![Four-Layer Model for describing Analytics Use Cases [2].](https://www.researchgate.net/publication/346345489/figure/fig1/AS:975095166144512@1609492212318/Four-Layer-Model-for-describing-Analytics-Use-Cases-2_Q320.jpg)
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
ScienceDirect
Available online at www.sciencedirect.com
Procedia Manufacturing 52 (2020) 350–355
2351-9789 © 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 5th International Conference on System-Integrated Intelligence.
10.1016/j.promfg.2020.11.058
10.1016/j.promfg.2020.11.058 2351-9789
© 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientic committee of the 5th International Conference on System-Integrated Intelligence.
Available online at www.sciencedirect.com
ScienceDirect
Procedia Manufacturing 00 (2019) 000–000
www.elsevier.com/locate/procedia
2351-9789 © 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review Statement: Peer-review under responsibility of the scientific committee of the 5th International Conference on System-Integrated Intelligence.
5th International Conference on System-Integrated Intelligence
Framework for Data Analytics in Data-Driven Product Planning
Melina Massmanna, Maurice Meyera,*, Maximilian Franka, Sebastian von Enzbergb, Arno Kühnb,
Roman Dumitrescua,b
aChair Advanced Systems Engineering, Heinz Nixdorf Institute, University of Paderborn, Fürstenallee 11, 33102 Paderborn, Germany
bFraunhofer IEM, Zukunftsmeile 1, 33102 Paderborn, Germany
* Corresponding author. Tel.: +49-5251-60-6227; fax. +49-5251-60-6268. E-mail address: maurice.meyer@hni.uni-paderborn.de
Abstract
While many companies are currently trying to use data generated and enabled by Industry 4.0 and networked systems to create data-based
services, this also results in new possibilities and potentials for product planning and product engineering. Product life cycle analytics plays an
essential role in data-driven product planning. In addition to the actual analysis, analytics projects must always take into account the use case,
the data collection and acquisition. In this paper we propose an extended framework for successful realization of data analytics solutions in
product planning. Starting from a thorough analysis of challenges in data-driven product planning, we derive requirements for structured data
analytics solutions in product planning. The proposed framework is based on standard models as CRISP-DM [1], the four-layer model for
Analytics Use Cases, and the Analytics Canvas [2] and offers an overview of structured solutions to fulfill the specialized requirements of data-
driven product planning. It consists of four phases “use cases”, “data sources”, “data acquisition & integration”, and “data analysis”, each
presenting corresponding approaches and methods. Based on a specific application example, we illustrate the application potential of using the
framework.
© 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review Statement: Peer-review under responsibility of the scientific committee of the 5th International Conference on
System-Integrated Intelligence.
Keywords: Data-driven Product Planning; Industrial Data Analytics; Data Inventory; Product Life Cycle
1. Introduction
The digital age is characterized by strong technical
developments. Starting from computer aided design, over the
concept of product life cycle management (PLM) to the digital
twin, which virtually represents the product, data is provided
and enables the analysis of product life cycle data including the
field data from the usage of the product.
This analysis allows drawing conclusions about the use of
the product and thus reveals potentials for optimization. It
allows receiving valuable and objective information about the
current state, failures and incorrect operations. While this is
common practice for software usage data, data from cyber-
physical systems (CPS) such as machines and plants widely
remains unused [3]. Using this information it is possible to plan
the next product generation or new products, which are better
adapted to the actual needs and less error-prone. This is most
relevant to hardware features, which require changes that are
more complex compared to software updates.
The corresponding field of research is called data-driven
product planning [3] and can be defined as the intersection of
strategic product planning, data analytics and (intelligent)
complex products. Products evolve from mechatronic systems
into more complex cyber-physical systems, which interact with
Available online at www.sciencedirect.com
ScienceDirect
Procedia Manufacturing 00 (2019) 000–000
www.elsevier.com/locate/procedia
2351-9789 © 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review Statement: Peer-review under responsibility of the scientific committee of the 5th International Conference on System-Integrated Intelligence.
