Proceedings of the ASME 2011 International Design Engineering Technical Conferences &

Computers and Information in Engineering Conference

IDETC/CIE 2011
August 29-31,
28-31, 2011, Washington, DC, USA

DETC2011-48125
DETC2011-4

AN ONTOLOGY OF CLASSIFICATION CRITERIA FOR FUNCTIONAL TAXONOMIES

Yoshinobu Kitamura Sho Segawa
Munehiko Sasajima Riichiro Mizoguchi

The Institute of Scientific and Industrial Research

Osaka University
8-1 Mihogaoka, Ibaraki, Osaka, 567-0047 Japan
{kita, segawa, msasa, miz}@ei.sanken.osaka-u.ac.jp

ABSTRACT
In order to facilitate sharing of functional models, some func- 1. INTRODUCTION
tional taxonomies each of which provides a set of verbs for Functionality is one of the key aspects of knowledge about arti-
representing generic functions (called functional terms here) facts. Thus, a functional model, i.e., a product model from the
have been developed. Their examples include some sets of gen- viewpoint of functionality, plays a crucial role in the conceptual
erally valid functions in the book written by Pahl and Beitz, design and in other engineering activities. Much research on the
Functional Basis (FB) developed by Hirtz et al. and FOCUS/Tx functional models such as [1-11] has been carried out to date. A
developed by the authors of this paper. The issue addressed in function of a component or a system in a functional model is
this paper is the implicitness of the criteria of classification of typically expressed as a pair of an active verb and its (gram-
functional terms in those taxonomies and thus unclearness of matical) object like in Value Engineering [1]. We here concen-
their definitions. This paper proposes an ontology of logical trate on such active verbs representing functions (called func-
criteria for classification of functional terms (called tional terms hereinafter).
FOCUS/View). Using the classes defined in FOCUS/View, the Sharing such functional models in engineering organiza-
classification criteria of a functional taxonomy can be explicitly tions facilitates engineering activities. One of the important
represented. These classes have been conceptualized based on approaches for this is to establish a taxonomy of general func-
deep investigation on FB and FOCUS/Tx. The benefits of the tional terms and to use it for representing functions in the func-
proposed FOCUS/View ontology include: (1) users of a taxon- tional models. Such a taxonomy provides a shared and con-
omy can easily understand differences of similar terms and se- trolled vocabulary for functional models and makes it easier to
lect an appropriate term out of them, (2) a developer of a tax- search them using the functional terms in the taxonomy.
onomy can check its logical classification structure and then Thus, some functional taxonomies have been proposed to
improve its logical clearness, and (3) we can compare different date [4][5][9][12][13][14]. For example, the book [5] by Pahl
taxonomies and establish more reliable mappings between their and Beitz shows some taxonomies of generally valid functions
terms for interoperability of functional models. In this paper, as in the German design methodology, which have been proposed
a demonstration of the benefit (1), the classification criteria of by Krumhauer, Roth and others independently. In US, Recon-
FB, FOCUS/Tx, Krumhauers and Roths generally valid func- ciled Functional Basis in the NIST Design Repository Project
tions are explicitly presented using FOCUS/View. As a demon- has been established [9] as a reconciliation of the original Func-
stration of the benefit (2), some logically problematic classifi- tional Basis [13] and other taxonomy [14]. This consists of a
cation structures of FB and the Krumhauers functions are dis- taxonomy of function (verb) and that of flow. We concentrate
cussed and modified for the logical clearness. For the benefit on the taxonomy of function as a functional taxonomy and call
(3), this paper demonstrates the mappings between FB and it FB in this paper.
FOCUS/Tx and a semantic interoperable document search sys- The authors have been involved in the research on func-
tem based on these mappings. tionality based on Ontological Engineering [15][16][17] and
have established a suite of functional ontologies named
KEYWORDS
Functional representation, ontology, functional design, de-
sign knowledge modeling

