//Logo Image
Author: Yeh-Liang Hsu, Chi-Yu Tai, Ting-Chin Chen (1999-09-03); last updated: Yeh-Liang Hsu (2001-02-16); approved: Yeh-Liang Hsu (2000-04-19).
Note: This paper is published in the Journal of Chinese Society of Mechanical Engineers, Vol. 21, No. 4, pp.369-377, 2000.

A design process model based on product states

Abstract

Design is a dynamic problem solving process. All designs start with a problem to be solved, and sub-problems are generated dynamically while solving the central problem. To reflect this dynamic problem-solving process, this paper presents a new design process model, which describes the ‘states’ of the product evolving during the design process, instead of describing the design activities. The design process is viewed as dynamically moving around a ‘product state space.’ A ‘product state’ is a representation of the product that the designers use to generate information to answer or evaluate the current design sub-problem, and it is usually represented by a design prototype. Mappings from design sub-problems to proper product states are important design decisions. This paper starts by describing design as a sequence of prototyping process. An industrial safety helmet design example is used throughout the paper to demonstrate the ideas. Then the product state space is defined. Finally this paper describes how the design sub-problems are mapped to the product state space. Using this model, designers are encouraged to focus more on “what sub-problem to solve,” “what tool to use,” instead of “what activity should I do next?”

Keywords: design process model, design prototype, product state.

描述產品態之設計程序模型-中文摘要

設計是一個動態的解決問題程序,所有的設計都起始於一個待解決的問題,而在解決這個問題時,子問題又不斷動態產生。為了反映這個動態的解決問題程序,本文提出一個新的設計程序模型,描述產品在設計過程中‘狀態’,而非描述設計的活動。一個‘設計狀態’是指設計者用來產生資訊以回答或評估目前子問題之產品表現方式,通常也就是一個設計原型,從設計子問題映射到適當的產品狀態是重要的設計決策。本文首先以一連串原型製作過程來描述設計程序,並以一工地用安全帽設計實例來闡述各種概念,接著定義了產品狀態空間,最後並描述如何由設計子問題映射到產品狀態空間。應用這個模型,設計者被鼓勵把思考焦點放在“應解決什麼設計子問題?”,“應使用什麼工具?”,而非“接下來應該做什麼設計活動?”

關鍵詞:設計程序模型,設計原型,產品狀態

Introduction

Design process plays an important role to the success of development of a product. Design process influences performance, quality, cost, and developing time of the product. Most researches in design process try to establish a structured model to describe or facilitate design. Design process models are usually categorized into descriptive models, prescriptive models, and cognitive models (Finger and Dixon, 1989; Cross, 1994). According to Cross, descriptive models simply describe the sequences of activities that typically occur in designing. These models usually emphasize the importance of generating a solution concept early in the design process. Prescriptive models attempt to prescribe a better or more appropriate pattern of activities. They usually offer a more algorithmic, systematic procedure to follow, and they emphasized the need for more analytical work to precede the generation of solution concepts.

Most design process models, descriptive or prescriptive models, divide the design process into several sequential phases based on the essential activities that the designers perform or should perform. These activities are often presented in a flow chart manner, with feedback loops showing the iterative returns to the earlier phases. Ulrich and Eppinger (1995) pointed out that structured design process models are valuable because they make the decision process explicit, they act as ‘checklists’ of the key steps in a development activity, and they are largely self-documenting. Moreover, one goal of design research is to build a computable design process model for realizing intelligent computer-aided design systems (Takeda et al., 1990). A well-structured design process model will be of great help for constructing such systems.

However, it is arguable that the design activities described in the design process models are general enough for designers of different experiences, and for various types of design. Designers at different experience levels may start a design from different phases. In the case study interviews, Baker et al. (1991) demonstrated that experienced design engineers tend to directly generate solutions to the problems given, rather than using structured design methodologies. Based on new knowledge involved in the product, Culverhouse (1993) classified design into four different types: repeat order, variant design, innovative design, and strategic design. Four different ‘design routes’ were proposed for the four different types of design.