5th International Conference on System-Integrated Intelligence
Framework for Data Analytics in Data-Driven Product Planning
Melina Massmanna, Maurice Meyera,*, Maximilian Franka, Sebastian von Enzbergb, Arno Kühnb,
Roman Dumitrescua,b
aChair Advanced Systems Engineering, Heinz Nixdorf Institute, University of Paderborn, Fürstenallee 11, 33102 Paderborn, Germany
bFraunhofer IEM, Zukunftsmeile 1, 33102 Paderborn, Germany
* Corresponding author. Tel.: +49-5251-60-6227; fax. +49-5251-60-6268. E-mail address: maurice.meyer@hni.uni-paderborn.de
Abstract
While many companies are currently trying to use data generated and enabled by Industry 4.0 and networked systems to create data-based
services, this also results in new possibilities and potentials for product planning and product engineering. Product life cycle analytics plays an
essential role in data-driven product planning. In addition to the actual analysis, analytics projects must always take into account the use case,
the data collection and acquisition. In this paper we propose an extended framework for successful realization of data analytics solutions in
product planning. Starting from a thorough analysis of challenges in data-driven product planning, we derive requirements for structured data
analytics solutions in product planning. The proposed framework is based on standard models as CRISP-DM [1], the four-layer model for
Analytics Use Cases, and the Analytics Canvas [2] and offers an overview of structured solutions to fulfill the specialized requirements of data-
driven product planning. It consists of four phases “use cases”, “data sources”, “data acquisition & integration”, and “data analysis”, each
presenting corresponding approaches and methods. Based on a specific application example, we illustrate the application potential of using the
framework.
© 2020 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review Statement: Peer-review under responsibility of the scientific committee of the 5th International Conference on
System-Integrated Intelligence.
Keywords: Data-driven Product Planning; Industrial Data Analytics; Data Inventory; Product Life Cycle
1. Introduction
The digital age is characterized by strong technical
developments. Starting from computer aided design, over the
concept of product life cycle management (PLM) to the digital
twin, which virtually represents the product, data is provided
and enables the analysis of product life cycle data including the
field data from the usage of the product.
This analysis allows drawing conclusions about the use of
the product and thus reveals potentials for optimization. It
allows receiving valuable and objective information about the
current state, failures and incorrect operations. While this is
common practice for software usage data, data from cyber-
physical systems (CPS) such as machines and plants widely
remains unused [3]. Using this information it is possible to plan
the next product generation or new products, which are better
adapted to the actual needs and less error-prone. This is most
relevant to hardware features, which require changes that are
more complex compared to software updates.
The corresponding field of research is called data-driven
product planning [3] and can be defined as the intersection of
strategic product planning, data analytics and (intelligent)
complex products. Products evolve from mechatronic systems
into more complex cyber-physical systems, which interact with
Melina Massmann et al. / Procedia Manufacturing 52 (2020) 350–355 351
2 M.Massmann et al. / Procedia Manufacturing 00 (2019) 000–000
the physical and digital world by integrating sensors, actuators
and advanced information processing technology [4]. The first
two underlying fields can be characterized as follows:
Strategic Product Planning is part of the process of Product
Engineering [5]. In this cycle the requirements for products are
systematically identified to capture the markets of tomorrow.
Three tasks are involved: foresight, product discovery and
business planning. The aim is to gain new product ideas and
approaches for product optimization.
Data analytics is the examination of data with the purpose
of knowledge discovery. Various standard models for
processing and analyzing data are known in the state of research
(see section 3). They take slightly different approaches to
structuring the data analytics process. To understand the
general problem domains of data analytics, the four-layer
model for Analytics Use Cases is given as an example for a
holistic standard model. It differentiates between “analytics use
case”, “data sources”, “data pools” and “data analysis” (see Fig.
1).
Fig. 1. Four-Layer Model for describing Analytics Use Cases [2].
Data-driven product planning offers many potentials
regarding the fulfilment of tasks in strategic product planning,
such as the planning of a new product generation where among
others ideas for new product features are explored. While
decisions in classical strategic product planning are often based
on subjective estimations and expensive methods or studies [6],
data-driven product planning makes use of existing data and
delivers more objective and reliable information for decision-
making. According to [7,8] companies have recognized these
advantages and want more data-driven decision making and
analysis of product usage in the field. This also reveals that the
approach is not yet widely applied in practice. We assume that
this is due to the challenges of data analytics and missing best
practices. Although analytics (tool) support is currently the
subject of research and first supporting functionalities exist (cp.
section 3), a simple company application is not yet possible.