1 Copyright 2011 by ASME

FOCUS1 [18]-[25]. The framework has been successfully de- 3 To integrate those criteria and organize them as an on-
ployed in industry [20]. The effects of sharing functional mod- tology, and
els in engineering organizations mentioned above have been 4 To apply the organized criteria for some taxonomies for
confirmed in this deployment. As a part of this suite of ontolo- suggesting their generality.
gies, an ontology of functional concepts (named FOCUS/Tx) In this paper, we firstly show an overview of the framework
defines functional terms and can be used as a functional taxon- with an example in Section 2. Section 3 presents the contents of
omy for functional models. FOCUS/View (as a result of the step 3 above) with its base-
The functional terms in the functional taxonomies are de- model for capturing functions. In Section 4, as demonstrations
fined and classified according to different criteria. The issue of how to use the ontology and of its benefits (a), (b) and (d)
addressed here is such criteria for classification are implicit in shown above, the classification criteria of some existing func-
the taxonomies. In many taxonomies, the terms are defined in tional taxonomies (i.e., FB [9], FOCUS/Tx [24], Krumhauers
natural language and then the criteria are, in many cases, im- generally valid functions [5], and Roths ones [5]) are presented
plicit. Thus, such definitions are sometimes ambiguous and it is using FOCUS/View. As a result of this analysis, some logically
difficult to distinguish similar terms. For instance, extract and problematic structures of classification in FB and the Krum-
remove in FB [9] are defined as draw, or forcibly pull out, a hauers one and their modification are discussed. These are
flow and to take away a part of a flow from its prefix place, examples of the benefit (d). Note that this is just a demonstra-
respectively. It might be difficult to catch exact difference be- tion of the benefit of clarifying classification criteria. We intend
tween them and to select an appropriate one for a device. In neither to evaluate these taxonomies nor to criticize them.
fact, Garbacz points out some problems of the classification of Aiming at a demonstration of the benefits (c), (e) and (f),
FB such as lack of principle of classification and non- Section 5 presents mappings between functional terms of FB
exhaustiveness from logical and ontological viewpoints [26]. and those of FOCUS/Tx and a semantic document search sys-
The goal of this research here is to propose an ontology for tem based on these mappings. FOCUS/View enables us to clar-
classification of functional terms (named FOCUS/View), which ify different levels of those mappings (the benefits (c) and (e)).
defines common criteria used in classification in functional We present an overview of the document search system based
taxonomies. For example, FOCUS/View defines the sameness on them which enables users to access documents about func-
of kinds of operands (target objects) as a classification crite- tions which are annotated with the terms defined in either of
rion. As a result of our investigation on FBs definitions, we those taxonomies. Section 6 discusses related work followed by
reveal that the terms extract and remove in FB seem to be the concluding remarks.
distinguished based on this criterion. The term extract is in-
tended to use for the case where a flow is separated from the 2. OVERVIEW OF THE FRAMEWORK
different kind of flow(s) and the term remove is intended to Figure 1 shows an overview of the framework proposed in this
use for the case where a flow is separated from the same kind research with FB [9] as an example. This figure shows a por-
of flow. In this way, the meaning of functional terms in a func- tion of each element for simple explanation. FOCUS/View con-
tional taxonomy can be explicated using FOCUS/View. sists of two main parts: the kinds of criteria for classification
The main aims and benefits of FOCUS/View include: (the upper left) and the values for classification (the upper
a) From the scientific point of view, to clarify how to clas- right). Each of them is organized as an is-a hierarchy (general-
sify functional terms in a comprehensive and logical way; specific relationship).
b) For users of a taxonomy, to facilitate understanding of the Generally, a level of classification in a taxonomy (or an on-
meaning of functional terms in the functional taxonomy tology) can be explained using a pair of a classification crite-
and selecting an appropriate term out of similar terms; rion and its values. For example, human class can be classi-
c) For users of multiple taxonomies, to facilitate comparison fied into man and woman sub-classes based on the classifi-
of functional terms in those taxonomies; cation criterion sex with its values male and female, re-
d) For a developer of a functional taxonomy, to help him/her spectively. Such a criterion of classification represents a view-
check the logical structure of the classification and to fa- point or a principle for that classification. The classification
cilitate improvement of its logical clearness; values for the sibling sub-classes explain the differentia among
e) For researchers of functional taxonomies, to help them them.
establish clearer and more reliable mappings between Using this FOCUS/View, we can explicate classification
functional taxonomies, and criteria of a functional taxonomy. In Fig. 1, the lower left
f) For engineers intending to share functional knowledge, to shows a portion of the original classification of FB in the table
enable them to access documents annotated with either of form [9]. The lower right part shows a result of applying
functional taxonomies using a semantic document search FOCUS/View, that is, the FB with explicit classification crite-
system based on those mappings. ria. A blue circle node shows a functional term. A red rectangle
In this research, for establishing FOCUS/View, we have con- shows a criterion for each classification. A yellow rectangle
ducted the research following the four steps shown below: shows a value for each term. For example, the branch is clas-
1 To investigate the definitions of functional terms in FB sified into separate and distribute according to the classifica-
and FOCUS/Tx as examples of functional terms, tion criterion of distinguishability of operands with its values:
2 To reveal the criteria of classification of functional possible and impossible, respectively. These classification
terms in these taxonomies, criteria and their values are defined in FOCUS/View. The
separate is further classified into three sub-terms based on two
1
This is an abbreviation of a Functional Ontology for Categorization, Utiliza- criteria pattern of focus on operands and the sameness of
tion and Systematization of functional knowledge. For an overview of this kinds of operands. We discuss the detail of such classification
project, please refer to http://www.ei.sanken.osaka-u.ac.jp/topics/Focus.

2 Copyright 2011 by ASME

 Criteria for classification Values for classification
FOCUS/View: is-a
an ontology of Operands (op. or ops.) Values for ops. Values for effects
classification is-a
criteria of Values for characteristics of ops.
Characteristics of ops. Focus on ops.
functional terms
(portion) Values for possibility of distinguishing
(meta- Possibility of Necessity of Pattern of
distinguishing ops. focused ops. focus on ops.
classification) Possible Impossible

FB-Function

Classification
Class Secondary Tertiary criterion Branch
Functional (Primary)
Branch Possibility of distinguishing operands
taxonomies Separate possible impossible
Divide Value for
Extract Applying classification
Remove FOCUS/View Separate Distribute
Distribute to a functional
The original classification of taxonomy Pattern of focus on operands
Reconciled Functional Basis (FB) for explicating Focus on all Focus on an operand
(portion) classification
criteria Divide The sameness of kinds of ops.
Different kinds The same kind
Original functional taxonomies
Extract Remove
FB with explicit classification criteria

Taxonomies with explicit criteria using FOCUS/View

Figure 1. An overview of FOCUS/View with Reconciled Functional Basis (FB) as an example of a target taxonomy.

criteria in FB in Section 4.1. Operand (O)

In this manner, FOCUS/View enables us to make criteria Operands
Input Agent (A) Output location (L)
for classification clearer. Each taxonomy uses specific criteria port port
Function
affect
in a specific order in its hierarchy. FOCUS/View is a classifica- Operand1 Operand3
context (Fc)
tion (hierarchy) of elements for classification of functional Operand2 flows Operand4 Function
context A
terms. In this sense, we can call this meta-classification.
Operands Operands Function A
state1 state2
3. FOCUS/VIEW Behavior/
This section presents the detail of the FOCUS/View ontology. effect (E)
Time
Function B

