Manufacturing Process Simulation in a Hybrid Cloud Setup

Gerhard Benedikt Weiß

, Dario Pietraroia

, Claudio Sassanelli

and Hugo Daniel Macedo

Dept-E, Co-Simulation & Software Group, VIRTUAL VEHICLE Research GmbH, Inffeldgasse 21a, 8010 Graz, Austria

TTS — Technology Transfer System, D’Ovidio 3, Milan 20131, Italy

Department of Management, Economics and Industrial Engineering, Politecnico di Milano,

Piazza Leonardo da Vinci, 32, 20132 Milan, Italy

DIGIT, Aarhus University, Department of Electrical and Computer Engineering, Finlandsgade 22, 8000 Aarhus, Denmark

Keywords:

Hybrid-cloud, Co-simulation, Model-based Design, Manufacturing, Tools.

Abstract:

Model-based design of manufacturing robotic systems involving the usage of different tools, models and the

co-simulation of the system behaviour beneﬁts from collaborative platforms enabling ready-to-use and cloud-

hosted tools and models. Nonetheless, due to market segmentation and the difﬁculty to deploy and support all

the existing tools and models in such a platform, it is, therefore, reasonable to consider a hybrid cloud-setup

where some tools run in the public cloud and other are only available in private clouds or dedicated machines

behind the walls of the licensed institution. In this paper, we report on a experiment of such scenario, where

a Matlab/Simulink

, LS-Dyna, and Model.CONNECT

powered co-simulation tool suite running in a

private cloud is combined with the DDD Simulation tool running inside a public cloud. Due to this setup it

was possible to combine a 1D hot stamping process simulation with a 3D visualisation. Finally the results

of the process simulation were improved by considering realistic movement of the robot. Our study elicited

several limitations and feature requests that need to be addressed to better support a hybrid cloud setup for

model-based design practitioners. We expect this initial contribution to trigger ground breaking research

encompassing all the community members interested in hybrid co-simulation setups.

1 INTRODUCTION

Systems gathering physical information through sen-

sors from the real world, processing them digitally

and autonomously affecting physical environment are

increasingly important in today’s world. These sys-

tems are called Cyber-Physical Systems (CPSs) and

are widespread in several industries, from transporta-

tion to smart cities and from smart manufacturing to

agriculture. As a consequence, manufacturers, al-

ready constrained by the need to provide a huge va-

riety and quantity of products, have also to embed

such systems in their products, making them smart

and connected (Porter and Heppelmann, 2014). To-

day, the change of production systems is gradually

becoming a need to sustain a position in the Indus-

try 4.0 domain, driven by computation, automation

and robotics, and declined in nine categories of inno-

vative digital technologies (autonomous robots, simu-

lation, horizontal and vertical system integration, IoT,

cybersecurity, the cloud, additive manufacturing, aug-

mented reality, big data analytics) (R

ußmann et al.,

2015). Data and observation become the strategic key

to support online decision-making in production sys-

tems reconﬁguration (Braun et al., 2012; Hasselbring

et al., 2019).

A signiﬁcant adoption of computation in pro-

duction is required enabling the achievement of the

higher-levels of the CPS 5C (connection, conver-

sion, cyber, cognition, and conﬁguration) architec-

ture (Lee et al., 2015). At this stage, systems op-

timise themselves and work at the full digital-twin

level. As a consequence, models of the system at

hand (i.e. Model-Based Design (MBD) tools, meth-

ods and competences) are required in manufacturing,

also boosting the role of software in manufacturing

composed by self-optimising CPSs and Digital Twins.

Also logistics, integrated with production systems,

have been strongly affected by digital technologies,

presenting several applications (autonomous robots

and vehicles, tracking and decision making systems,

smart products and cloud-supported networks, real-

time big data analytics), mainly gathered under the

umbrella of Internet of Things (IoT), to control the

Weiß, G., Pietraroia, D., Sassanelli, C. and Macedo, H.

Manufacturing Process Simulation in a Hybrid Cloud Setup.

