Vous êtes sur la page 1sur 16

Original Article

Proc IMechE Part C:


J Mechanical Engineering Science
A state–behavior–function model 2019, Vol. 233(7) 2302–2317
! IMechE 2018

for functional modeling of Article reuse guidelines:


sagepub.com/journals-permissions

multi-state systems DOI: 10.1177/0954406218791640


journals.sagepub.com/home/pic

Meng Zhao1 , Yong Chen2, Linfeng Chen3 and Youbai Xie1

Abstract
Since functional modeling can be very useful in conceptual design, it has received considerable attention from engineering
design community. However, existing functional models still cannot effectively assist designers in analyzing the function-
alities of multi-state systems during conceptual design. As a result, designers often have to manually carry out function-
ality analysis according to their experiences, which are not only tedious, but also error-prone. Therefore, this paper
proposes a state–behavior–function model for functional modeling of multi-state systems, which can provide designers
with automated functional analysis support. The approach involves not only a state–behavior–function model for
representing components, but also a model for representing the functionalities of multi-state systems. A prototype
software is then developed for demonstrating the proposed model, with the functional modeling of a peeler centrifuge as
an example to illustrate the proposed functional modeling approach.

Keywords
Conceptual design, computer-assisted design, mechanical design, modeling, system modeling, complex systems

Date received: 7 March 2018; accepted: 29 June 2018

merely employ a single functional diagram to repre-


Introduction sent the functionalities of a system, resulting in that
Functional modeling usually refers to the design activ- they cannot be used to establish explicit functional
ity of developing the models of products or systems models for multi-state systems. Hereby, a multi-state
based on their functionalities and the functionalities system refers to a multi-function system that can
of their components.1 It is widely acknowledged that achieve different functions at different operation
functional modeling can play a significant role in con- (working) states. A simple example of a multi-state
ceptual design,1–3 since functional models not only system is a hair-dryer, which usually has two oper-
allow engineers to share and communicate design con- ation states. One is the state of blowing air, which
cepts across various disciplines, but also can assist includes one function, i.e. the hair-dryer accelerates
them in achieving concept generation and design syn- air, and the other is the state of blowing and heating
thesis. Therefore, considerable functional modeling air, which includes another function, i.e. the hair-
research has been carried out in recent decades, with dryer accelerates and heats air. Note that a multi-
a result of many valuable outcomes.4–13 state system usually relies on different topological
It is found that most of the existing functional structures to achieve different functions.
modeling research has been focused on functional In this paper, we are primarily concerned with the
synthesis, e.g. how to generate solution concepts via development of a functional modeling approach that
functional reasoning. In recent years, researchers are
increasingly realizing the significance of how to pro- 1
School of Mechanical Engineering, Shanghai Jiao Tong University,
vide designers with effective support for function- Shanghai, China
based design analysis.12,14,15 However, existing 2
School of Aeronautics and Astronautics, Shanghai Jiao Tong University,
functional modeling approaches often represent the Shanghai, China
3
functional structure of a system with a static dia- College of Chemistry, Sichuan University, Sichuan, China
gram,15 resulting in that it is often difficult for a com-
Corresponding author:
puter-based tool to recognize the objects in the Yong Chen, School of Aeronautics and Astronautics, Shanghai Jiao Tong
diagram and to reason about the functionalities of University, 800 Dongchuan Road, Shanghai 200240, China.
the whole system. In addition, existing approaches Email: mechenyong@sjtu.edu.cn
Zhao et al. 2303

