Knowledge-based functional conceptual design: Model,representation,and implementation

Abstract

Conceptual design is thus crucial to define new creative products. During conceptual design process, knowledge derivate from different domains is required to achieve a creative design. However, most of the design models restrict their representation to a specific domain. Knowledge representation for multidisciplinary is still an incipient issue for engineering design researchers. Moreover, to increase the robustness of design resolutions is another problem obsessing designers. This article proposes an adapted Gero’s Function–Behavior–State model (Requirement–Function–Behavior–Principle solutions) to aid creative design activities during conceptual design stage by introducing an integrated cross-domain knowledge representation methodology, to increase robustness by reusing existing cases knowledge through retrieval algorithm, and to realize effectiveness by putting forward a design synthesis methodology. Finally, a prototype system named knowledge-based innovative conceptual design system is developed to implement our approaches, and effectiveness of the proposed model is illustrated with a detailed example related to the design of eye movement apparatus incorporated with oculomotor machine of vision prosthesis for acquired blindness, with an overall design procedure classified.

Keywords

Conceptual design modeling knowledge representation retrieval algorithm design synthesis

Introduction

Improving the quality of products and enhancing the innovation ability are goals of product design. In order to obtain a design resolution that satisfies the design demands, designers should turn to a more flexible and innovative design methodology, which will meet the demands of customers at an affordable level. Generally, product design activity is composed of conceptual design, embodiment design, and detail design (Pahl and Beitz, 1988). Compared with embodiment and detail designs, conceptual design plays a more critical role during the entire design process. Studies show that about 70% of the product life cycle cost is determined during conceptual design (Kota and Lee, 1993). This phase is thus crucial to define new, creative products. In addition, it is tough to smoothen a poor design decision of conceptual design in the subsequent stages.

Due to the importance of this stage, more and more researches have been conducted on conceptual design. Suh (1989, 1998, 2001) proposed axiomatic design (AD) model by dividing design process into four different domains, such as customer domain, function domain, physical domain, and the process domain, to systematize design effort. Kitamura et al. (2002) and Kitamura and Mizoguchi (2003) adopt functional ontology in conceptual design to integrate functional knowledge. Based on Bond Graph technique, Schemebuilder is proposed by Bracewell and Sharpe (1996) to represent a practical functional design platform to deal with a variety of dynamic systems, and the platform aims at support design process seamlessly.

As one of the earliest functional conceptual design models based on knowledge, Gero’s conceptual design model introduces three layers: function layer, behavior layer, and structure layer, and they are linked together by mapping and transformation method. The conceptual design process can mainly be divided into (Figure 1(a)) formulation, synthesis, analysis, evaluation, reformulation, and documentation processes, among which formulation transforms design functions into expected behavior to enable it, synthesis is involved to arriving at a solution structure to exhibit the desired behavior, analysis is presented to derivate the actual behavior, evaluation is hired to compare expected and actual behavior, reformulation process is employed to address the design change, and documentation is the process of design description.

Figure 1.

(a) Design process of Gero’s FBS model and (b) chronological order of conceptual design process of RFBPs model.

Since the FBS model was proposed, it has been the reference for constructing expert systems and it has been evolving; for example, Umeda et al. (1996) proposed the Function–Behavior–State (FBS) model, in which the state is proposed instead of structure; Gero and Kannengiesser (2004) proposed a situated FBS model introducing creative design environment aimed at supporting innovative design; Christophe et al. (2010) presented the RFBPs model to integrate methodologies with a model of knowledge for conceptual design consistent with model-driven engineering; and Deng et al. (2000) investigated contribution of environment effect to functions in the design process and proposed Function–Environment–Behavior–Structure (FEBS) model. The main characteristics of common FBS modeler are hire physical state transition and physical phenomena to represent the behavioral process, reasons the design functions by means of casual and function decomposition. Structure–Behavior–Function (SBF) model developed by Goel et al. (2009) takes behavior as an intermediate concept between the structure and function requirements and functions as the hypothetical link between structural behavioral states.

However, there still exist two aspects that the aforementioned works have not covered. First, these works only consider a specific domain knowledge and ignore across-domain knowledge reuse, which may lead to the loss of potential promising solutions. Second, the works ignore effective conceptual design tools to support innovative design process. To tackle these problems, we conducted the research on multidisciplinary knowledge representation and innovative conceptual design model based on functions. In this article, our research objectives focus on the following.

An adapted FBS model to aid innovative activities of conceptual design within a multidisciplinary knowledge environment.

Multidisciplinary knowledge representation to aid conceptual design process and construction of FBS knowledge cell as the smallest unit for knowledge reuse.

Usage of our approaches to develop an innovative design system to enable pushing design resolutions to designers automatically and usage of a design case to validate its effectiveness.

The rest of this article is organized as follows: our approach of adapted FBS modeling and multidisciplinary knowledge representation is described in section “Functional conceptual design modeling and representation.” Section “Techniques for model implementation” proposes key techniques for model implementation, describes knowledge cell retrieval algorithm and design synthesis methodology, and discusses the techniques that can be used to implement the proposed model properly. Section “Validation and implementation” focuses on a case study of the development of oculomotor machine of vision prosthesis for acquired blindness to validate and verify our approach to show that the designers can proceed with overall design processes to generate novel and promising design solutions. Section “Discussion and conclusion” discusses the proposed model and knowledge representation method and concludes with directions for future studies.