We have established this ontology by analyzing FOCUS/Tx

t1 t2 Function
and FB mainly. Firstly, we built a base-model for behavior and Time interval (T) context B
function, which is a basis for identifying the criteria. This base-
model is a generalized model which covers FOCUS/Tx, FB and Figure 2. The base-model of function for FOCUS/View
other device-oriented functional modeling. Secondly, we re-
trieved actual criteria used in FB or FOCUS/Tx. Lastly, we put port and time. A device can consist of sub-devices which
organized them into a hierarchy. are connected through their ports. So, a device can be a compo-
nent, an assembly, a sub-system and a system as a whole.
3.1. Base-model In the FOCUS framework, the notion of function is de-
Figure 2 shows a base-model assumed for FOCUS/View. This fined as a role played by the behavior under a teleological con-
is based on a device-oriented viewpoint, which is commonly text (called function context (Fc)) [20]2. A function context
adopted in many functional taxonomies. The behavior (or effect depends on either an intention of users (or designers) or the
(E)) of a device is defined as the objective (without designers overall function of the whole system. A function of the whole
intention) interpretation of its input-output relation as a black system is dependent on a users (or designers) intention, while
box. A device is connected to another device through its input a function of a component in a system is dependent on the sys-
or output ports. A device as an agent (A) changes states of tems function.
things input (called operands (O)) such as substance like fluid, Using this base-model, we do not intend to enforce this
energy, motion, force and information (In FB, they are called definition of function to other taxonomies. The base-model just
flows). The input-output relation of the behavior is, to be exact, 2
This definition is a precise version of our previous definition a function is a
the difference between the states of the operands at the input result of teleological interpretation under a goal [18]. This definition is of the
port (location, L) and time (time point, T) and those at the out- base-function for an operand. We define meta-functions for other function [19]
as well. In this paper, we concentrate on the base-function.

3 Copyright 2011 by ASME

 Classification criteria

(O)
is-a
(A) (E) (T)
(Fc)
Operand (op. or ops.) Agent Effect Function Condition / Way of (W)
Time
Context control achievement
(O-1) (O-4)
(O-2) (O-3) (E-1) (E-2) (Cd)
(O-5) (W-1)
Kind of Characteristics Change of Attribute Effect on Effect on
Focus on Use of Time Time
op. of ops. numbers of of ops. states processes
ops. intermediary ops interval point
ops.
(O-2-1) (E-1-1)
(O-5-1) (O-5-2) (T-1) (T-2)
Distinguishability The sameness (O-2-2) Pattern of Kinds of Location of (E-1-2)
Necessity
of ops. of kinds of ops. focus on ops. state change state change
of op.

Compositional (O-2-3)
(E-1-1-1) Quantitative change Categorical change (E-1-1-2)
relationship of ops.

Figure 3. Classification criteria defined in FOCUS/View (portion).

represents an assumption that a function is dependent on a pur-

pose, i.e., either the users intention or the systems function, 4. APPLYING FOCUS/VIEW TO THE EXISTING
which is represented as a function context. Such intention- TAXONOMIES
relatedness of function is captured in the literature as aims- This section presents the results of applying FOCUS/View to
means [2] means and ends [4], F-B relationship [6] and in some existing functional taxonomies.
value engineering [1]. We have investigated other definitions of
function and clarified the relationship among them in [22][25]. 4.1. Reconciled Functional Basis
Reconciled Functional Basis has been proposed by Hirtz et al.
3.2. Elements of FOCUS/View [9], which is a result of reconciliation of some previous tax-
Based on the base-model, we have investigated definitions of onomies and empirical generalization based on a great number
FB and FOCUS/Tx and then have conceptualized their classifi- of empirical studies. The functional taxonomy of FB consists of
cation criteria. Figure 3 shows a portion of FOCUS/View. Al- 52 terms in three levels of classification. Each of functional
most all of the top-level elements (classification criteria) corre- terms is defined in natural language with examples and corre-
spond to the basic elements in the base-model discussed above: spondents (synonyms). Table 1 shows a portion of those func-
agent (A), operand (O), effect (E), function context (Fc) and tional terms with their definitions [9]. As mentioned in Intro-
time (T). duction, such a definition in natural language is sometimes am-
For example, the node (O) represents a super-class of the biguous and it is difficult to distinguish similar terms. In addi-
criteria concerning the operand(s) (In the figures, it is abbrevi- tion, some logical issues of the classification have been found
ated to op or ops). Its sub-node (O-1) represents a criterion by explicating the classification criteria using FOCUS/View.
related with the kinds of the operand(s). For instance, transfer Figure 4 shows a portion of the result of applying
in FB is classified into transport and transmit based on this FOCUS/View to FB, i.e., FB with explicit classification criteria
criteria according to the values material and energy, respec- defined in FOCUS/View. The authors have identified them
tively. The node (O-5) represents the criteria concerning focus according to the definitions shown in Table 1 and the given
on operands. When we capture a function of a device, we examples in [9] as own interpretation of them. The top-level
sometimes focus on specific operand(s) among the operands of classification of FB has no unique principle for classification.
the device. Depending on its pattern (the criterion O-5-1), FB Among them, the terms branch and connect are distin-
classifies separate into divide, extract and remove as guished according to the (O-3) criterion change of numbers of
discussed in Section 4.1. FOCUS/Tx uses the same (O-5-1) operands. The both branch and connect are further classi-
criterion for classification of separate into take-out2 and fied according to the same criterion (O-2-1) distinguishability
divide as shown in Section 4.2. of operands (into separate/distribute and couple/mix,
The effect (E) node represents a super-node of classes of respectively). The separate is further classified into divide,
the classification criteria for the effect in the base-model. It has extract and remove. While the definitions of extract and
sub-classes such as quantitative change (E-1-1-1) and cate- remove mention a specific operand (to draw ... a flow and
gorical change (E-1-1-2) for classification based on the kinds to take away a part of a flow ..., respectively), the definition
of state-change of the operands. of divide does not (see Table 1). So, this classification of
The condition/control (Cd) criterion is concerned with separate seems to be based on the patterns of focus on oper-
the precondition of the effect and/or controlling the effects. The ands (O-5-1). This criterion alone, however, cannot distin-
way of function achievement (W) represents the background guish between extract and remove. So, this classification
knowledge such as physical principle in functional decomposi- needs extra criterion. It seems to be the sameness of kinds of
tion, in which micro-functions achieve a macro-function [20]. It operands (O-2-2) according to their definitions, examples and
represents how to achieve a function, the authors believe, the informal discussion with some of the developers of FB (see
which should be distinguished from function (what to achieve). the acknowledgement section). The remove seems to be in-
We will revisit this issue in Section 4.1. tended to use for the case where a flow is separated from the