can be used to achieve automated functional analysis alternatives can be retrieved from a knowledge base.
for multi-state systems. Note that functional analysis In addition, researchers have developed some auto-
is often a critical task for the development of complex mated approaches for assisting designers in generating
systems (e.g. civil aircrafts), since it can fundamentally solution concepts for electro-mechanical or mechan-
determine the safety characteristics of complex sys- ical systems,7,11,18–22 which are also regarded as a
tems. However, due to the lack of an automated ana- stream of functional reasoning research.13
lysis tool, designers often have to manually fulfill The analysis-oriented research is often aimed at the
such complex functionality analysis tasks, which are functional modeling approaches for analyzing and
not only time-consuming, but also error-prone. verifying a solution concept intended to fulfill a func-
In this research, a state–behavior–function-based tion. A pioneering work in this area is undertaken by
approach is proposed for the functional modeling of Stone et al.,23 who have developed an approach for
multi-state systems. A basic assumption is that the predicting failure modes based on a function-failure
solution concept (i.e. system concept) of a multi-state knowledge base. Kurtoglo and Tumer9 have proposed
system has been generated, and the aim of our func- a framework for identifying functional failures and
tional modeling research is to provide a methodology their propagations9 where a behavior simulation is
for designers to establish a formal model for represent- performed on the basis of function structures and con-
ing the functionalities of the whole system concept. figuration flow graphs. Aurisicchio et al.22 have pro-
This paper is organized as follows. The upcoming posed a function analysis diagram to model system
section reviews the related work. Next section pro- functions together with structure,22 which, however,
poses a state–behavior–function model for the func- cannot support automated functional analysis. Sen
tional modeling of system components. Subsequently, and Summers15 have proposed an approach for for-
a state–behavior–function model is developed for rep- malizing function structure graphs so that a computa-
resenting the functionalities of a system. With a peeler tional reasoning tool can be developed to support the
centrifuge as an example, later section illustrates how analysis of a solution concept. Aimed at addressing
the proposed approach can be used to achieve failures that originate from the interaction of physical
the functional modeling of a multi-state system. The and software subsystems, Mutha et al.12 have pro-
final section concludes this paper. posed a multi-domain failure analysis approach
through extending the functional-failure identification
and propagation framework. Jensen et al.24 have
Related work
applied data clustering methods to functional failure
Due to its significance, a large body of functional analysis, which allows a computational tool to reason
modeling-related research has been carried out in about system-level failure behaviors based on the fail-
recent decades, with a result of many valuable out- ures at the component level. In the systems engineering
comes, which can be found in some recent reviews.1,13 area, researchers have developed some SysML-based
Note that in some artificial intelligence approaches to assist engineers in modeling and verify-
research,10,16,17 functional modeling is also called ing the functionalities of complex systems.25,26
functional representation, which is focused on how It can be found that the synthesis-oriented
to model the working processes (i.e. functionalities) approaches are largely for generating solution con-
of a system for achieving desired functions. Due to cepts, which are not suitable for design analysis,
limited space, only some significant approaches while the analysis-oriented approaches have neglected
related to functional modeling are reviewed here. the fact that a multi-state system can change its topo-
Depending on different purposes, existing research logical structures to achieve different functions,
primarily falls into two categories, i.e. the synthesis- which, therefore, cannot effectively analyze function-
oriented research and the analysis-oriented research. alities of a multi-state system. In addition, it should be
The synthesis-oriented research on functional mod- noted that the concept of state or state transformation
eling is aimed at the approaches for transforming a in existing functional modeling research is largely dif-
desired function into a solution concept. Besides the ferent from that in our research. For example, in the
function–behavior–structure model,5 and the state– function–behavior–state model6 a state of an entity is
behavior–function model,10 most of such synthesis- defined by the attributes of the entity (e.g. weight) and
oriented research is rooted in the systematic approach its relations with external entities. In contrast, our
of engineering design,15 where functional modeling concept of state, as defined later in this paper, is
primarily refers to the process of decomposing an more concerned with the internal topology of an
overall function into sub-functions (and even sub- entity (e.g. component), i.e. how the internal elements
sub-functions if necessary), that of searching for suit- of an entity are connected with each other. Since
able solution principles for the sub-functions, and that multi-state systems are very common in complex sys-
of combining the solutions of sub-functions into a tems, it is then necessary to develop a formal
whole solution concept to address the overall function. approach for achieving the functional modeling of
For example, Umeda et al.6 have proposed the func- multi-state systems, so as to achieve automated func-
tion–behavior–state model, based on which solution tional analysis of multi-state systems.
2304 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

Functional modeling of components multiple states of a component as a set of state values.


A system can usually be regarded as an aggregate of For example, the switch of a hair-dryer often has three
components. To achieve the functional modeling of a states, i.e. off, on_1, and on_2. When the switch is in
system, it is indispensable to have a model for repre- state on_1, at which the hair-dryer merely blows air,
senting the components that constitute the system. In its state can then be represented as: sa(c_s-
this research, a component is modeled as an object witch) ¼ ‘‘on_1’’; when the switch is in state on_2, at
that may have multiple states, and at each state, the which the hair-dryer can both blow and heat air, its
object can achieve a specific function. Therefore, a state can then be represented as: sa(c_switch) ¼ ‘‘on_2’’.
state–behavior–function model is developed here for Note that the concept of state here is different from
representing the functionalities of a component. that in the function–behavior–state model,6 where a
state of an entity is defined by the attributes of the
entity (e.g. weight) and its relations with external enti-
Component
ties. In contrast, the concept of state in our research is
In our functional modeling research, a component is regarded as an abstraction of the internal configur-
regarded as a basic object of a system, which has a ation of an entity, which has no relation with external
name and a set of attributes to indicate what the com- entities. In addition, the concept of state in this
ponent is. Note that a component can be physically research is also different from that in the structure–
disassembled into some primitive elements, which, behavior–function model,10 where state is employed
however, are not discussed here since they are not to refer to the state of a functional flow, rather than
the concern of functional modeling. For example, a that of a component. Since a multi-state system relies
light bulb is often treated as a basic component in the on different states of components to achieve different
functional modeling process, which, however, can functions, it is necessary to represent the states of a
actually be disassembled into some primitive elements component in a system.
such as wire filament, glass cover, etc. In addition, it
should be noted that the physical form (i.e. appear-
ance) of a component is also not considered in this
Component behavior
research, since the form representation is also not In this research, a behavior of a component refers to a
necessary for functional modeling. change of a component from one state to another. In
To allow functional modeling, the representation a technical system, a component behavior is often
of a component also has some ports to receive input related to a control input, which can be called as a
flows and to release output flows. According to Pahl trigger. For example, a behavior of a solenoid (i.e.
et al.,27 the input or output flows here can be classified electromagnetic) valve (as shown in Figure 1) can
as material, energy, and (or) information flows. Note change from the state closed to the state left_open,
that a component probably does not have a real port when the left solenoid of the valve receives electricity
(such as a hole in a pipe), which means that a port of a as a trigger. Note that a trigger of a component
component may just be fictitious for the convenience cannot be treated as a functional input flow of the
of representing the functionalities of the component. component, since the trigger, instead of being changed
For example, a gear box does not have real ports for by the component, merely leads to the change of the
transmitting torque, where its input or output port state of component. In contrast, the traditional func-
actually corresponds to the input or output shaft. tional modeling approaches do not differentiate the
control inputs of a component from the functional
inputs, resulting in that they cannot explicitly repre-
Component state sent the state changes of the component.
A state of a component refers to the conceptualization Based on the above analysis, a component behav-
of a configuration of the elements in the component, ior, b, can then be represented as a triple, <t, sai, saf>,
which involves not only what elements the component where, t refers to the trigger, sai denotes the initial
has, but also the connection relations among these state of the component, and saf indicates the final
elements, as well as the physical states of these elem-
ents. Therefore, a state here can also be called a con-
figuration state. For example, an electrical switch can
have two states, i.e. on and off, while the internal con-
nection relation among the elements of the switch in
the on state is obviously different from that of the
switch in the off state.
In this research, a state of a component is repre-
sented as a state attribute of the component,
i.e. sa(c) ¼ val, where c refers to the component,
sa denotes the state attribute, and val is the state
value. As a result, it is then possible to represent the Figure 1. The behavior of a solenoid valve.
Zhao et al. 2305