Functional conceptual design modeling and representation

RFBPs processes of conceptual modeling

Conceptual design plays a central role in realizing the design requirements and innovative activities. In the case of Gero’s FBS model, the phase of aiding multidisciplinary innovation design is lacking, and it cannot be easily adapted to more flexible reasoning models (Camelo et al., 2007), and therefore, it needs to be updated according to current demands of the design. Therefore, we propose the RFBPs (Requirement–Function–Behavior–Principle solutions) model shown in Figure 1(b), in which the design process consists of seven steps with nine characters involved. The representation of characters is given in Table 1. The proposed model suggests that the conceptual design process starts with a description of design requirements, which states the actual demand functions and design constraints of customers and facilitates interactions between customers and design activities.

Table 1.

Representation of RFBPs characters.

Characters and definition
R is a set of design requirements to express what attributes the customer expects the designed system to possess.
Fe represents a set of expected functions extracted from R, the necessary knowledge to show what the design system should do according to design requirements.
Be is a set of expected behavior of the design system. Specifically, a set of variables that shows how the system should work in a given specific environment. Be is set according to the expected function.
Fs represents a set of principle solution functions derived from principles enabling the expression of effective functions of the specific principle, for example, its actual function.
Bp is a set of behaviors that plays a role of bridging/connecting link between Fs and S in the inner FBS knowledge cell, characterized with the embracing principle or process representation in order to aid innovation of the design process and to better understand the design.
Bs is a set of variables enabling the representation of the effective behavior of the design system. Bs is set according to the pre-synthesis structure, for example, its actual behavior.
S represents the embodiment of principle function, including the solid geometry (dimension, topology of subcomponents), features (surface roughness, accuracy grade), and so on.
C states the constraints of the design requirements that restrict the design process of the required system. It constitutes a group of one or more parameters or design rules.

Compared with Gero’s FBS model, our model possesses three main differences: introduction of set of FBS (Function–Behavior–Structure) knowledge cell abstracted from across-domain cases, involving design constraints derivate from former design processes, and synthesis process to produce creative solutions. These differences inject much convenience to the conceptual design system and help the designer to understand and solve design problems. Introduction of knowledge cell will facilitate the representation and usage of multidisciplinary knowledge. FBS knowledge cell not only provides designers with functions and effect of structure but also explains the essence of design method in favor of carrying innovation into practice. Moreover, it will benefit the understanding of how to improve the performance of structure to simulate innovative ideas. The design process of RFBPs model can be depicted as follows.

Step 1. Clarifying, analyzing, and representing design requirements

This step involves transformation of design requirements into the representation of expected functions and design constraints. During this step, design constraints should be clarified from the design requirements. Constraints that are introduced to restrict the design process and optimization of design solutions will be extracted. Expected functions are represented in the syntax Verb + Noun + Complement.

Step 2. Formulation and representation of expected behavior

This step involves transformation of design functions into expected behavior, while reference to design requirements to realize performance criteria is set by design requirements.

Step 3. Decomposing expected functions, searching for the FBS knowledge cell library corresponding to specific design functions, and pre-synthesizing FBS knowledge cell to construct a general structure

Functions cannot be decomposed into the next level of functional hierarchy without first going over to FBS knowledge cell library incorporated with restriction of design constraints. That is, we have to travel back and forth between the functional decomposition and the FBS knowledge cell library. By the retrieval of FBS knowledge cell library, principle solutions in the form of FBS knowledge cell are acquired to accomplish the mapping from the expected functions to physical structure, with the restriction of design constraints (such as design environment and geometry constraint) to support the proceeding design system. If the FBS knowledge cell is not found, then decompose the expected functions into subfunctions and continue the search process. Otherwise, change the state of the expected functions and turn to the next step. After all the subfunctions have been retrieved, collections of FBS knowledge cell will be pre-synthesized.

Step 4. Analyzing the actual behavior derived from the structured FBS knowledge cell

This step involves acquiring the actual behavior from the structured pre-synthesis system based on building blocks in terms of FBS knowledge cell by analyzing the corresponding actual functions and structure.

Step 5. Evaluating the expected behavior with actual behavior

This step involves comparing the expected behavior with the actual behavior to evaluate whether the decided design solution coincides with design constraints or will be accepted by the design system. Once the demands of the design are met, record the knowledge corresponding to FBS knowledge cell into a stack sequentially for further design process. Traverse the expected functions if all states are changed, and then terminate search and turn to a synthesis process or else return to step 2.

Step 6. Synthesizing FBS knowledge cell through backtracking the path of function decomposition to achieve conceptual design solutions

When overall corresponding FBS knowledge cells of decomposing branches of expected functions have been determined, design solutions can be further obtained by elaborately synthesizing collections of feasible FBS knowledge cell retrieved through backtracking the path of functional decomposition.

Step 7. Retaining design results into a document for further research

Once satisfying design synthesis is achieved, reserve design description for detail design.

Step 8. Reformulating the design process to address design changes

This step involves the case of design changes and ongoing adjustment of design requirements for getting better design solutions. Reformulation processes range from design requirements (step 8″′) and expected functions (step 8″) to FBS knowledge cell retrieval (step 8′) and resynthesis (step 8). Reformulation of structure, actual functions, expected functions, and design requirements address changes in variables and their range of values embodied in structures, functions, design requirements, and actual behavior separately.

