ArticlePDF Available

Analysis of High Frequency Data of a Machine Tool via Edge Computing

Authors:

Abstract and Figures

New technological capabilities of digitalization are enablers of processing a broad range of machine data. While so-called Low-Frequency Data (LFD) is captured at a sampling rate of several hundred milliseconds, High-Frequency Data (HFD) is based on a sampling rate in the single-digit millisecond range. In this paper, HFD is used to implement an edge-based analytics application for prediction purposes in a machine tool. This edge application leverages Siemens SINUMERIK Edge to capture HFD from a machine tool to recognize anomalies of any kind. The edge application is implemented as a show case in the Learning Factory of Graz University of Technology, the [email protected]
Content may be subject to copyright.
ScienceDirect
Available online at www.sciencedirect.com
Procedia Manufacturing 45 (2020) 343–348
2351-9789 © 2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientic committee of the 10th Conference on Learning Factories 2020.
10.1016/j.promfg.2020.04.028
10.1016/j.promfg.2020.04.028 2351-9789
© 2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientic committee of the 10th Conference on Learning Factories 2020.
Available online at www.sciencedirect.com
Procedia Manufacturing 00 (2019) 000–000
www.elsevier.com/locate/procedia
10th Conference on Learning Factories, CLF2020
Analysis of High Frequency Data of a Machine Tool via Edge
Computing
Stefan Trabesingera,, Andre Butzerina,b, Daniel Schallb, Rudolf Pichlera
aInstitute of Production Engineering, Graz University of Technology, Ineldgasse 25f, Graz 8010, Austria
bSiemens AG Austria, Siemensstraße 90, Wien 1210 , Austria
Abstract
New technological capabilities of digitalization are enablers of processing a broad range of machine data. While so-called Low-
Frequency Data (LFD) is captured at a sampling rate of several hundred milliseconds, High-Frequency Data (HFD) is based on a
sampling rate in the single-digit millisecond range. In this paper, HFD is used to implement an edge-based analytics application
for prediction purposes in a machine tool. This edge application leverages Siemens SINUMERIK Edge to capture HFD from a
machine tool to recognize anomalies of any kind. The edge application is implemented as a show case in the Learning Factory of
Graz University of Technology, the smartfactory@tugraz.
c
2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 10th Conference on Learning Factories 2020.
Keywords: Edge Computing; High-Frequency Data; Prediction of Tool Breakage; Anomaly Detection; Machine Learning
1. Introduction
By definition, edge computing is a distributed cloud-computing paradigm [1]. Edge computing, unlike cloud com-
puting, refers to distributed data processing in close proximity to devices, such as machine tools, that generate data
for further analysis [2]. By contrast, edge computing allows the analysis of HFD on its own platform - typically im-
plemented as an embedded computer in the automation network. The fields of application can be found preferably in
time critical industrial applications where this data stream must be stored and evaluated in a well-defined timeframe.
Time-critical applications in an industrial context, for example, are the timely shutdown of a machine or the reduction
of the feed rate of tools to avoid breakages. Whereas in traditional cloud computing environments, the data generated
by various assets and machines is typically passed to cloud services in an uncompressed way and without data pre-
processing. In the cloud environment, various software tools are used to analyze the data. However, the slowdown in
broadband expansion and delays in data transmission between central cloud servers and end devices at the edge of
the network prove to be an obstacle in growth. To combine benefits of both systems, edge computing is often used in
Corresponding author. Tel.: +43-316-873-7675 ; fax: +43-316-873-7678.
E-mail address: stefan.trabesinger@tugraz.at
2351-9789 c
2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 10th Conference on Learning Factories 2020.
Available online at www.sciencedirect.com
Procedia Manufacturing 00 (2019) 000–000
www.elsevier.com/locate/procedia
10th Conference on Learning Factories, CLF2020
Analysis of High Frequency Data of a Machine Tool via Edge
Computing
Stefan Trabesingera,, Andre Butzerina,b, Daniel Schallb, Rudolf Pichlera
aInstitute of Production Engineering, Graz University of Technology, Ineldgasse 25f, Graz 8010, Austria
bSiemens AG Austria, Siemensstraße 90, Wien 1210 , Austria
Abstract
New technological capabilities of digitalization are enablers of processing a broad range of machine data. While so-called Low-
Frequency Data (LFD) is captured at a sampling rate of several hundred milliseconds, High-Frequency Data (HFD) is based on a
sampling rate in the single-digit millisecond range. In this paper, HFD is used to implement an edge-based analytics application
for prediction purposes in a machine tool. This edge application leverages Siemens SINUMERIK Edge to capture HFD from a
machine tool to recognize anomalies of any kind. The edge application is implemented as a show case in the Learning Factory of
Graz University of Technology, the smartfactory@tugraz.
c
2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 10th Conference on Learning Factories 2020.