DOI: 10.5220/0010641700003062

In Proceedings of the 2nd International Conference on Innovative Intelligent Industrial Production and Logistics (IN4PL 2021), pages 49-58

ISBN: 978-989-758-535-7

inventory level, enhance information ﬂows and deﬁn-

ing products’ optimal routing transport (Atzeni et al.,

2021). Normally, production system lifecycle is com-

posed by long development phases and is structured

in sequential phases (conception and planning, design

and engineering, construction, validation, veriﬁcation

and commissioning, use and disposal). In the devel-

opment phase of the production systems lifecycle, it

is common to run manufacturing simulations without

simulating the properties of the material during each

production step. On the other hand, it is state of the art

to simulate the process during the production prepa-

ration phase (Wiﬂing, 2021).

Model-based design and simulation are powered

by development tools and different modeling lan-

guages, but it is expensive to invest in such tech-

nologies, because any reasonably complex system re-

quires different models and tools making the MBD

ﬁeld a heterogeneous and challenging community

(Robinson et al., 2021) for new adopters. Also, in

some cases consortia of partners developing a model

of a manufacturing process may have complementary

expertise, tools and licenses, but the interaction needs

to be orchestrated on a case-by-case approach. One

solution to this problem is to deploy the tools in a

single cloud environment, as the HUBCAP platform

(Badicu et al., 2021), yet it is not possible to guar-

antee a complete coverage of all potential tools and

models that an arbitrary user may need, thus a hybrid

cloud solution where public cloud resources are used

with additional private cloud machines is a possible

alternative but not well analysed.

In this paper, we report on work towards analyzing

the feasibility of hybrid cloud approaches. The results

gathered during a feasibility study on a hybrid cloud

approach to support a co-simulation of a metal hot

stamping process experiment by two complementary

partners. The process simulation was done in a lo-

cal machine available only to one of the partners and

the 3D simulation of the manufacturing was done in

a cloud hosted tool. The combination allowed to ﬁne

tune the production parameters in the design phase

of the production system; despite the fact that a suite

of the tools for the co-simulation was hosted locally

(unavailable to the other partner) and that the process

involved manual data transfers. We expect our re-

sults encourage similar experiments by members in

the manufacturing community, and that the hybrid

cloud setup becomes a valid approach in future ex-

periments with tool licensing issues.

The remainder of this paper is structured as fol-

lows: Section 2 introduces the background on the

HUBCAP cloud platform. Section 3 describes the

experiment, the modeled process and the joint co-

simulation between tools in a private cloud and a tool

in the HUBCAP cloud platform. Finally, Section 4

contains the result of our co-simulation results, Sec-

tion 5 presents a discussion of the results, and Section

6 contains our concluding remarks.

2 SIMULATION IN HUBCAP

The methodology behind Model-Based Design eases

the development of Cyber-Physical Systems (Fitzger-

ald et al., 2014). It prescribes the development of

mathematical models of the system’s components,

which are then used to reason, test, and simulate the

system behaviour in a agile virtual setup. Develop-

ing in a virtual setup is known to reduce costs and

to elicit requirements and risks in shorter time, yet

it is often the case that a CPS system under design

is highly complex and encompasses multiple models

developed in multiple tools.

Although a solution for coupling the different

models exists since long time ago in the form of co-

simulation, practice shows one faces several barriers

for new MBD adopters, because the different mod-

els and tools usually run in diverse software environ-

ments, require advanced expertise and setups, and in-

volve complex legal/licensing agreements.

To lower the complexity of potential adopters

of MBD, the HUBCAP project developed a cloud-

based platform (Macedo et al., 2021) featuring ready-

made software environments, where providers pre-

pare sandboxes (software environments, tools and

models) that adopters can explore via their browser

to “test before invest” with such models and tools at a

lower cost of entry.

Sandboxes as deﬁned in (Larsen et al., 2020) are

composed of several virtual machine instances expos-

ing a remote desktop connection via a web middle-

ware, catering for the different software environments

(Operating System, Libraries and other Dependen-

cies) required to run tools and models. In addition,

the middleware around the virtual machines enables

several users to interact and collaborate with the pre-

pared models and tool by featuring an “invite guest”

button, that allows a sandbox owner to invite other

users as guests to a sandbox environment facilitating

discussion, training and demo sessions.

