ArticlePDF Available

A Multimodeling Approach for the Simulation of Energy Consumption in Manufacturing

  • PHWT Private University of Applied Sciences

Abstract and Figures

In the design of manufacturing systems the consideration of resource usage, especially energy consumption, is getting more attention. However, the inclusion of all relevant physical processes in a unified modeling approach is a non-trivial task, if detailed analyses are required. The commonly used modeling approach for manufacturing systems is the discrete event modeling technique. However, models of physical processes are often continuous in nature and are modeled using ordinary differential equations or differential algebraic equations. Indeed, the investigation of such physical processes in manufacturing systems often demands a more specific consideration of process control operations, which are favorably modeled using state machines. To combine those different paradigms a multimodeling approach for manufacturing systems is proposed. The approach is illustrated by the example of a production line with an industrial furnace facility.
Content may be subject to copyright.
SNE 27(2) 6/2017 115
A Multimodeling Approach for the Simulation of
Energy Consumption in Manufacturing
Thorsten Pawletta1*, Artur Schmidt1, Peter Junglas2
1Research Group Computational Engineering & Automation, Wismar University of Applied Sciences,
Philipp-Müller-Straße 14, 23966 Wismar, Germany; *
2PHWT Vechta/Diepholz/Oldenburg, Rombergstraße 40, 49377 Vechta, Germany
Abstract. In the design of manufacturing systems the
consideration of resource usage, especially energy con-
sumption, is getting more attention. However, the inclusion
of all relevant physical processes in a unified modeling
approach is a non-trivial task, if detailed analyses are re-
quired. The commonly used modeling approach for manu-
facturing systems is the discrete event modeling technique.
However, models of physical processes are often continuous
in nature and are modeled using ordinary differential equa-
tions or differential algebraic equations. Indeed, the investi-
gation of such physical processes in manufacturing systems
often demands a more specific consideration of process
control operations, which are favorably modeled using state
machines. To combine those different paradigms a multi-
modeling approach for manufacturing systems is proposed.
The approach is illustrated by the example of a production
line with an industrial furnace facility.
The modeling and simulation of manufacturing systems
has been a subject of study for several decades. Accord-
ing to [1], the typical modeling approach is discrete
event modeling in this domain. This fact is reflected in
the popular simulation tools applied in this field today.
However, the situation has recently been changing,
because of new aspects that have been taken into ac-
count and increasing requirements for accuracy.
One of these aspects is the time-dependent energy
consumption of single-process operations, process
chains or an overall production system, which becomes
important in context with the increasing influence by
renewable energy sources and the associated volatile
energy availability and energy prices. Approaches for
single-process operations, such as in [2], are focused on
the energy consumption of single machine operations.
They are often based on differential algebraic equations.
However, approaches for investigating several machines
coupled to a process chain use more abstract models
mostly based on discrete event methods, such as in [3, 4].
Today most approaches related to energy processes in
manufacturing are only focused on the simulation of
energy consumption. In [5] it is emphasized that the
energy consumption of a production line (PL) has to be
considered in context with production planning and
scheduling operations. In fact, the energy consumption
then has to be examined together with all the other pro-
duction performance indicators, such as through-put
time, load factors, utilization etc. Hence, considerations
regarding the model design and permissible model sim-
plifications are important to master model complexity.
For instance, it is necessary to determine how finely
grained approximations for continuous energy con-
sumption processes should be. In [6] a discrete-event
approximation of those continuous behaviors is dis-
cussed, but depending on the research a more accurate
approximation can be required.
In [7] a simulator coupling is proposed to execute
manufacturing models with mixed discrete event and
continuous process behaviour, what we call hybrid
system dynamic. This approach is a customized solution
and it shows well-known problems of simulator cou-
SNE 27(2), 2017, 115 - 124, DOI: 10.11128/
Received: May 25, 2017,
Accepted: June 10, 2017 (Special Issue Review)
SNE - Simulation Notes Europe, ARGESIM Publisher Vienna
ISSN Print 2305-9974, Online 2306-0271,
Pawle tta
et al.
A Multimodeling Approach for Simulation of Energy Consumption
116 SNE 27(2) 6/2017
A hybrid modeling approach based on the Discrete
Event and Differential Equation System Specification
(DEV&DESS) in [8] is discussed in [9]. It uses an in-
line integration method that schedules the integration
time as discrete events. Thus, continuous processes can
be modeled using ordinary differential equations and are
solved within a discrete event-oriented simulation envi-
ronment. Both approaches are limited according to the
modularity and clear separation between model specifi-
cation and simulation execution.
This paper is a refined version of [10] and introduces
a multimodeling approach for manufacturing systems to
overcome those inadequacies. According to the theories
in [11, 12], multimodeling means breaking a system into
a network or hierarchy structure of individual models.
The models may be specified by different dynamical
behavior or are described using different methods [13,
14]. Hence, the overall multimodel is from the dynam-
ical point of view often a hybrid model.
The approach is illustrated by the example of a com-
ponent based PL with an industrial furnace facility that
is refined using multimodeling in different layers. Be-
side the classical production performance indicators, it
predicts the time-dependent energy consumption of its
main consumer, the furnace facility. The prototypical
example is implemented in the MATLAB/Simulink [10]
environment using different modelling methods, such as
entities, events, statecharts, ODEs and DAEs. Some
parts of the implementation, pitfalls and simulation
results will be presented to strengthen important parts of
the approach.
1 Multimodeling Approach for
Manufacturing Systems
In the past, methods of modeling and simulation were
mainly used for the planning and optimization of the
operation of manufacturing systems to determine pro-
duction performance indicators, such as through-put
time, facility utilizations, etc. The usage of discrete
event modeling approaches is typical for those investi-
gations. Figure 1 shows such a model of a simple PL
implemented using SimEvents within MATLAB/ Sim-
ulink. The PL is reduced for simplicity to a minimal
structure composed of: (i) a source component for gen-
erating the parts, called entities; (ii) a queue component
with FIFO policy; (iii) a server component named fur-
nace; (iv) a sink component for handled parts.
Statistical ports and signal scopes are omitted in the
figure. Such a discrete event-based simulation model
allows the prediction of the above mentioned perfor-
mance indicators.