Keywords: Edge Computing; High-Frequency Data; Prediction of Tool Breakage; Anomaly Detection; Machine Learning
1. Introduction
By definition, edge computing is a distributed cloud-computing paradigm [1]. Edge computing, unlike cloud com-
puting, refers to distributed data processing in close proximity to devices, such as machine tools, that generate data
for further analysis [2]. By contrast, edge computing allows the analysis of HFD on its own platform - typically im-
plemented as an embedded computer in the automation network. The fields of application can be found preferably in
time critical industrial applications where this data stream must be stored and evaluated in a well-defined timeframe.
Time-critical applications in an industrial context, for example, are the timely shutdown of a machine or the reduction
of the feed rate of tools to avoid breakages. Whereas in traditional cloud computing environments, the data generated
by various assets and machines is typically passed to cloud services in an uncompressed way and without data pre-
processing. In the cloud environment, various software tools are used to analyze the data. However, the slowdown in
broadband expansion and delays in data transmission between central cloud servers and end devices at the edge of
the network prove to be an obstacle in growth. To combine benefits of both systems, edge computing is often used in
Corresponding author. Tel.: +43-316-873-7675 ; fax: +43-316-873-7678.
E-mail address: stefan.trabesinger@tugraz.at
2351-9789 c
2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license https://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the scientific committee of the 10th Conference on Learning Factories 2020.
344 Stefan Trabesinger et al. / Procedia Manufacturing 45 (2020) 343–348
2Author name /Procedia Manufacturing 00 (2019) 000–000
combination with cloud computing [3]. This allows to implement the requirements for data protection, response time
behavior and autonomy [4].
2. Motivation for doing Edge Computing
With the technologies mentioned in the introduction, the example of the defined showcase for prediction of drill
breakage will be used to show how meaningful the results will turn out and if they provide benefits at all. In addition
to the demonstration of the interplay of edge and cloud computing, the networking integration of the edge devices
with all its IT security aspects is also very important. Likewise, the comparison of the LFD with the HFD is of
high interest. Aside the reading and saving of data another motivation lies in the correct identification of anomalies,
which is based on data analysis on approaches of machine learning algorithms. As a conclusion to this showcase, an
interactive display board will be created as part of the learning factory smartfactory@tugraz. It will be used to present
and explain the functionality of edge computing. This is a good starting point for further research and development
activities and learning opportunities.
3. System architecture for the Edge Computing Show Case
The already implemented network infrastructure between oce and shop floor will now be expanded by connecting
machines in the shop floor to the Internet [5]. The required components for the layout of the network are shown in Fig.
6a. The Siemens numerical control unit (NCU) SINUMERIK 840d sl and the HMI are basic features of the machine
tool Spinner U5-630. The edge device is a Siemens industrial computer IPC227E. Combined with the SITOP power
supply the edge device is operative. For integrating the edge device, the data network is retrofitted with a managed
Ethernet switch Scalance X208G inside the machine tools control cabinet. To guarantee a secure logic separation of
the two networks machine LAN and factory LAN VLANs are installed [6]. Configuration and management of the
edge device is transacted via the IoT operating system Mindsphere. For ensuring secure data exchange to the cloud
the Edge Device oers two independent network ports.
4. Setup of Experiments for Data Generation
4.1. Planning and execution
Many components of the wave gear – this is the demonstration part of the learning factory “smartfactory@tugraz” -
to be manufactured feature single drill holes with a diameter of 2.8 mm. Such a small-dimensioned drill hole exhibits
ActualAxisPositionX1 [mm]
10
8
6
4
2
0
ActualAxisPositionY1 [mm]
0
20
40
60
80
100
ActualAxisPositionZ1 [mm]
12
10
8
6
4
2
0
2
A
B
C
DE
F
F
F
1
12
13
14
15
Actual Axis Position Y1 [mm]
Actual Axis Position Z1 [mm]
Actual Axis Position X1 [mm]
Fig. 1. Drill travel length.
Stefan Trabesinger et al. / Procedia Manufacturing 45 (2020) 343–348 345
Author name /Procedia Manufacturing 00 (2019) 000–000 3
a high risk of drill breakage. This provoked the idea of using machine generated data as input for Edge Comput-
ing to investigate the settings for avoiding drill breakage. An NC program is written for automation of experiment
conduction and a better reproduction of the tool path. A high-speed camera is used to support the recognition of the
moment of breakage. Experiments are conducted on an aluminum cuboid with dimensions 20 mm x 15 mm x 125
mm. Experiment preparation is finished after cutting the raw material, performing the tool change and the clamping of
the work piece. The recording of data and high-speed video footage is recorded after start of the NC program. In order
to reconstruct the realistic eorts of the drilling process parameter sets for a drill breakage are built, see Fig. 1. The
whole drilling cycle includes the processing of parameter set A to E iterated until parameter set group no 12. Within a
parameter set sequence A to E spindle speed and drilling depth steadily increase while keeping a constant duration of
half a second. If there is no breakage, the NC program sets the drill to position no 13 where the NC program performs
the highest rotational speed to force drill breakage.