Figure 1 depicts the aspect of the sandbox web-

page. In the middle the user can interact with a re-

mote desktop of a virtual machine hosting the DDD

Simulator by TTS-Technology Transfer System S.r.l.

(a MBD tool used to provide solutions to manufactur-

ing companies and used in the experiments reported

in this paper). In the right-side we see some of the

IN4PL 2021 - 2nd International Conference on Innovative Intelligent Industrial Production and Logistics

additional controls/buttons a user resources to while

interacting with a sandbox.

Among the HUBCAP models and tools catalogue,

the users ﬁnd the INTO-CPS Application (Macedo

et al., 2020), with its co-simulation engine (Thule

et al., 2019), several models developed using the Vi-

enna Development Method and its associated tool

(Rask et al., 2020; Rask et al., 2021) the DDD Simu-

lator shown in Figure 1, but the catalogue is far from

covering the whole range of MBD tools. For instance,

the Model.CONNECT

co-simulation platform and

Matlab tools required for the experimental setup re-

ported in Section 3 are not available in the platform,

but are available and licensed for local usage inside

the institutions leading the experiments. This situ-

ation challenges the HUBCAP platform and might

hamper its wide adoption.

In addition, the HUBCAP sandbox concept was

designed in the cyber-security practice, where differ-

ent users’ resources run in disjoint and tightly con-

tained virtual machines ecosystems (operating sys-

tem, storage, networks,. . . ) to ensure a secure envi-

ronment (Kulik et al., 2021). Nevertheless, in some

cases, a user may be required to combine tools avail-

able in the public platform with tools running in a pri-

vate local server or neither available nor licensed to

run in the HUBCAP platform. For instance, in the

experimental setup detailed in Section 3 requires the

combination of a tool deployed as a HUBCAP sand-

box and a tool running a on-premises Desktop. Such

hybrid cloud setups are currently not a feature of the

HUBCAP platform, and this work clariﬁes the limits

and possible developments to support hybrid simula-

tion setups.

3 EXPERIMENT

Our experiment joins tools in a local machine with

a tool running in a HUBCAP sandbox that are used

together to jointly produce a open-loop co-simulation

of a metal hot stamping (press hardening) production

process which consists of three stages:

• furnace - modelling the heating of the unformed

metal plates,

• transfer - modelling a robotic manipulator moving

the plates from oven to press,

• forming press - modelling the draw and quenching

of the plate in the die.

The hybrid-cloud concept of the experiment is de-

picted in Figure 2, which highlights the dataﬂow be-

tween the public and private environments. The ma-

terial properties of the metal sheets are simulated on

a local PC while the 3D manufacturing simulations

runs in the HUBCAP sandbox environment. The ex-

periment was designed in that way for a simple rea-

son: the vendors of the tools used for the process

simulation are not part of the HUBCAP ecosystem.

An open loop approach is followed due the fact that

the Sandbox is an isolated (and secure) environment

which does not support a direct API (application pro-

gramming interface) interface. Nevertheless data can

be upload and downloaded. Thus, the following pro-

cedure has been followed:

1. 1D process simulation on a local PC (origin of the

use case) and upload of the results to the HUB-

CAP sandbox,

2. creating 3D - model of the manufacturing and

simulation based on timing of the 1D Simulation,

3. adding logic and kinematics to 3D-model,

4. DDD-Simulation of manufacturing which results

in new timing setting for the production steps,

5. adaptation of the process simulation according to

DDD-Simulation,

6. ﬁne tuning of the process parameters.

3.1 Process Simulation

The focus of the process simulation is the temperature

and yield stress behaviour of the metal plates at each

production stage. The simulation is based on existing

models from Virtual Vehicle Research Gmbh (VV)

and has been used for a digital twin simulation (Wi-

ﬂing, 2021):

• The furnace or oven model simulates the tempera-

ture increase of the plates and has as (production)

parameters the oven temperature T

oven

, number of

slots and minimum duration time in the oven t

min

• The transfer corresponds to the movement of a

robot which means a slow cool down phase for

the plates. There is only the transfer time as pa-

rameter t

trans

• Finally, the forming press includes two steps:

drawing and quenching which means a rapid cool

down of the plates. A key parameter of the press

model is the closing speed, which is actually not a

ﬁxed parameter but an input.