4 Copyright 2011 by ASME

 Table 1. Original organization and definitions of functional terms of Reconciled Functional Basis [9] (portion)

Primary Secondary Tertiary Definition (portion)

Branch to cause a flow to no longer be joined or mixed
Separate to isolate a flow into distinct component. The separated components are distinct from the origi-
nal flow, as well as each other.
Divide to split up a flow into parts or to classify distinct arts of a flow.
Extract to draw, or forcibly pull out, a flow.
Remove to take away a part of a flow from its prefix place.
Distribute to cause a flow to break up. The individual bits are similar to each other....
Connect to bring two or more flows ... together.
Couple to join or bring together flows such that the members are still distinguishable from each other
Join To couple flows together in a predetermined manner.
Link To couple flows together by means of an intermediary flow
Mix To combine two flows into a single, uniform homogeneous mass.
Change to adjust the flow in a predetermined and fixed manner
Increment to enlarge a flow
Decrement to reduce a flow
Shape to mold or form a flow
Condition to render a flow

Classification
FB Function
criterion (O-3) (E)

Value for Change of numbers of operands Effect

classification 1 -> n n -> 1
Change
Branch Connect (E-1-1)
(O-2-1) (O-2-1)
Kinds of state change
Distinguishability of operands Distinguishability of operands non-quantitative quantitative
possible impossible possible impossible

Distribute Mix Shape Q-change

Separate Couple
(O-5-1)
(W-1) (Cd)
Pattern of focus on operands (W-1)
Use of intermediary op. Use of intermediary op. Condition / Control
Focus on all Focus on an operand
not-used used not-used used prefixed conditional

Divide Separate2 Join Link Mix1 Mix2 Q-change2 Condition

(O-2-2)
The sameness of kinds of ops. Quantitative change (E-1-1-1)
The same kind Different kinds enlarge reduce

Remove Extract Increment Decrement

Figure 4. Reconciled Functional Basis (FB) with classification criteria defined in FOCUS/View (portion).
This is logically modified according to these criteria. (the gray nodes have been inserted)
same kind of flow, while extract seems to be intended to use use of intermediary operand (W-1 in Fig. 3) which is sub-
for the case where a flow is separated from the different kind of class of the way of function achievement (W). The authors
flow(s). believe that these functional terms that imply ways of function
This classification is problematic from the viewpoint of achievement do not represent pure functions and thus exclude
logic and ontological engineering. As a principle of ontological such terms from FOCUS/Tx as discussed in the next section.
engineering, classification of a concept into its direct sub- In addition, if we think the logical symmetry is important,
concepts should be based on a single criterion. If we apply this this criterion (W-1) should be applied to mix and then we
principle, an intermediate term is needed to be inserted as would have two sub-nodes (mix1 and mix2 in Fig. 4).
shown as separate2 with gray in Fig. 4 (The suffix number 2 In the original FB taxonomy shown Table 1, change is di-
in separate2 is added for distinguishing from separate). rectly classified into increment, decrement, shape and
The couple is classified into join and link. According condition. Obviously, this classification is based on multiple
to the definition and the example of link, the flows are cou- criteria. The right part of Fig. 4 shows a possible logical classi-
pled together by means of an intermediary operand (flow) such fication, in which a level of classification is based on a single
as a turnbuckle. So, this classification is based on how to criterion and then two intermediate terms (Q-change and Q-
achieve a function. We distinguish this from function (what to change2) have been inserted.
achieve) and then call the way of function achievement as In this manner, FOCUS/View helps us make classification
mentioned above. Thus, the criterion of this classification is criteria clearer and suggest possible logical improvement of

5 Copyright 2011 by ASME

 FOCUS/Tx
FOCUS/TxBase-function
Base-function
Change
Change Classification
Change composition
composition criterion
Change//affect
affect (O-3)
Change of numbers of ops. Value for
1 -> n n -> 1 classification
Change
Change//affect
affect Change
Change//affect
affect
relationship
relationshipbtw.
btw.ops.
ops. an
anop
op Separate
Separate Assemble
Assemble (O-2-2)
(O-5-1)
Change
Change Change
Change Change
Change The sameness of kinds of ops.
composition
composition position attribute Deform
Deform Convert
Convert Pattern of focus on ops. the same different different
position attribute Focus on an operand Focus on all kind portion parts ingredients

Change
Change Change
Change Unify
Separate
Separate Assemble
Assemble volume
volume temp.
temp. (O-5-2)
Take-out
Take-out22 Divide
Divide Unify Combine Compose
Combine Compose
Necessity of the operand (O-2-2)
The sameness of kinds of ops.
necessary unnecessary the same different different
Take-out
Take-out22 Divide Combine Compose
Divide Unify Combine Compose kind portion ingredients
Unify parts
Take-out
Take-out33 Remove
Remove Split Decompose
Split Detach
Detach Decompose
Take-out
Take-out33 Remove
Remove Split Decompose
Split Detach
Detach Decompose The sameness of kinds of ops. (O-2-2)
the same different different
kind portion part ingredient

