Department of Computer Science Technische Universitaet Muenchen
Chair XVI: Computer Aided Medical Procedures & Augmented Reality Home | Teaching | People | Publications | Labs & Locations| Applications | Open Positions | History | Partners | Internal
MQM  Model-Based Systems & Qualitative Reasoning            
Home | Projects | Teaching | People | Publications | Theses & Guided Research | Jobs

Research Projects

In progress..

Finished projects

  • INFIS  
  • Intelligent Fitness Studio

  • FAST  
  • Fault-model and specification-based test generation

  • MOM  
  • Model Manipulation Environment
  • LineMET  
  • Automatic Model-based Efficiency Analysis of Bottling Plants

  • Tragwerk-FMEA  
  • Computer-based Support for Quality Assurance of Structures (Structure FMEA)

  • LineMod [2007 - 2008] 
  • Diagnosemodelle für verkettete Abfüll- und Verpackungslinien in der Lebensmittelindustrie

  • AUTAS [2002-2005]
  • Automating FMECA for aircraft systems

  • MONET 2
  • The European Network of Excellence in Model-based Systems & Qualitative Reasoning

  • AGUA [2000 - 2003]
  • Assistant for Water Quality

  • IDD [2000-2003]
  • Integrated Diagnosis and Design for on-board Diagnosis

  • MONET [1997-2000]
  • The European Network of Excellence in Model-based Systems & Qualitative Reasoning

  • VMBD [1997-1999]
  • Vehicle Model Based Diagnosis

  • INDIA [1996-1999]
  • INDIA - Intelligent Diagnosis in Industrial Applications

  • QUALIDADE [1994-1999]
  • Qualitative Reasoning in the Domain of Automated Diagnosis and Ecology



Automated, expert-based training plan generation

Application problems: Configuration is the task of assembling a customized system from generic components. This is achieved by deducing suitable components from a component catalog based on user goals and properties and combining them appropriately. Within the scope of INFIS the concept of generic configuration, see [Kazmaier 2012], will be applied to generate user specific training plans. In this application vague user goals, such as "Muscle Gain" or "Losing Weight" will be applied along with biometric user properties, e.g. age or weight, to generate training plans. To this end, a knowledge base will be designed and implemented for representing knowledge from sport and training sciences. Training exercises are considered as components were as the training plan will be the desired configuration. The domain constraints, represented in the knowledge base, contain knowledge about how training methods and exercises affect certain muscle groups and how to achieve the training goals. Training Restrictions occur in form of the medical abilities and biometric parameters of the user.

INFIS is a cooperation between MQM, the TUM Sports and Health Department and the eGym GmbH, a Munich-based manufacturer or electronic fitness machines.

Research challenges

  • Knowledge Representation
  • Configuration based on soft and structural constraints
  • Diagnosis of the INFIS knowledge base
  • Links: For more information, see

    [Kazmaier 2012] Kazmaier F.: A Generic Configuration Tool and its Application to Vehicle Configuration. Master's Thesis, Tech. Univ. of Munich, Comp.Sci. Dept. 2012

    Contact: Prof. Dr. Peter Struss , Florian Grigoleit, M.Sc.


    LineMET TOP

    Automatic Model-based Efficiency Analysis of Bottling Plants

    Application problems: The project aims at extending the results of LineMod, which localizes the causes of downtime of the filler in terms of disturbances of single machines or conveyors causing gaps or tailbacks in the flow of objects through the plant. LineMET addresses reasons for reduced efficiency of the plant beyond such hard failures of individual machines:

    • The throughput is not only affected by a complete stop of the filler, but also in phases where the speed of the machine has to be reduced in an attempt to avoid stops. The model will be extended to cover such situations.
    • Sometimes, lack of input or tailback at the output cannot be blamed on a single cause, but results from a coincidence of several, distributed inadequacies. Besides a principled way to assess the contribution of several machines, this will require the representation of more complex behavior patterns, such as "stuttering" of machines. The current demonstrator has already been designed to support this by enabling the definition of complex features via more low level variables.
    • Another challenge arises from the fact that, sometimes, the ultimate cause of a problem lies beyond the boundaries of the physical plant, which is not captured by the previous model. For instance, the logistics (e.g. delivering pallets with improper bottles or with delay), manual intervention (sorting out improper bottles), bad production material or maintenance schedule may be inappropriate.
    • Finally, the project will provide means for automatically checking improper production data to avoid spurious diagnostic results.

