International Conference on Data Mining and Artificial Intelligence Methods for Automotive Systems
April 20, 2017
Information transfer through standardized data structures – Vision, Christian Sinzinger
The automotive industry is currently experiencing a big chance, production data gains more importance and to stay competitive in the future it is necessary to use methods of the digital factory. Two of these methods, AutomationML and the concept of cloud based data storage lead to the development of an application, which should make it more easy to organize data and accessible, thus the engineering process of new production stands becomes more efficient. To get a general overview the schematic structure between AML and the data storage as well as use cases should be shown, before the following presentations provide a more detailed view on these topics.
Information transfer through standardized data structures – AutomationML, Flavian Meltzer
For a decade, it has been officially known that the most cost-intensive part of a body-building project is in software engineering. The reason for this is the fact that in the engineering process many types of information from their perspective tool chains must come together and be combined. This situation is intensified by the heterogeneous engineering tool landscape that makes it difficult to use existing data and information from finished engineering steps without using a paper interface. For this reason many representatives of the automotive industry came together to solve these problems, which resulted in the AutomationML format. AutomationML is an independent data format that allows bridging the gap between the various engineering fields and tool chains, thereby improving the overall process. This means that tomorrow's engineering process can keep pace with the ever-brimming technological advances.
Information transfer through standardized data structures – Modeling, Sebastian Schrittenloher
Information was given, why a standardized data format is needed and the concept of AutomationML was explained within the last Abstract. Now the aim is to show how working with this new data format is possible. For this a tool prepared by the AutomationML Konsortium is used. It splits up the information into three main types, the roles, interfaces and elements (objects). AutomationML is in general an object oriented data format. This aspect is used to give a structure to the information, similar like it is done in XML.
When a model is built in AutomationML, the information gets its structure automatically by using the tool of AutomationML. In this case an example of a working station was modelled, that keeps information like pneumatics, mechanics and kinematics. One way how to structure an already existing machine should be shown as well as the actual used principles of the tool.
Model-View-Controller concept for the processing of AML model, Marek Drewniak
The process of gathering important technological aspects of an industrial installation as a consistent AML model is the basis for actions that can be performed within the Digital Factory concept. The next big step is to process that model in order to transfer the knowledge and support all actions related to design, virtual comissioning and maintenance of a production stand. The proposed idea for processing of the model bases on the use of .NET Framework Model-View-Controller (MVC). The Model part contains an hierarchy derived out of AutomationML which is translated to class-organised structure. The View takes role of the user interface, which links the technological description with external systems, specifically the human users using web applications. The Controller is responsible for mediation between the data stored in the Model and the View. The whole solution can be used as a platform which grants an access to specific field-oriented knowledge or to point out dependencies between installation components. This can be useful especially for all maintenance actions and virtual comissioning. Additionally, it can fill the space between other systems included in the Digital Factory concept, e.g. the shopfloor and production management systems.
Potential of a reference trajectory for energy comparability of robotics in the automotive industry, Michael Stempfle
The constant increase in the price of energy and more restrictive emission norms cause new challenges for automotive production sites. Especially the body shop has a high impact to factory costs and production firms are under pressure to lower operating costs on a regular basis. One of the largest consumer of energy is the 6-axis industrial robot. The high number of robots in the body shop provides an important lever for significant energy saving in the automotive body construction. This presentation discusses the development of a special reference trajectory and its potential to compare the energy efficiency of robots with different sizes or from different manufacturers. The results can help with the choice of energy efficient robots in an early production planning phase. Furthermore, it is shown how a reference cycle can be used for other studies as well.
A new approach to increasing energy efficiency by utilizing cyber-physical energy systems, Martin Bornschlegl
Cyber-Physical Systems and Industry 4.0 topics are in the focus of current research. Due to the increasing awareness for an optimal use of resources and the initiated turnaround in energy policy, it becomes more essential for a car manufacturer to increase transparency about energy consumption. Therefore, especially energy management systems can be improved with new opportunities due to interconnected systems. Cyber-Physical Energy Systems (CPES) will help to handle the use of volatile renewable energies and to realize intelligent shut down strategies. This presentation gives an overview about current CPES research topics and shows an approach, which helps to reduce the energy consumption. Thereby is the focus on smart industries and main aspects are avoidance of peak loads and reduction of base load from the perspective of a car manufacturer. This is one step towards the smart factory concept.