How can data be turned into useful information for strategic
planning? To address this question and to support companies
with the implementation, we propose an extended framework
for data analytics in data-driven product planning. The
framework is structured on the basis of the four layers of
analytics in order to address all important aspects of an
analytics project. For each level, possible solutions are provided
in the form of methods and procedures. These serve as an aid
for a systematic data-driven product planning.
The paper is structured as follows: in section 2 we analyze
problems and particular challenges in data analysis in data-
driven product planning and derive requirements for a
framework. In section 3 we give an overview of the state of
research. Afterwards, our framework and first steps towards an
application are presented (section 4 and 5).
2. Requirements for data analytics in data-driven product
planning
In this section, we sum up specific challenges in
implementing data analytics in data-driven product planning
and derive core requirements for our framework afterwards.
2.1. Key challenges
The application of data analytics methodology to data from
the product lifecycle plays an essential role in data-driven
product planning. In the following section we elaborate on key
challenges of the so-called product data analytics based on four
sub fields (cp. Four-Layer Model, Fig. 1): Use Case, Data
Inventory, Data Acquisition and Data Analysis.
The first step is to understand the problem and analyze the
domain. Objective and output of this problem analysis are the
analytics use cases. These are in general very important for
estimating the business value. Therefore, the definition of the
use case and the goal of the analytics case is an essential task,
but it is connected with some challenges. In data-driven
product planning different areas of interest can be focused, e.g.
product relevant use cases such as product optimization and
product personalization or customer oriented use cases such as
market understanding. Which of these fields are most relevant
for the company’s business? Based on the prioritization of use
cases, the question remains, which key analysis questions have
to be defined that specify the underlying goals for a subsequent
data analysis.
Data inventory in data-driven product planning addresses
the problem of identifying and collecting all the relevant
sources of data along the lifecycle in connection with data
properties, evaluation and structuring. This step is often the
first hurdle to the successful application of data analytics [9].
Especially when looking at data collection for data-driven
product planning, which is conducted along the whole product
life cycle, many challenges have to be addressed. One is the
large number of data sources and data, which have to be
discovered first and appear in very different formats and types
and thus are very heterogeneous [10,11]. This is often
accompanied by a missing or incomplete overview of the
needed and missing data. Fig. 2 shows a section of possible data
along the product life cycle and demonstrates the challenge of
heterogeneous data sources. This is the result of a workshop
with four different industrial companies
To be able to process the identified data efficiently,
solutions for data acquisition are prerequisite to organize, store
352 Melina Massmann et al. / Procedia Manufacturing 52 (2020) 350–355
M.Massmann et al. / Procedia Manufacturing 00 (2019) 000–000 3
and process the data. A data acquisition solution for data-driven
product planning is confronted with some challenges. Data
along the product life cycle is distributed in different systems
such as databases, IT systems or files. For a holistic analysis,
the data has to be integrated first in a data lake or a knowledge
repository [10] and data of varying formats have to be
harmonized in order to receive cohesive data sets. In many
cases relevant data is not available, thus sensors have to be
retrofitted to the CPS in order to acquire missing data. Added
data sources have to be integrated and harnessed. Another point
is the transformation from raw data to semantic events to be
able to uncover explainable correlations in the subsequent
analysis. This requires an architecture that considers all these
processes.