Design process has one important characteristic that may not be fully reflected in the structured design models: design is a dynamic problem solving process. All designs start with a problem to be solved, and sub-problems are generated dynamically while solving the original problem. Gero (1990) pointed out that designers begin designing with incomplete information and before all the relevant information is available. Actually design is an exploration process, as Gero further stated, “what is relevant only manifests itself as the design proceeds and varies with the decisions taken.” Most design process models model the design process from the designers’ point of view, as a structured pattern of a series of designers’ activities. Given the dynamic and unpredicted nature of the possible design activities, it may be worthwhile to re-examine the design process from how the design evolves during the design process.

This paper presents a design process model in an attempt to reflect this dynamic problem-solving process. Instead of describing a structured pattern of designers’ activities and tasks in the design process, this model describes the current ‘state’ of the product. A ‘product state’ is a representation of the product that the designers use to generate information to answer or evaluate the current design sub-problem. A product state is usually represented by a design prototype. The design process is viewed as dynamically moving around the ‘product state space.’

This paper starts by describing design as a sequence of prototyping process. An industrial safety helmet design example is used throughout the paper to demonstrate the ideas. Then the product state space is defined. Finally this paper describes how the design sub-problems are mapped to the product state space. Using this design process model, designers are encouraged to focus more on “what sub-problem to solve,” “ what tool to use,” instead of “what activity should I do next?”

Design as a Sequence of Prototyping Process

In the mechanical design process, building prototypes is a very effective way to communicate and evaluate design ideas, learning and solving design problems at hand, and discovering new design problems. A general comprehension of the word ‘prototype’ is a physical approximation of the product that is usually in full scale and fully functional, such as an a-prototype or a b-prototype. In design literatures, the word ‘prototype’ has a much broader meaning. Wall et al. (1992) stated that “a prototype is the first thing of its kind.” Their definition of a prototype includes “both electronic and physical representations of the part or product.” This definition has expanded the concept of a prototype from a physical artifact into an electronic representation of the design, such as a CAD model or a computer simulation.

Ulrich and Eppinger (1995) defined a prototype as “an approximation of the product along one or more dimensions of interest;” “any entity that exhibits some aspect of the product that is of interest to the development team can be viewed as a prototype.” This definition of a prototype has ranged from graphic ideas, system equations, computer simulation, to fully functional prototype product. They further classified prototypes along two dimensions: physical as opposed to analytical, comprehensive as opposed to focused. The design of a trackball on notebook computers is used as an example to illustrate the two dimensions. As shown in Figure 1, prototypes generated in the design process of the trackball are put on the two-dimensional plane. Ulrich and Eppinger stated that for physical products, fully comprehensive prototypes must generally be physical.  Therefore the fourth quadrant of the plot is labeled ‘not generally feasible.’ We think that using certain technology such as virtual reality, we can build a prototype that is comprehensive, while the prototype is still an analytical one. Therefore the fourth quadrant of this two-dimensional plane should be feasible.

Figure 1. Prototypes generated in the design process of the trackball (reproduced from Ulrich and Eppinger, 1995).

Gero(1990) defined a design prototype as “a conceptual schema for representing a class of generalized heterogeneous grouping of elements derived from alike design cases that provides the basis for the start and continuation of a design.” Gero pointed out that routine designs can be viewed as design prototype-instance refinement. Design prototypes are retrieved and selected, and instances are produced and refined. Under the broad definition of a design prototype, we also believe that the whole design process can be viewed as a procedure of building a series of prototypes, trying to solve a series of design sub-problems generated dynamically throughout the design process. A design may start with customers’ needs, an abstract description of what the product is expected to do. Then the design process is an evolutionary process that transfers from one prototype to another, gradually obtaining more detailed and more concrete descriptions of the product. The design process continues until a design solution, i.e., a real product, is obtained.

A project on redesigning the shell of an industrial safety helmet to improve its thermal properties is described here to illustrate this concept. At the beginning of this design project, the first question was, “what are the factors that affect the thermal properties of an industrial safety helmet?” For convenience, this question is denoted ‘Q1’ in the paper.

To answer this question, a two-dimensional heat transfer model simulating a worker wearing a hemisphere shell in a 30.0 sunny day, with wind blowing at 2.5 m/sec was built. Again for convenience, this computer simulation prototype is denoted ‘P1.’ From this prototype, we learned that radiant heat is the major heat source when wearing a helmet. Convection is the primary way to dissipate heat from beneath the helmet shell. Body heat only has a marginal effect on the temperature beneath the helmet shell.