Representation of multidisciplinary knowledge

Based on the above design process, all basic characters can be correctly represented as tabular form. In order to avoid different flows, naming taxonomy is derived from multidisciplinary or different domain experts; hereafter, we adopt the formal definition of flows proposed by Sheng et al. (2010), first given by Stone and Wood (1999). All design flows deal with three flow classes such as material, signal, and energy to standardize the expression of flow name. To facilitate the representation, some mathematical symbols are introduced: “{f}” denotes a set off, while f is any string of symbols; “{input flow}” denotes a set of input flow; “+” denotes join together of words to form an abstract syntax; and “||” denotes alternative.

Requirements (R) and constraints representation (C)

Scheme of design requirements encompasses a series of design attributes, aims at capturing actual details of customers’ specification and converting it into design constraints and expected functions. Essential information of expected functions is presented, as pairs of a transitive verb and a noun, even a complement is necessary occasionally, in the form of Pahl and Beitz’s (1988) syntax of verb + noun + complement; the verb attributes in the syntax adopts the transitive verb vocabulary as in Sheng et al. (2010), as proposed by Hirtz et al. (2002), or which allows the customers to define the design requirements and avoid ambiguity as in Camelo et al. (2007). Description of action related to a specific function is expressed by the transitive verb, the noun stands for the target of the function, and a complement is involved to give additional information belonging to the function, which can be used to convey the information more explicitly.

Constraints restrict design solutions that can be assumed by a group of one or more parameters or design rules. Six types of constraints are considered here: functional constraints, structural constraints, manufacturing constraints, assembly constraints, evaluation constraints, and optimization constraints. Functional constraints refer to environmental elements and variables required to drive the input and functional output; for example, transmission of voice depends on the medium, that is, functional constraints: if no medium exists in the surrounding area, then voice cannot be transmitted. Structural constraints relate to geometric restrictions among structures, such as topology, materials, and dimensions. For a key–slot connection, structural constraint is described as follows: For a key–slot connection, the exterior side of the key connects to the interior side of the slot. Manufacture and assembly constraints refer to requirements and variables of product manufacturability and assemblability. Evaluation constraints are direct attributes to evaluate design resolutions (such as weight, volume, dependent level, and distance), while optimization constraints are utilized to optimize evaluated products (such as maximize and minimize). Moreover, if constraints are fuzzy, it can be described by the mean of fuzzy linguistic, for example, machining accuracy can be described as“high, medium, low.” Annotation describes the additional information regarding design requirements. The final definition of design requirements is given in Figure 2.

Figure 2.

Representation of design requirements.

Functional representation (Fe, Fs)

Function reveals the abstract of specific capability the system can achieve, and the system can be a mechanical system/program in machine code. To facilitate the design process, distinctions between expected function (Fe) and actual function (Fs) are necessary and important. The expected function is the intention function extracted from design requirements, enabling wider and more flexible design synthesis by choosing a wider range of design solutions. The actual function states the function derivate by analyzing the principles from multidisciplinary knowledge to support the design activity across various disciplines and to facilitate interactions between the expected function and the FBS knowledge cell. It has a more direct and explicit relation with design solutions. The definition of functions is shown in Figure 3 in tabular form, and an example of function representation is given in Figure 4.

Figure 3.

Representation of functions.

Figure 4.

An example of function representation.

Behavioral representation (Be, Bp, Bs)

Behavioral representation pertains to the feature of a structural component and realizes the function. It describes intermediate states of the structure or mechanism and provides design process with more precise information. As in FBS approach, behavior is introduced to discover the inconsistency between the design requirements and design resolutions, such as unexpected input/output and functions that cannot be performed. Moreover, based on the fact that different behaviors point to different design results, difference between expected behavior and actual behavior should be evaluated here. The tabular representation of behavior is presented in Figure 5.

Figure 5.

Behavioral representation.

Behavioral_description describes the objective of the behavior in terms of semantic or mathematical symbols and reveals the intermediate state of embodiment and variables in different stages of states. In this way, the evolve procedure of design concept is directly brought forth to the designer. The finite set of Behavioral_variables states the intermediate state variable appears in the behavior, while the set of Casual_link indicates the relationship among variables affected or affected by others. Unexpected_I/O represents the type of flow mentioned earlier (material, signal, and energy), to demonstrate unexpected harmful inputs or side effects that hinder the required operation. actual_Functions and BehaviorOfStructure point to actual functions and structures that can be connected by this behavior separately. An example of behavioral representation is shown in Figure 6.

Figure 6.

An example of behavioral representation.

Meanwhile, scientific principles play an important part in developing an understanding of the design functions and pointing to potential solutions. A principle is a generalized directive to bring about a certain type of design functions. If any, it can play the role of a guideline that, when embodied, singly creates design functions. Principles can work as heuristic rules, which contain high-level concepts, which show the essence of why structure behaves this way and not that, can benefit the understanding of how to improve the performance of structures, assist in stimulating innovative ideas, and which finally lead to inventive solutions for conceptual design.

