Documente Academic
Documente Profesional
Documente Cultură
Authors:
Kilian Bächle
Account Manager IoT
CONTACT Software GmbH
kilian.baechle@contact-software.com
Stefan Gregorzik
Manager Corporate Development
CONTACT Software GmbH
stefan.gregorzik@contact-software.com
1Plattform Industrie 4.0, ZVEI (eds.): Details of Asset Administration Shell. Specification Part 1 – The exchange of information
between partners in the value chain of industry 4.0 (version 1.0). Berlin Nov. 2018
2 Industrial Internet Consortium, Plattform Industrie 4.0 (eds.): Architecture Alignment and Interoperability. White paper. 2017
form the practical foundation for our pro- the core concept of an “asset” and describes
posals. Sections 4 and 5 present our conclu- the following aspects:
sions in terms of the contents and behavior
Identification and description of an
of digital twin models. These give rise to sug-
asset
gestions about how to realize such digital
Submodels related to this asset
twin models, which we present in section 6,
and we conclude with an examination of the Property definitions to specify data
outlook for the future (section 7). from operations
Access Control
DATA MODELS FOR DIGITAL TWINS At least one crucial use case for the digital
twin data model mentioned above takes the
Probably one of the best-known data models
form of information exchange between part-
for digital twins is the “Asset Administration
ners in the supply chain, e.g. between “sup-
Shell” as formulated by Plattform Industrie
pliers” and “integrators.”
4.0. In short, this model is structured around
Figure 2: Applications for IIoT (x = % of matches from a cluster analysis of free text answers) (Smart Industrial Products, pg. 111)
3 Gartner Inc.: Innovation Insight for Digital Twins — Driving Better IoT-Fueled Decisions. 22 March 2017.
4 Müller, P.; Lindow, K.; Gregorzik, S.; Stark, R. (eds.): Smart Industrial Products. Berlin/Bremen/Düsseldorf 2019.
The study analyzed data collected from 183 experts representing different branches of industry, such as the automotive sector,
machine engineering and plant construction. 75% of these experts exercised development or product management functions.
requires sophisticated condition monitoring The manufacturer offers a full range of ser-
and maintenance control. vices from the design and manufacturing of
the equipment through to its installation,
As these vehicles are provided by different
maintenance and system overhauls. In most
suppliers, there is no common fleet-wide
cases, tasks are divided, with the manufac-
monitoring logic or even any common moni-
turer and customer working together to
toring mechanism, and the logistics supplier
commission the system while standard-
has had to implement fleet control on its
maintenance and small extensions are per-
own. Of course, weather conditions play a
formed by the customer, and more exten-
significant role here. The objective of such a
sive or complex modifications are under-
fleet control system is to monitor the differ-
taken by the manufacturer.
ent (groups of) vehicles and consolidate the
data for the joint reporting of general readi- At this point, a potential conflict arises when
ness information. Furthermore, a predictive the customer makes changes to the system
maintenance system based on historical without informing the manufacturer and
data from the vehicles' internal control sys- then subsequently requests a modification
tems is intended to proactively trigger or complains about a system failure. The first
measures to prevent malfunctions and challenge then facing the manufacturer is to
breakdowns. Automated spare parts provi- identify the up-to-date as-maintained state
sion might be the next logical step. of the system in order to have a reliable
baseline for any further activity.
This case gives rise to the following IIoT re-
quirements: The key abilities required of an IIoT system in
this context are to:
Consolidate key measures of system
behavior from different suppliers/as- Track control parameters (pressure,
sets speed, etc.)
