity. This chapter discusses computer-supported collaborative design and development and proposes a technique for supporting collaborative design. The chapter gives an application example relating to robot gripper conceptual design. 1 Introduction The design and development of products require the participation of individuals from different domains of expertise. Figure 1 shows the composition of a typical product development team for an electromechanical product of modest complexity. The members of such a team need to communicate with one another throughout the development cycle. They have to take the right decisions as early as possible in order to avoid unnecessary delays to the development process. Lack of communication between participants within or from outside the various sub-teams shown in the figure could result in tasks having to be revisited long after they have been first executed. doi: / /06

2 Finance Sales Team member Team member Legal Team member Manufacturing Engineer Team member Marketing Professional Industrial Designer TEAM LEADER Purchasing Specialist Electronics Designer Team member Core Team Mechanical Designer Team member Team member Extended Team Team member Team member Figure 1: A typical product development team for an electromechanical product of modest complexity (adapted from [1]). The product design and development cycle is shown in Figure 2. The cycle consists of two main processes, the design process and the manufacturing process. The figure also shows where computing technologies have been applied to support it. The manufacturing process has been assisted by the computer technology known as Computer-Aided Manufacturing (CAM). The activities in the design process can largely be classified as two types, synthesis and analysis. The analysis sub-process is supported by the technologies known as Computer- Aided Design (CAD) and Computer-Aided Engineering (CAE). However, as can be seen from the figure, the synthesis sub-process currently lacks computer support. Within that sub-process, there lies the early phase of design known as conceptual design. This critical phase often determines the success or failure of the product.

3 Design Process Synthesis Design need Design specifications Feasibility study & collection of design information Design conceptualisation Design documentation Design evaluation CAD + CAE Design analysis optimisation Design Analysis Analysis model Process planning Production planning Packaging Design and procurement of new tools Ordering of materials Quality control Production Shipping Sales and marketing NC, CNC, DNC programming Manufacturing Process Figure 2: Computer support in product design and development (adapted from [2]). Existing computing techniques employed to aid product design and development have two shortcomings. First, they do not address the needs of conceptual design which is the most critical phase of the development cycle. Second, the tools developed to date are more suitable for individuals than for groups working collaboratively. There is a need to develop computer techniques that aid the currently nonsupported activities in the product development cycle. The developed tools should be targeted at teams doing collaborative work.

4 Computer-supported co-operative work (CSCW) is the body of theory and practice that is concerned with the use of computers to assist and enhance the work activities of groups. Collaborative design and product development is an area of research that utilises CSCW. In the context of CSCW, collaborative design has two requirements: a shared representation of the design problem and a work space that the designers view and interact with [3]. This chapter focuses on the first requirement. It proposes a representation for supporting designers during the conceptual design phase. This will allow designers to view and dynamically modify design concepts. This representation can then be shared using some media techniques in order to develop real synchronous collaborative systems. In the remainder of this chapter, conceptual design is first reviewed. Next a technique for supporting conceptual design is presented. The implementation of the proposed technique using ICAD [4], a knowledge-based environment that allows the development of intelligent conceptual design systems, is then discussed. Finally, an example is shown to demonstrate the technique on the task of robot gripper design. 2 Conceptual design and collaboration Conceptual design is a complex, iterative and open-ended process. As previously discussed, it requires a great deal of decision making at a stage in the design process where knowledge about the design requirements and constraints is imprecise and incomplete. Furthermore, a great deal of expertise and creativity is required in order to conceive new design concepts. The conventional systematic approach for supporting conceptual design is shown in Figure 3. The design specifications constitute the input to the process. The design team starts by abstracting from these specifications to identify the essential problems. Then the team breaks down the design problem into simpler sub-problems, the most popular decomposition scheme being function decomposition. With function decomposition, the team establishes a function structure, in which the overall function of the design is decomposed into simpler sub-functions. When the function structure is complete, the team searches for sub-solutions to satisfy the lowest-level sub-functions that have resulted from the decomposition. Finally the sub-solutions are combined into a single solution, which will provide the overall function of the design. The above methodology is ideal when human designers are predominant in the design process, where the sequence of tasks can be successfully followed. Where computers are to be employed, however, this methodology will not work as effectively. Some of the tasks are currently too difficult for computers to carry out, due to the large amount of heuristic knowledge and creative thinking involved. Clarifying the essential problems, establishing function structures and combining sub-solutions into a final working solution are feasible tasks for human designers, but not yet for computers.