5. Predictive data analytics by means of machine learning
After successful capturing and import of data, it is most important to understand the physical mechanism that leads
to the generated data [7]. Afterwards that the eligibility of an appropriate machine-learning model follows.
5.1. Data capturing
The capturing of HF (High-Frequency) drive signals is driven by a sampling rate of 2 ms starting from the HF
probe as initial sensor inside the SINUMERIK 840d sl NCU [8]. An upload stream by a proprietary protocol with
a sampling rate of 100 ms transfers data to the SINUMERIK adapter. The edge app “AMW /capture” (Analyze My
Workpiece /capture) finally stores the data on a hard drive, see Fig. 2a. The capturing of LF (Low-Frequency) drive
signals is driven by a sampling rate of 100 ms by means of an encrypted OPC UA server-client connection shown in
Fig. 2b. The user triggers start and end of data capturing.
5.2. Processing of drill data
Two d ierent types of measurement patterns of data of one drill are important when regarding the parameter set A
up to F in its entirety. After the import, the measured data is stored in a single CSV file in the form of a time series.
The measurement data of the last drill hole is the one of breakage. The single CSV file is separated into single files
which characterize each single drill hole. Only the data where the drill is engaged with the material of the work piece
eectively is stored. All other data is discarded.
proprietary
protocol stream
(every 100 ms)
2 ms sampling rate
Adapter
Framework
amwcapture
HF probe
Spinner U5-630
Sinumerik
Adapter
Siemens
SINUMERIK
NCU 840d sl
Siemens
Edge Device
IPC 227E
hard
drive
Samba
Server USB port
server
b
aLaptop
Human
Machine
Interface
USB
sck
network share
machine LAN
Linux Debian Virtual Machine
OPC-UA Client
Table of
records
~100
ms
Hard drive
Spinner U5-630
Siemens SINUMERIK
NCU 840d sl
Human Machine
Interface
job control via
factory LAN
Publish-subscribe paern
Latest
value
Fig. 2. (a) HFD capturing architecture; (b) LFD capturing architecture.
5.3. Drill data assessment
The measurement data of a single drilling is shown in Fig. 3b. The red lined progress of current in case of breakage
shows major deviations compared to drills in idle mode (no contact to the work piece) respectively during a drilling
346 Stefan Trabesinger et al. / Procedia Manufacturing 45 (2020) 343–348
4Author name /Procedia Manufacturing 00 (2019) 000–000
5 4 3 2 1
drive current [A]
14
12
10
8
6
4
2
0
2
drilling depth [mm]
Tool Axis (SP1)
HF data (no contact drill - workpiece)
LF data (OPC-UA) (no contact drill - workpiece)
HF data (breakage)
OPC-UA data (breakage)
5.0 4.5 4.0 3.5 3.0 2 .5 2.0 1.5
drive current [A]
14
12
10
8
6
4
2
0
2
drilling depth [mm]
Z Axis (Z1)
HF data (no contact drill - workpiece)
LF data (OPC-UA) (no contact drill - workpiece)
HF data (breakage)
OPC-UA data (breakage)
2 1 0 1 2 3
actual speed [rpm] +1.33e4
14
12
10
8
6
4
2
0
2
drilling depth [mm]
Tool Axis (SP1)
HF data (no contact drill - workpiece)
LF data (OPC-UA) (no contact drill - workpiece)
HF data (breakage)
OPC-UA data (breakage)
ab
Fig. 3. (a) Drilling procedure; (b) Current and speed of single drillings.
process without breakage. Based on these results, current of the rotational z-axis (SP1) and current of the translational
z-axis (Z1) are chosen as evaluation parameter. To determine the drilling depth at the time of breakage without insights
into the drill hole a manual synchronization between camera frames and z-coordinates received from HFD enables
high benefit. The initial contact between drill and work is set as starting point and detected by use of the video software
“virtual dub”. Fig. 3b shows the coherence between current of SP1 and Z1 and the corresponding video frame. At the
instance of breakage, current SP1 suddenly sinks, the drill breaks and the remaining part of the drill staggers. The
eects described below always occur at constant feed. The drilling edge is getting dulled, less material is removed
and thus the set speed of 13 300 rpm is undershot. The machine control tries to retain the set speed, and increases the
current of SP1. The current of Z1 decreases since the drill pulls itself into the material. Thereby the cutting edge is
getting duller, material removal decreases while current of SP1 increases leading to an overstressing of the drill which
results in failure.