Figure 1: Simple discrete event-based production line
model with a furnace component in
However, a precise determination of time-dependent
energy consumptions of specific components or of an
overall system requires a more detailed modeling of
relevant components. In the case of the PL, the abstrac-
tion of the furnace facility as server in the sense of
queuing theory is insufficient. Its abstraction has to be
refined. According to [11, 14], states and events at this
level of abstraction have to be refined to more accurate
events and states at a next lower level. Such refinement
leads to a model or network of models at a next lower
level, which may be subject to refinement in subsequent
steps. The resulting model of such a refined component
is called a multimodel, consisting of interacting sub-
models. The multimodel of a component itself or the
overall model often combines several modeling para-
digms, and operates with different scales, or is a hybrid
system, according to [13], if it includes both continuous-
time and discrete-event behavior.
For a refinement of system components in discrete
event-based manufacturing models we suggest a layer
structure, as illustrated in Figure 2. Each layer repre-
sents a specific aspect of the component with well-
defined interaction relations between the layers. This
approach corresponds to the multilayered architecture, a
common design pattern, used in software development
[16]. Figure 2 suggests three general layers to refine a
manufacturing system component, in which the models
of different layers should specify the following charac-
The material flow layer describes as highest
abstraction the event-based flow of entities (i.e. parts)
into and out of the component. It is the basic layer for
connecting components to a production line or process
chain model, such as illustrated in Figure 1.
This layer may be refined in further steps using the
entity-based modeling method to map internal
material flows in more detail or to provide an
interface to the other layers.
et al.
A Multimodeling Approach for Simulation of Energy Consumption
SNE 27(2) 6/2017 117
The process control layer maps the local process
control operations of the component. This is
especially important for components with several
internal manufacturing operations, where the parts
are handled in several phases, which may be iterated
according to an internal control program or other
internal conditions.
The process physics layer implements details of
internal process operations that are relevant in the
different manufacturing phases, such as energy flows,
chemical reactions etc.
Figure 2: General layer structure to refine a manufacturing
system component.
Depending on the specific characteristics of a compo-
nent, the suggested layers can be arranged hierarchically
or as a network of models to form a multimodel. Of
course, each layer itself can be refined using models at a
next lower level or a layer may be omitted. It is im-
portant to adapt the level of abstraction to the questions
at hand and only include the processes that are needed
to answer them.
Generally, each layer should provide its specific
kind of information, which influences the interaction
relations between the layers. The material flow layer is
mainly concerned with logistic quantities such as wait-
ing times and utilizations; the process control layer
gives the order and timing of the different manufactur-
ing phases, which can be useful in the context of other
information. The kind of information provided by the
physics layer can vary widely depending on the specific
component characteristics and implemented details.
The various layers typically demand diverse model-
ing approaches and, depending on the level of abstrac-
tion, their scales are often different. For instance, in the
considered example of the furnace component energy
flows are accurately modeled based on physical laws,
which are described by ODEs or DAEs.
2 Hybrid System Modeling and
The suggested multimodeling approach for manufactur-
ing system modeling combines several modeling meth-
ods to describe different dynamical behaviour, called a
hybrid system. Subsequently, we want to highlight some
modelling methods and related simulation software.
2.1 Modeling methods
We will consider the modeling methods regarding the
suggested layer structure in Figure 2. The material flow
layer is often specified using a discrete-event modeling
method, which defines abstract entities moving between
stationary components and acting on them [8]. The
entities are identified in manufacturing systems with
workpieces or tools; temporary components move be-
tween production facilities.
A convenient method to describe the different manu-
facturing phases in a production facility, which is the
concern of the process control layer, is that of state
graphs [17]. The phases directly correspond to the states
and the transitions describe the internal process logic.
Alternatively, one could again use a process-based ap-
proach, wherein the entities denote abstract control
The actual manufacturing operations, here summa-
rized under the term process physics, are often modeled
based on natural or technical laws, e.g. from mechanics,
thermodynamics or chemistry. This results in continu-
ous models based on differential equations (ODEs). If
the description contains algebraic constraints or the
equations are constructed automatically using a physical
modeling approach [18], then the mathematical model is
enlarged to a system of differential algebraic equations
2.2 Related simulation tools
For the simulation of the logistic and process-oriented
aspects of a production system several discrete event-
based simulation environments exist and are in wide
industrial use, such as Arena [19] and Plant Simulation
[20]. Usually these programs lack algorithms such as
ODE or DAE solvers to cope with continuous system
specifications. This is why different simulators are cou-
pled to solve such problems, such as in [7]. The intro-
duced multimodel structure supports such simulator
couplings, but it cannot rectify its general problems.
Pawle tta
et al.
A Multimodeling Approach for Simul ation of Energy C onsum ption
118 SNE 27(2) 6/2017
Instead, one should use a software environment that
is capable of hybrid modeling and simulation. Such an
environment, increasingly used for manufacturing sys-
tem simulation, is AnyLogic [21]. It offers system dy-
namics, discrete event and agent-based methods. How-
ever, the mapping of complex ODEs to system dynam-
ics diagrams is quickly confusing and physical model-
ing techniques, according to [18], are not supported. As
a consequence, it may be only a useful choice for mul-
timodeling problems with relatively simple continuous
process physics.
Another widely used software supporting multimod-
eling, although less popular in the manufacturing simu-
lation domain, is the Matlab/Simulink environment.
Originally designed for the simulation of continuous
systems using the signal flow paradigm, it can be ex-
tended using additional toolboxes and blocksets to in-
clude discrete, discrete event or physical modeling fea-
tures: (i) The SimEvents blockset enables discrete event
modeling based on the entity approach; (ii) Stateflow
provides state chart modeling techniques; and (iii) Sim-
scape expands the continuous tool chest with physical
modeling features. This software environment provides
the widest range of features for multimodeling today
and it is already used and accepted in other engineering
domains. Hence, it will be used in the following to illus-
trate and validate the suggested multimodeling approach
by implementing some concrete examples.
An alternative choice could be to use a Modelica
based solution [18] with the additional packages de-
scribed in [22, 23] or Ptolemy [24] with the OpenModel-
ica extension according to [25]. However, both Modeli-
ca and Ptolemy are even more unknown than
MATLAB/Simulink in the manufacturing system com-
3 Basic Application to a
Manufacturing System
To illustrate the introduced approach the furnace com-
ponent of the PL model in Figure 1 will be analyzed for
multimodeling and a set of models with different levels
of abstraction will be designed.