In order to stimulate innovative ideology and help understanding the essence of structure, we suggest an attempt to give a definition of scientific principle in connection with the attribute of Behavioral_description in behavioral representation. The attribute of Principle_Name is given in the form of verb + noun. The verb represents the effect or mechanism of principle, and the noun states the property of principle, for example, fluid effect, Newton’s law, electromagnetism effect, and Pascal’s theorem. Contents of Principle_description consist of a semantic description of mechanisms to show the working principle and a graphical depiction to reveal the basic structure to perform the actual function. Driving_input and Principle_output reveal the input/output flow of principle to characterize the principle and its function. Similar to the functions’ expression, attributes of sub_Principle_of and sub_Principles are filled up by the high-level principle and low-level principles of this principle to facilitate the innovation design by giving more detailed information. One such sub-principle of the electromagnetic effect is the Biot–Savart law. principle_Functions is represented in terms of a set of finite functions that can be inferred from the principle. Constraints here point to variables of principle and environmental elements that probably affect the discharge of actual functions. Annotation shows the additional description, such as disciplinary information regarding the principle derivate or design cases. The representation of the principle is given in Figure 7, taking the principle of magnetic effect as an example. The main sub-principles of magnetic effect are magnetocaloric effect, hysteresis effect, eddy current effect, magnetoelectric effect, and magnetomechanical effect (Figure 8).

Figure 7.

Representation of the scientific principle.

Figure 8.

Subprinciples of magnetic effect and its representation.

Structural representation (S)

Composed of a series of subcomponents to realize specific functions, structure describes the basic geometry information. Based on the consideration to facilitate extension or adaptation of the structure, the structure is expressed in terms of parametric representation. A generic structural representation is given in Figure 9.

Figure 9.

Structural representation.

Structure is mainly represented in terms of Geometry_definition of component and subcomponents. Geometry_definition contains links with drawing and a 3D model to graphically illustrate the structure, where even the surface connected with other components will be marked. An example of electro motor is given in Figure 10.

Figure 10.

Structural representation of electro motor.

Integration of FBS knowledge cell

FBS knowledge cell works as a principle solution blocker for the building of design resolutions. Integration of FBS knowledge cell based on functional micro-knowledge cell (FMKC) (Sheng et al., 2010) is proposed in our former research. It can be taken as functional knowledge cell, which will guarantee reusability and robustness of design resolutions. Construction of FBS knowledge cell consists of three stages (Figure 11): functional knowledge cell establishment stage, function layer analysis stage to decouple coupled functions, and integration stage to form independent functional knowledge cells. However, there still exists a drawback, that is, the neglect of behavioral effect in formulation process. Therefore, we introduce behavior representation into functional knowledge cell after integration stage through retrieving input and output flow of functions in FBS knowledge cell to acquire satisfactory behavior, which can link functions and structure together, and then add it into the FBS knowledge cell library. In this way, we can arrive at functional decoupling to keep it independent while representing functional knowledge more explicitly, which will obviously support the design process better. As shown in Figure 12, FBS knowledge cell representation describes the essential knowledge and possesses links to the corresponding elaborate representation.

Figure 11.

Integration of FBS knowledge cell.

Figure 12.

Representation of FBS knowledge cell.

Techniques for model implementation

The proposed RFBPs model is based on reuse of existing cases by picking up the most similar cases, ignoring irrelevant cases, followed with further modification of the design variables inconsistent with design requirements, and ending with design synthesis. In this section, we discuss about knowledge cell retrieval algorithms and design synthesis methodology.

Knowledge cell retrieval algorithms

In order to facilitate reuse of existing cases, the concept of FBS knowledge cell is introduced. Therefore, knowledge cell retrieval algorithms are core techniques in design processes (Jin et al., 2009).

Considering possible contents of design knowledge, there are mainly three kinds of knowledge types involved in the design process: numerical value, textual value, and interval value. For example, say, the numerical value of key–slot connection is “125 MPa,” choice of “hydraulic cylinder” is the textual value, and the range of thermometer measurement represented as “[−100,500]” is the interval value. Considering retrieval of cross types, six kinds of similarity measurements are needed. However, as cross-type retrieval between textual value and interval/numerical value makes no sense, only four similarity measurements are needed: textual–textual, numerical–numerical, interval–interval, and numerical–interval. Corresponding similarity measurements are given in Table 4 in Appendix 1. After all the similarities between units of design requirements/functions and knowledge cell are calculated, the knowledge cell retrieval algorithm is used to gain the global similarity. On acquiring global similarity, $α$ is set as a threshold for controlling the number of similar knowledge cells and exclusion of the less valuable ones with low values, then ranking the knowledge cell according to the similarity. Finally, sets of FBS knowledge cells that satisfy the design requirements will be listed and pushed to the design synthesis process.

Design synthesis methodology

Computational design synthesis involves generation of innovative design solutions automatically or semiautomatically (Bergen et al., 2009). Research on design synthesis has attracted much interest in both mechanical engineering and artificial intelligence fields for many years. There are many design synthesis methods, such as improved morphological matrix (IMM) (Yong et al., 2006), A-design model (Campbell et al., 1999), function-to-form mapping method (Roy et al., 2001), behavior-assisted synthesis method (Hoover and Rinderle, 1989), bond graph–based design synthesis (Ulrich and Seering, 1980), and input–output-based functional block design synthesis approach characteristics of flow variables (type, spatial configuration, etc.) (Chakrabarti and Bligh, 1994, 1996a, 1996b).