5.4. Machine Learning approaches for HFD
There is the idea of searching for indicators of anomalies to prevent tool breakage. For that reason, two dierent
artificial intelligence approaches are used. LSTM (Long Short Term Memory) is a variant of recurrent neural networks
for accurate prediction of future values for repetitive procedures [9,10]. In this case the main influencing parameter
of LSTM is the forward and back-propagation from the time steps -10 to +10. The orange curves in Fig. 4shows
the prediction of values; the blue curves shows the real measurement values. To evaluate this result a benchmark
with the Isolation Forest (IsFo) model is of interest [11]. IsFo is a very fast and robust outlier detection approach
to classify values. In this case the main influencing parameter of IsFo is called “contamination” and it is defined as
the ratio between the number of expected anomalies related to all measurement values in the training set. The higher
the “contamination” the higher the anomalies. For detection, it is necessary to merge the individual data sets into a
single dataset. Fig. 5shows the identified anomalies and the anomaly density over the drilling depth. The isolation
forest model was chosen because the decision tree calculation method is easier for understanding and interpretation
than the neural network calculation method of LSTM. As consequence of the detection of anomalies, commonly
countermeasures are needed. If anomalies are detected the anomaly density value is calculated. In case of excessing
a given limit of anomaly density values, raw measurement data (HFD) is uploaded to the cloud. It is conceivable to
develop a combined approach, which allows an anomaly detection based on the predicted values. Such a combinations
contains first a prediction done via LSTM and secondly the detection of anomalies by using the isolation forest model
on the predicted values.
Stefan Trabesinger et al. / Procedia Manufacturing 45 (2020) 343–348 347
Author name /Procedia Manufacturing 00 (2019) 000–000 5
2.6 2.4 2.2 2.0 1.8 1.6 1 .4
10
8
6
4
2
4.0 3.5 3.0 2.5
10
8
6
4
2
Current SP1
Real values
Prediction
Current Z1
Real values
Prediction
drilling depth [mm]
drilling depth [mm]
Current [A] Current [A]
Fig. 4. Current prediction with LSTM (test dataset).
Fig. 5. Anomaly detection with isolation forest.
5.5. Machine Learning approaches for LFD
The amount of data based on the LFD capturing approach is too low for meaningful evaluation with the proposed
machine learning approaches. Both experiments using LSTM and isolation forest have not yielded reliable results for
accurate prediction nor detection of anomalies.
6. Edge App Development
The analyzed machine data is evaluated by means of the machine-learning model in the latest developed Edge App
”DADetection” (Drilling Anomaly Detection). If anomalies occur, they are sent to the MindSphere Cloud. For the
software development, Siemens provides an Edge App Software Development Kit (SDK). The Edge Device runs an
industrial OS based on Debian 9 with an execution environment for industrial apps. In order to make the Edge App
available for the Edge Device it is uploaded via the MindSphere functionality “App Publishing”. It is ready for use on
the Edge platform once the download and subsequent installation finishes by using another MindSphere functionality
called “App Management”. Fig. 6b shows the data capturing from the HF probe and how upload data is sent.
348 Stefan Trabesinger et al. / Procedia Manufacturing 45 (2020) 343–348
6Author name /Procedia Manufacturing 00 (2019) 000–000
proprietary
protocol stream
(every 100 ms)
2 ms sampling rate anomalies
Adapter
Framework
dadetecon
HF probe
Spinner U5-630
Sinumerik
Adapter
Siemens
SINUMERIK
NCU 840 d sl
Siemens
Edge Device
IPC 227E
Edge Apps
b
smartfactory@tugraz
Computer Networks
Machine LAN
Factory LAN
Internet/Cloud
Spinner U5-630
a
Siemens
SINUMERIK
NCU 840d sl
Factory LAN
Firewall
Factory LAN
Siemens
Scalance
Switch
Human
Machine
Interface
Siemens Edge
Device
IPC227E
Siemens
SITOP
PSU8200
Fig. 6. (a) Edge Computing Network Architecture; (b) Edge App Architecture.
7. Conclusion
In this paper Edge Computing was shown as a strong method for getting more knowledge about the actual condition
of a drilling tool while being in action. One of the most positive eects of edge computing are based on the fact that
a huge amount of data is generated without the need of additional sensor installations and the reliable and secure
exchange of data into the cloud. With such a data set and its computed results, the breakage of drill and other tools
can be predicted and when applying additional control systems even prevented. Tool usages can be increased and tool
costs can be cut down. A prerequisite of edge computing is a profound network knowledge for the alignment of data
streams and the ongoing experience at the handling of data analytics and machine learning models.
8. Acknowledgements
The project smartfactory@tugraz is funded by the Austrian Ministry for Transport, Innovation and Technology
and 19 industrial partners (see also: www.smartfactory.tugraz.at) and surveilled by the Austrian Research Promotion
Agency FFG. The project itself runs by the Institute of Production Engineering at Graz University of Technology.
References
[1] W. Shi, J. Cao, Q. Zhang, Y. Li, L. Xu, Edge computing: Vision and challenges, IEEE Internet of Things Journal, 3 (5) (2016) 637–646.