The calculation of the temperature is based on

(Weiß, 2013) chapter 6 and (Weiß et al., 2013), while

the calculation of the yield stress is based on a tem-

perature dependent stress strain table.

Figure 3 shows the simulation setup using

Model.CONNECT

as co-simulation platform. A

Python script is used to create (virtual) ID’s for the

Manufacturing Process Simulation in a Hybrid Cloud Setup

Figure 1: Snapshot of the DDD Simulator tool running inside a HUBCAP sandbox environment.

Figure 2: Hybrid cloud experimental setup: A local machine emulates on-premises private cloud hardware hosting

Matlab/Simulink

, LS-Dyna, and Model.CONNECT

; the HUBCAP platform serves as the off premises public cloud

infrastructure hosting the DDD Simulation tool; and the dataﬂow.

metal plates (source of the metal plates) in intervals of

15 seconds. The plates are taken by the oven model

as soon as a free slot is available. Both, the oven

and the transfer model are implemented in MATLAB

Simulink. Another Python script calculates the clos-

ing speed of the press depending on the temperature

after the transfer based on polynomial second order.

Finally, the drawing and quenching is modelled in LS-

Dyna. LS-Dyna uses ﬁnite element analysis to calcu-

late the sheet temperature while the Simulink models

are time based. The following co-simulation settings

have been used:

• coupling time (interval of data exchange): 10ms

(for all models)

• scheduling: sequential, with the following order:

ID generation, oven model, transfer model, press

model

The temperature and yield stress values of each step

have been saved in csv format including a time vector.

Also percentage status of the transfer and the posi-

tion of the press over time has been saved in csv ﬁles.

Figure 4 shows the results of the LS-Dyna model us-

ing LS-PrePost for visualisation. The stroke which is

coming down to form the the plate is colored in red.

The metal plate itself is colored in black (thin plane)

and the form/shell for the ﬁnal shape of the plate is

colored in green.

3.2 3D Manufacturing Simulation

The ﬁrst 3D environment realized for the HUBCAP

project is a visualization environment that represents

the results of the process simulation over time through

an animation. The tool used to create the 3D simu-

lation is DDDSimulator, a discrete event simulation

software with integrated 3D environment and DES

(discrete-event simulation) simulation, used to build

simulation models of production and logistic facilities

to simulate the system behavior and improve the per-

IN4PL 2021 - 2nd International Conference on Innovative Intelligent Industrial Production and Logistics

formances of existing or new systems. The csv output

ﬁles from the 1D model contain the time-series val-

ues of the variables representing the state of the pro-

cess. While running, the simulation model refreshes

the state of the system every 50 ms and parses all the

input ﬁles updating the environment according to the

new state. Some state changes affect the 3D environ-

ment while other, as the temperature of a plate or the

yield stress of the plate inside the press are plotted on

a dashboard. The 3D model consists of a sketched

geometry for the oven and the press while a 6 DOF

FANUC robot has been used to represent the trans-

portation system. The oven contains 9 slots and in

each slot one plate can be placed . Since there isn’t

any geometrical information about the oven, the posi-

tion of the 9 slots are modeled as frames which iden-

tify the coordinates in which the plates are placed in

space and each slot is placed approximately 100 mm

over the other along the Z axis. The transfer informa-

tion provided is the percentage of the movement from

the oven: 0 means the plate just left the oven and 100

means that the plate arrived into the press. So, when

the ID of a plate changes, the plate is moved from the

oven to a pick position at the base of the oven. Then

the plate is attached to the hand of the robot and the

angle of the base joint if the robot is updated.

The input data read from the data ﬂowing from the

process model are:

• Creation of a new plate: each time a new sheet ID

is generated, the model creates a 3D box repre-

senting the new plate,

• Plate ID for each plate in the oven: a ﬁle reports

the id of each plate in each slot and plates are po-

sitioned at the corresponding position,

• Temperature of each plate in the oven: the tem-

perature for each plant in the oven is plotted on a

dashboard and the color of the plate changes ac-

cordingly,

• Transfer state: the transfer state is refreshed and

the robot position updated,

• Press position: the z position press die is red and

the position of the die is updated in the 3D model,

• The yield stress of the plate in the press is plotted

on the dashboard.