An Aggregated Cyber Physical System – A Holistic “Model of Models” based on OPC UA, Suprateek Banerjee
Information modelling is one of the important aspects in today’s automation scenario. Currently there are numerous information modelling standards available which focus on specific independent aspects of the whole system. In case information is needed from multiple information models each modelling a different aspect of the system, it results in a mesh like scenario. This paper discusses an approach to resolve the resulting mesh scenario. The initial sections discuss the problem in greater detail, thereafter introducing the concept of Aggregation and how it could be a possible solution. Finally, the paper presents an architecture which can be taken as a guideline for implementation purposes.
The current measurement of the Demonstration Vehicle, Dariusz Zonenberg, Adam Ziebinski, Jakub Nowowiejski
The main objective of this project is to implement the measuring system in the Demonstration Vehicle in a close cooperation with AIUT company. The Demonstration Vehicle is a holonomic mobile robot platform with Siemens S7-1200 PLC as a main controlling unit. The communication process between the Demonstration Vehicle and the measuring application is maintained with the use of Snap7 library, which has a freeware license. This measurements are used for monitoring and diagnosing purposes. The maximum measured current value due to hardware specification is equal 2.4 [A]. There were designed two methods for calibrating the current sensors. The final method was selected after the series of tests. Also, the communication process with the use of Snap7 library was verified. For estimating the average power consumption of the Demonstration Vehicle there was designed an algorithm with the featured assumptions (can calculate the approximated consumption in [Ah] unit). Thanks to the measurements, not only the current consumption and worktime of the mobile robot can be estimated, but also the behaviour of other vehicle’s components can be analysed in the diagnostic processes. As a result, it may be possible that robot stop performing the actual task in case of detected irregularity in measurements.
Energy efficiency analysis by PLC - results of OKO production line research work at AIUT, Dariusz Zonenberg, Chłopaś Łukasz
The presentation focuses on the implementation of a system capable in a diagnosis of a OKO i5 production line. It mainly emphasizes on the solution of a problem including methods of the implementation. The solution consists of several steps. The first task is an analysis and characterisation of the production line, its work conditions and mediums which supply individual end-point devices. Additionally, the requirements for the measurement data are formulated in order to create the methodology for their collection. The second task is the processing of signals from the PLC and the production process. The purpose of this task is to aggregate and filter only the needed data, so the processing is maximally efficient and functional. This is especially important due to the amount of stored production data and limiting the time needed for its processing. The final part of the work is the practical use of the information extracted from the measurement data. In order to obtain reference values the post-processing mechanisms evaluate measurement data gathered from the production line. The results are forwarded to an on-line processing system which calculates the average consumption values, verifies the results and compares them to master values modified by the standard deviation. The results of the comparison are used for notification of line’s work, e.g. by setting alarms in case of abnormal energy consumption. Notifications are displayed on an operator’s panel and forwarded to OPC server for statistical analysis.
Energy efficiency analysis for OPC UA and Rapid Minier - results of OKO production line research work at AIUT, Rafal Cupek, Dariusz Zonenberg, Jakub Duda
This presentation of research work performed at AIUT is related to energy efficiency analysis of OKO production line. This work was a direct continuation of previous secondment at AIUT. Initial goal was gather and analyzed energy consumption signals from real, production robotic assembly line used in second stage (backend) production of GPRS telemetry devices. Initial task was to prepare data mining process: data preparation, gathering, cleaning and transformation. Signals (energy measures and PLC status bits) important for energy efficiency analysis were identified and gathered - production line cycles were distinguished and energy consumption per cycle computed. This allowed to process and analyze data in Rapid Miner tool using various data mining techniques, including data clustering based on energy consumption, cross validation (energy efficiency and actual PLC program status) and computation of energy alarm thresholds for line. Research team looked also for anomalies related to energy consumption and for ”markers” for abnormal energy consumption that could be identified on-the-fly in production environment. This allowed do compute and implement energy alarm thresholds and apply energy efficiency warnings into PLC itself. One of the goals was try to abstract analysis from this particular line - findings and tools developed should be universally applicable to discrete production lines.