[2] Y. Ai, M. Peng, K. Zhang, Edge computing technologies for Internet of Things: a primer, Digital Communications and Networks, 4 (2) (2018)
77 – 86.
[3] J. Pan, J. McElhannon, Future Edge Cloud and Edge Computing for Internet of Things Applications, IEEE Internet of Things Journal, 5 (1)
(2018) 439–449.
[4] C. Esposito, A. Castiglione, F. Pop, K. R. Choo, Challenges of Connecting Edge and Cloud Computing: A Security and Forensic Perspective,
IEEE Cloud Computing, 4 (2) (2017) 13–17.
[5] S. Trabesinger, R. Pichler, D. Schall, R. Gfrerer, Connectivity as a prior challenge in establishing CPPS on basis of heterogeneous IT-software
environments, Procedia Manufacturing, 31 (2019) 370 – 376.
[6] S. A. Alabady, F. Al-Turjman, S. Din, A Novel Security Model for Cooperative Virtual Networks in the IoT Era, International Journal of
Parallel Programming, (2018) 1 – 16.
[7] C. I. Noshi, J. J. Schubert, The Role of Machine Learning in Drilling Operations; A Review, Society of Petroleum Engineers, Pittsburgh (2018).
[8] S. Kehne, T. Berners, A. Epple, C. Brecher, Automatic system identification of forward feed drives in machine tools, Advances in Production
Research: Proceedings of the 8th Congress of the German Academic Association for Production Technology (WGP), (2019) 144 – 152.
[9] R. C. Staudemeyer, E. R. Morris, Understanding LSTM – a tutorial into Long Short-Term Memory Recurrent Neural Networks, (2019).
[10] D. Park, S. Kim, Y. An, J.-Y. Jung, LiReD: A Light-Weight Real-Time Fault Detection System for Edge Computing Using LSTM Recurrent
Neural Networks, Sensors, 18 (7) 2018.
[11] F. T. Liu, K. M. Ting, Z. Zhou, Isolation Forest, 2008 Eighth IEEE International Conference on Data Mining, (2008) 413-422.
... The importance of the high-frequency data (HFD) results from the very fine temporal sampling interval of 2 ms. This means that much more high-resolution data is available for evaluation than is the case for example, with OPC UA (sampling intervals of 100 ms is possible here) (Trabesinger et al., 2020). The HFD are then transferred to the edge device via a proprietary protocol stream, stored and made available for further analysis (Shi and Dustdar, 2016). ...
... Both HF and LF-data can be used to calculate the necessary parameters under consideration; but, since HF-data has a higher sampling rate, the results obtained from this are more enhanced and precise. A use case developed by Trabesinger et al. in Trabesinger et al. (2020) depicts the difference between these data types. The high-frequency machine data generated while machining the two parts have been published on Mendeley Data in raw format (Abdul Hadi, 2021). ...
... Transmission: As described by a flowchart in Fig. 5 and in Trabesinger et al. (2020), the HF-data is captured using the edge device. The transmission of this data can be done tethered via OPC-UA client or wireless via edge computing environment. ...
Article
Full-text available
The reduction of CO2 by moving from fossil to renewable energy sources is currently high on the agenda of many governments. Simultaneously these governments are also forcing the reduction of energy consumption. The primary focus of these agendas is on mobility, building, and industrial sectors. For the latter, energy-efficient shop floors and machining processes assist the reduction of energy consumption. Previous research has focused on energy-efficient machining strategies during machining processes. However, an energy-efficient start-up of these machines or their spindle axis start-up has been neglected until now. This paper focuses on this neglected issue by comparing the energy-efficiency, production time, and cost-efficiency of the CNC (computer numeric control) machine by varying the power input at the spindle axis. This is done by analysing the high-frequency data (500Hz) of the machine from machining operations that is retrieved via the edge device. Concepts of data analytics and especially EDA (exploratory data analytics) were used to interactively visualize the inter-dependencies and develop results. It is shown that optimized reduction of spindle power input value leads to both: peak power smoothing from 20kW to 10kW and lowering of overall energy consumption by approximately 1.4%. Moreover, the costs and production time are marginally affected (0.518% and 0.523% respectively) by this optimized reduction of spindle power input value. Thus, this paper highlights a novel method from data acquisition to process improvement towards energy-efficient and sustainable machining.
... Thus, to track even comparatviely tiny changes requires high frequency data. Therefore, a custom protocol is needed that allows much smaller update intervals [13]. However, since the NCU's primary task is to control the machine, it can be used to collect and transfer this data, but not to process and analyze it. ...
... The first station of the raw data once it is transferred from the NCU is an Edge Device. These devices are relatively low-powered, when compared to a full-blown server infrastructure, but are still high-powered appliances in the context of a shopfloor device [13]. They usually have several gigabytes of RAM, a multi-core processor, and at least two network adapters to properly keep the operational technology (OT) and the IT network separated. ...