To complete the interaction, a pure-simulation

conveyor which carries away the worked plate has

been added, this additional module doesn’t depend on

any parameter from the ﬁrst simulation but moves the

piece with a constant velocity. This ﬁrst phase of the

experiment worked as a proof of concept to interface

the two tools through the HUBCAP platform.

To provide a more signiﬁcant case, an experiment

which combined the usage of the two simulations was

created. In the new experiment, a kinematic model

of the transfer between the oven and the press using

the 6 DOF (degree of freedom) robot was created, the

model is used to estimate transfer times. Figure 5

shows the 3D environment with the robot extracting

a plate from the oven. The resulting values are then

used as inputs by the 1D simulation model to evalu-

ate the thermal characteristics of the process and the

deformation of the metal sheet in the press. Such re-

sults can then be transfer back to the 3D simulation

model and visualized in the 3D animation. The previ-

ous model didn’t also take in account the approaching

stroke of the press which is needed to bring the die

in contact with the plate, once the metal sheet was

put into the press the process started instantly. In the

kinematic model an approaching stroke was added to

simulate the additional operation and thus the addi-

tional transfer time for the plate.

4 RESULTS

Although the process involved manual steps to

transfer data between the tools, the open-loop co-

simulation of the metal hot stamping process using

the hybrid cloud setup was feasible and in the fol-

lowing we highlight the improvement achieved by the

combination of the tools.

Figure 3: The connection diagram for the different components of the co-simulaton.

Manufacturing Process Simulation in a Hybrid Cloud Setup

(a) Initial state: plate (plane between red and green shapes)

arrives at the press, stroke (in red) at initial position

(b) Intermediate state: stroke going down

ﬁnished

Figure 4: Press states calculated by LD-Dyna and visual-

ized in a post process using LS-PrePost.

4.1 Process Simulation

Figure 6 shows a comparison of the temperature

traces of the initial simulation (step 1 according to the

enumeration in section 3) and the adapted and ﬁne

tuned simulation (step 6) at each production stage.

New sheets are create in intervals of 15 seconds,

which is the bottleneck in the initial simulation since

trans

was assumed with 5s. However, the DDD-

Simulation showed that a realistic movement of the

robot from the oven to the press takes 14 seconds

plus 3 seconds more for grasping (see table 3), which

means a new sheet can be taken after 34s. That new

bottleneck can be clearly seen in the ﬁgure. While the

start-up phase (time span till all oven slots are occu-

pied) is the same for both simulations it can be seen

Figure 5: 3D simulation with kinematic simulation of trans-

fer.

that the throughput f

t p

is much lower in case of the

adopted simulation.

Figure 7 shows a one to one comparison of a sin-

gle sheet after the startup phase (time shifted to zero

for comparison purposes). The ﬁrst row shows the

sheet temperature trace for all production stages while

the seconds one shows the yield stress. In case of the

adopted simulation (dashed) it can be seen that the

sheet remains much longer in the oven, while in case

of the initial simulation it remains for the minimum

time of 150s. Another consequence of the increased

transport time is that T

oven

had to be raised to 850°C in

order to reach more or less the same ﬁnal yield stress

value at the end of the production. The change of the

oven temperature belongs to the ﬁne tune step (step

6). The results of the adopted and ﬁne tuned results

can be displayed again in the 3D simulation.

Table 1 gives an overview of the simulation pa-

rameters and table 2 shows how long it takes to cre-

ate one piece t

sheet

, the throughput for an hour f

t p

af-

ter the start up phase and the ﬁnal yield stress value

f inal

Table 1: Simulation parameters.

oven

min

slots t

trans

◦

C s s

initial sim 800 150 9 5

modiﬁed sim 850 150 9 17

Table 2: Performance and outcome (steady state).

sheet

t p

f inal

s pieces/h MPa

initial sim 157.4 240 197.8

modiﬁed sim 341 61 181.1

4.2 Visualization of Results from 1D

Simulation

As previously stated, model based design techniques

often involve the usage of multiple tools and involves

different expertises. The dynamic visualization of

IN4PL 2021 - 2nd International Conference on Innovative Intelligent Industrial Production and Logistics

Figure 6: Comparison of the process simulation: the left column shows the initial simulation while the right column shows

results of the adapted and ﬁne tuned simulation. First row: the temperature values of the metal sheets at each slot of the oven.

