Hochschule Ostwestfalen-Lippe

Open Master und Bachelor Theses

Please contact Prof. Niggemann for further information.

Diagnosis and optimization of industrial processes

The Fraunhofer Application Center Industrial Automation (IOSB-INA) develops a software solution for distributed diagnosis and optimization of industrial processes.
The process data is collected on an OPC UA server and centrally evaluated with an analysis computer. The data processing comprises data reduction, automatic diagnosis of process anomalies and intelligent optimization of process parameters. Dependent on the use case, the results can be displayed in different graphical user interfaces.
In the context of these investigations, there are several possibilities for praxis projects, Bachelors and Masters theses.
A possible topic is the development of diagnosis methods which allow for automatic error detection and the assignment of these errors to possible error sources. Such diagnosis algorithms have great practical importance as non-operation periods and error recovery in the industrial sector can cause high costs. A further topic which gains increasing importance is the energy efficiency of industrial processes. In the context of current research projects of the leading-edge cluster itsowl – Intelligent Technical Systems OstwestfalenLippe – it is possible to contribute to the development of modern optimization methods. These methods aim the automatic optimization of energy efficiency of technical systems. Alternatively or additionally it is possible to develop graphical user interfaces in the described application fields.
Thesis Type: Project work / Bachelor / Master
Begin: Immediately
Using QR-Codes for Location Based Plant Visualization

Using available software for QR-code scanning, a software has to be implemented to display location dependent diagnosis information of the Lemgo Model Factory. This should be implemented for a mobile device based on iOS or Android. This Thesis will be supported by a local company.

Thesis Type: Bachelor
Creation of a Mechatronic Plant Model using WinMOD

For the Lemgo Model Facory, a plant model should be created using the tool WinMOD. This model should used to simulate the plant (including PLCs) on a PC and late in a Hardware-in-the-Loop scenario.

Thesis Type: Bachelor
Runtime Estimation Algorithm for Programs

Using a given compiler suite, the worst-case and average runtime of programs (e.g. C programs) should be anayzed by means of analysis of the compiled code. For this, a model of the runtime features of a specific processor has to be created. For the modeling of the program's runtime, modeling techniques developed in Lemgo are used.

Thesis Type: Master oder Bachelor (limited topic)
Learning of Plant Behavior

To identify faults in complex plants, first of all a model of the normal behavior is needed. Using this model, computers can detect faults by comparing the model predictions to plant measurements. In this work, such models are learned using machine learning algorithms. For this, the plant structure and the model structure must be known beforehand. Here machine learning algorithms such as Support Vector Machines and Regression approaches are applied.

Thesis Type: Master and Bachelor (implementation of an algorithm)
Implementation of Monitoring Agents

An important problem for the control of complex plants and factories is the acquisition of all necessary data from the sensors. Due to the complexity of such plants and due to the spatial distribution of the sensors, this acquisition and aggregation of the data is a problem on its own. By integrating software agents into the PLCs and I/O devices, all data can be collected and aggregated on a dedicated server. In this work such an agent structure is implemented prototypically.

Thesis Type: Bachelor or Master (including time synchronization of the agents)
Implementation of an OPC UA client for process monitoring and diagnosis on Ipad and IPhone

To connect manufacturing execution systems and enterprise resource planing to the devices on manufacturing level, several technologies are used today. To indroduce an industrie wide standard OPC UA was developed to meed special requirements and provide properties necessary for industrial communication. In this work an OPC UA client for IPhone and/or IPad shall be developed. Using this tool it should be possible to monitor and diagnose the plant on a mobile device. The implementation shall be done using Objective C and Xcode. The Lemgo-Model-Factory located in the CIT building can be used for testing and demonstration.

Machine Learning, Data Mining, and Diagnosis
Learning the normal behavior using Selforganizing Maps

Self Organizing Maps (SOMs) are used to model sequences of structured procedures. The objective in this thesis is to implement the algorithm and adapt it to the learning of the normal behavior of automation systems. The Lemgoer Modellfabrik (LMF) can be used for testing the implementation. The learned model shall be used for diagnosis and novelty detection.

Prerequisites: A general understanding of algorithms and software engineering
Integration of a real-time capable protocol into the OPC UA standard for the data exchange distributed systems

The specially in the Fraunhofer Institut developed middleware for industrial automation (mINA) gives automation systems the possibility to provide an automatically reconfigurable behaviour. An important component within the software architecture is based on the OPC UA standard. The aim of this master thesis is to integrate a real-time capable communication protocol in the OPC UA standard. Furthermore the protocol shall be tested and a comparison to other protocols shall be made.

Thesis Type: Master

Thesis proposals in Artificial Intelligence, Machine Learning, Data Mining, Data Analysis

Project environment: Due to the high degree of automation in production, the complexity and the error rate increase. In order to manage the growing complexity self-learning diagnosis systems are used. For this purpose, methods are developed using approaches of Artificial Intelligence that mimic human behavior and automatically learn the normal behavior of the system. This knowledge base is used in the subsequent diagnosis phase, in order to detect deviations from normal behavior (anomalies) during runtime.