In order that conceptual design can arrive at a possible novel and promising resolution through retrieving a multidisciplinary FBS knowledge cell library, exhaustive search strategy is hired here to support design synthesis methodology. As depicted in Figure 13, we propose a design synthesis methodology that starts from analysis and representation of expected functions abstracted from design requirements, which follows closely with retrieval of satisfied FBS knowledge cells, and ends with synthesis of feasible FBS knowledge cells through backtracking of the path of functional decomposition as well as considering the “or/and” relationships among subfunctions (Zhang et al., 2001).

Figure 13.

Design synthesis algorithm of FBS knowledge cell.

The methodology is briefly described as follows:

Step 1. Analyzing all possible expected functions from design requirements. Specifying iteration number as maximum search depth for a specific expected function (initial search depth as zero) and threshold similarity coefficient $α$ .

Step 2. Representation of the jth expected function (F(i,j)) according to flow table and transitive verb table following the representation format given in section “Functional representation (Fe, Fs).”

Step 3. According to F(i,j).function_name, FBS knowledge cell library searches all possible knowledge cells, if successful, pushes them to step 6, else continue.

Step 4. According to F(i,j).input_flow, FBS knowledge cell library searches all satisfied knowledge cells, if successful, pushes them to a set of satisfied principle solutions {Ps(i,0), …, Ps(i,k), …, Ps(i,n)}, else, further decomposes F(i,j) into subfunctions.

Step 5. Evaluating Ps(i,k).output_flow with F(i,j). output_flow; if satisfied, generate combinational principle solutions through chaining principle solutions together in accordance with a sequence of input/output flows, else, assign the value of Ps(i,k).output_flow to a new function F(i,k,p).input_flow and the value of F(i,j).output_flow to F(i,k,p).output_flow, and return to step 4.

Step 6. According to knowledge cell retrieval algorithms, design constraints and expected behavioral criterion, weighing attributes and behavior of F(i,j) with Ps(i,j), once Ps(i,j) meets the set threshold similarity coefficient, push it to the functional morphological matrix, else, further decompose F(i,j) into subfunctions.

Step 7. Judging whether there are any expected functions in Step 1 that were never searched before; if false, synthesise FBS knowledge cell in the morphological matrix and list all the possible design solutions by backtracking the functional decomposition path (Figure 14) and exit, else, turn to Step 1 and pick up one that has not been searched before.

Figure 14.

Decomposition and synthesis path of design process.

Validation and implementation

Until recently, many research groups engage in artificial vision apparatus (Cha et al., 1992; Dobelle, 2000; Dowling et al., 2005) for blind people with damage to eyes and optic nerves; our project is motivated by remedying a defect ignored by most artificial vision apparatus researchers, that is, the visual reception sets are always fixed. Due to this reason, blind users have to track moving objects through head rotation instead of eye movement to keep the object in the center of vision as people with normal vision would do. In order to validate the usage of proposed RFBPs model and techniques, conceptual design of artificial vision apparatus incorporated with oculomotor machine is employed to demonstrate the performance of the method and its innovative mechanism.

Design requirements definition

The whole vision formulation mechanism of human and the constitution the of vision reflex arc (Figure 15(a)) are analyzed. Corresponding summary diagram of functions that should be embraced by the oculomotor machine is given (Figure 15(b)).

Figure 15.

(a) Sketch map of human vision reflex arc and (b) whole flow path of the artificial vision system.

Clarifying design requirements

The flow path of artificial vision system, which states the whole information flow and functional requirements corresponding to human vision system, consists of the basic functions to form vision and the control functions to adjust definition and direction of vision. By processing the visual signal to form instructions to control the incident angle of light and the incident light intensity, even the focus can be adjusted to form an explicit image. The consulted key design requirements are listed in Table 2.

Table 2.

Design requirements of eye apparatus.

No.	Item of design requirement	Subitem	Value	Unit	Type
1	Range of rotate angle	Left/right	>30	Degree (°)	Numerical
		Up/down	>70	Degree (°)	Fuzzy
2	Distance of visibility		>5	m	Fuzzy
3	Response time		<1	s	Fuzzy
4	Image stationary time		1	s	Numerical
5	Endure temperature		[−30,+70]	Degree (°)	Interval
6	Capability demanding	Reliability	High	None	Fuzzy
		Transmission noises	Low	None	Fuzzy
		Transmission ability	High	None	Fuzzy
		Pollution resistance ability	High	None	Fuzzy
7	Structural demanding	Compact	Yes	None	Textual
		External diameter (total)	[+23,+24]	mm	Interval
8	Materials demanding	Biological materials	Yes	None	Textual
		Material weight	Light	None	Fuzzy
		Working strength	Medium	None	Fuzzy

Formulation and representation of design functions

According to the analysis of vision formulation mechanism of human and aforementioned functional representation, design functions are expressed in the type of transitive verb + noun + complement and input/output flow. For example, input energy flow of function: Adjust incident light is default and its output flow is Mechanical energy, then adhere the expected behavior to function by searching behavioral input flow, output flow, and keyword. The expected behavior has behavioral variables as {rotate angle, light intensity, response time, distance of visibility}. According to section “Design synthesis methodology,” all suitable principle solutions meet the output flow Mechanical energy would be retrieved from FBS knowledge cell library and pushed to morphological matrix by analyzing the design constraints and assessment of behavior. However, if none of the FBS knowledge cells are retrieved directly, then further decompose the function into subfunctions: adjust focus, regulate rotate angle, regulate light intensity, and continue searching for feasible FBS knowledge cells. This process will never be stopped until all the feasible FBS knowledge cells of subfunctions are acquired sequentially. Illustration of the eye movement apparatus’ functional decomposition, which presents the expected functions of design requirements, is given in Figure 16.