Consider environmental influences Describe the as-maintained-state in
Define limit values for parameters terms of the components and assem-
and related measures bly situation
Provide spare parts orders Integrate a description of third party
components (modifications by cus-
Case 3: Maintenance and modification
tomer)
of assets
Describe the interaction with other
The third case involves a manufacturer of surrounding production assets
large-scale plant systems for handling food
Case 4: Monitoring of overall systems
products. Based on customer requirements,
and predictive maintenance
the manufacturer combines a number of
modules and functional units to create an in- The fourth case relates to train operation.
tegrated system (configure-to-order). This Since trains are operated for many hours
system is then interlinked with the custom- every day, often in tough environments, the
er's production environment.
requirements in terms of material, compo- Based on this use case, the most critical IIoT
nent and condition monitoring are ex- requirements are to:
tremely high. A carriage breakdown causes
Consolidate key measures of system
major difficulties. Furthermore, the with-
behavior
drawal of a train from service has serious
Consider environmental influences
timetabling implications since replacing it
may take hours. During operation, the repair Describe the as-maintained-state in
options for trains are limited as the railway terms of the actual components in-
depot is often the only site where proper stalled, including third party spare
maintenance and repair can be performed. parts
Describe the interaction with sur-
Carriage doors, in particular, are a crucial rounding components of the overall
component of a train. Any door failure can system
extend the stopping time in stations and
cause delays. Furthermore, the safety buffer CONTENT OF A DIGITAL TWIN DATA
for evacuating trains in the event of an emer-
MODEL
gency is reduced if an exit is blocked; and if
there are too many such exits, the train may Regarding the requirements placed on digi-
not be allowed to operate at all. At the same tal twin data models, we need to start from
time, doors are sensitive to contamination, a general view. In the survey mentioned
and, as regular moving parts, mechanical prior (“Smart Industrial Products”), the three
and drive elements suffer from high levels of most important elements of a digital twin
wear. identified by the industry experts were: 1)
As trains have a long product lifecycle, the the software version installed on the asset,
replacement of parts and the refurbishment 2) the asset configuration in terms of prod-
of subsystems or even complete carriages is uct-specific parameters and 3) the unique ID
a regular activity. In many cases, such modi- of the asset. Additional requirements just
fications to trains, regarded as a system in slightly outside of the top three were 4)
themselves, may influence the behavior of views of components from the perspective
the doors. To predict their functionality or of different disciplines, 5) related documen-
the risk of breakdowns based on sensor tation and audit information and 6) 3D visu-
data, it is therefore necessary to possess an alization (see Figure 3).
overall up-to-date as-maintained-state of
the carriage.
Software version 78
Unique ID 71
Audit Trail 68
Simulation models 61
Vendor information 58
3D visualization 57
0 10 20 30 40 50 60 70 80 90 100
Figure 3: Requirements for contents of Digital Twins (x = % of positive answers, multiple selections possible) (Smart Industrial
Products, pg. 68)
behavior. As the analysis of asset behavior is model if a fully-fledged basis for IIoT applica-
mostly based on time series data, the config- tions is to be achieved.
uration history of an asset's component
Behavioral Data
structure is of great relevance (e.g. for audit
purposes). Data describing the behavior and operation
of assets takes the form of values from sen-
Environment
sors (such as temperatures, flows) and mes-
Assets are influenced by their environment sages from controllers (e.g. error warnings,
in terms of surrounding systems (e.g. car- ready states). Both types of data may appear
riages), installation sites (such as drilling in different technical protocols and may
platforms) and environmental conditions have significantly different characteristics:
(such as hot or cold weather, presence of Sensor data usually takes the form of a con-
dust, etc.). Thus, the proper recording of tinuous stream of values while messages are
such environmental data is a relevant as- discrete and intermittent.
pect.
Environmental Data
Models and descriptions
Data describing the situation and the envi-
A simple verbal description coupled with the ronment in which the asset operates (e.g.
component structure may not suffice for all temperature, humidity) may originate from
the data analytics requirements relating to sensors in the asset itself or its environment
an asset. Additional aspects such as geome- or can be provided by web services. In light
try or the operating or maintenance situa- of this variety of possible data sources, it is
tion (e.g. motor within a vehicle) are an es- obvious that different technical protocols
sential part of the digital twin data model. may be used to transfer this data (possibly
even in the same way as for the behavior
Control Parameters (including history)
data). It therefore seems reasonable to con-
To define the operating profile of an asset, sider this data separately in a digital twin
the software programs and parameters used data model.
(such as settings for motor speed, pump
Finally, another type of data needs to be in-
pressures) must be recorded and stored.
cluded within a digital twin data model:
Once more, the data history is of great rele-
vance because changing these parameters Connectivity Parameters
modifies the behavior of the asset.