Topics: We offer theses in the field of artificial intelligence. Methods should be developed and evaluated that learn a model of the normal behavior automatically (without the need of expertise).
Furthermore, we offer theses in the field of model-based diagnosis. Here, algorithms will be developed for more effective detection of anomalies. Also, methods should be developed to identify the error cause.
All developments can be tested on realistic example systems.

Begin: immediately

Contact Person: M.Sc. Alexander Maier (alexander.maier(at)hs-owl.de)

Implementation and Evaluation Of A Time-Synchronization Ap-proach for Distributed Dataloggers

Background: Process monitoring is a more and more important task for automation systems today. In terms of condition and energy monitoring, lots of data has to be collected accurately. It is desirable, for machine diagnosis and process optimization, to acquire data from distributed probes in an industrial network in a synchronized way. Together with an industry partner Fraunhofer IOSB-INA implemented an architecture for datalogging which should be enhanced with time-synchronization. The architecture is able to monitor process data exchanged via industrial Ethernet protocols like PROFINET, EtherCAT or Modbus/TCP.

Approach to implement and evaluate should be based on the standardized IEEE-1588 time-synchronization protocol (Precision Time Protocol).

Research Questions

• What time accuracy can be achieved with this approach?

• For which typical industrial processes is this approach suitable?

If you are interested in this topic and heard of one of the following before you could be the perfect fit for us: C, C++, and (Industrial) Ethernet: Modbus/TCP, PROFINET, EtherCAT, IEEE 1588 (PTP)

Thesis Type: Project Work, Master

Dynamic Data Resampling

Background: Multiple data sources, especially when taken from different machines are often synchronized, but the sampling rate is different for each machine and type of data. If those data is saved or processed by anomaly detection algorithms for example, it is very important to briefly resample this data, especially if different types of continuous data streams are merged.

In statistics multiple types of resampling algorithms like Jackknife or cross-validation are used. The aim of this project work is to identify, test and implement a suitable algorithm for automatically
acquired data from automation machines.

Design Problems to Solve

1) Identify a suitable method and compute the statistical inference of given data.
2) Implementation of a Dynamic Data Resampling class in C# which supports merging and resampling of at least two data streams of different types with a suitable resample methods.  

Research Questions

• Which type of resample method is more suited in the context of automation  industry?
• Which type of model could be computed with the inference methods?
• Are there other, more suitable algorithms?

The project work will be conducted according to Master of Information Technology program regulations.

Thesis Type: Project Work, Master

Project work supervisor: Prof. Dr. rer. Nat. Oliver Niggemann

M² Sensorweb – Nutzbarmachung von Fahrzeugdaten zur Hochwasserprognose

Der Klimawandel führt zu neuen Phänomenen und Bedrohungsszenarien, auch in unseren gemäßigten Breiten. So werden Starkregenereignisse aus Sicht der Klimaforscher zunehmen. Hierdurch ergeben sich zunehmende Gefährdungen durch Hochwasserszenarien. Ein Schwerpunkt liegt in diesem Projekt auf Prognosen für kleine Regenwassereinzugsgebiete, in denen bislang nur eine geringe Zahl von Niederschlags- und Wasserpegelsensoren im Einsatz sind und damit für kleinere Gemeinden, die bislang von den Warnsystemen nicht erfasst werden bzw. nicht von ihnen profitieren können.
Im Forschungsvorhaben Mobile Moving Sensorweb (M² Sensorweb) soll untersucht werden ob die mangelnde Sensorik durch die Verwendung von den in Fahrzeugen integrierten Sensoren ausgeglichen werden kann. Fahrzeugdaten, wie z.B. Wischerfrequenzen und Regensensoren sowie Außentemperatur, fallen in modernen Fahrzeugen allemal an und können somit auch für kleine Gemeinden verfügbar gemacht werden. Durch die zunehmende Integration von Kommunikationstechnik in Fahrzeuge (Car2Car, mobile Internetanbindung, Smartphone-Integration) können diese Daten heute in Echtzeit kommuniziert werden.
Im Rahmen dieser Arbeit sollen erste Vorarbeiten geleistet werden, die die Nutzbarkeit der Fahrzeugdaten zeigen sollen. Hierfür soll der inIT E-Smart genutzt werden, der die  Außentemperatur und Scheibenwischerfrequenz über die Diagnoseschnittstelle bereitstellt. Diese Daten sollen ausgelesen und mit Metadaten zur Position und Höhe verknüpft werden. Anschließend sollen die Daten geeignet aufbereitet und visualisiert werden. Wünschenswert wäre eine Integration der Informationen in das bestehende elektronische Fahrtenbuch.

Typ: Praxisprojekt/Bachelorarbeit

Ansprechpartner: Jens Dünnermann, M.Sc. E-Mail: jens.duennermann(at)hs-owl.de