In data analysis for product planning, two approaches can
be distinguished: 1) the validation of predefined hypotheses by
experts using statistical methods and 2) the data-driven
identification of new patterns, dependencies, correlations and
hypotheses by explorative data mining. For data-driven product
planning both approaches should be considered. A hypotheses-
or domain-driven approach is common, however a large
potential for added value lies in the identification of previously
unknown hypotheses and findings. The challenges in this layer
lie in the variety of possible analytics methods. For the analysis
of correlations alone, a large number of solution classes can be
considered as correlation, regression, association or causal
analyses. Moreover, for all these many different algorithms can
be used. The selection of the appropriate method requires
expert knowledge of the techniques and their data requirements
as well as their advantages and disadvantages. In summary, the
following challenges can be identified:
1. Large number of or unknown possible use cases
2. Key analysis questions for use case specification
3. Including a large number of data sources and data
along the life cycle and within the company
4. Heterogeneous data formats and types
5. Missing or insufficient overview of needed and
missing data
6. Merging of distributed data in different forms
7. Analytics methods selection requiring expert
knowledge
2.2. Key requirements
Based on the challenges in chapter 2.1 we derive the
following requirements Ri which can be met by applying the
framework and its processes and methods:
(R1) Easy identification of possible use cases
(R2) Systematic derivation of key analysis questions
(R3) Systematic inventory of data along the product life
cycle
(R4) Specification, understanding and classification of
heterogeneous data
(R5) Good overview about needed and missing data
(R6) Clear process of data acquisition and necessary steps
for integration
(R7) Structuring and selection of relevant analytics
methods and procedure model
3. State of research
We group existing approaches, which cover single or some
of the requirements, according to their focus on different
aspects of the issue at hand:
3.1. General analytics frameworks
There are several process models for processing and
acquiring knowledge from data. One of the most widespread
models is the Cross-Industry Standard Process for Data Mining
(CRISP-DM) [1]. The modified CRISP-DM according to
Reinhart et al. [9] also takes into account the step of data
acquisition. However, CRISP-DM does not explicitly address
the primary strategic goals of an analytics project.
Kühn et al. [2] present the Analytics Canvas, a semi-formal
specification technique for describing an analytics use case and
the necessary data infrastructure during the early planning and
specification of an analytics project. It is based on the four-
layer model to describe an Analytics Use Case (see section 1).
According to that, when implementing an analytics project the
first step is to understand the problem with the objective to
define concrete analytics use cases. In the second step data
sources have to be examined. Domain experts are necessary to
identify and further specify their data. Afterwards, data needs
to be stored and processed. This is essential for data integration
and preprocessing.
The Canvas uses five layers: analytics use case, data analysis,
data pools, data description and data source. Within each layer,
the necessary elements for an analytics project are described.
The Analytics Canvas provides a framework for data analytics
projects, but does not specify the individual layers of the
canvas. Here, specific methods, models and procedures are
missing that help to overcome the challenges.
Looking at the analysis more in detail, applications may be
characterized by four types of data analytics: descriptive,
diagnostic, predictive and prescriptive [12]. Based on this
characterization, Steentrup et al. [13] propose a stage model for
describing analytics use cases. This model does not consider
any techniques or methods that solve the underlying tasks.
Fig. 2. Extract of data along the product life cycle.
Melina Massmann et al. / Procedia Manufacturing 52 (2020) 350–355 353
4 M.Massmann et al. / Procedia Manufacturing 00 (2019) 000–000
3.2. Management of product life cycle data
Besides these general frameworks, we want to present some
approaches considering product development or product life
cycle data.
Kassner et al. propose the product lifecycle analytics
approach for the holistic integration and analysis of data from
multiple data sources around the product lifecycle [10]. They
derive a reference architecture ApLAUDING (An Architecture
for Product Life cycle Analytics with Unstructured Data
INteGration), which consists of three layers: an integration
layer, an analytics layer and a presentation layer and provides
components and tools. The architecture does not address a
holistic data inventory, where data sources are identified and
structured. Goals of analytics project with their use cases in
data driven product planning are not in focus.
Demminger et al. regard the application of data in product
development, optimization and new product generations in the
context of Technical Inheritance [14]. The focus is on
collection, storage and exchange of the data. The concept
disregards concrete recommendations for the implementation
of data analysis.
Abramovici et al. propose a concept for leading feedback
into product development and present a knowledge-based
feedback assistant for the decision support. They consider
structured feedback from the product usage phase, and do not
address data along the whole product life cycle [15].
Another approach is the framework for Big Data driven
product life cycle management by [16]. The framework
consists of four components: data sensing and acquisition, data
processing and storage, developing the model and performing
data mining and Big Data application for PLM. The authors
focus on big data generated by product embedded information
devices and provide an overall solution for optimizing the
decision-making processes in different phases of the life cycle.