While trying to identify these factors, more questions were raised: “how do we redesign the helmet shell to improve its thermal properties? (Q2)”  “how do other helmets do on these factors? (Q3)”

An experiment was established to simulate the conditions of a head wearing a helmet (P2). The average temperatures beneath the helmet shell, the speed of heat dissipation through convection, and the temperature contour beneath the helmet shell, were used to describe the thermal properties of a helmet. Helmets of different types and makes were tested, various design concepts were examined, and some design suggestions for improving thermal properties of industrial safety helmets were made. According to these design suggestions, a new helmet design concept that should provide better ventilation and better insulation against solar radiation was generated, in the form of a concept drawing as shown in Figure 2, another prototype (P3).

Figure 2. The concept drawing of a new helmet shell design. The arrows indicate airflow.

Very soon many questions were raised when this concept drawing is presented: “Can we make it look better without hurting its thermal properties? (Q4)” “Will there be stress concentration around the ventilation holes? (Q5)” “Is it manufacturable? (Q6)” “Will it pass the standard impact and penetration test? (Q7)” “How do you prove that the new design have better thermal properties? (Q8)” “Will water leak into the helmet through the ventilation holes if it rains? (Q9)”

Answering these questions formed more sub-problems for the design. Prototypes were built to answer or evaluate these design sub-problems. Prototypes built with paper (P4) and clay (P5) were used to give a three dimensional feel of the appearance of the helmet. The shape of the helmet shell was further rounded. Different colors and coating were tried on the prototype to give it a better look.

To answer the sub-problem about the mechanical properties of the helmet, a finite element model of the helmet shell was built (P6). Stress analysis showed that with a static load on top of the helmet shell, there was stress concentration around the ventilation holes. Reinforcement wall was added around the ventilation holes to cope with this problem. On the manufacturablity of the new helmet shell design, another computer simulation prototype using plastic mold flow analysis software was built (P7) to evaluate how a plastic injection molding can be built at a moderate cost.

To answer the rest of the design sub-problems, physical prototypes had to be built. Several helmets were made by vacuum molding (P8). These helmets were used in the impact and penetration test, the thermal experiment, and the watering test. Cracks were found on the helmet shell during the impact test, and a new sub-problem occurred: “how to redirect the impact energy? (Q10)” Further observation of the path of the cracks showed that the impact energy flow was blocked by the ventilation holes. Ridges were added on the helmet shell accordingly, and new helmets were made by vacuum molding again (P9). This time the new design passed the impact test.

Another question was raised during this modification process, “how to modify the shape of the shell to further improve convection? (Q11)” A special helmet prototype made of semi transparent plastomer (P10), with small stripes of paper attached throughout the inside of the shell, was used to provide detail information on how well the air flows under the shell. Finally a final engineering drawing (P11) with all the detailed dimensions was presented to the manufacturer.

In spite of all the test data, the manufacturer still wished to know “do the workers really feel better wearing this helmet? (Q12)” To answer this sub-problem, a-prototypes (P12) were built and a human description test, in which testers compared the new helmet with other helmets, was performed. After this sub-problem was confirmed, the manufacturer built the b-prototypes (P13) to answer the next sub-problem “how this helmet can be built using the current facilities of the manufacturer? (Q13)” And finally, the product (P14) was produced.

In this design project, a total of 13 design sub-problems were raised, and a total of 14 prototypes (including the final product) were built. The design process of the helmet can be viewed as a mapping of a series of design sub-problems to a sequence of prototyping process, as shown in Figure 3. Actually a lot more questions were asked during the design process, such as, the mechanical properties of plastic materials, how to measure the impact loads, where to buy certain components, etc. Many questions can be answered directly by designers’ knowledge, their previous design experience, or by information searching, and are not listed as design sub-problems.

Figure 3. Mapping of a series of design sub-problems to a sequence of prototyping process.

From this example we can observe that design sub-problems occurs dynamically during the design process. Some of the sub-problems are expected in the structured design models, but many sub-problems are not. They occur only under the special situation the design team encounters, because a specific design concept is presented, or certain information is obtained.