Split
Splitoff
off Take
Takeoff
off Extract
Extract The original classification FOCUS/Tx with explicit
of FOCUS/Tx (portion) Split
Splitoff
off Take
Takeoff
off Extract
Extract classification criteria (portion)
Figure 5. Classification criteria of FOCUS/Tx (Portion)

taxonomies. Note that these classification criteria are the au- FOCUS/Tx as shown in Fig. 5. In addition, FB also uses the
thors interpretation of the original definitions in [9] as an ex- same criterion for the classification of separate (precisely
ample of application of FOCUS/View. Thus, the authors do not speaking, the distinction between extract and remove) as
claim their correctness. In addition, the suggested modifications discussed in the previous section. In this manner, FOCUS/View
here are from purely logical point of view. The authors also do provides clues for easy understanding of a taxonomy and easy
not claim that appropriateness of those inserted terms and their comparison among different taxonomies.
usefulness from the engineering point of view. These are dif- The criteria used in FOCUS/Tx cover almost all of the cri-
ferent issues and should be in nature verified by empirical study teria used in FB with some exceptions. These exceptions can be
and/or practice in industry as discussed in Section 6. explained by the fact that some criteria imply different ones and
the policy that FOCUS/Tx excludes the functional terms that
4.2. FOCUS/Tx are classified according to one of a way of function achieve-
FOCUS/Tx defines generic types of the base-functions (called ment criteria (W criteria in Fig. 3) as discussed above. Thus,
functional concepts in the ontology. They correspond to func- these W criteria are never used in FOCUS/Tx. From this obser-
tional terms in this paper). The left part of Fig. 5 shows its por- vation, we can say that FOCUS/Tx covers FB sufficiently. This
tion. A functional concept (a class of function) is defined onto- fact is very interesting, considering the following backgrounds
logically using constraints on the cardinality of operands, rela- of these taxonomies. They have been developed independently
tionships among them and/or designers intention to change from each other using different natural languages for terms
(focus of intention). For example, a function to divide an op- (FOCUS/Tx is designed firstly in Japanese, while FB is de-
erand is defined by the following semantic constraints: (1) the signed for (and defined by) English). We revisit this result with
cardinality of the input focused operand must be 1, (2) the car- the mapping result in Section 5.1.
dinality of the output focused operands must be greater than 1,
(3) there must be material-product relationship between the 4.3. Krumhauers functions and Roths ones
input operand and the output operands and (4) all the output The classification of the Krumhauers generally valid functions
operands are equally focused. The first three are inherited from is explained in the book [5] that it is based on differences be-
the super-concepts such as separate. The fourth one enables tween input and output of (1) type, (2) magnitude, (3) number,
us to distinguish the divide function from the sibling function (4) location and (5) time as shown in the left part of Fig. 6.
take-out2. This ontology has been implemented using an on- Their classification criteria can be represented using
tology editor Hozo3 in its own language and in OWL and FOCUS/View as shown in the right part of Fig. 6 with interme-
SWRL languages. In these implementations, although those diate nodes (depicted with gray) for logical clearness. As you
definitions are clear, the classification criteria and its values are can see, the original one-level classification implies many dif-
not explicitly conceptualized. ferent criteria.
Using FOCUS/View, we can make the classification crite- The Roths generally valid functions shown in [5] use al-
ria clearer as shown in the right part of Fig. 5 (Only subclasses most the same criteria as ones used in the Krumhauers func-
of change composition are shown). Thanks to FOCUS/View, tion. The difference is that the criterion quantitative change is
we can easily understand that the same O-2-2 criterion the not used in the Roths classification. So, in a manner similar to
sameness of the kinds of operands is used for the classification Fig. 6, we can explicate its classification criteria.
of three functional terms: assemble, divide and take-out3 in Consequently, FOCUS/View can explain these taxonomies
3 other than FB and FOCUS/Tx which we have investigated for
http://www.hozo.jp

6 Copyright 2011 by ASME

 Krumhauers
Krumhauersfunction
function
Krumhauers
Krumhauersfunction
function
(Type) (Time) (T)
Time
(T-1) (T-2)
Change Store
Store time interval time point
Change
TI-Change Store
Store
Increase
Increase Connect
Connect Channel
Channel (E-1-1)
Kinds of state change
Decrease Branch
Branch Stop
Stop
Decrease non-quantitative quantitative
(Magnitude) (Number) (Place)
C-Change Q-Change
Characteristics of input-output (E-1-1-2) (O-4)
Categorical change Attribute of ops.
type composition magnitude location

Change
Change Change Change
(convert)
(convert) composition magnitude Move
(E-1-1-1) (E-1-1-1)
(O-3) Change of numbers of ops. Qualitative change
Qualitative change
n -> 1 1 -> n enlarge reduce -> 0 n1 -> n2

Connect
Connect Branch
Branch Increase
Increase Decrease Stop
Stop Channel
Channel
Decrease

Figure 6. Classification criteria of Krumhauers generally valid functions.

(Left) Original classification [5], (Right) Re-organization using FOCUS/View with inserted terms (the gray nodes)

FOCUS/Tx
FOCUS/TxFunction
Function

FB Function Change
Change
composition
composition
(O-3) (O-3)
correspondence
Change of numbers of operands Change of numbers of operands
1 -> n n -> 1
1 -> n

Branch Separate
Separate Assemble
Assemble
(O-2-1) mapping1 (O-5-1)

Possibility of distinguishing operands Pattern of focus on operands

possible impossible Focus on an operand Focus on all

Separate Distribute Take-out

Take-out22 Divide
Divide
(O-5-1) Necessity of the operand (O-5-2)
Pattern of focus on operands necessary unnecessary
Focus on all Focus on an operand
(O-2-2) Take-out
Take-out33 Remove
Remove
Divide
The sameness of kinds of ops. (O-2-2)
The same kind Different kinds The sameness of kinds of ops.
different different the same
Remove Extract part ingredient kind portion
mapping2
Take
Takeoff
off Extract
Extract Split
Splitoff
off
mapping3

Figure 7. Mappings between Reconciled Functional Basis (FB) and FOCUS/Tx based on FOCUS/View.