4. Framework for data analytics in data-driven product
planning
As mentioned in section 3, existing approaches and
frameworks do not solve all challenges of data analytics in
data-driven product planning; a gap exists. Therefore, we
propose a framework with methods, concepts and procedures
for data analytics in data-driven product planning in this
section. It is structured based on the four-layer model (see Fig.
1) and two structuring approaches for data analytics, the
differentiation of “hypothesis-driven” and “explorative”
approaches and the four types of data analytics (see section 3).
For each layer of the framework, we refer to and announce
methods, tools, and procedure models to solve the key
challenges and describe their interplay in context of data-driven
product planning. It is summarized in Fig. 3.
4.1. Analytics use cases
The goal is to identify business-relevant analytics use cases
in the context of strategic product planning. In this context,
different main tasks are involved, which in turn involve various
activities. Typically, for example, the activities "Analyzing the
Fig. 3. Framework with methods, concepts and procedures for Data Analytics in Data-driven Product Planning.
354 Melina Massmann et al. / Procedia Manufacturing 52 (2020) 350–355
M.Massmann et al. / Procedia Manufacturing 00 (2019) 000–000 5
initial situation" and "Performing foresight" are carried out in
the task foresight (cp. section 1). Several potentials for using
data analytics are hidden along these activities. These need to
be uncovered within the layer and evaluated in terms of
strategic relevance and business value, this means that they
make a contribution to achieving the company's objectives.
Examples for typical value creating potentials are cost
reduction or growth increase [17]. We propose a process-based
potential analysis using OMEGA, a process modelling method,
for defining such business-relevant use cases systematically
[18]. Once the process under consideration has been mapped
with the help of domain and OMEGA experts, problematic
steps can be easily identified and possible analytics solutions
derived. As an alternative or supportive approach, we develop
a use case catalogue that provides a structured overview about
possible applications in strategic product planning. It is a
collection of standard use cases, with a concise description,
information on typical use, analytics target and relevant
business objectives/potentials. Additionally, key analysis
questions are given for each use case. The catalogue will be the
result of an extensive literature research and several potential
analysis workshops with different industrial companies based
on a reference process for strategic product planning. The use
case catalogue enables an easy identification process (R1) as
well as derivation of concrete questions, assumptions and
analytics goals (R2).
4.2. Data sources
The use case definition is followed by data understanding,
which is gained by analysis of data sources and their data. We
propose the development of a procedure model that enables a
systematic data inventory with method support (R3).
In the case of hypothesis-driven use cases and where
specific assumptions about known variables are considered,
concrete data points can be derived and thus a data requirement
can be identified. Data requirements can be summarized in a
hypothesis profile. Such a profile contains a defined hypothesis
and information on its variables and necessary data points.
Next, relevant data sources and their data are discovered and
recorded. If we have a concrete data requirement, we already
know quite well which data are relevant and what to look for.
But if a more open explorative question is to be considered, it
is important to have an overview about existing data in the
company. In order to obtain such an overview, a structuring
frame may support which allows parties involved to analyze
relevant life cycle phases and functional areas of the company
by providing two dimensions (R3) [19]. This enables a holistic
identification of data and data sources giving a precise
overview about the sources and places of origin of data in a
company. Based on this frame, a next step is a knowledge base,
which provides typical data sources and data in a company
along those two dimensions. It serves as a tool for companies
in the form of a checklist (R5). After discovering the data, third
step is to specify them by describing them with the help of
analytics relevant meta-data. This is necessary for finding the
appropriate analysis workflow/pipeline subsequently. For this
task, data profiles and predefined data clusters can be used
(R4). These tools provide a structured overview about relevant
properties, such as data structure, and their features/
specifications (e.g. data row or text) and allow the users to
classify their data accordingly to derive important information
for further procedure.