    Research challenges: Like the LineMod demonstrator,a consistency-based model-based approach [Struss 08] will be pursued. A major research problem to be tackled in this project is temporal diagnosis with highly uncertain and arbitrary delays. As a result, the temporal indexing of model-based predictions and diagnoses cannot be based on uniform delays and a pre-defined, finite set of time points as in previous solutions. 

    Links: For more information (in German), see

    [Struss-Ertl 09] Peter Struss, Benjamin Ertl : Post-mortem Diagnosis of Bottling Plants Based on Recorded Data In: Safeprocess'09, 7th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, Barcelona, 2009. ISBN: 978-3-902661-46-3

    [Struss 08] P. Struss: Model-based Problem Solving In: van Harmelen, F., Lifschitz, V., and Porter, B. (eds.). Handbook of Knowledge Representation, Elsevier, 2008, ISBN-13: 978-0-444-52211-5, pp. 395-465

    Contact: Prof. Dr. Peter Struss


    Tragwerk-FMEA TOP

    Computer-based Support for Quality Assurance of Structures (Structure FMEA)

    Application problems: Today, design planning processes are performed under tight time constraints, iterative and therefore liable to errors. With the increasing complexity of modern structures planned and designed with the aid of computers, the existing regulatory systems of preventive quality control are becoming increasingly inadequate. To mitigate the occurrence of planning mistakes and the resulting damages during construction and use of the structure due to insufficient robustness, a new approach to quality management of the process of structural design is, especially for the early design stages, is urgently needed.

    The project builds upon FMEA (Failure Mode and Effect Analysis, see DIN EN 60812), a quality assurance method that is applied in other engineering areas and, for instance, mandatory in the automotive and aeronautics industries. The project aims at providing a standardized method based on a model of the structure, which is transformed to an appropriate level of abstraction, exploiting results from a related project of the European aeronautics industries (AUTAS ). It enables the user to objectively estimate the risk of failure of a structural part(s) due to collapse and to take necessary counter-measures at early stages of the process.

    The system helps ensuring the quality of structural safety, independent of the size of the structure, in which possible planning errors and resulting risks are analyzed and enumerated. Of particular interest are a quick and specific identification of critical structural parts and a clear definition of priorities in the quality of design calculations.

    Research challenges: FMEA is an inherently qualitative analysis, since it has to cover classes of failures and also assess the impact in terms of its significance. Since, on the other hand, the basis are numerical computations (using a FE-model), the model for performing the analysis has to be abstracted to a qualitative level. This is a challenge to our work on automated model abstraction [Struss 02], [Fraracci 09]. 

    Links: For more information (in German), see

    [Eisfeld-Struss Bautechnik] (in German) M. Eisfeld, P. Struss : Qualitätssicherung im konstruktiven Ingenieurbau In: Jahrbuch 2009 Bautechnik. VDI Verlag, Düsseldorf: 2008. pp.179-188. ISBN: 978-3-18-401660-9

    [Fraracci 09] Alessandro Fraracci: Model-based Failure-modes-and-effects Analysis and its Application to Aircraft Subsystems.Dissertationen zur Künstlichen Intelligenz DISKI 326, AKA Verlag, ISBN 978-3-89838-326-4, IOS Press, ISBN 978-1-60750-081-0

    [Struss 02] P. Struss: Automated Abstraction of Numerical Simulation Models - Theory and Practical Experience. In: Sixteenth International Workshop on Qualitative Reasoning, Sitges, Catalonia, Spain, 2002.

    Contact: Tesfaye Regassa



    Fault-model and specification-based test generation

    Application problem: Passenger vehicles turn more and more into electronic platforms and the application of embedded systems affects safety-critical areas. With the increasing number of functions in cars, electronic systems have become more complex. In current practice, requirements are not specified in a formal way. This is unlikely to change in the near future and prevents the application of formal verification methods. To ensure quality and compliance with the requirements to some extent despite this fact, testing of embedded systems is necessary and crucial. Existing approaches to automated test generation are rarely used in practice. One reason is that models that are formal, consistent and complete, are required. Creating them not only consumes significant effort and time and also requires special training in formal. Secondly, existing approaches to test generation do not fulfill the test engineers?demands for transparency and flexibility. This is because they are based on abstract and structural model properties and not to the user requirements. In addition, test criteria are not easily adaptable to match the engineer’s perspective and knowledge. The aim of FAST is to address these issues and, therefore, to establish the application of automated test generation in today’s practice:

    • Test generation is based on (functional) user requirements, rather structural properties of the code.
    • Requirements are entered in a structured, but natural-language-oriented manner not too far from the current practice.
    • It allows the test engineer to explicitly express fault hypotheses, which are used as test ideas and which determine the detail and focus of the generated test suite.
    • The generated tests can be justified and inspected by the test engineer and exported to EXAM, a tool which is well established in the automotive sector and which allows the automatic execution of a test suite on a test bench.

    Research challenges: The project builds on qualitative fault-model based test generation [Struss94]. While the original work is based on state-based relational models, requirements on embedded systems contain metric temporal constraints. Furthermore, the model has to be automatically generated from the functional requirements stated by the user. To this end, natural language sentence templates are transformed into statements in a formal requirement language (FRL) with a precise semantics. They are then translated into timed constraint automata, providing the model for test generation. In order to limit state-space explosion, requirements that are irrelevant for an individual test task are ignored, and, thus, only a partial model rather than the entire set of requirements is used for computation. The test generation algorithm computes tests by performing a search in the state-space of the timed automata incrementally (by expanding the input sequence). The correctness of the translation of FRL into timed constraint automata and of the test generation algorithm, as well as its completeness is formally proven. In addition, a proof is given for the completeness of test generation based on the partial models. The stepping stone for FAST is an evaluation on a real-world case-study, the adaptive cruise control system (ACC).

    [Struss 94] Struss, P.: Testing Physical Systems. In: Proceedings of AAAI-94, Seattle, USA, 1994.

    [Esser-Struss 07]  M. Esser, P. Struss: Fault-model-based Test Generation for Embedded Software In: International Joint Conference on Artificial Intelligence (IJCAI). 6th January through 12th January 2007, Hyderabad, India

    [Esser-Struss DX07]  M. Esser, P. Struss: Obtaining Models for Test Generation from Natural-language-like Functional Specifications In: G. Biswas et. al. (eds.), DX'07, 18th International Workshop on Principles of Diagnosis. May 29-31, 2007, Nashville, USA, pp. 75-82.

    [Esser-Struss IICAI07]  M. Esser, P. Struss: Automated Test Generation from Models Based on Functional Software Specifications In: 3rd Indian International Conference on Artificial Intelligence (IICAI-07), December 17-19 2007, Pune India.

    Contact: Prof. Dr. Peter Struss


    Model Manipulation Environment

    Application Problems: As industrial plants and products become more complex and economical and environmental constraints become more tightened, the need for using models to support various tasks during the life-cycle of a product or plant (such as design, control, on-board and off-board diagnosis, testing, and maintenance) increases. Also in other areas, such as biology, ecology, or economics, models are used for testing hypotheses and planning interventions.

    The degree of re-use of models for different systems, scenarios, and purposes is very low, even when they are quite similar, and there are few libraries containing models that are context-independent. Models used for successful projects often have specificities of the task, domain, or even device compiled into them, so that they cannot be integrated into different projects even of similar nature. Another obstacle is the use of different modeling paradigms and tools that cannot be integrated. 

    The Model Manipulation Environment (MOM) project has been initiated to develop foundations and a software tool for managing, manipulating, transforming, and integrating models. It should support the user in building and using model libraries with a higher re-use of its elements and make the modeling process more transparent.

    Research Challenges:

    • Building an ontology for a layered representation that distinguishes and explicitly represents alternative modeling decisions (an appropriate structural representation, as well as choices of relevant quantities, variables, and domains) that enables the re-use of those layers that are shared among different models
    • Supporting and automating the integration of different fragments of a behavior model, e.g. by mappings between different domains
    • Designing abstract interfaces to support the integration of various existing modeling environments, such as Matlab/Simulink, Modelica, Spice, etc
    • Developing the theoretical foundations and specification for model operators (e.g. for automated model abstraction, aggregation, structural transformations) that use the abstract interfaces and, hence, are applicable to classes of modeling environments.
    • Developing a model of the modeling process, including problems of distributed development of models and model libraries.