state. Note that it is possible that a component may However, the above component function represen-
allow multiple triggers, each of which corresponds to tation has not taken into consideration the fact that a
a different component port to receive an independent component can have multiple states for achieving dif-
control input. Therefore, it is more reasonable to rep- ferent functions. In view of this, all possible functions
resent a behavior as a quadruple, i.e. < t, pb, sai, saf > , of a component should then be represented as a func-
where pb refers to a behavior-related port for receiving tion set, i.e. {fj}. Hereby, each function, fj, can be fur-
a trigger (i.e. a control input). With the aforemen- ther represented as a triple, (saj, Ij, Oj), where, saj
tioned solenoid valve as an example, a behavior of denotes the jth value of the state attribute sa (i.e.
the component can then be represented as <t_DC, the jth state) of the component, Ij represents a set of
p_EleLeft, s_Closed, s_LOpen>, which means that input flows at the corresponding state, i.e. {(ijk, pjk)},
when the valve receives a trigger t_DC from port and Oj denotes a set of output flows at this state, i.e.
p_EleLeft (i.e. the electrical port at the left side), its {(ojt, pjt)}. Still with the solenoid valve as an example,
state will change from the state s_Closed to s_LOpen a possible function f can then be represented as
(i.e. left open). (s_LOpen, {(f_liquid, p_A)}, {(f_liquid, p_L)}), which
Note that the concept of behavior here is funda- means that when the valve is in the state of s_LOpen,
mentally different from those in the existing func- the input flow f_Liquid at port p_A will be transferred
tional modeling research. For example, in the to the output flow f_Liquid at port p_L. Another func-
structure–behavior–function model,10 a behavior tion of this valve can be represented as (s_ROpen,
refers to a state change of a functional flow, which {(f_liquid, p_A)}, {(f_liquid, p_R)}), which means
is not directly related to a component. In contrast, the that when the valve is in state s_ROpen, the input
concept of behavior in our research refers to the state flow f_Liquid at p_A will be transferred to the
change of a component itself, which does not involve output flow f_Liquid at port p_R. It can be found
a functional flow. With an electrical switch as an that a salient feature of the proposed functional rep-
example, the behavior concept in the structure–behav- resentation approach consists in that it allows the rep-
ior–Function can be represented with the location resentation of different functional models of a
change of electricity (i.e. the functional flow), while component at different states, which is different
the behavior of the switch in our research is repre- from the previous research, where a component
sented as the state change from off to on or from on merely has one function.
to off. It is evident that our behavior concept allows
the representation of the state transition of a compo-
Schema for representing components
nent, which is necessary for the functional modeling
of multi-state systems. Based on the above analysis, a schema can then be
proposed to represent the state, behavior, and func-
tion of a component, as shown in Figure 2. In the
Component function figure, the aforementioned solenoid valve is also
According to Pahl et al.,27 a component function can employed as an example to illustrate the component
be regarded as an abstract action that a component representation schema. Figure 3 shows the state–
imposes on one or multiple flows to change the flow(s) behavior–function model of a component, where the
from input to output. The (functional) flows here can component can be in the state of A, B, or C, and can
often be classified as material flows, energy flows, or change its state to deliver different functions (i.e. func-
information flows. Note that unlike a material flow, tion1 and function2).
an energy or information flow may often be an Note that the concept set for the functional model-
abstract flow and, therefore, cannot be directly ing of components here is largely different from that
observed. For example, although the function of a in SysML. Although our concept of state is similar to
gear box is often regarded as to transmit torque, that in SysML, the concepts of behavior and function
what can be directly observed is the rotation of the are fundamentally different from those in SysML. Our
input shaft and that of the output shaft, rather than concept of behavior is more like the concept of state
the torque. transition in SysML, while SysML does not have a
Based on the above understanding, a function of a formal definition for function, which, though, is neces-
component can then be represented as a binary, sary for functional modeling. In addition, SysML fails
i.e. < I, O>. Here, I represents a set of functional to provide a mechanism to deal with the change of
flows at the input ports, i.e. {(ik, pk)} (k ¼ 1, 2,. . ., configuration relations (discussed in the ‘‘System
m), where ik is the kth input flow at port pk, while O state’’ section), which is often necessary for represent-
refers to a set of flows at the corresponding output ing the functionalities of a multi-state system.
ports, i.e., {(ot, pt)} (t ¼ 1, 2,. . ., n), where ot is the tth
output flow at port pt. Note that it is often necessary
to further represent a flow with some qualitative attri-
Functional modeling of systems
butes and quantitative parameters, which is not ela- A system is an assembly of different components that
borated here for conciseness. work together to achieve a desired function. It should
2306 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

Figure 2. The schema for representing components.