... Figure 5 illustrates the connectivity which is needed for the presented edge computing use case. Similar applications are realized using this connectivity approach, as presented in [13]. A Network Switch inside the machine's cabinet is used to separate the Machine LAN and the Factory LAN (the OT and the IT network, respectively). ...
Preprint
Full-text available
An important prerequisite for determining whether a certain product is pro-ducible in any given production facility is an accurate assessment of which production lines and/or the machines are able to execute the necessary production steps. Not only the static information about the capabilities of the machines, but also the conditions of machines and tools are significant for this analysis. Because of the deviation of machine capabilities with increasing deterioration and weary of the equipment, it is also necessary to continuously monitor the status of the machine and analyze the machine conditions. In this paper, we present an approach for generating production plans across multiple factories, considering both static information and dynamic data analysis. Edge devices constantly monitor high frequency machine data and report condensed machine states to an Industrial IoT platform (IIoT). A marketplace within the cloud-application MindSphere enables us to integrate the requirements of the products and the capabilities of the production sites. Customers are be able to evaluate these production plans based on duration, energy consumption, CO 2 footprint etc.
... When device users approach connected machines, the application displays machine information without the need for inspection. Trabesinger et al. collected information using an industrial device called SINUMERIK Edge to collect, process and send information via the internet making it easily accessible to other members of the office with access to the network [19]. ...
Conference Paper
Full-text available
Additive manufacturing (AM) is emerging within many industrial applications due to inherent advantages such as rapid prototyping and production. However, the correlation of process parameters across modules and their impacts on product quality are not yet fully understood. This article presents a system built out of Internet of Things (IoT) and edge computing technologies to collect and analyze AM process in-situ. An IoT thermal camera platform was developed, and integrated within an Laser Based Additive Manufacturing (LBAM) system to collect information that could be used to characterize the thermal distribution surrounding the melt pool. Machine learning techniques were utilised to identify the occurrence of defects using the collected low-resolution thermal images.
... This high frequency software sensor installed onto the NCU provides high resolution data with a sampling interval Dt of 2ms. This has be shown by [32] to be beneficial compared to conventional low frequency data. The NC code used in this study is acquired in a retroactive readout process from the NCU. ...
Article
Nowadays, the reduction of CO2 emissions by moving from fossil to renewable energy sources is on the policy of many governments. At the same time, these governments are forcing the reduction of energy consumption. Since large industries have been in the focus for the last decade, today also small and medium enterprises with production lot size one are increasingly being obliged to reduce their energy requirements in production. Energy-efficient CNC machine tools contribute to this goal. In machining processes, the machining strategy also has a significant influence on energy demand. For manufacturing of lot size one, the prediction of the energy demand of a machining strategy, before a part is manufactured plays a decisive role. In numerous previous studies, analytical models between the energy demand and the machining strategy have been developed. However, their accuracy depends largely on the parameterization of these models by dedicated experiments. In this paper, different machine learning algorithms, especially variations of the decision tree (’DecisionTree’, ’RandomForest’, boosted ’RandomForest’) are investigated for their ability to predict the energy demand of CNC machining operations based on real production data, without the need for dedicated experiments. As shown in this paper, the most accurate energy demand predictions can be achieved with the ’RandomForest’ algorithm.
... This requires specific solutions since this is very control-specific. Trabesinger et al. [86], provides an interface description for an Sinumerik Edge which, unfurtunatly, cannot be transferred to other numerical control systems (e.g., Fanuc, Heidenhain etc.). ...
Article
Full-text available
This paper presents a brief introduction to competition-driven digital transformation in the machining sector. On this basis, the creation of a digital twin for machining processes is approached firstly using a basic digital twin structure. The latter is sub-grouped into information and data models, specific calculation and process models, all seen from an application-oriented perspective. Moreover, digital shadow and digital twin are embedded in this framework, being discussed in the context of a state-of-the-art literature review. The main part of this paper addresses models for machine and path inaccuracies, material removal and tool engagement, cutting force, process stability, thermal behavior, workpiece and surface properties. Furthermore, these models are superimposed towards an integral digital twin. In addition, the overall context is expanded towards an integral software architecture of a digital twin providing information system. The information system, in turn, ties in with existing forward-oriented planning from operational practice, leading to a significant expansion of the initially presented basic structure for a digital twin. Consequently, a time-stratified data layer platform is introduced to prepare for the resulting shadow-twin transformation loop. Finally, subtasks are defined to assure functional interfaces, model integrability and feedback measures.