    [Fraracci 09] Alessandro Fraracci: Model-based Failure-modes-and-effects Analysis and its Application to Aircraft Subsystems.Dissertationen zur Künstlichen Intelligenz DISKI 326, AKA Verlag, ISBN 978-3-89838-326-4, IOS Press, ISBN 978-1-60750-081-0

    [Regassa09] Tesfye Regassa: Specification, Implementation and Evaluation of a Tool for Multiple Modeling. Master thesis, Technische Universität München, Comp. Sci. Dept., 2009

    [Struss-Regassa 10] Peter Struss, Tesfye Regassa: MOM-An Environment for Multiple Modeling, 24th International Workshop on Qualitative Reasoning, Portland, USA, 2010

    [Struss et al. 2011] Peter Struss, Alessandro Fraracci, D. Nyga: An Automated Model Abstraction Operator Implemented in the Multiple Modeling Environment MOM, 25th International Workshop on Qualitative Reasoning, Barcelona, Spain, 2011.

    Contact: Tesfaye Regassa

    LineMod - Diagnosemodelle für verkettete Abfüll- und Verpackungslinien in der Lebensmittelindustrie TOP

    Technische und organisatorische Schwachstellen führen dazu, dass Abfüll- und Verpackungsanlagen heute nur mit Verfügbarkeiten von 50-75 % betrieben werden. Um Stillstandszeiten zu minimieren und dadurch die Effizienz zu steigern, müssen die Hauptstillstandsverursacher (Maschinen, Fördermittel und Peripherieschnittstellen) möglichst rasch identifiziert werden. Die Wechselwirkungen der Aggregate innerhalb der veketteten Linien machen dies schwierig. Aufgrund der Entkopplung der Maschinen durch die zwischengeschalteten Gebindepufferstrecken führt nicht jeder Fehler an einer Maschine zwangsläufig zu einem Stillstand des Zentralaggregats. Auf der anderen Seite können sich Maschinenstillstände verschleppen und sich erst zeitversetzt (auch lange nach der Fehlerbehebung) auswirken. Auf der Grundlage automatischer Betriebsdatenerfassung (BDE) kann eine statistische Auswertung von Maschinenstillstandszeiten erfolgen. Diese scheitert jedoch häufig an uneinheitlich definierten Zuständen bei Maschinen verschiedenen Fabrikats und ist aufgrund der Wechselwirkungen entlang der Linie nicht ausreichend. Es werden deshalb, zusätzlich zu einer standardisierten Datenbasis, Modelle benötigt, die die zeitlichen Abhängigkeiten innerhalb von Abfüll- und Verpackungsanlagen berücksichtigen und auf deren Grundlage Lösungen für eine unterstützte, weitgehend automatiierte Fehlerlokalisierung entwickelt werden können. Für die Lebensmittelindustrie existieren diese bisher nicht. Das Ziel des Projekts LineMod ist der Aufbau einer Bibliothek von Diagnosemodellen, mit welchen Betriebsdaten in Kombination mit der jeweiligen Anlagenstruktur auf verfügbarkeitsmindernde Störungen hin untersucht werden können. Hierbei sollen sowohl das stillstandsverursachende Aggregat im Rahmen einer horizontalen Diagnose innerhalb der Anlage identifiziert als auch die Ursache innerhalb dieses Aggregats im Rahmen einer vertikalen Diagnose diagnostiziert werden. Den zeitlichen Abhängigkeiten, die bei der Datenauswertung bisher vernachlässigt wurden, kommt eine besondere Bedeutung zu.

    Links: For more information (in German), see


    AUTAS - Automating FMECA for aircraft systems TOP

    Failure modes and effects criticality analysis (FMECA) is concerned with the assessment of effects of potential faults in a designed artifact, their criticality, detect ability and possible remedies.
    The top-level objective of the project is the improvement of the reliability analysis process through the development of a software environment that will enable specialists doing reliability analysis, to perform their job in a reduced time span, and to improve the quality of their output. Today the available support mainly concerns the possibility of storing and organizing data about the failure states of each component, while for the reliability analysis itself poor or no support at all is provided.