3.1 Multimodeling of furnace component
Our objective of multimodeling is the refinement of the
furnace component to investigate its time-related energy
consumption. Industrial furnaces are widely used in
metalworking processes and are one of the most exten-
sive energy consumers in manufacturing systems. In
addition, their internal operation is generally rather
complex, making this an ideal example for refinement
using different modeling approaches. The operation of
such a furnace is patterned in the following after the
descriptions in [9, 26].
When parts arrive at the furnace, they are collected
until a given batch size is reached. A complete batch
then enters the furnace, is processed and leaves the
furnace. Then, the batch will probably be resolved for
processing the parts by other facilities. The refinement
of such internal processes of a component is part of the
material flow layer, as demonstrated in Figure 2.
Moreover, the heat treatment itself consists of sever-
al phases that are implemented by a local control; these
are mapped and refined at the process control layer.
Figure 3 shows an example of such a control with six
operation phases.
Figure 3: State graph describing a local control of
furnace operations.
The idle phase spans the times before parts have entered
and after all parts have left the furnace. During the load
phase parts enter the furnace. In the heat-up phase the
furnace is heated until a given temperature is reached;
this is then held constant during the hold phase. Accord-
ing to the requirements, the heat-up and hold phases can
be iterated several times with different temperatures.
Finally, the parts leave the oven in the unload phase. An
additional cool-down phase may be included either to
make sure that the parts leave the furnace with a moder-
ate temperature or to describe a shutdown of the fur-
et al.
A Multimodeling Approach for Simulation of Energy Consumption
SNE 27(2) 6/2017 119
Since the focus of our example study lies on the en-
ergy consumption, the heat flows in the furnace have to
be considered in more detail, and must be refined at the
process physics layer (Fig. 2). The energy source is the
power supply of the actual heater. From here the heat
flows mainly through convection and radiation process-
es to the parts and to the internal structures and the
casing of the furnace. During the heat-up and hold phas-
es losses are mainly due to conduction through the cas-
ing into the environment, while in the load or unload
phases additional losses are caused by the open doors.
Because of the complicated geometry, the physical
details, especially of the convection processes, are also
rather complicated. However, for the estimation of the
total heat flows common approximative methods usual-
ly give quite accurate results. Concrete mathematical
models for specifying the process physics will be con-
sidered afterwards in context with their prototypical
Based on the previous considerations, Figure 4
shows the multimodel structure with defined interfaces
for the refinement of our furnace component based on
the layers introduced in Figure 2. In this abstraction it
consists of three interacting models: (i) MF for the ma-
terial flow; (ii) PC for the process control; and (iii) PP
for the process physics. The labels (B) and (C) are not
of interest at this point.
The models have to communicate in several ways:
MF receives parts from the external input port <1>
and sends to the PC the number of parts that have
entered the MF.
When the batch size is reached, the PC starts the PP,
which models the different manufacturing phases
during the operation of the furnace.
During the operation of the furnace the PC signals
the current manufacturing phase to the PP, which
adapts the internal heat flows accordingly. This can
mean changing the supplied heat between the heating
and holding phases or increasing the losses due to the
doors being open while loading or unloading.
In return the PP sends the current temperature values
of the furnace and the parts to the PC, which uses
them to determine whether the heat-up phase or the
optional final cool-down phase of the furnace is
complete (Fig. 3).
When the manufacturing phase unload (Fig. 3) is
finished, the PC sends a leaving signal to the MF,
which accordingly forwards the processed parts to its
external output port <2>.
Figure 4: Basic multimodel structure with defined
interfaces for the refinement of the furnace
Each model has its specific set of parameters and output
quantities: The MF defines the batch size and logistic
properties such as average waiting time and machine
utilization; the PC gets the heating program and outputs
the time in the different manufacturing phases. The PP
needs a lot of physical parameters for the calculation of
the heat flows and provides the power requirements
during the process phases as well as the total energy
3.2 Design of a model library for the furnace
Based on the basic multimodel structure in Figure 4, a
set of models for each layer has been implemented and
organized in a library (Fig. 5). The several models use
different modeling methods or have varying levels of
detail. They are labeled with two letters denoting the
layer and a third letter in brackets giving the level of
detail in ascending order, with (A) being the simplest
The basic material flow model MF(A) uses only a
simple server, while MF(B) explicitly contains an
input tray, where the batch is compiled.
The process control subsystem PC(A) uses a simple
entity-based model, implemented in SimEvents, to
describe one pass through the four basic process
phases and ignoring idle and cool-down phases. PC(B)
enables a repetition of heat-up and hold phases
according to its heat program parameter, implemented
as an entity-based model in SimEvents. Additionally,
PC(C) adds a cool-down phase, but because of the
more complex control logic it is implemented using
the state machine approach with Stateflow.
Pawle tta
et al.
A Multimodeling Approach for Simulation of Energy Consumption
120 SNE 27(2) 6/2017
The process physics model PP(A) uses only the
internal oven temperature and a simple formula for the
global losses, while PP(B) adds the temperature of the
parts, the heat transfer between oven and parts and
additional losses during the load and unload phases.
Both use standard Simulink blocks to implement the
corresponding differential equations. Finally, PP(C)
employs physical modeling, which is modeled using
Simscape for the same physical processes as PP(B).
This makes the physical model structure more trans-
parent and easier for engineers to expand.
Figure 5: Library with models for composing several
multimodels with different levels of detail for the
furnace component.
4 Some Modeling &
Implementation Details
In the following, three multimodel variants for the fur-
nace will be described in more detail: (i) a very basic
multimodel named ovenBAA the capital letters stand
for the composition of MF(B), PC(A) and PP(A) mod-
els; (ii) a medium complex multimodel ovenBCA; and
(iii) the most complex multimodel variant ovenBCC.
While this section is devoted to some implementation
details of the single models used in the three multimod-
els, the next one will discuss some simulation results.
The implementation was carried out using the
MATLAB/Simulink environment and related tools.
4.1 Material flow model
All of the examples considered here use the extend-
ed model MF(B) for mapping the internal material flow.
The entity-based model structure of MF(B) (Fig. 6)
consists of two simple servers: the first for the input
tray; and the second (N-Server) for the furnace proper.