Note that the ‘prototypes’ here are not solutions to the design sub-problems. They are the representations of certain elements of the product, the means or design tools designers use trying to solve the design sub-problems. The ‘generation-evaluation-modification’ feedback loops exist within each design sub-problem and between design sub-problems. To effectively and efficiently solve a design sub-problem, designers must consider the characteristics of the current sub-problem to decide what representations or design tools to use in order to solve this problem. Designers’ capability of using the design tools has to be considered also. Therefore, mappings from design sub-problems to proper product states are important design decisions. As mentioned earlier, most design process models try to provide a structured pattern of designers’ activities. Instead of describing the design activities, we can turn to the other side to describe the states of the product evolving through the design process, and discuss how these decisions are made.

The Product State Space

As illustrated in the previous section, the design process can be viewed as a sequence of prototyping process. A design prototype represents a ‘state’ of the product. Formally, a ‘product state’ is a representation of the product that designers use to generate information to answer or evaluate the current design sub-problem.

Figure 4 shows a three-dimensional product state space. This figure is adapted from the two-dimensional space in Figure 1, which is used to describe the characteristics of prototypes. The first axis is still ‘comprehensive versus focused.’ The second axis is changed from ‘physical versus analytical’ to ‘physical versus virtual,’ because the word ‘analytical’ has some flavor of ‘quantitative.’

Figure 4. The three-dimensional product state space.

The third dimension added is ‘qualitative versus quantitative.’ This is an auxiliary axis since a product state can be simultaneously quantitative and qualitative. A representation of the product that can provide quantitative information may also be used for qualitative evaluation. But a design sub-problem can be strictly a quantitative problem or qualitative one. With this auxiliary axis, it would help the designer to consider the characteristic of a design sub-problem, and decide what design tools or representation to use in order to answer this design sub-problem.

A point in this three-dimensional space is a product state. To put a product state into the three-dimensional product state space, designers should consider (a) Is this a physical or virtual representation of the product? (b) Does it implement most attributes or just a few attributes of the product? (c) Does it generate quantitative or qualitative information (or both) about the product?

There are several extreme points in this product state space. The initial state may be a list of needs of the customers, or a mission statement from upper management. It is usually an abstract description of what the product is expected to do. The customers’ needs or the mission statement are usually the most virtual, and the most comprehensive representation of the product (while it may contain both qualitative and quantitative descriptions of the product). So it represents the extreme of the product state space in the fourth quadrant of the horizontal plane in Figure 4. On the other hand, the final product is the most physical, and the most comprehensive representation of the product. So it represents an extreme of the product state space in the first quadrant. Note that the final product appears in both the upper and lower half of the product state space because the final product can generate the most detailed quantitative information, while it can also provide all the qualitative feel. So the final product also represents the upper and lower extremes on the qualitative versus quantitative axis.

A product state is usually represented by a design prototype. All prototypes used in the helmet design project described in the previous section are product states of this design, and are plotted in the product state space as shown in Figure 5. Note that product state ‘P0’ is the mission statement of the design project. The position of a product state on a given axis represents the level of the state on that axis. For example, P5 (a clay helmet prototype) is ‘more physical’ than P4 (a paper helmet prototype). Therefore the position of P5 is higher than that of P4 on the vertical axis. However, the distance between these two product states may not have exact quantitative meaning. Similarly, P1 (a 2-D heat transfer simulation model), P6 (a finite element stress analysis model), and P7 (a mold flow analysis model) are all computer simulation models for a single attribute of the product. Thus the three product states occupy the same position on the focused & virtual plane. The 2-D heat transfer model only gives a trend of the effect of various heat sources, while the finite element gives rather precise quantitative stress analysis results. Thus they have different positions on the quantitative vs. qualitative axis. Finally, P12 (a prototype), P13 (b prototype) and P14 (the final product) appear in both upper and lower half of the product state space because they can generate both quantitative and qualitative information.

Figure 5. Design prototypes of the helmet design project.

Analogous to Figure 3, the ‘design path’ is also plotted in Figure 6 to connect the product states. The design path provides a clear picture of how the design problem is solved. In the helmet design problem, the design path started from the extreme point P0 in the fourth quadrant of the horizontal plane (Figure 6(a)), then moved left and diverged into several focused product states in order to solve the more focused design sub-problems (Figure 6(b)). Towards the end of the project, the path converged again, moved up and right toward the first quadrant for more physical and more comprehensive product states (Figure 6(c)). Finally the path ended at the other extreme point P14, the final product. From this figure, the design process can be viewed as dynamically moving around the product state space.