be noted that in the functional modeling research, it is where sys represents a system of interest and sa
usually assumed that the components in a system do refers to the state attribute. For example, when a
not directly interact with each other, but act on the Lighting System (LS) is in the normal electrical con-
flows between these components. Although such an nection state, its state can be represented as:
assumption probably does not accord with the situ- sa(sys_LS) ¼ ‘‘normal’’.
ation in the physical world, it is often acceptable for However, a state attribute alone is often insufficient
functional modeling and analysis. From the perspec- to explicitly represent the complete state of a complex
tive of functional modeling, the concepts of state, system, which is often an assembly of components and
behavior, and function of a multi-state system are sub-systems. For the purpose of functional analysis in
elaborated as below. conceptual design, a detailed system state should deal
with not only the composition of a system, but also
the state attributes of its components and (or) subsys-
System state
tems, as well as the configuration relations among
Similar to the state concept of a component, a state of these components and subsystems. Therefore, we add-
a system refers to a configuration of the system’s itionally employ a triple to represent the detailed state
internal objects (i.e. components or subsystems) in a of a system, i.e. (C, SA, R). Hereby, C stands for the
specific situation. object composition of the system, SA represents the
Just like a component state, a system state can also state attribute set of the related objects, and R refers
be represented with a state attribute, i.e. sa(sys), to the configuration relation set.
Zhao et al. 2307

Figure 3. The state–behavior–function model of components.

The object composition (C) of a system means functional modeling research has dealt with.5,15
what objects the system has, i.e. what components With a hair-dryer as an example, Figure 4 gives
and (or) subsystems are in the system. Note that the both a physical structure diagram (the upper one)
material/energy/information objects that flow and a functional diagram (the lower one) of a hair-
through a system are not treated as a part of the com- dryer. It can be found that in the upper (i.e. physical)
position of a system. Therefore, the composition of a diagram, the power source (i.e. the plug) is directly
system can then be represented with the union set of a connected with the heater (R) and the motor (M),
component set and a subsystem set, i.e. C ¼ {ci}[{sj}, and the heater is connected with the motor in a par-
where ci is a component, and sj is a subsystem, which allel manner; however, in the lower (i.e. functional)
can be further decomposed into sub-subsystems and diagram, the plug is not directly connected with the
(or) components. Note that a subsystem probably heater and the motor, and the heater is just connected
does not exist for a simple system. For example, it is with the motor in a sequential manner. Therefore, the
often not necessary to decompose a hair-dryer into configuration relations in a system are often largely
some subsystems for functional analysis. Therefore, different from the structural or physical relations in a
the composition of a hair-dryer can simply be repre- system, which means that the (configuration) state of
sented as a component set, i.e. {c_plug, c_switch, a system is also largely different from a structural
c_motor, c_heater, c_fan, c_wire_1,. . .}. state of a system.
The state attribute set (SA) is an aggregate of the In this research, a configuration relation is repre-
state attributes of the objects in a system at a time. sented as an association, i.e. (<oi, pik>, < oj, pjt>),
Assuming that a system is decomposed into some where oi denotes the source object that outputs a
components and subsystems, where the components flow, pik refers to the source object’s port where the
do not belong to the subsystems, an object state set of output flow goes out, oj represents the destination
the system can then be represented as a union set, i.e. object of the flow coming from pik, and pjt is the
SA ¼ {sa(ci)}[{sa(sj)}, where sa(ci) denotes the state port of the destination object that the flow comes in.
attribute of component i, and sa(sj) refers to the Here, an object can be a component, a subsystem or
state attribute of subsystem j. Note that the state attri- an environmental object of a system. With the config-
bute of a subsystem here, should be associated with a uration diagram shown in Figure 4 as an example,
detailed state representation of the subsystem, which some configuration relations can be represented as:
allows a design engineer to understand the detailed (<environment, null>, < c_plug, pin>); (<c_plug,
state of the subsystem. Therefore, the state attribute pout>, < c_switch, pin>); (<c_switch, pout>, < c_motor,
of a subsystem can also be regarded as a state index of pin>); (<c_switch, pout>, < c_heater, pin>);
the complete state representation of the subsystem. (<c_motor, pout>, < c_fan, pin_rotation>); (<e_air,
The configuration relation set (R) is a set of rela- null>, < c_fan, pin_air>); etc. It can be found that the
tions that describe how the components of a system proposed model can explicitly represent different con-
are functionally connected with flows, i.e. from which figuration relations of a multi-state system. In con-
component the input flows come, and to which com- trast, the existing approaches (e.g. SysML) can
ponent the output flows go. Note that a configuration merely represent a single configuration relation for a
relation here is largely different from a structural or system, regardless of the system being a single-state
physical connection relation, which the previous system or a multi-state system.
2308 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

Figure 4. A comparison of the physical diagram and the functional diagram of a hair-dryer.