Both hold incoming entities (i.e. parts) up to the given
batch size, until their succeeding gates open to transfer
the entities to the next stage. The intermediate Gate
guarantees that a full batch is always delivered to the
furnace; the final Release Gate is triggered by the
model at the process control layer (Figs. 7, 8) at the
end of the unload phase.
Figure 6: Discrete event-based model of MF(B) using
4.2 Process control models
The process control model starts when the number of
parts that have entered at the material flow layer is equal
to the batch size. In the process control model PC(A)
(Fig. 7), implemented using SimEvents, this leads to the
creation of a control entity that passes through a line of
servers denoting the different phases. Except for one, all
servers simply have a fixed processing time; only the
heat-up phase is different. Here the entity is held in a
server until the current oven temperature, which is com-
puted in the model at the physics layer (Figs. 9, 10), has
reached the given temperature Tset. After the unload
phase, the control entity is destroyed and the wake-up
signal (trigger) is generated, which opens the Release
Gate in the model at the material flow layer (Fig. 6).
Figure 7: Discrete event-based model of PC(A) using
et al.
A Multimodeling Approach for Simulation of Energy Consumption
SNE 27(2) 6/2017 121
The much more elaborate process control model
PC(C) (Fig. 8) incorporates all six phases, pictured in
Figure 2, as well as possible repetition of the heat-up
and hold phases. According to the theory in [8], it is a
hybrid DEV&DESS model. It is modeled based on the
state machine approach, but it includes continuous state
event handling and has been mainly implemented using
Stateflow. The two auxiliary subsystems (schedTEvent,
checkTemperature) create events when the waiting time
has changed or the heat-up or cool-down temperature
has been reached. The model PC(C) is essentially an
adapted version of a model described in [9].
Figure 8:
Discrete event-based model with continuous
state event handling of PC(C) mainly using
4.3 Process physics models
The basic process physics model PP(A) (Fig. 9) uses a
simple power balance to compute the change of the
oven temperature :
   (1)
where is the total heat capacity of the oven. The
power loss is computed with Newtons simple law of
   (2)
where is the temperature of the surroundings and
a constant that subsumes all convective and conductive
processes. The heating power is assumed to be constant
during the heat-up phase and to match the losses during
the hold phase:
    
 (3)
For the computation of the total power demand of
the furnace, a constant power is added, which sub-
sumes all non-heating processes, such as a base load or
the power electronics.
Figure 9: ODE-based model of PP(A) using Simulink.
The model PP(C) (Fig. 10) uses physical modeling
based on Simscape to incorporate much more physical
details, such as convection and radiation from the fur-
nace during load and unload phases to the environ-
ment. The governing differential algebraic equations are
not built up explicitly here; instead, the physical com-
ponents such as heat capacities and various kinds of
heat flow are represented directly, which makes the
physical structure of the model much clearer.
Figure 10: Physical model (DAE) of PP(C) using
4.4 Implementation pitfalls
The extension of MATLAB/Simulink by several packag-
es (blocksets) is necessary to make a multi-paradigm
model possible, but it also leads to small inconsisten-
cies, which have to be overcome. A minor nuisance here
is the large number of different ways used to express
values and signals in the packages used: Simple scalar
values can be time-based in Simulink, event-based in
SimEvents or physical-valued in Simscape. An event can
be defined by a sample time hit, a rapid value change
(edge), a special trigger signal or a function call.
Pawle tta
et al.
A Multimodeling Approach for Simulation of Energy Consumption
122 SNE 27(2) 6/2017
To clearly define a common interface for the inter-
acting models, their external values are fixed in the
following way: The temperature and phase values are
time-based; the number of parts is event-based; and the
leaving signal is a function call. If the internal imple-
mentation of a block generates differing types, then one
has to use one of the many converter blocks to get the
proper kind of signal, e.g. the round Event to Timed
Signal blocks that can be seen in Figures 7 and 8.
This problem is especially annoying in the physical
modeling environment. The large number of necessary
converters and reference points clutters the model and
destroys the clear physical structure. Hiding them in
subsystems is an obvious way to regain an ordered visi-
ble representation of the underlying physics model.
5 Exemplary Simulation Results
Using the introduced approach, a large number of simu-
lation studies is possible to calculate multifaceted per-
formance indicators. We will restrict our exemplary
consideration to the energy consumption of the furnace
that is the main consumer in our simple production line.
Moreover, we will investigate the costs of refinement
relating to the simulation run time by comparing differ-
ent multimodels, which use different modeling para-
digms or are based on a different level of detail.
5.1 Results relating to the energy aspect
In the exemplary experiment 24 parts were processed
using a batch size of six. After the heat-up phase their
temperatures were held first to 400 for 40 minutes,
then to 800 for 30 minutes. Figure 11 shows some
simulation results related to the energy consumption,
calculated using our most complex multimodel variant
ovenBCC: the temperatures of the furnace and parts, the
power consumption of the furnace, the accumulated
used energy and the current internal manufacturing
phase, each as functions over time.
The results are similar to those presented in [9], but
ovenBCC incorporates more details, especially of the
physical model layer. The other two example models
ovenBAA and ovenBCA basically reproduce the results
from [9], since they are based on the same physical
model assumptions. For comparison, Table 1 shows the
total energy use E1 at the exit of the last part and E2 at
the end of the simulation; Figure 12 displays the power
consumption for the three models.
Model E
[kWh] E
ovenBAA 71.3 108.8
ovenBCA 69.0 106.6
ovenBCC 82.6 117.6
Table 1: Comparison of the total energy consumption.
Figure 11: Simulation results of model variant ovenBCC.
The higher energy needs in variant ovenBCC are due to
the heating of the parts, which enter the oven with the
low temperature of the environment an effect that has
not been taken into account in the other models. The
small difference between the two simpler multimodel
variants results from the different timing of the phases,
as can be seen from Figure 12.
It is interesting to note that the total energy results
only differ by 10
% - 15
%. If this level of accuracy is
sufficient for the question at hand, e.g. for a global as-
sessment of a complex production line, then one can use
one of the simple physics models. However, Figure 12
shows that for a detailed examination of the power
needs during the individual phases one has to stick to
the complexities of multimodel variant ovenBCC.
et al.
A Multimodeling A pproach for Simul ation of E nergy C onsum ption
SNE 27(2) 6/2017 123
Figure 12: Comparison of the power results.