the construction of FOCUS/View. This suggests a level of gen- lishment of mappings between functional terms of them. The
erality of FOCUS/View for representing classification criteria mappings between functional taxonomies enable us to realize
of functional terms. Of course, we do not claim its comprehen- interoperability between them. In our previous papers [23][24],
siveness for all functional taxonomies. Applying it to other ex- we presented mappings between FB and FOCUS/Tx. In this
isting functional taxonomies such as those in [4][12] and fur- paper, we suggest FOCUS/View makes mapping easier and
ther investigation on its generality remain as future work. more accurate than previous mappings.
Figure 7 shows some examples of the mappings between
5. USING FOCUS/VIEW FOR MAPPINGS BETWEEN FB and FOCUS/Tx terms based on FOCUS/View. We have
TAXONOMIES established a set of a correspondence mapping between a term
5.1. Mapping between FB and FOCUS/Tx of FB and a term of FOCUS/Tx, which have similar meanings.
The explication of the classification criteria facilitates compari- Finding corresponding terms has been manually done based on
sons between different functional taxonomies and then estab- the explicit classification criteria using FOCUS/View. For ex-

7 Copyright 2011 by ASME

ample, as shown in the upper part of Fig. 7, we can establish a These success rates are still regarded as very high, consid-
mapping (denoted by mapping1) between branch of FB and ering the background of these taxonomies mentioned in Section
separate of FOCUS/Tx, since their classification criteria and 4.2. Consequently, both of these success rates and the high cov-
their values are the same. We can say that this type of mappings erage ratio of the classification criteria discussed in Section 4.2
shows exact correspondence. strongly suggest the validity of the content of both FOCUS/Tx
On the other hand, in the lower part of Fig. 7, both the and FB from their commonality. The suggested validity is sup-
mapping2 between extract of FB and take off of FOCUS/Tx ported by their applications as well. FB is widely used and has
and the mapping3 between extract of FB and extract of many empirical studies. FOCUS/Tx has been deployed in
FOCUS/Tx are not exact one. The classification criterion for manufacturing companies in Japan [20].
these terms is the same (the sameness of kinds of ops. (O-2-
2)). The values for the classification are, however, different. 5.2. Interoperability using mappings of taxonomies
Extract of FB is based on the different kind as the value, The mappings between functional taxonomies can improve
while take off and extract of FOCUS/Tx are based on more interoperability of functional knowledge. We have developed a
fine-grained values: different part (such as change by struc- semantic document search system (named Funnotation Search
tural decomposition) and different ingredient (such as change System) [21] which can provide engineers with interoperable
by a chemical reaction), respectively. So, we can say that these access to annotated technical documents by searching for func-
mappings represent rough correspondence between them, tional terms based on the mappings between FB and
which are different from the exact correspondence such as FOCUS/Tx [24]. On the basis of the Semantic Web technology,
mapping1 discussed above. technical documents are annotated with metadata using func-
In this manner, the explicit classification criteria using tional terms defined in either FB or FOCUS/Tx. By translating
FOCUS/View facilitate more reliable mappings between func- the functional terms in the query and the metadata annotated
tional taxonomies. In addition, it enables us to show the differ- with documents, the search system can access both documents
ence of the degrees of the exactness of the mappings. that are annotated based on either FB or FOCUS/Tx. Figure 8
As reported in [24], in the mappings from FB to just shows a search result when a user gives split of
FOCUS/Tx, the terms in FB cover (have mappings to) 33 terms FOCUS/Tx as a search word. It includes not only documents
in FOCUS/Tx out of the total of 89 (37%). Its success rate was annotated with split of FOCUS/Tx but also those documents
calculated about 80% when the paper was written (please refer annotated with distribute of FB which has a mapping to split
to the paper [24] for the criteria for calculation of this success of FOCUS/Tx. Please refer to the papers [21][24] for the detail
rate). In the mappings from FOCUS/Tx to FB, the terms in of the Funnotation framework.
FOCUS/Tx cover 43 terms in FB out of the total of 52 (83%). Using FOCUS/View, the user can check the exactness of
Its success rate was calculated about 70% excluding the terms the mappings and select documents with the exact correspon-
in the different grain-sizes. dence only according to his or hers intention.
Using FOCUS/View, we can analyze more accurate suc-
cess rate of mappings based on the degree of the exactness of 6. DISCUSSION AND RELATED WORK
the mappings. For example, among the successful mappings The ultimate goal of this research is to enumerate possible clas-
from FB to FOCUS/Tx reported above, it turned out that 7 sification criteria of functional terms. On the other hand, Rec-
mappings are rough correspondences. So, if we regard only onciled Functional Basis is a result of merging two existing
exact correspondences as successful mappings, the accurate taxonomies aiming at a standardized taxonomy [9]. We aim at
success rate would be 75%. clear comparison between different taxonomies based on the

Figure 8. Interoperable search result by the Funnotation system based on mappings between functional taxonomies.