5 Specification Abstract to identify essential problems Establish function structures Overall function - sub-functions Search for working principles to fulfil the sub-functions Combine working principles into working structures Select suitable combinations Firm up into principle solution variants Evaluate variants against technical and economic criteria Principle solution (concept) Figure 3: Steps of conceptual design [5]. Furthermore, human designers like to be able to visualise and interact with their developing design ideas, as well as communicating them to others. Simple sketching, for instance, continues to be the most important tool for the designer, alongside computer-based geometric modelling. Design is collaborative by nature. Collaboration means working together in order to achieve a common goal. There are four different types of collaboration [3]. The first is where collaboration takes place at the same time and location. An example of this is when a team meets in a room to accomplish a particular task. The second type is where the collaboration is at different times but at the same place. An example of this is a design workstation which can be accessed asynchronously by different users. The third type of collaboration involves different times and places. An example of this is the use of to communicate with other members who have access to shared distributed databases. The fourth type is when the collaboration takes place at the same time but at different locations. The latter type is currently the target of research on

6 collaborative design and product development. Designers working in a team need to share their ideas and expertise. This type of collaboration can be achieved by providing a representation of the design problem and solution that will allow designers to interact with one. This representation has to be made accessible to designers located apart geographically, using currently available multimedia and communication technologies. In the following section, the representation issue will be addressed. A technique for automating the concept generation process is proposed. This will allow designers automatically to generate many design concepts based on different requirements. Every time a designer enters his requirements, an instance of a design solution will be generated. Other designers can enter other requirements or modify current input values. This will alter that instance accordingly. The solution is finalised when the team is satisfied with it. 3 Automated concept generation The quality of automatic design synthesis tools is dependent on the quality of the knowledge they contain [6]. This section describes a knowledge-based approach for generating design concepts for customisable products. Any engineering system comprises a number of components. These components are arranged spatially in a specific way to deliver a required overall function. An automobile, for example, primarily consists of an engine, body, drivetrain and wheels. Each of these components is employed to deliver some function. Collectively, when arranged together, they deliver the overall function provide transport. The spatial layout is fixed in almost all customisable products (for example, automobiles, electric appliances and personal computers). For a product to be customisable, it should have a stable configuration. It has been very successful at delivering a specific function so revolutionary changes in its design are rarely necessary. Instead, such products are subject only to slight modification in one or more of their components to satisfy new customer requirements. An example of a customisable product that has undergone a relatively major change is the Motorola MicroTAC cellular phone, which replaced the DynaTAC model in the mid 1980s. The new design only involved the introduction of the folding flap, but was still considered an innovative revolutionary design [1]. It is also widely recognised that much of design work consists of applying previous known solutions to current problems. Designers look back at how previous problems were solved and, if possible, use those solutions or some of their innovative characteristics to tackle new problems. This attempt to develop a computer support tool for conceptual design includes three primary elements. The first is the provision of existing successful engineering components (sub-solutions) which have previously been used to provide a specific function. The second element is the ability to adapt these subsolutions, to enable them to be used again as part of new solutions. The third lllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllll

7 element is the ability to configure the (adapted) sub-solutions into a final solution to fulfil the required overall function. These three elements are dealt with in the following sub-sections. 3.1 Representing sub-solutions A design department will normally hold details of many components that were originally created to deliver a specific function within a particular system. When confronted with a new design problem, the novice designer may not be aware of their existence while the experienced designer may perhaps have forgotten about them, even though these old solutions might have been a result of creative thinking and might exhibit a degree of novelty. They might also hold much design knowledge within their definition. These old solutions might be suitable as sub-solutions to new design problems, but need to be represented thoroughly and indexed effectively to allow them to be searched for and selected during the conception of new products. The sub-solutions may be in-house designs, or external designs developed in other domains, or designs that have been registered as patents. The representation scheme, however, should allow these sub-solutions to be modelled reliably and independently of their source. All the physical, geometrical and functional attributes of the sub-solutions should be represented independently of any bias towards a specific product. They should be labelled according to the function(s) they can provide (output function) as well as any function that they require in order to deliver that output function (input function). The output functions may be intended or unintended, as when a lamp gives light as an intended function but also produces heat as an unintended function. Both types of function should be represented separately. During the generation of a design concept, the search for sub-solutions is governed by their output functions. In practice, more than one sub-solution may deliver the same function. When represented, these sub-solutions should therefore also be assigned preference values according to how well they satisfy the function. 3.2 Ensuring adaptability of sub-solutions Any sub-solution that is to be used as part of a new design needs to be represented in a flexible way which will allow it to change according to the requirements of the new environment that it will be in. A sub-solution, therefore, is represented using two types of attribute, enduring and non-enduring. An attribute of a sub-solution is an enduring one if it remains unchanged during the design life of the sub-solution. If it is modified, the very concept of the design changes and therefore the sub-solution becomes a different entity. A non-enduring attribute, on the other hand, is an attribute of a sub-solution that, if changed, does not alter the design concept, but only creates another instance of it. To illustrate these attribute types, consider a gear. The fact that it is a circular element with teeth at its outer circumference and a hole or a shaft through its

8 centre represents the very concept of the gear and therefore these attributes are its enduring attributes. A gear will always be of circular shape and have a central hole or shaft and a toothed circumference. On the other hand, the number of teeth, the radius of the hole, the material of the gear, the thickness of the material and all other dimensions are non-enduring attributes which do not affect the concept of the gear even if they change. The enduring and non-enduring attributes of a component are all important. However, non-enduring attributes matter particularly in conceptual design, where the designer is still uncertain about many parameters. Non-enduring attributes are represented as inputs to the sub-solutions. Their values need to be specified before the sub-solution can be utilised. Sometimes these values will be provided by the end-user, while in other cases they will be passed on by other subsolutions with which they will be interacting. 3.3 Configuring the sub-solutions Geometric configuration refers to the total geometric structure of the product, consisting of the approximate or precise geometry of its components and their overall spatial arrangement [7]. There are two primary requirements for configuration problem solving. These are: A set of predefined components, and A description of the desired configuration. The first requirement has been dealt with in the previous sub-section. The representation of the components using enduring and non-enduring attributes allows them to be adaptable enough so as to cope easily in new configuration environments. The second requirement for solving configuration problems involves two activities. The first is choosing the right components. Function reasoning is employed to select components based on the functions they provide. The second activity is known as associating, in other words, establishing relationships between the selected components. This is achieved by defining the product skeleton a priori. A skeleton is an abstract model of the layout of the product. The system starts, first, by selecting what is known as the central component of the product the jaws are the central components of a robot gripper around which other components are positioned. The layout of customisable products is stable. That is, modification to the product may occur in one or more of its components, but this rarely affects the overall layout in any significant way. 4 Implementation The technique described above has been implemented using ICAD [3], a knowledge-based environment that allows the development of intelligent conceptual design systems. The development language is the ICAD Design Language (IDL), which is based on Common Lisp.

9 4.1 Representing the components (sub-solutions) As shown in Figure 4, a mechanical gripper can be decomposed into three primary components, namely, the jaws, the mechanism and the actuator. The jaws are the main component (central component) that will provide the most important function which is to grasp the object that needs handling. The actuator supplies mechanical energy and the mechanism transmits it to the object via the jaws. Actuator Mechanism Jaws Figure 4: Components of a robot gripper with a front view for the mechanism. Every component is defined as an independent part. All the required design information about the component is encapsulated within the definition of that part (defpart). This includes the geometry, the material types and the function that the component delivers. Figure 5 shows the defpart for a component called disc-jaw. An instance of the part disc-jaw is depicted in Figure 6. The disc-jaw defpart consists of three sections, headed by defpart keywords, shown in bold in Figure 5. The first section includes all the nonenduring attributes (:modifiable-optional-inputs). These attributes take their values from the end-user, or from other components in the final solution, and can