5.2 Comparison of run times
How much do we pay for the additional level of detail
in the complex multimodel? For a comparison of run
times, the number of parts was raised to 600 and three
runs were performed for each multimodel. To get rid of
loading and compile times, the means of only the last
two results were taken, leading to the values in Table 2.
Model Run Time [s]
ovenBAA 25.1
ovenBBA 27.7
ovenBCA 20.7
ovenBCB 39.3
ovenBCC 49.8
Table 2: Comparison of run times.
The results show that the complex physical model using
Simscape in multimodel ovenBCC (line 5) costs a factor
of two in execution time, which can be reduced in this
case to a factor of 1.6 by replacing DAE-based physical
modeling with standard ODE methods using Simulink
(ovenBCB, line 4). However, we see essential differ-
ences in execution time for multimodels with the same
simple process physics model PP(A) (line 1-3), which
are surprising.
The multimodels ovenBBA (line 2) and ovenBCA
(line 3) use process control models of nearly the same
complexity, but in ovenBBA implemented with Si-
mEvents and in ovenBCA with Stateflow. Moreover, the
multimodel ovenBCA (line 3) uses the more complex
process control model PC(C), than ovenBAA (line 1)
with the PC(A) model that was implemented with Si-
mEvents. Apparently, the implementation of SimEvents
has some potential for optimization.
6 Conclusion
The introduced multimodel approach supports the com-
ponent-oriented refinement of manufacturing system
models, using various modeling methods and models
with different levels of abstraction. Generally, it sug-
gests a refinement of components based on the three
logical layers material flow, process control and process
The first layer maps the internal material flow. It de-
livers an external input and output interface for connect-
ing with other components in a manufacturing system
model and an internal interface for communication with
the second layer. The dynamic behavior at this layer is
generally discrete event-based. In the second layer local
process control operations are modeled. This layer pro-
vides event-based control inputs for the other two lay-
ers, but it could be necessary to handle state events of
continuous values as well. In the third layer specific
process operations should be mapped. The dynamic
behavior at this layer can vary significantly and particu-
larly depends on the level of abstraction. Of course,
depending on the characteristics of a system component
and the intended level of abstraction, layers may them-
selves be omitted or refined using several interacting
The paper illustrated the approach using the example
of a production line with an industrial furnace as the
main component. According to the layers, models with
different levels of abstraction have been implemented
for the furnace, which could be aggregated to a multi-
model. Then, the energy consumption of the furnace
was investigated with three different multimodels and
the simulation run times were measured. Hence, accura-
cy effects and computing costs could be compared.
The approach opens many new ways for investigating
manufacturing systems, but the complexity is increasing.
Hence, in the next step it should be combined with met-
amodeling techniques such as those considered in [27].
Pawle tta
et al.
A Multimodeling Approach for Simul ation of Energy C onsum ption
124 SNE 27(2) 6/2017
Acknowledgments. Thorsten Pawletta and Artur
Schmidt thank the German Research Foundation for
funding this research (no. PA 631/2-2). Peter Junglas is
grateful for the hospitality extended to him by Tom
Schramm at the Department of Geomatics, HCU Ham-
burg, during his work on this project.
[1] Lefeber E, Rooda JE. Modeling and Analysis of Manu-
facturing Systems. In: Fishwick PA. editor. Handbook of
Dynamic System Modeling, Boca Raton: Cpaman &
Hall/CRC, 2007, p. 34_1-34_20.
[2] Eisele C. Simulation-based optimization of the electrical
consumption of metal-cutting machine tools (German).
[dissertation], TU Darmstadt, Germany, 2014.
[3] Weinert N. An approach for the planning and operation
of energy-efficient production systems (German). [disser-
tation], TU Berlin, Germany, 2010.
[4] Larek R. Resource-efficient design of manufacturing
process chains by simulation and numerical optimization
(German). [dissertation], Univ. Bremen, Germany, 2012.
[5] Schrems S. A method for the model-based integration of
machine-specific energy demand in production planning
(German). [dissertation], TU Darmstadt, Germany, 2014.
[6] Larek R, Brinksmeier E, Meyer D, Pawletta T, Ha-
gendorf O. A discrete-event simulation approach to pre-
dict energy consumption in machining processes. Pro-
duction Engineering Research and Development, Ber-
ling: Springer Pub., 2011, p. 575-579.
[7] Peter T, Wenzel S. Simulation-based Planning and Eval-
uation of Energy Efficiency for Production Systems in
Car Manufacturing (in German). Simulation in Produc-
tion and Logistics, Fraunhofer Pub., Stuttgart, Germany,
2015, p. 535-544.
[8] Zeigler BP, Kim TG, Prähofer H. Theory of Modelling
and Simulation. SanDiego: Acad. Press, 2nd ed., 2000,
510 p.
[9] Schmidt A, Pawletta T. Hybrid modeling of manufactur-
ing process chains and their energy consumption in a
discrete event simulation environment (German). In:
Wittmann, J. editor. Proc. ASIM 2014-22. Symp. Simula-
tionstechnik, Sep 2014: Berlin, Germany, p. 109-116.
[10] Schmidt A, Pawletta T, Junglas P. A layered structure
for modeling manufacturing processes with the inclusion
of energy consumption. In: Proc. ASIM Symp. STS/
GMMS, Mar. 2016: Lippstadt, Germany, p. 155-164.
[11] Fishwick PA. Simulation Model Design and Execution:
Building Digital Worlds. Upper Saddle River: Prentice-
Hall Inc., 1995, 432 p.
[12] Park M, Fishwick PA, Lee J. Multimodeling. In: Fish-
wick P.A. editor. Handbook of Dynamic System Model-
ing, Boca Raton: Cpaman & Hall/CRC, 2007, p. 14_1-
[13] Mosterman PJ. Hybrid Dynamic Systems: Modeling and
Execution. In: Fishwick PA. editor. Handbook of Dy-
namic System Modeling, Boca Raton: Cpaman &
Hall/CRC, 2007, p. 15_1-15_26.
[14] Shephard MS, Seol ES, Dale BF. Towards a Multimodel
Hierarchy to Support Multiscale Simulation. In: Fish-
wick PA. editor. Handbook of Dynamic System Model-
ing, Boca Raton: Cpaman & Hall/CRC, 2007, p. 12_1-
[15] The MathWorks. Simulink: Simulation and Model-Based
Design. Online:
(called 2015-11-11).