8 Copyright 2011 by ASME

enumerated classification criteria and establishing mappings discussed as the base-model in Section 3.1. In fact, the base-
(ontology matching in the terminology of [27]) rather than model of FOCUS/View is a simplified version of more generic
merging (ontology merging), in order to allow the diversity of model discussed in [22]. In that paper, we have discussed dif-
conceptualization of functions and their classification. Thus, ferent conceptualizations (definitions) of the notion of function
FOCUS/View provides not a super-set (logical sum) of the ex- (such as environment function [7]) other than device-oriented
isting taxonomies but generic and common classification crite- one. Then, we have proposed an ontology of definitions of
ria used in functional taxonomies. function in a hierarchical manner, which is called a reference
As pointed out in [28], a shared ontology can facilitate ontology of function (FOCUS/Ref) [22]. Other function repre-
semantic integration. The top-level generic ontologies such as sentation frameworks such as [6][8][11] would be explained
DOLCE4 and YAMATO5 can be used as the shared ontology. not at the level of FOCUS/View but at the level of FOCUS/Ref.
Our FOCUS/View also can be regarded as a kind of such a The top-level of FOCUS/View can be used for explanation of
shared ontology for matching concepts in ontologies, though a the classification of FOCUS/Ref as well. The integration of
functional term of a functional taxonomy is not a subtype of a FOCUS/View and FOCUS/Ref remains as future work.
class defined in FOCUS/View but is used as criteria for classi- In this paper, a functional taxonomy is analyzed from a
fication. In this sense, the top-level ontologies are at super- logical point of view, though as we noted in Introduction nei-
level, while FOCUS/View is at the meta-level. ther evaluation of the existing taxonomies nor criticism on them
ONIONS methodology [29] is pioneering work to integrate is our aim. Generally, a functional taxonomy can be evaluated
terminologies based on formal and generic ontologies. It in- by empirical way or by industry practice. The examples of such
cludes the conceptual analysis phase, in which the entities of evaluation can be found in [32][33].
a source terminology are represented in a formal way. Although
our approach is not based on formal and generic (top-level) 7. CONCLUDING REMARKS
ontologies for integration, the explication of classification crite- In this paper, we proposed an ontology of classification criteria
ria using FOCUS/View corresponds to a kind of the conceptual of functional terms in functional taxonomies based on investi-
analysis. gation on two existing taxonomies (i.e., FB and FOCUS/Tx).
Some logical, semantic, formal, or mathematical defini- We applied the ontology to other existing taxonomies as well as
tions of functional taxonomies have been proposed in the litera- those taxonomies. As shown in Section 4, the ontologys appli-
ture (e.g., [30][31]). In [30], the FB terms are semantically de- cation to those taxonomies clarifies the classification criteria
fined using the Semantic Web technologies such as OWL and implicit in their original classification. In addition, based on the
SWRL. In [31], some of FB terms are defined in a set-theoretic analysis, some intermediate terms have been inserted, which
notation. Both research efforts aim at automatic reasoning such contribute to logical clarity of each taxonomy.
as consistency checking. FOCUS/Tx also has ontological defi- From the viewpoint of design knowledge management, the
nitions of its functional terms. The main aim of FOCUS/View ontological clarification of the classification criteria of func-
is to uncover classification criteria of taxonomies which have tional taxonomies contributes to interoperability between them
been left implicit rather than to define each functional term for by establishing more reliable mappings between terms defined
such automatic reasoning. In addition, the definitions in in them. We presented the interoperable document search sys-
[30][31] capture only differences between input-flows and out- tem based on such mappings.
put-flows (the operands in our terminology). FOCUS/View The complete demonstration of the benefits of this kind of
includes rich concepts organized in an is-a hierarchy as catego- the ontology is in nature difficult. Among the possible benefits
rization criteria from other aspects for capturing functions and of the ontology listed in Introduction, the forth one ((d) check-
thus enables us to clarify deep conceptual criteria for categori- ing the logical structure and its improvement) was demon-
zation more richly. strated in the application to FB and Krumhauers functions. The
Garbacz proposes an ontologically-refined FB based on fifth and sixth ones ((e) mapping and (f) interoperability) were
some ontological upper-level distinctions defined in DOLCE demonstrated using the mappings between FB and FOCUS/Tx
[26]. The proposed taxonomy has clear ontological classifica- and the implemented document search system. The first one
tion criteria at the upper-level. FOCUS/View aims at more con- ((a) scientific contribution to understanding of function) is in
crete-level classification for engineering practice. nature difficult to be proved. The proposed ontology, however,
Ideally (please refer to the following paragraph for a limi- explicates (at least) some criteria, which have been implicit in
tation), any functional taxonomies can be built by selecting a the some existing taxonomies. This ontology is the first and
classification criterion and their value sets from those defined important step towards comprehensive understanding of classi-
in FOCUS/View for each level of classification and by deter- fication of functional terms.
mining a specific order of applying these classification criteria
for organizing a hierarchy. ACKNOWLEDGMENTS
Of course, the authors have no intention to claim that the The authors are most grateful to Robert B. Stone (currently
comprehensiveness and/or completeness of the classification Oregon State University) and his colleagues of Missouri Uni-
criteria defined in FOCUS/View. Firstly, the comprehensive- versity of Science and Technology for the informal discussion
ness of such an ontology can be evaluated in nature not by a especially for clarifying some portions of the semantics of FB,
theoretical way but by an empirical way. Secondly, as a limita- though this does not imply their agreement on the result of our
tion of FOCUS/View, its elements shown in Section 3.2 are analysis. The authors thank Masanori Ookubo and Naoya
mainly for a device-oriented modeling of function, which is Washio for their contribution.
4
http://www.loa-cnr.it/DOLCE.html
5
http://www.ei.sanken.osaka-u.ac.jp/hozo/onto_library/upperOnto.htm