To be able to receive data from and send
In our opinion, the definition of a digital twin data to an asset, information on how to ad-
data model is not complete without these dress the asset in the (inter)net is required.
reference characteristics. As a smart asset This means unique addresses such as IP’s or
delivers data in an individual format, the de- MAC declarations are required together with
scription of these data streams (not the field a description of the authorization method.
data in terms of “payload” itself but the
meta-information) must be part of the data
As mentioned above, data can appear in dif- suppliers may not work with the same soft-
ferent protocols. Another aspect of connec- ware.
tivity parameters therefore relates to the
Therefore, one initial requirement for main-
definition of these protocols and the formats
taining digital twin model data is the ability
to be interpreted in the IIoT system.
to consolidate partial models from different
parts of the supply chain into a single model
MAINTAINING A DIGITAL TWIN DATA that is associated with the final asset.
MODEL
Maintenance and extensions
The major challenge associated with all the
While the tasks described above are already
references and meta-descriptions men-
complex, the main challenge actually arises
tioned in Section 4 relates to how to gener-
during operation over the lifecycle of the as-
ate them and keep them up to date. The fol-
set. Maintenance and refurbishment can
lowing requirements apply to the mainte-
substantively change the way an asset be-
nance process:
haves, and new or replacement components
Consolidation will be installed in it—including from suppli-
ers that were not part of the original OEM
As we have seen in the use cases (Section 3),
supply chain. Third party maintenance pro-
a digital twin may relate to a product con-
viders may modify assets without reference
taining components that originate from mul-
to the OEM if they possess corresponding
tiple sources. Industrial products typically
service-level agreements with the operating
consist of a number of third party compo-
company.
nents assembled in deep supply chains with
tier-1, tier-2 or other suppliers. To make it This leads to the obligation on the part of the
possible to create a digital twin of a final owner or operator of the asset to keep the
product, these sources have to be connected digital twin up-to-date so that it describes
and consolidated into a common model to the asset's as-maintained-state.
reflect the complete, fully updated as-deliv-
A second requirement for digital twin mod-
ered-state.
els therefore refers to the possibility of
Hence, the concerns and challenges involved maintaining and extending them—even if
in keeping the digital twin up to date also ap- the operator is not part of the manufacturing
ply to the OEMs in the supply chain. The supply chain. This also includes the friction-
OEM will require up-to-date twin descrip- less handover of the digital twin data from
tions of all the components in a way that per- the manufacturing process to the operating
mits the (automated) compiling of partial process owner.
models from the different suppliers in the
Granularity
supply chain (or a number of pre-consoli-
dated models from the tier-1 suppliers). It is The task of consolidating and maintaining a
clear that models have to be interoperable digital twin’s data during operation raises
across system boundaries since OEMs and the question of the level of detail the partial
- 10 - November 2019
Digital Twin in Industrial Application – Requirements to a Comprehensive Data Model
models and the overall model should pos- of transfer mechanisms, such as dedicated
sess. From the point of view of the final as- services or infrastructures (e.g. cloud plat-
set, a detail in a given component (e.g. a re- forms). Stakeholders must be able to choose
sistor in a motor) might not be of interest to their own way to communicate with each
the operator, whereas it could be a critical other.
part for the manufacturer of the component
Another aspect of independence concerns
(motor).
the portability of a digital twin: one of the
The way in which digital twins are handled players involved in the overall lifecycle of the
throughout the supply chain must therefore asset may want to replace the existing IIoT
include methods that make it possible to se- system with a new one. It is therefore critical
lect and filter contents from one stage to an- to be able to transfer the complete digital
other. As far as the top-level view of the final twin data model(s), e.g. including data dic-
asset is concerned, only components that tionary, connectivity etc.
have an influence on the asset's behavior or
functioning might be relevant in terms of CONCEPTS TO REALIZE A DIGITAL
monitoring and analytics. However, in terms TWIN DATA MODEL
of a closed loopback to product design or
spare parts management, other components When considering the requirements placed
may play a role in the asset's business pro- on the reference data of the digital twin
cesses. model and its maintenance, we need to ad-
dress the question of the complex data
Obviously the required level of detail is a
structures that have to be described along
question of product design and the business
the supply chain. This question relates to
case associated with the asset. As compo-
both the data sources for the twin data
nents are often used in several different
model and the format used to describe and
products, the OEM’s data requirements for a
transport it.
specific component in no way constitute a
generally applicable component model—an- Data Sources of Digital Twins
other OEM using the same component might
In our view, the only reasonable source for
have a totally different usage scenario and
the essential structural data is to be found in
therefore a different view of the component.