(a) In the beginning, the design path moves left and diverges.

(b) The design path gradually converges.

(c) The design path ends at the final product.

Figure 6. Design path shown in top view of the design state space.

Note that many product states along this design path are not planned beforehand. For example, the output of each phase of a typical design process model can be plotted in the product state space, as shown in Figure 7. In the beginning of the helmet design project, only the mission statement (P0), the design concept (P3), the final engineering drawing (P11), and the a-prototype, b-prototype, and the final product (P12, P13, P14) were planned. There was a given schedule for these product states. The rest of the product states were generated dynamically to answer the design sub-problems occur during the design process. A CPM (Critical Path Method) type of method can also be introduced for project management.

Figure 7. The output of each phase of a typical design process model, shown in top view of the design state space.

Mapping from Sub-problems to Product States

The design process is now viewed as a mapping of a series of design sub-problems to the product state space. This mapping actually represents important design decisions during the design process. Designers must consider the characteristics of the current sub-problem to decide what design tools or what kind of representation to use in order to solve this problem. Note that the sub-problems and the product states may not be a one-to-one mapping.

In research literatures on prototyping or rapid prototyping techniques, many researchers report the factors to consider when selecting a prototyping tool (Wall et al., 1992; Barkan and Lansitif, 1993; Ashley, 1995; Lange and Bhavnani, 1995; Sorovetz, 1995). In summary, levels of approximation to the final product (appearance, mechanical properties, accuracy) and design flexibility (time, cost) are the factors frequently mentioned. Similar factors should be considered when mapping a design sub-problem to a product state.

A product state is used to generate information about the design sub-problem. To decide its level of approximation to the final product, the designer should consider what is the information required to answer the current design sub-problem. The following questions should be asked when selecting a product state:

l  What is the information required? Is it about certain attributes of the product, or is it about the whole product? How much detail do I need?

l  What is the type of information required? Is it qualitative information or quantitative information? How accurate do I need?

l  How is the product state used to generate the information? Is it through physical experiments or virtual simulation? How close in forms to the final product do I need?

Obviously, these three questions directly relate to the three axes of the product state space. After considering these questions, the designer should be able to locate regions in the product state space to which they can map the current design sub-problem, as shown by the gray regions in Figure 8. Note that for a design sub-problem, several regions may be identified. Further decision has to be made on selecting one or several design tools to build proper product states, considering the design flexibility required.

Figure 8. Possible regions in the product state space to map the current design sub-problem.

For certain design sub-problems, especially in the early stage of design, various design concepts have to be evaluated, some of the design concepts will be discarded, and it is possible that the design will have major changes. For this type of design sub-problems, we should consider product states that have great design flexibility, that is, the cost and time required for generating and modifying the product state is low.

Design flexibility of a product state depends on the nature of the design tools used to construct the product state. It also depends on the availability of the design tool and capability of the design team. For example, in general a product state represented by computer simulation model has better design flexibility than a physical one. However, if the simulation software is not available to the design team, or the design team is not capable of using the software, the cost and time required to construct such a product state will be much higher.

In Figure 9, all design tools available to the helmet design team are plotted in the product state space according to their characteristics. This figure can be called a ‘design tool space.’ Note that here ‘design tools’ can be hardware or software that can help designers in designing. ‘Design tools’ can also mean designers’ capability to perform certain theoretical analysis or experiments. The design flexibility is also ranked in a 1~5 scale for each design tool according to the capability of the design team. The design tool space is specific to a design team. It can be updated whenever a new design tool is available to the design team, or the capability of using certain design tools improves. The design team can then compare Figure 8 with their own design tool space (such as Figure 9), and choose the most proper design tools to construct the product state, considering the characteristics of the design tools and the design flexibility required.