9 Copyright 2011 by ASME

 [20] Kitamura, Y., Koji Y., and Mizoguchi, R., 2006, An ontological
REFERENCES model of device function: industrial deployment and lessons
[1] Miles, L. D., 1961, Techniques of value analysis and engineering, learned, Applied Ontology, 1(3-4), pp. 237-262.
McGraw-hill. [21] Kitamura, Y., Washio, N., Koji, Y., Sasajima, M., Takafuji, S.,
[2] Hubka, V. and Eder, W.E., 1988, Theory of Technical Systems, and Mizoguchi, R., 2006, An Ontology-Based Annotation
Springer-Verlag. Framework for Representing the Functionality of Engineering
Devices, In Proc. of ASME IDETC/CIE 2006, DETC2006-99131.
[3] Chandrasekaran, B., Goel, A. K. and Iwasaki, Y, 1993, Func-
tional representation as design rationale, Computer, 26 (1), pp. [22] Kitamura, Y., Takafuji, S. and Mizoguchi, R., 2007, Towards A
48-56. Reference Ontology for Functional Knowledge Interoperability,
In Proc. of the ASME IDETC/CIE 2007, DETC2007-35373.
[4] Lind, M., 1994, Modeling goals and functions of complex indus-
trial plants, Applied Artificial Intelligence, 8, pp. 259-283. [23] Ookubo, M., Koji, Y., Sasajima, M., Kitamura, Y., Mizoguchi, R.,
2007, Towards Interoperability between Functional Taxonomies
[5] Pahl, G. and Beitz, W., 1996, Engineering Design - a Systematic using an Ontology-based Mapping, In the Proc. of 16th Interna-
Approach, Springer-Verlag. tional Conference on Engineering Design (ICED 07), 154.
[6] Umeda, Y., Ishii, M., Yoshioka, M., Shimomura, Y., and Tomi- [24] Kitamura, Y., Segawa, S., Sasajima, M., Tarumi S., and Mizogu-
yama, T., 1996, Supporting conceptual design based on the func- chi, R., 2008, Deep Semantic Mapping between Functional Tax-
tion-behavior-state modeler, Artificial Intelligence for Engineer- onomies for Interoperable Semantic Search, In Proc. of the 3rd
ing Design, Analysis and Manufacturing, 10, pp. 275-288. Asian Semantic Web Conference (ASWC 2008), LNCS 5367, pp.
[7] Chandrasekaran, B. and Josephson, J.R., 2000, Function in De- 137-151.
vice Representation, Engineering with Computers, 16 (3/4), pp. [25] Kitamura, Y. and Mizoguchi, R., 2009, Some Ontological Dis-
162-177. tinctions of Function based on the Role Concept, In Proc. of the
[8] Gero, J.S. and Kannengiesser, U., 2002, The Situated Function- ASME IDETC/CIE 2009, DETC2009-87168.
Behaviour-Structure Framework, In Proc. of Artificial Intelli- [26] Garbacz, P., 2006, Towards a standard taxonomy of artifact
gence in Design 02, pp. 89-104. functions. Applied Ontology, 1(3/4), pp. 221-236.
[9] Hirtz, J., Stone, R.B., McAdams, D.A., Szykman, S., and Wood, [27] Euzenat J., Shvaiko, P., 2007, Ontology Matching, Springer, Hei-
K.L., 2002, A Functional Basis for Engineering Design: Recon- delberg.
ciling and Evolving Previous Efforts, Research in Engineering
Design, 13, pp. 65-82. [28] Noy, N., 2004, Semantic integration: A Survey of Ontology-
based Approaches. ACM SIGMOD Record archive 33(4).
[10] Stone, R. B. and Chakrabarti, A. (eds.), 2005, Special Issues:
Engineering applications of representations of function, Artifi- [29] Gangemi, A., Pisanelli, D. M., Steve, G., 1999, An overview of
cial Intelligence for Engineering Design, Analysis and Manufac- the ONIONS project: Applying Ontologies to the Integration of
turing, 19 (2 and 3). Medical Terminologies, Data and Knowledge Engineering, 31.
[11] Goel, A. K., Rugaber, S., and Vattam, S., 2009, Structure, be- [30] Yang, S. C., Patil, L., Dutta, D., 2010, Function Semantic Rep-
havior, and function of complex systems: the structure, behavior, resentation (FSR): A Rule-Based Ontology for Product Func-
and function modeling language, Artificial Intelligence for Engi- tions, Journal of Computing and Information Science in Engi-
neering Design, Analysis and Manufacturing, 23 (1), pp. 23-35. neering, 10(3), 031001.
[12] Kirschman, C.F. and Fadel, G.M., 1998. Classifying Functions [31] Sen, C., Summers, J.D., and Mocko, G.M., 2010, Toward a For-
for Mechanical Design, Journal of Mechanical Design, 120 (3), mal Representation of the Functional Basis Verbs, In Proc. of
475-482. the Eighth International Symposium on Tools and Methods of
Competitive Engineering (TMCE 2010).
[13] Stone, R.B. and Wood, K.L., 2000. Development of a Functional
Basis for Design, Journal of Mechanical Design, 122 (4), 359- [32] Ahmed, S., 2005, Encouraging reuse of design knowledge: a
370. method to index knowledge, Design Studies 26(6), 565-592.
[14] Szykman, S., Racz, J.W. and Sriram, R.D., 1999. The Represen- [33] Caldwell, B.W., Sen, C., Mocko, G.M., Summers, J.D. and Fadel,
tation of Function in Computer-Based Design, In Proc. of 1999 G.M., 2008, Empirical Examination of the Functional Basis and
ASME Design Engineering Technical Conferences. Design Repository, In the Proc. of the 3rd International Confer-
DETC99/DTM-8742 ence on Design Computing and Cognition.
[15] S. Staab and R. Studer (eds.), 2003, Handbook on Ontologies,
Springer.
[16] Mizoguchi, R., 2003, 2004, Tutorial on ontological engineering -
Part 1-3, New Generation Computing, 21(4) and 22(1-2).
[17] McMahon, C., van Leeuwen, J. (eds.), 2009, Special issue on
Developing and Using Engineering Ontologies, Artificial Intelli-
gence for Engineering Design, Analysis and Manufacturing,
23(1).
[18] Sasajima, M., Kitamura, Y., Ikeda, M., and Mizoguchi, R., 1995,
FBRL: A Function and Behavior Representation Language. In
Proc. of IJCAI-95, pp. 1830-1836.
[19] Kitamura, Y., Sano, T., Namba, K., and Mizoguchi, R., 2002, A
Functional Concept Ontology and Its Application to Automatic
Identification of Functional Structures. Advanced Engineering
Informatics, 16(2), pp. 145-163.

10 Copyright 2011 by ASME