Product Lifecycle Management (PLM),
Digital twin data models therefore need to where consolidation capabilities, references
provide options for controlling the granular- to the product structure and, in particular,
ity of the specific models that are processed. change information (including versions of
structures) are available. We therefore re-
Distribution and porting
It is obvious that unproblematic data trans-
fer must be possible if the exchange of digital
twins between stakeholders is planned. This
means that it is necessary to be independent
gard PLM as the mechanism capable of es- with the resulting digital twin. This digital
tablishing digital twins with solid founda- twin then reflects real-world data and makes
tions and consistent contents5. it possible to utilize the data upstream in the
process in order to optimize design and pro-
PLM deals with product information in a very
gress (“closed-loop engineering”).
detailed and granular manner due to its de-
sign focus (“Engineering-BOM”). Applying Describing and transporting Digital Twins
this information to a digital twin makes
The next aspect is to describe a technical
downsizing (quantitative reduction) neces-
means that makes it possible to capture and
sary: as described above, not every screw or
distribute digital twins.
resistor is relevant for the monitoring and
analysis of operations. In fact, quite the op- As an analogy, one could say that a STEP-
posite is true: too many details would cause model7 for digital twins is required, i.e. a for-
friction and overhead in the processes. mat allowing for the exchange of digital
twins over system boundaries in high pro-
While reducing the structures involved in
cess quality. IIoT systems are mostly and—as
digital twins is highly beneficial, it is crucial
far as we can see—will generally remain pro-
not to lose the link to the design perspective.
prietary, at least in their specific customiza-
This is vital for channeling feedback from the
tions, due to the business cases they have to
field into the engineering processes. To im-
cover: we need only consider the above-
plement such a closed-loop engineering pro-
mentioned different scenarios and system
cess, it is necessary to take a retrospective
requirements that apply to OEM and compo-
view from the downsized structure of the
nent suppliers as an example. If we bear
digital twin to the engineering-BOM.6
these differences in mind, it seems normal to
Up to this point, the data provision process assume that there will be friction and data
is very similar to that found in collaboration loss (e.g. due to improper matching, incom-
scenarios in engineering, e.g. in the case of patible categorization, etc.) at the interfaces
joint ventures or consortia. In engineering, between the source and target systems
this is the point at which we regard the “ab- (based on a comparison of the quality of the
stract” product in terms of a design. How- internal data and the transferred data). By
ever, if we move away from these typical making the transfer format more compre-
PLM-based approaches, we have to consider hensive, it will be possible to reduce these
the “thing” in terms of serialization, together losses and achieve high process quality in
5See also: Malakuti, S.; Schlake, J.; Grüner, S.; Schulz, D.; Gitzel, R.; Schmitt, J.; Platenius-Mohr, M.; Vorst, P.: Digital twin – a key
software component of Industry 4.0. ABB Review 12/2018.
6Dickopf, T.; Apostolov, H.; Müller, P.; Göbel, J.; Forte, S.: A Holistic System Lifecycle Engineering Approach. 29 th CIRP Design
2019.
7 Wikipedia: ISO 10303 (https://en.wikipedia.org/wiki/ISO_10303), visited 6 September 2019
- 12 - November 2019
Digital Twin in Industrial Application – Requirements to a Comprehensive Data Model
Return to IIC Journal of Innovation landing page for more articles and past editions
The views expressed in the IIC Journal of Innovation are the contributing authors’ views and do
not necessarily represent the views of their respective employers nor those of the Industrial In-
ternet Consortium.
© 2019 The Industrial Internet Consortium logo is a registered trademark of Object Management
Group®. Other logos, products and company names referenced in this publication are property
of their respective companies.
- 14 - November 2019