    MONET is a Network of Industrialists, Academics and Researchers with a common long-term technological development objective in Model Based Systems & Qualitative Reasoning (MBS & QR). The MONET European Network of Excellence is aimed at providing a long-term framework for research co-operation and integration that will assist in the coordination of European research in model-based systems and qualitative reasoning through technology transfer into industry. The Network is co-ordinated by the Centre for Intelligent Systems within the Department of Computer Science, University Wales, Aberystwyth, UK. The second phase of this network has started in January 2002.

    AGUA - Assistant for Water Quality TOP

    The project aims at building the prototype of a tool that supports operators of water treatment plants in assessing the state and evolution of both the plant and its natural resources (reservoirs), in identifying conflicts with quality goals and regulations, and in determining and evaluating potential remedial actions.

    [StrussHeller99a] Struss, P. and Heller, U.: Model-based Support for Water Treatment. In: Milne, R. (ed.): Qualitative and Model Based Reasoning for Complex Systems and their Control, Workshop KRR-4 at the 16th International Joint Conference on Artificial Intelligence, Stockholm, 1999, pp. 84-90

    IDD - Integrated Diagnosis and Design for on-board Diagnosis TOP

    The importance of diagnosis in on-board automotive systems is constantly growing together with the complexity of the systems. The average dimension of the diagnostic code inside a modern electronic control unit (ECU) is now more than 50% of the whole code. The diagnostic code has the important role of detecting deviations of the system from the correct (or the optimal) behavior and identifying all possible and foreseen problems to which the identified situation could lead.
    This project aims at introducing a new approach to the problem. The industrial goal is to formalize and standardize the diagnostic design process, anticipating the introduction of diagnosis in the chain. This methodological goal has to be combined with another important goal: giving to the designers tools that can help them in evaluating and understanding the effects (and, specifically, the effects on diagnostic procedures) of each choice on the system being designed.

    [VDI-01] R. Brignolo & others: Integration of Design and Diagnosis Into a Common Process  In: Electronic System for Vehicles, VDI-Berichte, Düsseldorf 2001

    VMBD - Vehicle Model Based Diagnosis TOP

    The objective of the project is the development of an integrated environment (tool kit) that will allow car manufacturers and component suppliers to rapidly develop powerful on-board and off-board diagnostic systems able to reach the goal of 50% reduction of down times with 5 years advance. This will give a competitive edge to European Automotive Industry, leading at the same time toward the development of European regulations in the field of vehicle  diagnostics.

    [Sachenbacher/Struss/Carlen 00] M. Sachenbacher, P. Struss, C. Carlén: A Prototype for Model-based On-board Diagnosis of Automotive Systems  In: AI Communications, Vol 13(2), 2000

    INDIA - Intelligent Diagnosis in Industrial Applications TOP

    INDIA aims at a substantial contribution to the industrial application of knowledge-based - in particular: model-based-diagnosis of technical devices. To this end it wants to improve the possibilities to tailor a diagnostic system to the demands of a certain application (e. g. with respect to safety, ecology, variety, integration into development and runtime environment).
    It approaches these aims from two sides. On the one hand it transfers results of research into industrial applications. On the other hand it analyses the problems of this process and use them to drive further research. Hence, the consortium consists of research institutes, vendors of diagnostic systems and users of such systems.

    [Cunis/Struss 01] Cunis R., P. Struss: INDIA - Intelligente Diagnose in der industriellen Anwendung, pages 1-9. In: Hotz L.,Struss P.,Guckenbienl T.,(editors). Intelligent Diagnosis in Industrial Applications.Shaker Verlag, Aachen,Germany, 2001.

    QUALIDADE - QUALItative Reasoning in the Domain of Automated Diagnosis and Ecology TOP

    The project QUALIDADE aims at developing reasoning techniques and implementing tools for ecological and environmental management support.
    The application domain is the support of hydro-ecological resource management as carried out by the municipal department of water and sewage (DMAE) in Porto Alegre. The ecosystem of Rio Guaiba, the water supply of the city, is endangered by repeated occurrences of local algal blooms. The research department of DMAE aims at understanding, predicting and preventing these occurrences.
    These tasks will be supported by a model-based approach, which forms the foundation for a coherent framework for model building, situation assessment, monitoring, simulation, diagnosis and therapy planning.

    Last updated: Gulnar Mehdi, November 21st, 2013

    Computer Science XVI | Technische Universität München