[16] Buschmann F, Meunier R, Rohnert H, Sommerlad P,
Stal M. Pattern-Oriented Software Architecture: A Sys-
tem of Patterns. vol. 1. New York: John Wiley & Sons,
1996, 476 p.
[17] Harel D. Statecharts: A Visual Formalism for Complex
Systems. Science of Computer Programming, 8, 1987, p.
[18] Fritzson PA. Principles of Object-Oriented Modeling
and Simulation with Modelica 3.3. New York: Wiley &
Sons, 2015, 1256 p.
[19] Kelton WD, Sadowski R, Zupick N. Simulation with
Arena. 6th ed. New York: McGraw-Hill, 2014, 656 p.
[20] Bangsow S. Tecnomatix Plant Simulation: Modeling and
Programming by Means of Examples. Springer: Berlin,
2016, 713 p.
[21] Grigoryev I. AnyLogic 7 in Three Days. 2nd ed. Noth
Charleston: CreateSpace Independent Pub., 2015, 256 p.
[22] Elmqvist H, Gaucher F, Mattsson SE, Dupont F. State
Machines in Modelica. In: Otter M, Zimmer D, editors.
Proc. 9th Int. Modelica Conf., Sep 2012: Munich, Ger-
many, p. 37-46.
[23] Sanz V, Urquia A, Cellier FE, Dormido S. System mod-
eling using the Parallel DEVS formalism and the Model-
ica language. Simulation Modelling Practice and Theory,
18(7), 2010, p. 9981018.
[24] Ptolemaeus C. System Design, Modeling, and Simulation
using Ptolemy II., 2014, 690 p.
[25] Mirzaei M. Integration of OpenModelica into the Multi-
paradigm Modeling Environment of Ptolemy II. Master
Thesis, Univ. of Linköping, Sweden, 2014.
[26] Heuer V, Löser K. Energy optimization of thermal-
chemical vacuum processes and aggregates in large-scale
productions (German). Elektrowärme International, 69,
2011, p. 261-268.
[27] Pawletta T, Schmidt A, Zeigler BP, Durak U. Extended
Variability Modeling Using System Entity Structure On-
tology within MATLAB/Simulink. In: Proc. SCS Int.
SpringSim/ANSS, Apr. 2016: Pasadena/CA, USA, 2016,
p. 62-69.
... With regard to practical applications, the method introduced in this paper will be applied in the next step to M&S of a production system family of Industry 4.0 as shown in [45]. Variants of flexible production must be examined, taking into account the energy consumption of the production facilities. ...
... Variants of flexible production must be examined, taking into account the energy consumption of the production facilities. In [45], the modeling of the individual manufacturing facilities was done modularly on three levels (material flow, process control, process physics). In particular, for the process control and process physics levels, models were developed with varying degrees of detail, which resulted in considerable differences in accuracy and runtime. ...
System entity structure has been used since the 1970s as a formal ontology framework, axiomatically defined, to represent the elements of a system of systems and their hierarchical relationships resulting in a family of hierarchical models. One challenge with this approach is the process of exploring a family of hierarchical models, and selecting a particular composition from which a fit-for-purpose discrete-event simulation model can be automatically synthesized and executed. This paper deals with the definition of performance metrics which are used to guide the Modeler to select the most practical model of a real-world system from among a potentially large set. The discrete-event simulation model is selected among a set of system entity structure family models by using model level activity metrics in order to be able to have a prediction of the accuracy and the performance of the models. A case study is presented to illustrate and validate the relation between activity metrics and system entity structure concepts.
... Eine mögliche Variante hierzu kann die kombinierte Simulation sein, also die Kombination eines ereignisdiskreten und eines kontinuierlichen Modellteils [9]. Wenn es die Natur des technischen Systems zulässt, kann dessen Energiebedarf über Differentialgleichungen modelliert werden [10]. Dies ist jedoch bei vielen Produktionsressourcen, wie z.B. ...
Conference Paper
Full-text available
The price of a commodity, as electricity, is determined on a commodity market. A market is efficient when the supply and demand in the market are at an equilibrium. Efficient markets run on information. Information can cause a spontaneous and instantaneous change within the supply and demand in a market. The market communicates this new equilibrium through the change of the price of a commodity. In the electricity market the supplier and consumer communicate through electrical load profiles. A load profile signals when and how much energy should be consumed within a certain time frame without causing a change in the price of elec-tricity. Creating such load profiles is commonly done by the supplier of energy by means of standard load profiles. Here we propose a data-driven simulation-based method that allows for the consumer to create its own specific load profile, which potentially will bring down the cost of energy consumed.
... Thiede 2012) or Matlab (e.g. Pawletta et al. 2017). The alternative is to couple several tools and enable communication through a middleware interface or directly (e.g. ...
Conference Paper
Within the Section "Simulation in Production and Logistics" of the association for simulation in the German-speaking area (ASIM), the relevance of energy-related considerations led to the foundation of a dedicated scientific workgroup. One of their primary results is a comprehensive literature overview of related approaches from research and industrial perspectives. The experts cooperated to identify, evaluate, and classify relevant literature. Interim findings have been published with special emphasis on logistics and manufacturing. Recently, work was focused on converting the findings into a structured database, to foster the involvement of other experts by providing consistent access to all results in the public domain and to provide means to submit new contributions proactively to the group. Thus, the workgroup intends to raise awareness and motivate other researchers to join the effort. This paper summarises the rationale and methodology of the activities as well as the provision as a public database.
... Blended modeling and simulation have been reported for the planning and optimization of the operations of manufacturing system by other researchers. The focus was to optimize system performance such as energy consumption, facility and staff utilization rate, and through-put rate of a production line [8]. [9] presents a hybrid modeling methodology for enterprise resource planning in a manufacturing system. ...
Conference Paper
Full-text available
Being able to quickly detect anomalies and reason about their root causes in critical manufacturing systems can significantly reduce the analysis time to bring operations back online, thus reducing expensive unplanned downtime. Machine learning-based anomaly detection approaches often need significant amounts of labeled data for training and are challenging to scale for manufacturing deployments. A robust blended system dynamics and discrete event simulation physics-based modeling methodology is proposed for the task of automated anomaly detection. The blended model consists of discrete event simulation (DES) components for the discrete manufacturing process modeling, and system dynamics (SD) components for continuous variables. The methodology strikes a balance between the computational overhead for online monitoring and the level of details required to perform anomaly detection tasks. The implementation of models takes an object-oriented approach, allowing multiple components of a smart factory to be robustly described in a modular, extendable and re-configurable manner. The proposed methodology is applied to and validated by data collected from a real commercial manufacturing plant. A production line is modeled with DES components and heat transfer is modeled with SD. The blended model is then utilized for anomaly detection. It is demonstrated that the model-based approach is effective not only for detecting but also explaining particular types of anomalies in a commercial discrete manufacturing system.