Article
Full-text available
The high-frequency (HF) machine data is retrieved from the Spinner U5-630 milling machine via an Edge Device. Unlike cloud computing, an Edge Device refers to distributed data processing of devices in proximity that generate data, which can thereby be used for analysis [1], [2]. This data has a sampling rate of 2ms and hence, a frequency of 500Hz. The HF machine data is from various experiments performed. There are 2 experiments performed (parts 1 and 2). The experimented part 1 has 12 .json data files and part 2 has 11 .json files. In total, there are 23 files of HF machine data from 23 experiments. The HF machine data has vast potential for analysis as it contains all the information from the machine during the machining process. One part of the information was used in our case to calculate the energy consumption of the machine. Similarly, the data can be used for retrieving information of torque, commanded and actual speed, NC code, current, etc.
Article
Full-text available
Additive manufacturing (AM) is emerging within many industrial applications due to inherent advantages such as rapid prototyping and produc- tion. However, the correlation of process parameters across modules and their impacts on product quality are not yet fully understood. This article presents a system built out of Internet of Things (IoT) and edge computing technologies to collect and analyze AM process in-situ. An IoT thermal camera platform was developed, and integrated within an Laser Based Additive Manufacturing (LBAM) system to collect information that could be used to characterize the thermal distribution surrounding the melt pool. Machine learning techniques were utilised to identify the occurrence of defects using the collected low-resolution thermal images.
Article
An important prerequisite for determining whether a certain product is producible in any given production facility is an accurate assessment of which production lines and/or the machines are able to execute the necessary production steps. Not only the static information about the capabilities of the machines, but also the conditions of machines and tools are significant for this analysis. Because of the deviation of machine capabilities with increasing deterioration and weary of the equipment, it is also necessary to continuously monitor the status of the machine and analyze the machine conditions. In this paper, we present an approach for generating production plans across multiple factories, considering both static information and dynamic data analysis. Edge devices constantly monitor high frequency machine data and report condensed machine states to an Industrial IoT platform (IIoT). A marketplace within the cloud-application MindSphere enables us to integrate the requirements of the products and the capabilities of the production sites. Customers are be able to evaluate these production plans based on duration, energy consumption, CO2 footprint etc.
Article
Full-text available
More and more elements in a modern production system are becoming smart devices and it is reasonable to integrate them into a Cyber Physical Production System (CPPS) in a consistent manner. There is the broad goal of many industrial players to open the rich potential of the Industrial Internet of Things (IIoT), which has to be fast, secure and adaptive. The growing number of “things” in a production system requires satisfying connectivity solutions that are different to an easy coming “Peer-to-Peer architecture (P2P)”. The approach followed at the [email protected] – the Learning Factory of Graz University of Technology in Austria – to achieve such an integration is to deploy an Enterprise Service Bus (ESB), which is at the core of Service Oriented Architecture (SOA). The goal is to integrate three main software applications including the Product Lifecycle Management (PLM), the Enterprise Resource Planning (ERP) and the Management Execution System (MES). The additional challenge in this project is that these mentioned main applications are all provided by different suppliers. The selected SOA approach provides the basis for a scalable and extensible solution via Connectivity Modules and standardized interfaces. The [email protected] validates this SOA approach by applying it to a real and fully operational manufacturing line. The [email protected] is not only a learning factory but rather an open ecosystem by offering both students and researchers as well as industrial partners the ability to perform research on this subject. It is thus the ideal place to study the challenges and to understand the benefits of pushing a CPPS to a mature level in terms of connectivity in a manufacturing context.
Article
Full-text available
The Internet of Things (IoT) has particular applications in public safety as well as other domains such as smart cities, health monitoring, smart homes and environments, smart industry, and various types of pervasive systems. The attacker can simply attack the IoT device in such applications, because it is randomly distributed, dynamic topology and not reliable due to energy and communication limitation. Moreover, the threat to confidentiality and security is increasing as the number of devices connected in IoT is increasing. As the numbers of devices connected to the Internet is expanding, the threat to confidentiality and security is increasing. The aim of this paper is design a typical network security model for cooperative virtual networks in the IoT era. This paper presents and discusses network security vulnerabilities, threats, attacks and risks in switches, firewalls and routers, in addition to a policy to mitigate those risks. The paper provides the fundamentals of secure networking system including firewall, router, AAA server and VLAN technology. It presents a novel security model to defense the network from internal and external attacks and threats in the IoT Era. A testbed is built to investigate the proposed model, and the performed assessment show an effective security performance with a good network performance.
Article
Full-text available
The Internet is evolving rapidly toward the future Internet of Things (IoT) which will potentially connect billions or even trillions of edge devices which could generate huge amount of data at a very high speed and some of the applications may require very low latency. The traditional cloud infrastructure will run into a series of difficulties due to centralized computation, storage, and networking in a small number of datacenters, and due to the relative long distance between the edge devices and the remote datacenters. To tackle this challenge, edge cloud and edge computing seem to be a promising possibility which provides resources closer to the resource-poor edge IoT devices and potentially can nurture a new IoT innovation ecosystem. Such prospect is enabled by a series of emerging technologies including Network Function Virtualization (NFV) and Software Defined Networking (SDN). In this survey paper, we investigate the key rationale, the state-of-the-art efforts, the key enabling technologies and research topics, and typical IoT applications benefiting from edge cloud. We aim to draw an overall picture of both ongoing research efforts and future possible research directions through comprehensive discussions.