Note that it is usually necessary to combine the The trigger set (T) of a system is composed of a set
state attribute representation and the detailed model of triggers. Since a system may have multiple compo-
representation together to represent a state of a nents and (or) subsystems, and not all components or
system, i.e. (sai, < Ci, SAi, Ri > ). Hereby, sai refers subsystems can accept control inputs (i.e. triggers),
to the ith value of the state attribute of a system, the destination of a trigger needs to be made clear
which can be treated as a description of the system in a trigger representation. Therefore, a trigger of a
state at a higher level, while Ci means the object com- system can be represented as a triple, i.e. (ti,, oi, pBij),
position of the system at the state sai, SAi represents where, ti refers to the trigger, oi represents a system
the corresponding state attribute set, and Ri denotes object (a component or a subsystem) that receives the
the corresponding configuration relation set. With trigger, and pBij is the control port of the system object
such a combined representation, it is then possible to receive the trigger. As a result, the trigger set (T) of
for a computational system to reason from the state a system can be represented as {(ti,, oi, pBij)}. For
attribute representation of a system to the detailed example, to start the engine of a car, a driver usually
state representation of its components, and vice versa. needs to step on the brake and turns the key to the
start position. Here, the two triggers can be repre-
sented in a set: {(f_force, c_brake, pB_foot),
System behavior
(h_momentum, c_key, pB_hand)}, where f_force refers
Similar to the concept of component behavior, a to the force imposed by the foot of the driver, and
system behavior refers to that a system changes h_momentum is the momentum imposed by the hand
from one state to another, which is usually induced of the driver.
by some triggers (i.e. control inputs). Since a system Theoretically speaking, the initial state (SI) and
may have multiple components and (or) subsystems the final state (SF) of a system should be repre-
that have multiple (configuration) states, a system can sented with the state attribute (i.e. sa) and the
have multiple triggers to allow these components and detailed state model (i.e. < C, SA, R>). However,
(or) subsystems to change their states. Therefore, a since there is a one-to-one corresponding relation
behavior of a system can also be represented as a between the state attribute representation and the
triple, <T, SI, SF>, where T refers to a set of triggers, detailed state model, and all possible detailed
and SI and SF represents the initial state and the final states should be represented with the detailed state
state of the system, respectively. representation approach, it is then not necessary to
Zhao et al. 2309

employ the detailed state representations in the


System function
system behavior representation, which means that Similar to a component function, a system function is
the initial state (SI) and the final state (SF) of a regarded as an abstract action of a system at a specific
system can then be simplified as two states, i.e. saI state, which transforms one or multiple input flows
and saF. Therefore, the representation of a system into some output flows. Therefore, a system function,
behavior can then be simplified as a triple, i.e. (T, fj, can also be represented as a triple, (sj, Ij, Oj), where,
saI, saF), where T is a set of triggers of behaviors of sj denotes a specific state of a system, Ij represents a
the internal objects of a system. A trigger in T can set of input flows at this state, i.e. Ij ¼ {(ijk, pjk)}, and
be represented as (ti,, oi, pBij), which means that the Oj denotes a set of output flows at this state, i.e.
object oi can change its state if it receives the trigger Oj ¼ {(ojt, pjt)}. Similar to the system behavior repre-
ti at port pBij. When a system receives a set of trig- sentation, the system state (i.e. sj) in the above
gers, i.e. {(ti,, oi, pBij)}, objects in the system can representation can also be simplified as a state attri-
receive the triggers. These objects can then change bute (i.e. saj). Note that the input and output flows of
their states according to the component behaviors, a system, i.e. Ij and Oj, need to be further allocated to
i.e. < ti, pBij, sai_Initial, sai_Final > . Figure 5 illustrates the corresponding ports of the internal objects of the
the relation between a behavior of a system and system, which will be elaborated later in the subse-
behaviors of the objects in the system. quent section.
Based on the above analysis, it is then possible to For the analysis of a complex system, the above
explicitly represent a behavior of a system. Still with functional representation approach allows a design
the hair-dryer mentioned in the previous subsection as engineer to simplify the representation of a subsys-
an example, when a person turns on a hair-dryer (i.e. tem’s function as some input flows and output
change the position of the switch from off to on_1, the flows, and thus can avoid representing the complex
behavior of the hair-dryer can then be represented as: interactions among the components in the subsystem.
({(h_force, c_switch, pB_hand)}, off, on_blowing), where If a subsystem in itself is very complex and is not the
{(h_force, c_switch, pB_hand)} refers to a trigger (i.e. primary concern of system analysis, such a system
h_force) by a hand force on the switch, off means function representation can also alleviate unnecessary
that the initial state of the hair-drying system is off, system modeling work. For example, since the power
and on_blowing indicates that the final state of the source of a system often refers to a complex electricity
hair-drying system is on_blowing. With the system grid, it is neither necessary nor possible for a system
behavior representation, it is then possible to repre- engineer to explicitly represent the electricity grid;
sent a behavior of a complex system in a hierarchical therefore, the electricity grid can often be simplified
manner, and thus can simplify the behavior represen- as a subsystem that merely outputs electricity to a
tation of a complex system. system of interest.

Figure 5. A behavior of a system and behaviors of the objects in the system. (a) The behavior of a system, (b) The behaviors of the
internal objects of the system.
2310 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

System function allocation the allocation relations from the system level to the
Although the function of a system can be represented component level can be represented as: {(sys_HD, p1,
with the model illustrated in the previous section, a ele_energy, c_Plug, p1), (sys_HD, p2, air, c_Inlet, p1),
system engineer may also be required to analyze a (sys_HD, pc1, h_force, c_Switch, pc1), (sys_HD, p2,
system (or a subsystem) at a more detailed level (e.g. acce&heated_air, s_HD, p2)}. Note that there is also
at the component or subsystem level). In such a situ- a simplified system behavior included in the diagram,
ation, it is then necessary to break down the system i.e. the switch changes from the state of ‘‘off’’ to the
into components and (or) subsystems (i.e. open the state of ‘‘on’’, which is necessary for the hair-dryer to
black box of a system), and then model the functional work.
processes about how these components or subsystems With a typical system as an example, Figure 7
interact with the input flows. Since the representation shows the state–behavior–function model of the
of a system function has merely stated the input and system, where it is assumed that the system can be
output flows at the system level, it is then necessary to in the state of S0 or S1. As a result of the state tran-
hierarchically allocate the inputs and outputs of a sition from S0 to S1, the function that the system in
system to the component or subsystem level, so that Figure 7 fulfills can change, i.e. transforming the input
it is possible for the components or subsystems of the flow (flow_x) of the system to flow_z, instead of
system to act on the flows. flow_y.
In this research, such a functional allocation rela-
tion is represented as a binary group, i.e. (sys, ps, f,
obj, po), which means that the flow (f) of the system
Test
(sys) at port (ps), is allocated to the object (obj) at port Based on a commercial software, some customization
po, i.e. f(ps (sys)) ¼ f(po (obj)). Still with the hair-dryer development has been carried out to implement the
mentioned before as an example, the system configur- state–behavior–function-based approach for func-
ation and its breakdown are shown in Figure 6. Here, tional modeling, with the result of a prototype