Figure 16.

Illustration of the functional decomposition of the eye movement apparatus.

Design resolutions

Based on the functional decomposition processes of eye movement apparatus and calculating the similarity, the morphology matrix of subfunctional principle solutions is given in Table 3.

Table 3.

Morphology matrix of subfunctional principle solutions.

No.	Function_name	Driving_input—Functional_output	Principle_Name	Structure
1	Receive instructions	Electrical energy—electrical energy, Electrical/analog signal—digital signal	Voltage’s law	Voltage signal receiver
		Electromagnetic energy—electrical energy, digital signal—digital signal	Ohm’s law	Current signal receiver
2	Adjust focus	Electrical energy—mechanical energy	Magnetomechanical effect, electrostriction, magnetostriction, capacity effect, piezoelectric effect	Push–pull electromagnet, rotary electric motor, line electric motor, micro-motor
3	Regulate rotate angle	Mechanical energy—mechanical energy	Newton’s second law	Universal coupling
4	Regulate light intensity	Thermal energy—mechanical energy, hydraulic energy—mechanical energy	Thermal expansion effect, fluid effect, Pascal effect	Micro-driver, hydraulic cylinder, pneumatic rod, artificial muscle
5	Form image	Optical energy—optical energy	Lambert–Beer law, refractive law, reflection law, image-forming principle	Concave lens, convex lens, plane lens
6	Convert light signal	Optical energy—electrical energy, analog signal—digital energy	Photoelectric effect, photoconductive effect	CCD, CMOS

CCD: charge-coupled device; CMOS: complementary metal-oxide semiconductor.

When the decomposition and retrieve processes end, then proceed with the design synthesis process to get a feasible design solution by backtracking the path of the functional decomposition. For example, the structures micro-motor, push-pull electromagnet, and micro-motor will combine together for the structure to realize the function adjust incident light. Finally, after evaluation of generated design resolutions, one of the corresponding design solutions, blue tooth receiver → micro-motor → push-pull electromagnet → micro-motor → micro-camera, is chosen. Afterward, retain the design resolutions in the document. Further detailed design of design solutions, such as the structural arrangement, still depends on the effort of designers.

Implementation of the proposed model

Employing MyEclipse 7.1 as developer and SQL 2005 as the database system, a prototype system with architecture of Client/Server named knowledge-based innovative conceptual design system (KICDS) has been developed. About 196 commonly used FBS knowledge cells and 65 scientific principles derivate from multidisciplinary fields, including mechanical engineering, physics, hydraulic, electronics, and electrical, are built in the system. The system consists of mainly three subsystems: functional design subsystem, FBS knowledge cell management subsystem, and basic data management subsystem. Typical interfaces of KICDS are given in Figure 17, including defining of design constraints/design functions/design behavior, showing of design resolutions, and displaying of tree view of functional decomposition.

Figure 17.

Typical interfaces of KICDS.

Discussion and conclusion

The proposed RFBPs model is a combination of Gero’s FBS model and case-based reasoning (CBR) model. CBR model has been well formalized for reasoning and reusing of existing cases, while it is hard to arrive at explicit resolutions. Gero’s FBS model can provide a more implicit signal for design process by introducing function/behavior/structure. The proposition of FBS knowledge cell is to better represent design knowledge from different disciplines in the view of the function domain. It makes the representation of existing multidisciplinary cases integral and reusable, which will facilitate the design process. Such modification brings exciting changes as many innovations spring from knowledge of different domains. Using the FBS knowledge cell retrieval algorithm, it is possible to retrieve all the possible knowledge cells. The expected functions can be thus be decomposed into subfunctions that can be fulfilled by the FBS knowledge cells, which is domain independent and can provide designers related knowledge in conceptual design. Finally, KICDS system is built according to the RFBPs model and techniques. By making use of KICDS system, designers can employ knowledge from different disciplines in order to stimulate innovative conceptual design. The developed system is not to create knowledge automatically but to provide human designers with various kinds of formalized functional knowledge across domains, which can hint at possible novel and promising design resolutions. The FBS knowledge cell library and design constraints database should be dynamic, which can be expanded by adding new elements into it.

The contributions of this research can be summarized as an adapted RFBPs model and its implementation techniques. The FBS knowledge cell is proposed to construct a consistent knowledge representation model for conceptual design processes. In this study, we also propose the design synthesis methodology and FBS knowledge cell retrieval algorithm.

Note that all discussions and conclusions drawn should be considered together with the specific experimental design. There are still some limitations of this approach in the industrial application: (1) The scale of RFBPs library is not enough to tackle development of the complicated system. Therefore, much effort should be expended to enrich the FBS knowledge cell library in the future. (2) Considering there exist conflicts between existing FBS knowledge cell and the newly adding FBS knowledge cells, conflicts resolution mechanism of FBS knowledge cells should be done in our future research (Xiao and Jie, 2011). (3) Transformation of natural language into formal representation through natural language modeling interface (Lee et al., 2012). (4) Compatibility of working principles to achieve design functions in terms of technical feasibility.