The second row: temperature decrease of the metal sheets during the transfer. Last row: the temperature trace of the sheets

during stamping.

Figure 7: One to one comparison of a single sheet: the dashed line shows the trace of the initial simulation while the

continues line shows the trace of the adapted and ﬁne tuned simulation. The different colours of the sections indicate the

current production stage.

the results of process simulation evolving over time

inside a sketched 3D environment greatly helps the

analysis. In the case of our experiment, one was able

to follow the evolution of temperature parameters and

yeld stress for all the plates in the system. The anal-

ysis perfrormed by both experts of material process

simulation and DES event expert simulation was the

basis for the second part of the experiment.

Manufacturing Process Simulation in a Hybrid Cloud Setup

4.3 3D Kinematic Simulation of System

The use of 6DOF kinematics simulation allowed to

better estimate the transfer times. Table 3 shows the

resulting moving time for different phases of the pro-

cess, the resulting interval became the new transfer

time data.

Table 3: Times for each phase of transfer operation.

Movement phase (s)

extraction from oven 3

transfer to intermediate position 6

transfer to press 2

move to safe position before press starts 3

press rapid stroke 3

Total 17

The model also allowed to study the time that

the plates are stored in the oven, table 4 shows that

the system reaches a steady-state state for the heating

time after 10 plates.

Table 4: Oven time during ramp up.

plate id oven time (s)

0 153

1 173.998

2 194.996

3 215.994

4 236.992

5 257.99

6 278.988

7 299.986

8 320.984

9 320.982

10 320.982

11 320.982

For the purpose of designing the production line,

the model shows that the bottleneck is the robot, since

it’s used both to load and unload the press. A possi-

ble solution to increase the productivity would be to

use another robot to unload the press, this could allow

not only an increase of the throughput but also some

energy saving reducing the oven temperature and the

new temperature could be estimated using the process

simulation.

5 DISCUSSION

The 1D process simulation is an abstract simulation

of the production. Figure 6 shows indeed important

material properties of the manufacturing stages but it

is not really easy to understand what happens, while

a 3D animation gives a much better impression. By

simulating a 3D model of the production line and dis-

playing the results of the 1D process simulation both

kind of simulation have been combined. The ﬁrst 3D

simulation was purely based on the 1D results, allow-

ing the audience easily to follow the manufacturing

steps and to see the impact on the material properties

of the sheets. However, this was done without consid-

ering the kinematics of the robot.

The simulation has changed from a pure visuali-

sation to a DDD-simulation by adding logic and kine-

matics’s to the 3D model. It turned out that the as-

sumed transfer time of 5 seconds was not a realis-

tic parameter value in the initial process simulation.

As a consequence the robot becomes a bottleneck in

the production and the throughput decreases signif-

icantly to 25% of the initial simulation. Moreover,

going back to the process simulation and adapting the

the transfer time parameter showed that the ﬁnal yield

stress values was not satisfyingly anymore due to the

temperature decrease during the transfer. Thus, the

oven temperature had to be raised.

Another aspect of the demonstrated use case is the

collaboration between VV and TTS. Data generated

by VV could be easily uploaded to the shared sand-

box, where TTS further proceeded it by simulating a

3D environment. Moreover, TTS added it’s own sim-

ulation logic and kinematics which in turn, changed

relevant simulation parameters for the process simu-

lation. Finally, VV had to ﬁne tune the simulation and

generated again new data for the 3D simulation.