Figure 6. The function allocation of a hair-dryer.


Zhao et al. 2311

Figure 7. The state–behavior–function model of systems.

software for achieving functional modeling. The


Functional modeling
prototype software not only allows designers to Peeler centrifuges are very common in manufacturing
achieve function modeling of multi-state systems, and waste-processing industries, which can separate
but also supports functional simulations based on solid particles from liquid. A typical peeler centrifuges
the models of the systems. It is primarily composed can be in five working modes, i.e. feeding, filtrating,
of three subsystems, i.e., the composition manage- washing, spinning, and discharging. After some liquid
ment subsystem, which allows engineers to manage with solid particles is fed into a peeler centrifuge, the
the composition of a multi-state system; the func- peeler centrifuge will start filtrating, through the high-
tional modeling subsystem, which is for engineers to speed rotation of the centrifugal subsystem, which
model the states, behaviors, and functions of compo- pushes the liquid towards a filter layer (medium). As
nents, as well as those of subsystems that design a result, the solid particles can be deposited on the
engineers are concerned with; and the functional filter layer, but the remaining liquid on the surface
simulation subsystem, which can perform functional of the solid particles still needs to be eliminated.
simulation based on the functional models. A peeler Thereafter, the state of washing follows, where deter-
centrifuge is employed here as an example to demon- gent is fed into the peeler centrifuge to flush the filter
strate how to achieve functional modeling and simu- layer. After the detergent is drained, the peeler centri-
lation with the aid of the prototype software. fuge starts spinning to eliminate the detergent.
2312 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

Finally, in the state of discharging, the solid particles conciseness. As shown by the red ellipse in this
on the filter medium are scraped with the peeler and figure, the connector between p_PC_Elec and
discharged from the peeler centrifuge. In this example, p_HCS_Elec indicates the allocation relation between
it is assumed that the solution concept of a peeler the peeler centrifuge at a higher level and the hydrau-
centrifuge has been generated. The primary compos- lic control subsystem at a lower level. This allocation
ition of the peeler centrifuge is shown in Figure 8. relation dictates that electrical energy flowing into
Among the six subsystems, the hydraulic control peeler centrifuge through p_PC_Elec would be allo-
subsystem plays a crucial role in the state transitions cated to hydraulic control subsystem at p_HCS_Elec,
of the peeler centrifuge. Figure 9 shows the allocation which can be represented as (Peeler_centrifuge,
and the configuration relations in which the hydraulic p_PC_Elec, f_Elec, Hydr_control_subsystem, p_HCS_
control subsystem is involved, while the other rela- Elec). As shown by the blue ellipse Figure 9, the con-
tions of the peeler centrifuge are not displayed for nector between p_HCS_Kinetic and p_DS_Kinetic indi-
cates a configuration relation, i.e. (<Hydr_control_
subsystem, p_HCS_Kinetic>, < Discharging_subsys-
tem, p_DS_Kinetic>). This configuration relation
allows the hydraulic control subsystem to provide the
discharging subsystem with kinetic energy, so that the
peeler in the discharging subsystem can work. The con-
figuration relations between the hydraulic control sub-
system and the feeding subsystem indicate that
hydraulic control subsystem can deliver triggers to
the control ports of feeding subsystem, which allows
feeding subsystem to change its state. Likewise, the
configuration relations between the electric control
subsystem and the hydraulic control subsystem indi-
cate that electric control subsystem can deliver triggers,
which set off state transitions of hydraulic control
subsystem.
Within the hydraulic control subsystem, there are
three solenoid valves, each of which has a control port
to receive triggers allocated from the subsystem, as
shown in Figure 10. Each valve fulfills a different
function, as the state of the valve changes as a result
of the trigger. According to the function illustrated in
Figure 2, solenoid valve for inlet can deliver a hydrau-
lic flow through either p_SVI_Mix or p_SVI_Dtg,
Figure 8. The primary composition of the peeler centrifuge. depending on the trigger it receives at port

Figure 9. The allocation and configuration relations in which the hydraulic control subsystem is involved.
Zhao et al. 2313

Figure 10. The allocation and configuration relations of the hydraulic control subsystem.