10 be re-specified at any stage of the design process. The enduring attributes are all included in the second section (:attributes). All the attributes that belong to the first two sections are parameters of the component. The geometric structure of the component is represented in the third section of the defpart (:parts). (defpart disc-jaw (subtracted-solid) :modifiable-optional-inputs (:disc-jaw-width 4 :disc-jaw-height 3 :disc-jaw-thickness 0.25) :attributes (:height (the :unfinished-disc-jaw :height) :width (the :unfinished-disc-jaw :width) :length (the :unfinished-disc-jaw :length) :display-controls (merge-display-controls (:hidden-line-removal t)) :solid (the :unfinished-disc-jaw) :subtract-solids (list (the :hinge-opening) (the :screw-hole))) :parts ((unfinished-disc-jaw :type unfinished-disc-jaw :jaw-thickness (the :disc-jaw-thickness) :jaw-width (the :disc-jaw-width) :jaw-height (the :disc-jaw-height)) (hinge-opening :type box-solid :height (* 0.06 (the :unfinished-disc-jaw :height)) :length (the :unfinished-disc-jaw :jaw-thickness) :width (the-child :height) :position (:top (the-child :height) :right 0)) (screw-hole :type cylinder-solid :radius 0.8 :length (the :length) :position (:top (- (* 0.09 (the :height)) (the-child :radius)) :right (- (* 0.2 (the :width)) (the-child :radius)))))) Figure 5: The definition of the part disc-jaw. This representation of the disc-jaw component is in accordance with the general structure of a defpart. Other components might be represented slightly differently.

11 Figure 6: An instance of the part disc-jaw. 4.2 Adaptive components The adaptability of sub-solutions is essential for the successful generation of new design concepts. A component might satisfy a particular sub-function but its design may not allow it to be adapted for combining with other components. Versatility in a component is achieved where it possesses many non-enduring attributes. The more non-enduring attributes a component has, the more flexible and adaptable it will be, and hence the more opportunity there will be for subsequent refinement during conceptual design Dimension refinement Consider the vacuum-cup component shown in Figure 7. When used as part of a new solution, some of its dimensions may need to be modified. The height of the cup, the top hole and bottom hole radii and the thickness of the material are all non-enduring attributes and hence can take different values depending on the new environment in which they will work. In Figure 8, the vacuum cup is shown with some of its dimensions re-specified while the concept remains unaltered. If the vacuum cup is to form part of the vacuum gripper solution shown in Figure 9, the value of its top hole radius will determine the value of the bottom hole radius of the mating mechanism, or vice versa. The two radii will nominally be the same, providing the necessary sealing action.

12 Top hole φ Thickness Height Bottom hole φ Figure 7: An instance of the vacuum-cup part with non-enduring attributes. (b) (a) (c) Figure 8: Three instances (a), (b) and (c) of the part vacuum cup with modified values for height, top radius and bottom radius.

13 Figure 9: Vacuum cups as part of a vacuum gripper system Ease of manipulation Manipulating the position, orientation and quantity of components is an essential aspect in conceptual design. An automobile engine needs to be positioned in a specific place with a specific orientation. The wheels need to be of a certain number. At the conceptual design stage, when developing a component to solve a specific sub-problem, the designer should not be overly concerned about whether the component will fit, in what orientation, and how many of them are needed. The representation of the component should allow for easy manipulation of these aspects later on. Consider, for example, the disc jaw shown in Figure 6 which can be used as part of a gripper for grasping discs. The gripper depicted in Figure 10 is a solution for a particular design problem. If the weight and size of the disc increase, the current gripper may no longer be suitable. The layout of the jaws and their number may need to change and, as a result, the mechanism and actuator sizes will have to adapt to these changes. The designer, at the conceptual design stage, might not be aware of the weights of the different objects that the gripper will be handling, and ideally should then not pay too much attention to such an issue. A good representation for adaptability should provide scope for this type of situation. Hence, in this example, the number of jaws and spacing between them are left as variables, i.e. inputs for the disc jaws. Figure 11 shows how, by modifying the values of these input parameters, a new gripper is automatically generated to suit a new situation.