Power management module for mobile platform, Adam Ziebinski, Paweł Rybka
Mobile platforms just like any other battery-powered applications, should be designed in a way which would enable them to run as long as possible. For the sake of saving power in the autonomous vehicle project there was created a power management system which consists of a transistor power key (which allows powering on and off 16 independent devices) and current measuring circuits thanks to which power consumption can be monitored. Such system enables saving energy – unused devices can be powered down and turned back on when their readings are essential. The module can be also used for diagnosing malfunctions or even recognizing what kind of device is connected to the key basing only on their supply current waveforms (for example with the use of neural networks). Moreover, not every device mounted on the platform runs on 14.8 V (2x7.4 V batteries connected in serial), so additionally there are several DC-DC converters to be added (for 5 V and 3.3 V). The final module (including PCB project) is based on STM32 ARM microcontroller (STM32F303), 4 shift registers (74HC595), 32 N-Channel MOSFET transistors (IRLML2502), 6 current shunt monitors (INA199A3), 3 step-down converters (ST1S10) and several passive components (resistors, capacitors and coils).
Indoor localisatio, Krzysztof Tokarz, Łukasz Szczepański, Krzysztof Wosik
Nowadays, in digital age people are accustomed to that they can easily get information about where they are and how they can reach any point on the globe. That all is thanks to localization systems like global positioning system (GPS), GSM operators data etc. These systems has one important disadvantage. They cannot work in closed areas ( buildings, undergrounds etc.). Localization in closed areas can make people life much easier, for example we can find straight path to shop of our interest in big shopping center, free place to park a car in underground parking, place where before we park a car or any location in chosen building. In AutoUniMo project we try to develop in-door localization system for our autonomous platform, but nothing stay in a way to use it for other purposes. Many concepts and ideas have been developed for the in-door localization systems. Researchers try to base their projects on ultrasound waves, 802.11 wireless network (WiFi) , inertial measurements and other. We had chosen two ways to achieve our goal. Firstly, we used strength of WiFi signal. If target device (user) take the information about signal strength of Wifi transmitter in his vicinity his position can be estimated based on special map on which every specific point is described by strength of signals of all Wifi transmitters in this point. Secondly, one will base of a Doppler effect which is the change in frequency or wavelength of a wave for an observer moving relative to its source. Thanks to that phenomena we can measure target velocity and direction of move depending on source. Enough many sources make possibility to estimate target position. Both solutions have advantages and disadvantages, but we think it is possible to develop working in-door localization system based on them.
XML data acquiring over the Ethernet for the processing in CANoe, Adam Ziebinski, Rafal Cupek, Marek Drewniak, Michał Kruk
A growing number of formats and interfaces used in various sensors induce using universal formats like XML. Data formatted in XML and encapsulated into TCP frames can be easily distributed for further use. From the other hand, there is a need of receiver that depending on need can process this data or forward it. CANoe software from Vector company includes those functionalities and offers a wide range of tools for data analysis and the graphical interpretation (user interface, waveform graphs and data analysis dedicated windows).
This data can be also used for logs creation and allowing the measurement recreation. The purpose of this work connects solutions mentioned above. The mobile platform is equipped into RaspberryPi and sensors (accelerometer, magnetometer, gyroscope, encoders, ultrasound distance sensor). Measurements from all the sensors are gathered and buffered by RaspberryPi in ASCII coding using specially designed XML format. Passive TCP server built in RaspberryPi is being periodically queried for the measurements by client on PC. Later on data is received, decoded in CAPL node configured in CANoe and displayed on the user interface. It is possible to create logs, analyze measurements and plot the charts basing on decoded data.
Lidar CAN bus communication with Raspberry Pi PiCAN module, Adam Ziebinski, Rafal Cupek, Marek Drewniak, Michał Rędziński
The simultaneous development of automotive, electronics and necessity of protection all road users has led to the introduction of Advanced Driver Assistance Systems. ADAS consists of many components that are usually connected by CAN bus, commonly used bus in automotive. Individual items are specialized and prepared to work with other ADAS elements. Using the ADAS component to work with another system requires the preparation and proper operation of the CAN bus. It is usually necessary to simulate the original environment to which this module was adapted. Raspberry Pi as a versatile computer platform and in combination with the PiCAN module allows to create communication with dedicated ADAS modules. One of the ADAS module is Lidar - Light Detection and Ranging. Lidar allows to measure the distance and speed of approaching obstacles in front of the vehicle. Researchers using Raspberry Pi and PiCAN have created a program that makes use of Lidar separately from the rest of ADAS. They also solved problems encountered while preparing communication with Lidar. The last step was to measure the performance and stability of a built system, depending on various factors, such as the CAN bus load. Alternative communications solutions such as CANoe and Ethernet/CAN module have been used for research purpose.