Article
Full-text available
With the rapid development of mobile internet and internet of things (IoT) applications, the traditional centralized cloud computing is encountering severe challenges, such as high latency, low spectral efficiency (SE), and non-adaptive machine type of communication. Motivated to solve these challenges, new technology is driving a trend which shifts the function of centralized cloud computing to the edge devices of the network. Several edge computing technologies originating from different backgrounds have been emerging to decrease latency, improve SE, and support the massive machine type of communication. This paper comprehensively presents a tutorial of three typical edge computing technologies, including the mobile edge computing, cloudlets, and fog computing. In particular, the standardization efforts, principles, architectures, and applications for these three technologies are summarized and compared. From the viewpoint of radio access network, the difference between mobile edge computing and fog computing are highlighted, and the characteristics of fog computing based radio access network are discussed. Finally, open issues and future research directions are identified as well.
Conference Paper
Drilling problems such as stick slip vibration/hole cleaning, pipe failures, loss of circulation, BHA whirl, stuck pipe incidents, excessive torque and drag, low ROP, bit wear, formation damage and borehole instability, and the drilling of highly tortuous wells have only been tackled using physics-based models. Despite the mammoth generation of real-time metadata, there is a tremendous gap between statistical based models and empirical, mathematical, and physical-based models. Data mining techniques have made prominent contributions across a broad spectrum of industries. Its value is widely appreciated in a variety of applications, but its potential has not been fully tapped in the oil and gas industry. This paper presents a review compiling several years of Data Analytics applications in the drilling operations. This review discusses the benefits, deficiencies of the present practices, challenges, and novel applications under development to overcome industry deficiencies. This study encompasses a comprehensive compilation of data mining algorithms and industry applications from a predictive analytics standpoint using supervised and unsupervised advanced analytics algorithms to identify hidden patterns and help mitigate drilling challenges. Traditional data preparation and analysis methods are not sufficiently capable of rapid information extraction and clear visualization of big complicated data sets. Due to the petroleum industry's unfulfilled demand, Machine Learning (ML)-assisted industry workflow in the fields of drilling optimization and real time parameter analysis and mitigation is presented. This paper summarizes data analytics case studies, workflows, and lessons learnt that would allow field personnel, engineers, and management to quickly interpret trends, detect failure patterns in operations, diagnose problems, and execute remedial actions to monitor and safeguard operations. The presence of such a comprehensive workflow can minimize tool failure, save millions in replacement costs and maintenance, NPV, lost production, minimize industry bias, and drive intelligent business decisions. This study will identify areas of improvement and opportunities to mitigate malpractices. Data exploitation via the proposed platform is based on well-established ML and data mining algorithms in computer sciences and statistical literature. This approach enables safe operations and handling of extremely large data bases, hence, facilitating tough decision-making processes.
Article
A key benefit of connecting edge and cloud computing is the capability to achieve high-throughput under high concurrent accesses, mobility support, real-time processing guarantees, and data persistency. For example, the elastic provisioning and storage capabilities provided by cloud computing allow us to cope with scalability, persistency and reliability requirements and to adapt the infrastructure capacity to the exacting needs based on the amount of generated data.
Article
The proliferation of Internet of Things and the success of rich cloud services have pushed the horizon of a new computing paradigm, Edge computing, which calls for processing the data at the edge of the network. Edge computing has the potential to address the concerns of response time requirement, battery life constraint, bandwidth cost saving, as well as data safety and privacy. In this paper, we introduce the definition of Edge computing, followed by several case studies, ranging from cloud offloading to smart home and city, as well as collaborative Edge to materialize the concept of Edge computing. Finally, we present several challenges and opportunities in the field of Edge computing, and hope this paper will gain attention from the community and inspire more research in this direction.
Automatic system identification of forward feed drives in machine tools
  • S Kehne
  • T Berners
  • A Epple
  • C Brecher
S. Kehne, T. Berners, A. Epple, C. Brecher, Automatic system identification of forward feed drives in machine tools, Advances in Production Research: Proceedings of the 8th Congress of the German Academic Association for Production Technology (WGP), (2019) 144 -152.
Understanding LSTM -a tutorial into Long Short-Term Memory Recurrent Neural Networks
  • R C Staudemeyer
  • E R Morris
R. C. Staudemeyer, E. R. Morris, Understanding LSTM -a tutorial into Long Short-Term Memory Recurrent Neural Networks, (2019).
  • D Park
  • S Kim
  • Y An
  • J.-Y. Jung
D. Park, S. Kim, Y. An, J.-Y. Jung, LiReD: A Light-Weight Real-Time Fault Detection System for Edge Computing Using LSTM Recurrent Neural Networks, Sensors, 18 (7) 2018.