6 CONCLUDING REMARKS

In this paper, we report on a experiment of a combina-

tion of co-simulation tools running in a hybrid cloud

setup. The locally licensed Matlab/Simulink

LS-Dyna, and Model.CONNECT

powered co-

simulation tool suite was hosted on-premises and

combined with the DDD Simulation tool running in-

side a off-site cloud. The experiment shows the com-

bination is feasible and enables the production and

improvement of co-simulation results in the cases

where licensing issues require a hybrid cloud ap-

proach. Moreover our study provides evidence of the

feasibility of such hybrid cloud setups in supporting

model-based design. Nevertheless, there are known

drawbacks/challenges associated with hybrid cloud

setups, mainly due to the fact that those setups in-

volve different institutions, and we conﬁrmed some of

the challenges, in particular we were required to intro-

duce manual steps to perform dataﬂow, which would

be easy to automate in case the experiment was com-

IN4PL 2021 - 2nd International Conference on Innovative Intelligent Industrial Production and Logistics

pletely hosted on-premises. To realize the full poten-

tial of a hybrid cloud co-simulation solution based on

the HUBCAP platform the following need to be ad-

dressed in future work:

First and foremost, automated orchestration of the

cloud environment via an API or a federated cloud

approach where users can add their private resources

as computation nodes (virtual machines) to a particu-

lar HUBCAP sandbox would ease the dataﬂow of our

experiment that could be automated instead of a man-

ual data iteration (upload new timing settings from a

conﬁguration ﬁle, run experiment, download result...)

Also, the improvement of intellectual property protec-

tion and licensing makes the platform more suitable

for adoption by industrial partners. The co-simulation

data and environment shared between TTS and VV

was protected under existing contractual agreements

between the two partners. To be useful, the coupling

of the private resources and hybrid cloud should also

be available to be used by other partners in the MBD

community without any legal binding. Both by guar-

anteeing the privacy of the platform user data and by

allowing external partners to host their IP protected

components locally.

ACKNOWLEDGEMENTS

The work presented here is partially supported by the

HUBCAP Innovation Action funded by the European

Commission’s Horizon 2020 Programme under Grant

Agreement 872698.

The publication was partly written at Virtual Ve-

hicle Research GmbH in Graz, Austria. The au-

thors would like to acknowledge the ﬁnancial support

within the COMET K2 Competence Centres for Ex-

cellent Technologies from the Austrian Federal Min-

istry for Climate Action (BMK), the Austrian Federal

Ministry for Digital and Economic Affairs (BMDW),

the Province of Styria (Dept. 12) and the Styrian

Business Promotion Agency (SFG). The Austrian Re-

search Promotion Agency (FFG) has been authorised

for the programme management.

REFERENCES

Atzeni, G., Vignali, G., Tebaldi, L., and Bottani, E. (2021).

A bibliometric analysis on collaborative robots in lo-

gistics 4.0 environments. Procedia Computer Science,

180:686–695.

Badicu, A., Iordache, G., Suciu, G., Macedo, H. D., Sas-

sanelli, C., Terzi, S., and Larsen, P. G. (2021). De-

ploying the Smart Energy Tool for Investment Simu-

lation inside the HUBCAP Sandbox. To appear at the

9th International Workshop on Simulation for Energy,

Sustainable Development and Environment.

Braun, S., Bartelt, C., Obermeier, M., Rausch, A., and

Vogel-Heuser, B. (2012). Requirements on evolution

management of product lines in automation engineer-

ing. IFAC Proceedings Volumes, 45(2):340–345.

Fitzgerald, J., Larsen, P. G., and Verhoef, M., editors

(2014). Collaborative Design for Embedded Systems

– Co-modelling and Co-simulation. Springer.

Hasselbring, W., Henning, S., Latte, B., M

obius, A.,

Richter, T., Schalk, S., and Wojcieszak, M. (2019).

Industrial devops. In 2019 IEEE International Confer-

ence on Software Architecture Companion (ICSA-C),

pages 123–126. IEEE.

Kulik, T., Macedo, H. D., Talasila, P., and Larsen, P. G.

(2021). Modelling the HUBCAP sandbox architec-

ture in VDM: A study in security. John Fitzgerald,

Tomohiro Oda, and Hugo Daniel Macedo (Editors),

page 20.

Larsen, P. G., Macedo, H. D., Fitzgerald, J., Pfeifer, H.,

Benedikt, M., Tonetta, S., Marguglio, A., Gusmeroli,

S., and Jr., G. S. (2020). A Cloud-based Collaboration

Platform for Model-based Design of Cyber-Physical

Systems. pages 263–270. INSTICC, Proceedings of

the 10th International Conference on Simulation and

Modeling Methodologies, Technologies and Applica-

tions - Volume 1: SIMULTECH.