4.3. Data acquisition
After discovering and specifying relevant data sources and
their data, they have to be acquired and converted into a usable
structure and format. This step requires a holistic data
integration across heterogeneous sources and data. Different
steps and concepts are necessary for the task of integration,
including meta data management, harmonization and semantic
annotation (R6). A good meta data management is the basis for
integration solutions since it allows the assessment of data
provenance and quality and offers more transparency. The
process of combining data of varying formats and columns and
transforming it into one cohesive data set is called data
harmonization. Through the use of semantic technology data
can be consolidated into meaningful and valuable information.
Besides these tasks, further necessary subtasks can be defined.
Further research is needed here.
4.4. Data analysis
The last step is to analyze the data and to specify a data
analysis approach, which supports value generation as
described in the use case. An adequate analysis pipeline must
be prepared for this, which includes necessary pre-processing,
modelling and validation steps to reach the analytics target
defined in the use case layer with the integrated relevant data.
To support this process we build upon a structuring for
analytics approaches which uses meaningful properties and
features. Categories such as analytics problem, input, output,
required data volume etc. and concrete features such as
analyzing correlations or clustering for the former category
enhance the understanding about the methods and allow even
non-experts to make an informed decision. Based on the
structuring approaches for data analytics, the framework
proposes a method catalogue and a map that clusters and sorts
the methods along these properties and features. A selection
system using a tree structure for example captures the method’s
requirements, prerequisites, advantages and disadvantages and
allows the selection of adequate analytics techniques and
algorithms based on the analytics goals/tasks and input data
examined in the layer Data Sources (R7). These methods are
brought into a process by using adapted procedure models.
5. Towards an application
The following describes one application example. This
demonstrates the application potential and use of our
framework without focusing on implementation details.
In the example, the goal was to plan and optimize the next
product generation of a hot-forming machine. This use case
derived from the company’s strategic objectives
competitiveness through product leadership and innovations.
To achieve product leadership, the ability to improve own
products is necessary. Hence, the use case catalogue
Melina Massmann et al. / Procedia Manufacturing 52 (2020) 350–355 355
6 M.Massmann et al. / Procedia Manufacturing 00 (2019) 000–000
recommends the use of data-driven generation and retrofit
planning. As concrete assumptions in form of hypotheses about
the product are available, hypotheses profiles can be used
within the data inventory procedure. One profile contains for
example the hypothesis “If the pressure at standstill falls below
the system pressure by more than 3 bar, there is a leakage in the
system.” This also results in the need for data on system error
messages (leakage), machine states (standstill) and pressure
measurements (system pressure). The knowledge base can
provide information about where these data points are typically
found, e.g. service reports or machine data acquisition systems
often provide information about reported system errors;
pressure values can be found in measurements from pressure
sensors. Now these data points can be further specified. In
profiles they can be described by various characteristics, e.g.
degree of structure, format and quality and thus they can be
assigned to different data clusters. One data cluster could be
structured, time-series continuous sensor/signal data. Finally,
it can be checked whether the required data points exist in the
company by interviewing the respective data responsible
persons. The first point of contact can be typical data sources
from the knowledge base. It is also possible to use an already
existing overview of the data for this purpose. The knowledge
base provides the necessary framework for this. Afterwards the
required data must be brought together in order to be able to
evaluate them coherently. If error messages from unstructured
service reports are used, they have to be extracted from the
reports and transformed into a numeric format. When all data
points (columns) are available in one cohesive data set, they
can be analyzed using statistical methods and advanced
analytics. The needed approach can be found quickly in the
structuring: descriptive hypotheses-driven analysis. Based on
this and the assigned data clusters, the appropriate techniques
and algorithms can be selected and it can be proceeded
accordingly. For example, our case requires statistical
regression methods and causality analyses in order to verify
correlations. Depending on the scale level and the number of
independent and dependent variables, different methods are
selected. The analysis is followed by the interpretation of the
results.
6. Conclusion
Data-driven product planning promises to unlock so far
unknown potentials by integrating data analytics into the
process of strategic product planning. Due to major challenges
in the context of analytics activities, however, this approach is
rarely used in practice. We presented a framework to address
these challenges and the resulting requirements. Along the four
layers of analytics projects, the framework proposes
procedures and methods that support the planning and
implementation of a successful data analysis in product
planning. Future research will concentrate on the development
and validation of the methods and integration of further
necessary tasks. The presented concept for the framework is
tested together with four companies of the manufacturing
industry and will be further elaborated.