Point

Design Tool

Flexibility

T1

Heat transfer analysis software

****

T2

Thermal Experiment

**

T3

Hand drawing

*****

T4

Paper model

****

T5

Clay model

***

T6

Finite element analysis software

****

T7

Mold flow analysis software

****

T8

Vacuum mold

**

T9

Final engineering drawing

*****

T10

Steal mold

*

Figure 9. The design tool space

The Procedure of Using the Model

Finally Figure 10 is a flow chart of using this design process model. To initialize the model, the design team is presented with a design problem. The design team makes an initial design plan for all necessary product states and their schedule of the check points of the design project. These product states are plotted in the product state space, considering their required characteristics. The design tool space for the design team is also ready. As shown in Figure 10, a design sub-problem occurs as the design team explores the initial design problem. The design team then considers the level of approximation required to answer the design sub-problem, and maps the design sub-problem to several possible regions in the product state space.

 

Figure 10. Flow chart of the new design process model

In the third step, the design team selects a proper design tool that has the required characteristics and design flexibility from the design tool space. Using this design tool, the design team constructs a product state and generates information using this product state, till the current design sub-problem is solved. Then the designers go to the next design sub-problem. This procedure iterates until all design sub-problems are solved.

In this design process model, designers focus on two things: to explore the design problem and raise new design sub-problems, and to construct product states in order to solve these design problems. There are also two major types of design decisions in this model: to decide what design tools to use to construct proper product states, and to decide on the design sub-problems using information generated by the product states. After the design project is finished, design experience can be recorded in the form of the sequence of design sub-problems and product states evolving through the design process. This sequence represents the thinking logic to the design team. The design team can also update their design tool space as their capability improves, or they should try to learn new design tools if they realize necessary.

Conclusions

Design is a dynamic problem solving process. The logic flow of the design process should be dominated by the design problem itself, not by a given pattern of design tasks. This paper presents a new design process model based on product states. This new model describes the ‘states’ of the product evolving during the design process, instead of describing the design tasks. The design process is viewed as dynamically moving around the ‘product state space.’ Using this model, designers are encouraged to focus more on “what sub-problem to solve,” “ what tool to use,” instead of “what activity should I do next?”

References

Ashley, S., “Rapid Prototyping is Coming of Age,” Mechanical Engineering, Vol. 117, No.7, pp. 62-68 (1995).

Baker, K.D., Ball, L.J., Culverhouse, P.F., Dennis, I., Evans, J., Jagodzinski, A.P., Pearce, P.D., Scothern, D.G.C., and Venner, G.M., “A psychologically based intelligent design aid,” in Intelligent CAD Systems III: Practical Experience and Evaluation, Hagen, P.J.W. and Veercamp, P.J. (eds), Eurographics Seminars, Springer-Verlag, Berlin, pp. 21-41 (1991).

Barkan, P., and Lansitif, M., “Prototyping: a Tool for Rapid Learning in Product Development,” Concurrent Engineering: Research and Applications, No.1, pp. 125-134, (1993).

Cross, N., Engineering Design Methods: Strategies for Product Design, 2nd edition, Wiley, New York (1994).

Culverhouse, P.F., “Four Design Routes in Electronics Engineering Product Development,” Journal of Design and Manufacturing, Vol. 3, pp. 147-158 (1993).

Finger, S., and Dixon, J. R., “A Review of Research in Mechanical Engineering Design,” Research in Engineering Design, Vol.1, No.1, pp.51-67 (1989).

Gero, J.S., ”Design prototype: A Knowledge Representation Schema for Design,” AI Magazine, pp. 27-36 (1990).

Lange, A.E., Bhavnani, M., “Rapid Prototyping Technologies,” Aerospace Engineering, Vol. 15, No. 2, pp. 24-26 (1995).

Sorovetz, T., “Justifying Rapid Prototyping,” Manufacturing Engineering, Vol. 115, No. 6, SME 3, pp. 0361-0853 (1995).

Takeda, H., Veerkamp, P., Tomiyama, T., and Yoshikawa, H., “Modeling Design Processes,” AI Magazine, Vol.11, No.4, pp. 37-48 (1990).

Ulrich, K.T., and Eppinger, S.D., Product Design and Development, McGraw-Hill, New York (1995).

Wall, M.B., Ulrich, K.T., and Flowers, W.C., “Evaluating Prototyping Technologies for Product Design,” Research in Engineering Design, Vol. 3, pp. 163-177 (1992).