Footnotes

Appendix 1

Table 4.

Similarity measurements for different types of content.

Similarity measurement	Algorithmic notation
Textual–textual: 0–1 algorithm (Jin et al., 2009)	$si m_{T} (a, b) = (a = b) ? 1 : 0$
Numerical–numerical: Euclidian distance function (Jin et al., 2009)	$\begin{matrix} d (a, b) = {(\sum_{i = 1}^{n} w_{i} \| a_{i} - b_{i} \|^{2})}^{1 / 2} \\ si m_{n} (a, b) = 1 - \frac{d (a, b)}{Max - Min} \end{matrix}$
Interval–numerical: interval similarity measurement (Bryant, 1997)	$\begin{array}{l} s i m_{I} (a, b) \\ = s i m ([a_{1}, a_{2}], b) \\ = {\begin{matrix} \frac{b \times (\ln a_{2} - \ln a_{1})}{a_{2} - a_{1}} b \leq a_{1} \\ 1 - \frac{1}{a_{2} - a_{1}} (\frac{{(b - a_{1})}^{2}}{2 f} + a_{2} - b - b \times (\ln a_{2} - \ln b)) a_{1} < b < a_{2} \\ \frac{a_{2} + a_{1}}{2 b} b \geq a_{2} \end{matrix} \end{array}$
Interval–interval: interval similarity measurement (Bryant, 1997)	$\begin{matrix} si m_{II} (a, b) \\ = sim ([a_{1}, a_{2}], [b_{1}, b_{2}]) \\ = \frac{\int_{a_{1}}^{a_{2}} \int_{b_{1}}^{b_{2}} (1 - \frac{d (x, y)}{Max - Min}) dydx}{(a_{2} - a_{1}) (b_{2} - b_{1})} \end{matrix}$
Total: knowledge cell retrieval algorithm	$si m_{W} (Fe, FBSc) = \frac{\sum_{i = 1}^{N} w_{i} \times sim {(a, b)}_{i}}{\sum_{i = 1}^{N} w_{i}}$

$sim (a, b)_{i}$ stands for the similarity between $a$ and $b$ ; $[a_{1}, a_{2}]$ is the interval from $a_{1}$ to $a_{2}$ ; $w_{i}$ represents the weight of $sim (a, b)_{i}$ .

Acknowledgements

We are grateful to the reviewers and the editors for their constructive suggestions. The authors are also grateful to the anonymous reviewers for their valuable comments.

Declaration of conflicting interests

The authors declare that there is no conflict of interest.

Funding

This research is supported by the National Basic Research Program of China (No. 2011CB707503, 2011CB013305), National Natural Science Foundation of China (Nos 51075262, 51275293, 50575142, 50775140, 60304015, 51005148), Program for New Century Excellent Talents in University (No. NCET-08-0361), National High Technology Research and Development Program of China (No. 2008AA04Z113), Shanghai Committee of Science and Technology (No. 11JC1406100, 11BA1405300, 08JC1412000, 09DZ1121400, 10DZ1960102), National Key Scientific Instruments and Equipment Development Program of China (No. 2011YQ030114), National Key Technology Research and Development Program of China (No. 2010BAK69B18), Program of Introducing Talents of Discipline to Universities (No. B06012), and 2010 Shanghai Jiaotong University Polytechnic Cross-Fund.

Author biographies

Jin Ma is a PhD candidate at the Institute of Knowledge-Based Engineering, Mechanical Engineering Department, Shanghai Jiao Tong University, China, since 2011. He has participated in Research on Multidisciplinary Creative Collaborative Design based on Knowledge Fusion, funded by National 863 High Technology Research and Development Program of China (no. 2008AA04Z113) and Research on Collaborative Creative Design, funded by National Natural Science Foundation of China (No. 51075262). His current research areas include concurrent and collaborative design and creative conceptual design.

Jie Hu received his PhD from Zhejiang University, Hangzhou, China, in 2001. He worked at Tsinghua University for postdoctoral research from 2001 to 2003. He is currently a professor at Shanghai Jiao Tong University. He is leading the National Natural Science Foundation of China under grant nos. 51075262, 50775140, 50575142, and 60304015. He has developed methods for areas, such as concurrent and collaborative design, knowledge fusion, and knowledge-based design. He is interested in concurrent and collaborative design and manufacturing. He has published approximately 50 articles in journals and conferences.

Kai Zheng is a PhD candidate in the Institute of Knowledge-Based Engineering, Mechanical Engineering Department, Shanghai Jiao Tong University, China, since 2011. His current research areas include concurrent and collaborative design and Model Validation.

Ying-Hong Peng received the PhD degree from Central South University, Changsha, China, in 1993. He worked at Shanghai Jiao Tong University for postdoctoral research from 1993 to 1995. He is currently a professor at Shanghai Jiao Tong University. He has been director at the Institute of Knowledge-Based Engineering of Shanghai Jiao Tong University since 2002. He has authorized and edited two books and has published approximately 80 articles in journals and conferences.

References

Bergen

Kristina

Frank

(2009) A framework for computational design synthesis based on graph-grammars and function-behavior-structure. In: Proceedings of the ASME 2009 international design engineering technical conferences & computers and information in engineering conference, San Diego, CA, 30 August–2 September.