Lee, J., Bagheri, B., and Kao, H.-A. (2015). A cyber-

physical systems architecture for industry 4.0-based

manufacturing systems. Manufacturing letters, 3:18–

23.

Macedo, H. D., Rasmussen, M. B., Thule, C., and Larsen,

P. G. (2020). Migrating the into-cps application to

the cloud. In Sekerinski, E., Moreira, N., Oliveira,

J. N., Ratiu, D., Guidotti, R., Farrell, M., Luckcuck,

M., Marmsoler, D., Campos, J., Astarte, T., Gonnord,

L., Cerone, A., Couto, L., Dongol, B., Kutrib, M.,

Monteiro, P., and Delmas, D., editors, Formal Meth-

ods. FM 2019 International Workshops, pages 254–

271, Cham. Springer International Publishing.

Macedo, H. D., Sassanelli, C., Larsen, P. G., and Terzi,

S. (2021). Facilitating model-based design of cyber-

manufacturing systems. To appear in the proceedings

of the 54th CIRP Conference on Manufacturing Sys-

tems.

Porter, M. E. and Heppelmann, J. E. (2014). How

smart, connected products are transforming competi-

tion. Harvard business review, 92(11):64–88.

Rask, J. K., Madsen, F. P., Battle, N., Macedo, H. D., and

Larsen, P. G. (2020). Visual Studio Code VDM Sup-

port. In Fitzgerald, J. S. and Oda, T., editors, Pro-

ceedings of the 18th International Overture Workshop,

pages 35–49. Overture.

Rask, J. K., Madsen, F. P., Battle, N., Macedo, H. D., and

Larsen, P. G. (2021). The Speciﬁcation Language

Server Protocol: A Proposal for Standardised LSP

Extensions. In Proenc¸a, J. and Paskevich, A., edi-

tors, Proceedings of the 6th Workshop on Formal In-

tegrated Development Environment, Held online, 24-

25th May 2021, volume 338 of Electronic Proceed-

Manufacturing Process Simulation in a Hybrid Cloud Setup

ings in Theoretical Computer Science, pages 3–18.

Open Publishing Association.

Robinson, C., Bagnato, A., Pastrone, C., Weyer, T., Popov,

P., Macedo, H. D., Sassanelli, C., Larsen, P. G., Luz,

C. H., Henshaw, M., Buron, C., Akerkar, R., and

orfﬁ, M. (2021). Bridging the stakeholder commu-

nities that produce cyber-physical systems. In et al,

M. D., editor, HiPEAC Vision 2021, volume 12, pages

20–29.

ußmann, M., Lorenz, M., Gerbert, P., Waldner, M., Justus,

J., Engel, P., and Harnisch, M. (2015). Industry 4.0:

The future of productivity and growth in manufactur-

ing industries. Boston Consulting Group, 9(1):54–89.

Thule, C., Palmieri, M., Gomes, C., Lausdahl, K., Macedo,

H. D., Battle, N., and Larsen, P. G. (2019). To-

wards Reuse of Synchronization Algorithms in Co-

simulation Frameworks. In Co-Sim-19 workshop.

Weiß, W. (2013). Thermische Auslegung von Werkzeugen

ur Pressh

artprozesse. PhD thesis.

Weiß, W., Machhammer, M., Krall, F., and Kolleck, R.

(2013). Simulation of the hot forming process in con-

sideration of the thermal state of the tools. In Pro-

ceedings CHS2, Hot Sheet Metal Forming of High-

Performance Steel, volume 4, pages 91 – 98, Ger-

many. Verlag Wissenschaftliche Scripten. 4th Inter-

national Conference on Hot Sheet Metal Forming of

High-Performance Steel CHS² ; Conference date: 09-

06-2013 Through 12-06-2013.

Wiﬂing, M. (2021). A lean approach to foster zero

defect and ﬁrst-time-right manufacturing con-

trol in a smart factory. https://www.v2c2.at/

building-the-functional-digital-twin-in-a-smart-factory.

[Online; accessed 7-April-2021].

IN4PL 2021 - 2nd International Conference on Innovative Intelligent Industrial Production and Logistics