... Schmidt and Pawletta [244,212] also developed a hybrid simulation approach for describing resource consumption in industrial processes. Instead of using co-simulation, their approach is based on the Discrete-Event and Differential Equation System Specification (DEV&DESS) modeling formalism (see also Section 5.2) and the implementation is embedded into MATLAB/SimEvents, for which they develop a custom model library. ...
... Als Beispiel führt das Bundesministerium für Bildung und Forschung [103] unter anderem die Anwendungsbereiche des Maschinen-und Anlagenbaus sowie der Automatisierungs-und Fertigungstechnik auf. Letztere stehen gegenwärtig durch die sogenannte Energiewende und die damit einhergehende Variabilität bei der energetischen Auslegung fertigungstechnischer Prozessketten beim Produktionsplanungsprozess vor ganz neuen Herausforderungen [87,112,140,141,144]. Zur Beherrschung der Variantenvielfalt fordert Holdschick bereits ein Variantenmanagement in der modellbasierten Produktlinienentwicklung [67] und damit den Einsatz der M&S. ...
... Als Beispiel führt das Bundesministerium für Bildung und Forschung [103] unter anderem die Anwendungsbereiche des Maschinen-und Anlagenbaus sowie der Automatisierungs-und Fertigungstechnik auf. Letztere stehen gegenwärtig durch die sogenannte Energiewende und die damit einhergehende Variabilität bei der energetischen Auslegung fertigungstechnischer Prozessketten beim Produktionsplanungsprozess vor ganz neuen Herausforderungen [87,112,140,141,144]. Zur Beherrschung der Variantenvielfalt fordert Holdschick bereits ein Variantenmanagement in der modellbasierten Produktlinienentwicklung [67] und damit den Einsatz der M&S. ...
Full-text available
Modeling and Simulation (M&S) forms the basis of the modern development and planning process in engineering. Modeling deals with the abstraction of essential aspects of technical systems and their implementation in a model. During simulation the implemented model is executed aiming to gain knowledge about the system behavior. A single model execution is called a simulation run. A simple simulation experiment comprises many simulation runs with varied input variables. Complex simulation-based experiments are characterized by the integration of simulation with numerical methods such as optimization, screening or sensi-tivity analysis. Highly complex simulation experiments investigate the different model structures in addition to the parameter space. The classical M&S approach is reaching its limits in engineering. One reason is the growing customers demand for customized products. This increases the variety and often the complexity of technical systems as well as of the resulting models and investigation objectives. This means that the specification and execution of simulation experiments includes not only a set of model variants with their parameter settings, but also combinations with numerical methods and their configuration. Furthermore, the sequence of variants to be investigated or numerical methods to be applied often arise from previously determined experiment results. This thesis contributes to the development of general methods for variant management in M&S up to the simulation experiments’ level and their automated execution. The proposed solutions are developed based on the ASIM procedure model, the modular-hierarchical modeling approach and the System Entity Structure/Model Base (SES/MB) framework. The variant management comprises the phases variant analysis, variant formalization, variant implementation and variant generation. Furthermore, a modular-hierarchical structuring for simulation experiments is presented. By extending the SES/MB framework, not only model variants but complete experiment variants can be specified, implemented and generated. Moreover, a framework for automated experimentation based on the extended SES/MB framework is proposed. Finally, the methods are demonstrated using an application example from the field of production and logistics.
... The build method can be configured via MATLAB scripts depending on the specific execution unit (EU). Currently, the toolbox supports the generation of Simulink models, including SimEvents, Simscape and Stateflow components (Pawletta, Schmidt, & Junglas, 2016) and MATLAB DEVS models (Deatcu, Freymann & Pawletta, 2017). Moreover, first experiments for generating Dymola and OpenModelica models have been carried out. ...
Versatile systems are characterized by a high degree of variability, regarding configuration and customization for employment in a particular context, and reactive adaptability during operation. In engineering, cyber-physical systems and flexible, reconfigurable, or adaptable production systems are typical examples. Their development requires enhancements of methods and tools for effective modeling and simulation of such systems. Based on the System Entity Structure and Model Base (SES/MB) approach, a novel infrastructure will be introduced that supports modeling and simulation of versatile and adaptable systems. The infrastructure extends the classical SES/MB framework by an experiment control unit and an execution unit. Moreover, the classical SES theory is extended by some new concepts. After the new components and concepts are introduced, they are then illustrated through the example of an industrial robot control application. Finally, a prototype implementation with MATLAB/Simulink is discussed briefly.
Full-text available
Effective and efficient planning of energy efficiency measures is of great importance to manufacturing companies. Material flow simulation that has been extended to also include energy is increasingly used in this context as a tool for the analysis of complex interactions between material flows and energy flows in factories. This thesis deals with the conception and the systemization of a methodology for designing such simulation studies specifically in re-planning projects that aim for energy efficiency improvements. Taking basic approaches from decision theory into particular consideration, it is intended to provide guidance in deciding on the project-specific manifestation for relevant characteristics of a simulation study in this problem area (e.g. the manner to model energy consumption), regardless of the utilised simulation solution. The developed solution comprises a 13 steps spanning process model as well as nine solution modules to support decisions concerning the choice of manifestation for selected characteristics. In this way, the entire process from the development of an energy efficiency measure through the actual application of simulation (following VDI guideline 3633 Part 1:2014) to the eventual investment decision is assisted. The results of this thesis were exemplarily tested in three case studies. All initially defined requirements could thereby be positively verified.