Bracewell

Sharpe

JEE

(1996) Functional descriptions used in computer support for qualitative scheme generation—schemebuilder. AI EDAM 4(10): 333–346.

Bryant

(1997) A case-based reasoning approach to bankruptcy prediction modeling. Intelligent Systems in Accounting, Finance and Management 6: 195–214.

Camelo

Mulet

Vidal

(2007) Function and behavior presentation for supporting flexible exploration and generation in a functional model for conceptual design. In: International conference on engineering design, Paris, France, 28–31 August.

Campbell

Cagan

Kotovsky

(1999) A-design: an agent based approach to conceptual design in a dynamic environment. Research in Engineering Design 11: 172–192.

Cha

Horch

Normann

(1992) Mobility performance with a pixelized vision system. Vision Research 32: 1367–1372.

Chakrabarti

Bligh

(1994) An approach to functional synthesis of solutions in mechanical conceptual design. Part I: introduction and knowledge representation. Research in Engineering Design 6: 127–141.

Chakrabarti

Bligh

(1996a) An approach to functional synthesis of solutions in mechanical conceptual design. Part II: kind synthesis. Research in Engineering Design 8: 52–62.

Chakrabarti

Bligh

(1996b) An approach to functional synthesis of solutions in mechanical conceptual design. Part III: spatial configuration. Research in Engineering Design 2: 116–124.

10.

Christophe

Bernard

Coatanea

(2010) RFBS: a model for knowledge representation of conceptual design. CIRP Annals: Manufacturing Technology 59: 155–158.

11.

Deng

Britton

Tor

(2000) Constraint based functional design verification for conceptual design. Computer-Aided Design 32: 889–899.

12.

Dobelle

(2000) Artificial vision for the blind by connecting a television camera to the visual cortex. ASAIO Journal 46(1): 3–9.

13.

Dowling

Boles

Maeder

(2005) Mobility assessment using simulated artificial human vision. In: 1st IEEE workshop on computer vision applications for the visually impaired in conjunction with the IEEE conference on computer vision and pattern recognition, San Diego, CA, 20–26 June.

14.

Gero

Kannengiesser

(2004) The situated function-behavior-structure framework. Design Studies 25:373–391.

15.

Goel

Rugaber

Vattam

(2009) Structure, behavior and function of complex systems: the structure, behavior, and function modeling language. Artificial Intelligence for Engineering Design, Analysis and Manufacturing 23: 23–35.

16.

Hirtz

Stone

McAdams

. (2002) A functional basis for engineering design: reconciling and evolving previous efforts. Research in Engineering Design 13: 65–82.

17.

Hoover

Rinderle

(1989) A synthesis strategy for mechanical devices. Research in Engineering Design 1: 87–103.

18.

Jin

Jie

Ying

(2009) A case retrieval method combined with similarity measurement and multi-criteria decision making for concurrent design. Expert Systems with Applications 36: 10357–10366.

19.

Kitamura

Mizoguchi

(2003) Ontology-based description of functional design knowledge and its use in a functional way server. Expert Systems with Applications 24(2): 153–166.

20.

Kitamura

Sano

Namba

. (2002) A functional concept ontology and its application to automatic identification of functional structures. Advanced Engineering Informatics 16(2): 145–163.

21.

Kota

Lee

(1993) General framework for configuration design. Part 1: methodology. Journal of Engineering Design 4: 277–294.

22.

Lee

Fenves

Bock

(2012) Semantic product modeling framework and its application to behavior evaluation. IEEE Transactions on Automation Science and Engineering 9(1): 110–123.

23.

Pahl

Beitz

(1988) Engineering Design: A Systematic Approach. London: Springer-Verlag.

24.

Roy

Pramanik

Sudarsan

. (2001) Function-to-form mapping: model, representation and applications in design synthesis. Computer-Aided Design 33: 699–719.

25.

Sheng

Jie

Ying

(2010) Representation of functional micro-knowledge cell (FMKC) for conceptual design. Engineering Applications of Artificial Intelligence 23: 569–585.

26.

Stone

Wood

(1999) Development of a functional basis for design. Journal of Mechanical Design 122(4): 359–370.

27.

Suh

(1989). The Principles of Design. Oxford: Oxford University Press.

28.

Suh

(1998) Axiomatic design theory for systems. Research in Engineering Design 10(4): 189–209.

29.

Suh

(2001) Axiomatic Design: Advances and Application. Oxford: Oxford University Press.

30.

Ulrich

Seering

(1980) Synthesis of schematic descriptions in mechanical design. Research in Engineering Design 1: 3–18.

31.

Umeda

Ishii

Yoshioka

. (1996) Supporting conceptual design based on the function-behavior-state modeler. Artificial Intelligence for Engineering Design, Analysis and Manufacturing 10(4): 275–288.

32.

Xiao

Jie

(2011) A fuzzy knowledge fusion framework for terms conflict resolution in concurrent engineering. Concurrent Engineering: Research and Applications 19: 225–234.

33.

Yong

Peien

Bin

. (2006) Automated conceptual design of mechanisms using improved morphological matrix. Journal of Mechanical Design 128: 516–526.

34.

Zhang

Tor

Britton

(2001) EFDEX: a knowledge based expert system for functional design of engineering systems. Engineering with Computers 17(4): 339–353.