Acknowledgements
This research and development project is funded by the
German Federal Ministry of Education and Research (BMBF)
within the “Innovations for Tomorrow’s Production, Services,
and Work” Program and implemented by the Project
Management Agency Karlsruhe (PTKA). The author is
responsible for the content of this publication.
For more details about the project, visit www.dizrupt.de
References
[1] Shearer C. The CRISP-DM Model: The new blueprint for data mining.
Journal of Data Warehousing, volume 5; pp. 13-22, 2000.
[2] Kühn A, Joppen R, Reinhart F, Röltgen D, von En zberg S, Dumitrescu R.
Analytics Canvas–A Framework for the Design and Specification of Data
Analytics Projects. Procedia CIRP 2018, 70, 162-167.
[3] Massmann M, Meyer M, Frank M. Significance and Challenges of Data-
driven Product Generation and Retrofit Planning. Procedia CIRP 2019, 84,
992 - 997.
[4] Broy M. Cyber-Physical Systems – Innovation durch Software-intensive
eingebettete Systeme, Springer, Berlin, Heidel berg 2010.
[5] Gausemeier J, Czaja A, Dülme C. Innovationspoten tiale auf dem Weg zu
Industrie 4.0. In: Wissenschafts- und Industrieforum Intelligente
Technische Systeme 2015, Heinz Nixdorf Institut, (2015).
[6] Buerke G. Marktforschungsmethoden für Innovationen und
Produktverbesserungen. Wirtschaftswiss. Schriften. Heft 03/ 2016.
[7] Lueth K, Patsioura, C, Williams Z, Kermani Z. Industrial Analytics
2016/2017 – The current state of data analytics usage in industrial
companies. IoT Analytics, Hamburg, 2016.
[8] Harvard Business Review Analytic Services Survey, August 2018.
[9] Reinhart F. Indust rial Data Science – Data Science in der i ndustriellen
Anwendung. Industrie 4.0 Management, 32:27-30, 2016.
[10] Kassner L, Gröger C, Mitschang B, Westkämper E. Product life cycle
analytics–next generation data analytics on structured and unstructured
data. Procedia CIRP, 33, 35-40, 2015.
[11] Li J, Tao F, Cheng Y, Zhao L. Big Data in product lifecycle
management. The International Journal of Advanced Manufacturing
Technology. 2015.
[12] Reinhart F, Kühn A, Dumitrescu R. Layer Model for the Development of
Data Science Applications in Plants and Machines. Wissenschaftsforum
Intelligente Technische Systeme (WInTeSys) Paderborn 2017.
[13] Steenstrup K, Sallam R L, Eriksen L, Jacobson S F. Industrial Analyti cs
Revolutionizes Big Data in the Digital Business. Gartner Research, 2014.
[14] Demminger C, Mozgova I, Quirico M, Uhlich F, Denkena B, Lachmayer
R, Nyhuis P. The Concept of Technical Inheritance in Operation: Analysis
of the Information Flow in the Life Cycle of Smart Products. Procedia
Technology, 26, 79-88, 2016.
[15] Abramovici M, Lindner A, Walde F, Fathi M, Dienst S. Decision
support for improving the design of hydraulic systems by leading feedback
into product development. ICED 11 - 18th International Conference on
Engineering Design. 2011.
[16] Zhang Y, Ren S, Liu, Y, Sakao T, Huisingh D. A Framework for Big
Data driven product lifecycle management . Journal of Cleaner Production
2017.
[17] Mohr N, Hürtgen H. Achieving business impact with data. Digital
McKinsey 2018.
[18] Fahrwinkel U. Methode zur Modellierung und Analyse von
Gecshäftsprozessen zur Unterstützung des Business Process R eengineering.
Dissertation, Universität Paderborn, HNI-Verlagsschriftenreihe 1, 1995.
[19] Massmann M, Meyer M, Frank M, von Enberg S, Kühn A, Dumitrescu
R. Method for data inventory and classification. 53rd CIRP Conference on
Manufacturing Systems 2020.