14 Figure 10: A disc gripper to handle discs of diameter up to 50 mm Shape refinement Modifying the shape of a component during concept generation is not as easy as refining its dimensions. Dimension refinement can be achieved by the emerging technology of parametric design. Shape refinement, however, will give more opportunities for creativity during concept generation [8]. Both simple and complex shape refinement might take place. Refining a box, for example, to take a different rectangular shape is a simple refinement. A complex shape refinement is illustrated in the following example. The jaw, shown in Figure 12, is a basic rectangular jaw, but within its definition it contains inputs which allow it to have certain novel features. The end-user, during the conceptual design process, can quickly explore such features. The input (flat-jaw-needed?), for instance, asks the end-user if he/she requires the picking surface of the jaw to be flat. The default answer is yes. If the answer is no then the surface will no longer be flat and the end-user, by changing various input values, can have different picking surfaces, as depicted in Figure 13.

15 Figure 11: Disc gripper to grasp discs of diameter up to 80 mm after automatic modification by the conceptual design system. Picking surface Figure 12: A parallel jaw with a flat picking surface.

16 (a) (b) (c) (d) Figure 13: A parallel jaw with a variety of possible picking surfaces. 4.3 Configuring the components Design-a-gripper-1 is the generic gripper definition which models the skeleton for the different configurations that will be generated by the system. In other words, it controls the generation of design concepts. It comprises the three gripper components (:parts) mentioned previously, namely, the jaws, mechanism and actuator. Gripper concepts are configured according to the types of components the system selects for the specific inputs provided by the end-user in each concept generation session. The mechanism for configuring new design concepts is described here. The most important component of a robot gripper is the jaws that are to grasp the object. This component is selected first according to the type of object to be handled. The function being embodied here is grasp object. The end-user does not need to know anything about this required function, which has already been pre-programmed. He/she only has to give information about the geometric

17 structure of the object and its material, and the system will search for jaws that have an output function called grasps (X), where X is the geometric structure of the object. This structure could be cylindrical, conical, rectangular, flat or any other geometric shape. The jaws have an input function, which needs to be realised so that it can deliver its output function. The system automatically selects components that will provide this function. This process continues until the system finds all necessary components. These components are then attached to the central component. Although, as will be seen later, in this work the positioning of the components was precisely achieved, in some design problems this may be more difficult. However, the components could be positioned using qualitative direction keywords (such as above, below and on the right side of). The full solution is then configured, and any further refinement is possible after the product has been instantiated. The block diagram shown in Figure 14 illustrates this process. Concept generator Selection sub-system Selection order 1. select jaws 2. select mechanism 3. select actuator Configuration sub-system Input: Material & geometric structure Information: Jaws & mechanism input functions Components data base Components Selected components Arrange the selected components Output: Gripper design Figure 14: Concept generation process. In the following section, an example is given to describe the automatic generation of a robot gripper concept according to the requirements of the end user. 5 An illustrative example The end-user is first required to enter the material and geometric structure of the object to be grasped, as shown in Figure 15.

18 Figure 15: Selection of the material and geometric structure of the object to be grasped. The system chooses a suitable type of jaws based on these inputs. It then asks additional questions according to the type of jaws selected. Figure 16 shows some of the information requested for those particular jaws, which have been picked specifically for the given object (Figure 17 depicts the selected jaw-type). This indicates the versatility of the jaws during the concept generation process. For this example, the default values are used. At any level of the conceptual design process, the designer may go back and refine any of the inputs. Finally, when all the required information has been provided, an instance of a gripper concept is generated as shown in Figure Discussion The conceptual design process has been described as a complex, and difficult to automate, product development phase. A great deal of expertise and creativity is required during this phase, particularly during the concept generation stage. Some research has been undertaken to support that phase of design. However, most of this work has focused on the functional aspects of the process and therefore the developed tools apply to the later stages of task clarification and earlier stages of conceptual design. In the research reported in this chapter, an approach for automating the central activity in conceptual design has been presented. The approach involves employing a representation scheme that facilitates collaboration during design.