Figure 11. The user interface for functional modeling of solenoid valve for inlet.

p_SVI_Control. The user interface for functional mod- a multi-state system is provided, functional simulation
eling of solenoid valve for inlet is illustrated in Figure can then assist designers in predicting how the system
11. In this manner, solenoid valve for inlet is intended will operate and when the states of some components
to control the feeding subsystem to allow the input of will change. Based on the functional model of the
either the mixture or the detergent. Likewise, solenoid peeler centrifuge developed above, functional simula-
valve for outlet is intended to allow the output of tion can be carried out to demonstrate how the proto-
either the liquid or the detergent in the discharging sub- type software can support automated system
system. Solenoid valve for peeling delivers a hydraulic functional analysis.
flow to Cylinder at either p_C_Forward or The functional simulation begins with setting the
p_C_Backward, so that the cylinder can output a bidir- initial state of the system. In this example, the peeler
ectional translation to enable the peeler in the dischar- centrifuge is assumed to be in the state of filtrating,
ging subsystem to engage or retract. which means that the centrifugal subsystem rotates
liquid mixed with solid particles and then the dischar-
ging subsystem expels the liquid from the peeler cen-
Functional simulation
trifuge. Figure 12 shows the initial state of the peeler
As noted by Tomiyama et al.,14 functional simulation centrifuge, where the blue rectangles refer to the
can be a practical application of functional modeling components of a system or subsystem that are in
research, since computer-based functional simulation operation, while the white rectangles refer to the com-
tools can be employed to perform validation during ponents that are not working. Note that a system or
conceptual design and to identify overlooked design subsystem in operation is not filled with blue color,
failures. When the state–behavior–function model of which is different from a component in operation. In
2314 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

Figure 12. The initial state of the peeler centrifuge.

Figure 13. The intermediate state of the peeler centrifuge.


Zhao et al. 2315

Figure 14. The final state of the peeler centrifuge.

this figure, the black lines with arrows indicate how trigger of its behavior as shown in Figure 11. In the
the flows are transferred from the output ports of pre- state DtgOpen, the valve delivers hydraulic energy
vious objects (components or subsystems) to the input f_Hydr through the port p_SVO_Dtg. The configur-
ports of the following objects. Note that the electric ation and allocation relations can then be employed to
control subsystem and some components have been transfer this hydraulic flow to the control port of the
omitted in this figure for conciseness. In this state, component hydraulic valve for out_Dtg. As a result,
hydraulic control subsystem receives the electric the component hydraulic valve for out_Dtg changes its
signal (f_LiqOpen) as a trigger that leaves solenoid state from Closed to Open, thus activating the func-
valve for outlet in the state LiqOpen. As a result, the tion of transferring a liquid flow from input to output.
hydraulic flow (f_Hydr) is sent to the control port In this manner, the signal f_DtgOpen makes the peeler
p_DS_contLiq of discharging subsystem, so that the centrifuge to change to an intermediate state between
liquid flow (f_Liq) can be expelled from port filtrating and washing, as shown in Figure 13.
p_PC_Liq. Thereafter, the functional analysis engineer can also
After the initial state of the peeler centrifuge has send another trigger to enable the prototype software
been set, a functional analysis engineer can then to simulate a subsequent state transition of the peeler
release some triggers to observe how the peeler cen- centrifuge. For example, when the engineer sends an
trifuge changes its state. In the prototype software, for electric signal flow f_DtgOpen to the port
example, a designer can simulate the process that the p_HCS_contInlet of hydraulic control subsystem,
state of the peeler centrifuge changes to the state of which is then allocated to the control port of the com-
washing, through sending the electric signal ponent solenoid valve for inlet, the state of the solenoid
f_DtgOpen to the control port p_HCS_contOutlet of valve is then changed as DtgOpen. As a result, the func-
the hydraulic control subsystem. Note that the tion of this valve is activated, which means that
signal LiqOpen at port p_HCS_contOutlet shown in hydraulic control subsystem then outputs hydraulic
Figure 12 has been removed prior to sending signal flows through the port p_HCS_inDtg, as well as the
f_DtgOpen. The allocation relation between the sub- port p_HCS_outDtg. Thereafter, the component
system and the solenoid valve enables the signal hydraulic valve for in_Dtg, whose trigger is a hydraulic
f_DtgOpen to arrive at the control port of the com- flow, changes its state from Closed to Open, so that
ponent solenoid valve for outlet. Once receiving the centrifugal subsystem receives the flow of detergent.
signal, solenoid valve for outlet changes its state from Finally, the component hydraulic valve for out_Dtg ful-
Closed to DtgOpen, since the signal f_DtgOpen is the fills the function of transferring flows from input to
2316 Proc IMechE Part C: J Mechanical Engineering Science 233(7)