Conference Paper
Full-text available
Software product-lines are designed to tackle the development of systems that are characterized by a high degree of variability. They define variation points where different solutions can be derived for different products. Such variability mechanisms can be defined at different levels of abstraction, ranging from requirements specification to source code implementation. Several efforts have been made regarding variability modeling on the system model level, which have been mostly inspired by approaches that have been developed by the software engineering community. Regarding the rising synergy between the systems engineering and modeling and simulation, this paper presents an approach for variability modeling based on the System Entity Structure (SES) ontology. Several system configurations are specified using an SES or a set of SES's on a metamodel level, while configurable models are implemented as basic models; these are organized in a Model Base (MB). Pruning is then utilized to specify a particular system variant. We propose some general extensions of the baseline SES approach with a focus on a new concept called SESFunctions. The extensions have been implemented within an SES toolbox for the MATLAB/Simulink environment. The proposed approach is exemplified using an engineering example in order to present its applicability.
Conference Paper
Full-text available
In the design of manufacturing processes the consideration of resource usage, especially energy consumption , is getting more attention. However, the inclusion of the relevant physical processes in a unified modeling approach is often a non-trivial task. The commonly used simulation tools in this domain usually only support discrete event modeling techniques, while models of physical processes are often continuous in nature and are modeled using ordinary differential equations (ODEs) or differential algebraic equations (DAEs). Moreover, the investigation of such physical processes demands a more specific consideration of process control operations, which are favorably modeled using state machines. To combine the diverse paradigms in a clear-cut way, a layered structure is proposed, which encapsulates the different levels of the model and defines a precise interface between them. The example of an industrial furnace serves as a proof of concept and demonstrates its versatility by integrating a transaction-based material flow model with a state machine describing the process control and physical modeling of the energy flows.
Full-text available
Part I: Introduction. Chapter 1: Introduction to Modeling and Simulation. Chapter 2: A Quick Tour of Modelica. Part II: The Modelica Language. Chapter 3: Classes, Types, and Declarations. Chapter 4: Inheritance, Modifications, and Generics. Chapter 5: Components, Connectors, and Connections. Chapter 6: Literals, Operators, and Expressions. Chapter 7: Arrays. Chapter 8: Equations. Chapter 9: Algorithms and Functions. Chapter 10: Packages. Chapter 11: Annotations, Units, and Quantities. Part III: Modeling and Applications. Chapter 12: System Modeling Methodology and Continuous Model Representation. Chapter 13: Discrete Event, Hybrid, and Concurrency Modeling. Chapter 14: Basic Laws of Nature. Chapter 15: Application Examples. Chapter 16: Modelica Library Overview. Part IV: Technology and Tools. Chapter 17: A Mathematical Representation for Modelica Models. Chapter 18: Techniques and Research. Chapter 19: Environments. Appendix A: Modelica Formal Syntax. Appendix B: Mathematica-style Modelica Syntax. Appendix C: Solutions for Exercises. Appendix D: Modelica Standard Library. Appendix E: Modelica Scripting Commands. Appendix F: Related Object-Oriented Modeling Languages. Appendix G: A Modelica XML Representation. References. Index.
This book systematically introduces the development of simulation models as well as the implementation and evaluation of simulation experiments with Tecnomatix Plant Simulation. It deals with all users of Plant Simulation, who have more complex tasks to handle. It also looks for an easy entry into the program. Particular attention has been paid to introduce the simulation flow language SimTalk and its use in various areas of the simulation. The author demonstrates with over 200 examples how to combine the blocks for simulation models and how to deal with SimTalk for complex control and analysis tasks. The contents of this book ranges from a description of the basic functions of the material flow blocks to demanding topics such as the realization of a database-supported warehouse control by using the SQLite interface or the exchange of data by using XML, ActiveX, COM or DDE.
Saving natural resources by limitation of the applied energy amounts is more and more an issue in discrete production. The origin of energy-efficient production execution could be found during production system development, where by the determination of production processes, equipment and organizational structures the minimal achievable amount concerning energy consumption is specified. Despite this relevance of the system planning phase, appropriate instruments for analyzing alternative process chains and system variants are more or less non-existing. Overcoming this gap is the objective of the EnergyBlocks-Planning-System, which has been developed at the Chair of Assembly Technology and Factory Management of the TU Berlin.
The energy optimization of thermoprocessing equipment is of great ecological and economical importance. Thermoprocessing equipment consumes up to 40 % of the energy used in industrial applications. Therefore it is necessary to increase the energy efficiency of thermoprocessing equipment in order to meet the EU's targets to reduce greenhouse gas emissions. To exploit the potential for energy savings, it is essential to analyze and optimize processes and plants as well as operating methods of electrically heated vacuum plants used in large scale production. The process can be improved by accelerated heating through the application of "convective heating". In addition higher process temperatures can be applied in diffusion-controlled thermochemical processes to accelerate the process significantly. Modular vacuum systems prove to be very energy-efficient because they adapt to the changing production requirements step-by-step. An optimized insulation structure reduces thermal losses considerably. Energy management systems installed in the plant-control optimally manage the energy used for start-up and shutdown of the plants while preventing energy peak loads. The use of new CFC-fixtures also contributes to reduce the energy demand.
The scope of Modelica has been extended from a language primarily intended for physical systems modeling to modeling of complete systems by allow-ing the modeling of control systems including state machines. This paper describes the state machines intro-duced in Modelica 3.3. Any block without conti-nuous-time equations or algorithms can be a state of a state machine. Transitions between such blocks are modeled by a new kind of connections associated with transition conditions. The paper gives the details for building state ma-chines and includes several examples. In addition, the complete semantics is described using only 13 Modelica equations.
We present a broad extension of the conventional formalism of state machines and state diagrams, that is relevant to the specification and design of complex discrete-event systems, such as multi-computer real-time systems, communication protocols and digital control units. Our diagrams, which we call statecharts, extend conventional state-transition diagrams with essentially three olements, dealing, respectively, with the notions of hierarchy, concurrency and communication. These transform the language of state diagrams into a highly structured&apos; and economical description language. Statecharts are thus compact and expressive--small diagrams can express complex behavior--as well as compositional and modular. When coupled with the capabilities of computerized graphics, statecharts enable viewing the description at different levels of detail, and make even very large specifications manageable and comprehensible. In fact, we intend to demonstrate here that statecharts counter many of the objections raised against conventional state diagrams, and thus appear to render specification by diagrams an attractive and plausible approach. Statecharts can be used either as a stand-alone behavioral description or as part of a more general design methodology that deals also with the system&apos;s other aspects, such as functional decomposition and data-flow specification. We also discuss some practical experience that was gained over the last three years in applying the statechart formalism to the specification of a particularly complex system.