19 Please enter the right-jaw width 1.2 Please enter the right-jaw length 1.5 Please enter the right-jaw height 2 Please enter the right-jaw front radius 0.8 Please enter the right-jaw rear radius 0.4 Please enter the distance required between the parallel jaws 2.2 Figure 16: Information requested for the selected jaws.

20 Length Width Height Rear radius Front radius Figure 17: The selected jaw and some parameters that the system requires. Implementing the proposed representation scheme, the developed KBS automates the concept generation process for robot grippers. Interaction with the system is very simple and the end-user of the system does not require much expertise in gripper design. The research contributes directly to the area of configuration design, in particular, with respect to the representation of knowledge concerning predefined components. Two aspects overlooked by previous researchers have been addressed. The first is the comprehensive representation of predefined components, including information about their geometry, material and functionality. The second aspect is the adaptability of these predefined components, which will allow them a degree of flexibility during the association stage of the configuration process. Hayman [9] classified design problems, according to the opportunities they present for innovative and creative solutions, into four categories: selection design, configuration design, parametric design and original design. Rosenman and Gero [10] state the more efficient a system is in controlling the selection of structural elements and the more adept it is at manipulating such elements the greater the opportunity there will exist for creative design. The tool described here has achieved the automation of the majority of the creative levels mentioned by Hayman [9].

21 Figure 18: The final gripper solution. The KBS will not always produce a complete solution since solutions are configured using only the components available in the knowledge base. Solutions are not selected off the shelf, which is the traditional way for producing gripper designs in research on robot grippers. However, the system will in most cases give at least part of a solution, therefore facilitating the job of the design team. A new part may be developed and then integrated into the component database. The addition of new sub-solutions is simple and can be done by the designer and developer of the system. Altogether, the research reported in this chapter shows that it is possible to automate the concept generation process at least in the case of a robot gripper, which is a moderately complex design task. 7 Summary This chapter has presented a novel approach for automating the concept generation activity which is carried out during the early phase of design known as conceptual design. The approach is suitable for supporting collaborative

22 design work. The chapter started by indicating the motivations for, and objectives of, the work presented. The proposed approach was then discussed. Three requirements were addressed, namely: the availability of sub-solutions, the adaptability of the sub-solutions and the ability to configure sub-solutions into final solutions. The implementation of the proposed approach using the ICAD system was then presented, and finally an example of the automatic generation of a gripper design concept was given. References [1] Ulrich, K.T. & Eppinger, S.D., Product Design and Development, McGraw-Hill: New York, [2] Lee, K., Principles of CAD/CAM/CAE Systems, Addison Wesley: London, [3] Saad, M. & Maher, M., Shared understanding in computer-supported collaborative design, Computer-Aided Design, 28(3), pp , [4] KTI: Knowledge Technologies International, The ICAD System (Release 7.2.2) User s Manual, Lexington, MA, [5] Pahl, G. & Beitz, W., Engineering Design: A Systematic Approach, Springer-Verlag: London, UK, [6] Potter, S., Darlington, M.J., Culley, S.J. & Chawdhry, P.K., Design synthesis knowledge and inductive machine learning. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 15, pp , [7] Guan, X. & MacCallum, K.J., Handling of positional information in a system for supporting early geometric design. Proceedings of the 1995 Lancaster International Workshop on Engineering Design, Springer- Verlag: London, pp , [8] Mitchell, W.J., A computational view of design creativity, Modelling Creativity and Knowledge-Based Creative Design, eds. J.S. Gero & M.L. Maher, Lawrence Erlbaum: Hillsdale, pp , [9] Hayman, B., Fundamentals of Engineering Design, Prentice-Hall: Upper Saddle River, [10] Rosenman, M. & Gero, J., Creativity in design using a design prototype approach, Modelling Creativity and Knowledge-Based Creative Design, eds. J.S. Gero & M.L. Maher, Lawrence Erlbaum: Hillsdale, pp , 1993.