output, as shown in Figure 14. At this time, peeler component can change over time if the component
centrifuge outputs the flows of detergent and liquid has a timer in it. To address this issue, our future
(f_Dtg & f_Liq), which means that the state of the work will incorporate a time-dependent model into
peeler centrifuge changes to washing as a result of the the state–behavior–function model reported in this
signal f_DtgOpen sent by the functional analysis research.
engineer.
From the above example, it can be found the Declaration of Conflicting Interests
prototype software can simulate how a multi-state The author(s) declared no potential conflicts of interest with
system changes its state to deliver different functions respect to the research, authorship, and/or publication of
when the system receives some external triggers this article.
(i.e. control signals). It can also be found that the
proposed state–behavior–function approach allows Funding
designers to model the functionalities of a system in
The author(s) disclosed receipt of the following financial
a hierarchical way, so that the functional simulation
support for the research, authorship, and/or publication
process can either work on the system level or on the of this article: This work was supported by the Ministry
component level. of Industry and Information Technology of the People’s
Republic of China.
Conclusions
Since functional models can be very useful for con- ORCID iD
ceptual design, a large amount of research has been Meng Zhao http://orcid.org/0000-0002-5302-8272
undertaken in the functional modeling area. However,
little research has been carried out on how to assist References
designers in achieving the functionality analysis of 1. Erden MS, Komoto H, Van Beek TJ, et al. A review of
multi-state systems during conceptual design. function modeling: Approaches and applications. AI
Therefore, this research attempts to develop a state– EDAM 2008; 22: 149–169.
behavior–function model for achieving the functional 2. Kuttig D. Potential and limits of functional modelling
in the CAD process. Res Eng Des 1993; 5: 40–48.
modeling of multi-state systems.
3. Maarten Bonnema G and Van Houten FJ. Use of
Since a system can be regarded as an assembly of
models in conceptual design. J Eng Des 2006; 17:
components, it is indispensable to achieve functional 549–562.
modeling of the components first. With the proposed 4. Hunt JE, Pugh DR and Price CJ. Failure mode effects
approach, the states, behaviors, and functions of analysis: A practical application of functional modeling.
components can be explicitly represented, establish- Appl Artif Intell 1995; 9: 33–44.
ing functional models that allow components to be 5. Qian L and Gero JS. Function–behavior–structure
in different states and achieve different functions. paths and their role in analogy-based design. AI
Based on the functional models of components, con- EDAM 1996; 10: 289–312.
figuration relations between components (subsys- 6. Umeda Y, Ishii M, Yoshioka M, et al. Supporting con-
tems), along with allocation relations between ceptual design based on the function-behavior-state
modeler. AI EDAM 1996; 10: 275–288.
systems (subsystems) and components have been
7. Chakrabarti A and Bligh TP. A scheme for functional
employed to establish functional models of multi-
reasoning in conceptual design. Design Stud 2001; 22:
state systems, so that the states, behaviors and func- 493–517.
tions of the systems can be explicitly represented. As 8. Campbell MI, Cagan J and Kotovsky K. The A-Design
demonstrated by the example of the peeler centri- approach to managing automated design synthesis. Res
fuge, the state–behavior–function models above Eng Des 2003; 14: 12–24.
allow a computer-aided tool to perform a functional 9. Kurtoglu T and Tumer IY. A graph-based fault identi-
simulation to support functionality analysis of multi- fication and propagation framework for functional
state systems. design of complex systems. J Mech Des 2008; 130:
A potential use of the state–behavior–function 051401–8.
models involves functional failure analysis during 10. Goel AK, Rugaber S and Vattam S. Structure, behav-
ior, and function of complex systems: The structure,
conceptual design. If a functional failure occurs to a
behavior, and function modeling language. AI EDAM
component (subsystem) within a system, a functional
2009; 23: 23–35.
simulation can be conducted based on the functional 11. Chen Y, Liu Z-L and Xie Y-B. A knowledge-
model of the system, so as to support the analysis of based framework for creative conceptual design of
the impact of the failure on the functionality of the multi-disciplinary systems. Comput-Aid Des 2012; 44:
whole system. Note that the proposed model does not 146–153.
consider the fact that the states of a component can 12. Mutha C, Jensen D, Tumer I, et al. An integrated multi-
change as a result of physical phenomena that occurs domain functional failure and propagation analysis
within the component, which is also a drawback of approach for safe system design. AI EDAM 2013; 27:
the current work. For example, the state of a 317–347.
Zhao et al. 2317

13. Eisenbart B, Gericke K and Blessing L. An analysis of 22. Aurisicchio M, Bracewell R and Armstrong G. The
functional modeling approaches across disciplines. AI function analysis diagram: Intended benefits and coex-
EDAM 2013; 27: 281–289. istence with other functional models. AI EDAM 2013;
14. Tomiyama T, Van Beek TJ, Alvarez Cabrera AA, et al. 27: 249–257.
Making function modeling practically usable. AI 23. Stone RB, Tumer IY and Van Wie M. The function-
EDAM 2013; 27: 301–309. failure design method. J Mech Des 2005; 127: 397–407.
15. Sen C, Summers JD and Mocko GM. Physics-based 24. Jensen DC, Bello O, Hoyle C, et al. Reasoning about
reasoning in conceptual design using a formal represen- system-level failure behavior from large sets of func-
tation of function structure graphs. J Comput Inform tion-based simulations. AI EDAM 2014; 28: 385–398.
Sci Eng 2013; 13: 011008–12. 25. Mhenni F, Choley J, Riviere A, et al. SysML and safety
16. Iwasaki Y and Simon HA. Causality in device behavior. analysis for mechatronic systems. In: 9th France_Japan
Artif Intell 1986; 29: 3–32. & 7th Europe_Asia Congress on mechatronics and
17. Chandrasekaran B and Josephson JR. Function in research and education in mechatronics (REM), 2012,
device representation. Eng Comput 2000; 16: 162–177. pp.417–424. New York: IEEE.
18. Bracewell RH and Sharpe J. Functional descriptions 26. Cao Y, Liu Y, Fan H, et al. SysML–based uniform
used in computer support for qualitative scheme behavior modeling and automated mapping of design
generation—‘‘Schemebuilder’’. AI EDAM 1996; 10: and simulation model for complex mechatronics.
333–345. Comput–Aid Des 2013; 45: 764–776.
19. Campbell MI, Cagan J and Kotovsky K. Agent-based 27. Pahl G, Beitz W, Feldhusen J, et al. Conceptual design.
synthesis of electromechanical design configurations. In: Engineering design: A systematic approach. 3rd ed.
J Mech Des 2000; 122: 61–69. London: Springer, 2007, pp.159–225.
20. Sridharan P and Campbell MI. A study on the gram-
matical construction of function structures. AI EDAM
2005; 19: 139–160.
21. Kurtoglu T, Swantner A and Campbell MI. Automating
the conceptual design process: ‘‘From black box to
component selection’’. AI EDAM 2010; 24: 49–62.

Vous aimerez peut-être aussi