Citation
Ilge Akkaya, Patricia Derler, Edward A. Lee. "Aspect-Oriented Fault Modeling and Anomaly Detection in Ptolemy II". Talk or presentation, 7, November, 2013; Presented at the 10th Biennial Ptolemy Miniconference, Berkeley. .

Abstract
Fault modeling has been a primary focus of the well-established fields of circuit design and fault-tolerant control system research. For the emerging research field of cyber-physical systems, this modeling aspect plays an even more fundamental role due to the increased complexity in the networked-real time system. As concurrency and distributed behavior become ubiquitous in system design, faults due to networks, latency and cyber-physical interactions gain importance and detecting faults in pre-deployment stages becomes essential. In the domain of modeling and simulation, behavioral modeling often focuses on implementing the correct behavior that takes no account of the faults that might occur in system execution. However, numerous fault scenarios exist in data content, communication, analog and digital component failures, all of which can drive system behavior beyond the model context. Cyber-Physical System models can be refined continuously to address more possible behavior under different fault conditions, however, at the expense of increasing model complexity beyond usefulness. Aspect-oriented modeling is a means to represent different concerns related to system design as separate entities that are evaluated in concurrency with a functional representation. The separation of concerns in this sense has several benefits in fault modeling such as enabling modular and flexible representations for fault aspects while maintaining model simplicity and the models that define of correct operation. In Ptolemy, an aspect-oriented modeling formalism is developed to inject families of deterministic, nondeterministic and stochastic faults and to apply anomaly detection techniques to monitor anomalies in system behavior. The Modal model domain has been extended to enable probability expressions at transitions, and in turn, to represent Markov chain and Hidden Markov Model behavior. On the anomaly detection side, a component library has been developed for Hidden Markov and Mixture Model parameter estimation and classification.

Electronic downloads

• PtMiniconferenceTalk.pdf · application/pdf · 3012 kbytes

• HTML

  Ilge Akkaya, Patricia Derler, Edward A. Lee. <a
  href="http://chess.eecs.berkeley.edu/pubs/1038.html"><i>Aspect-Oriented
  Fault Modeling and Anomaly Detection in Ptolemy
  II</i></a>, Talk or presentation,  7, November,
  2013; Presented at the <a
  href="http://ptolemy.org/conferences/13" >10th
  Biennial Ptolemy Miniconference</a>, Berkeley.
  .

• Plain text

  Ilge Akkaya, Patricia Derler, Edward A. Lee.
  "Aspect-Oriented Fault Modeling and Anomaly Detection
  in Ptolemy II". Talk or presentation,  7, November,
  2013; Presented at the <a
  href="http://ptolemy.org/conferences/13" >10th
  Biennial Ptolemy Miniconference</a>, Berkeley.
  .

• BibTeX

  @presentation{AkkayaDerlerLee13_AspectOrientedFaultModelingAnomalyDetectionInPtolemy,
      author = {Ilge Akkaya and Patricia Derler and Edward A. Lee},
      title = {Aspect-Oriented Fault Modeling and Anomaly
                Detection in Ptolemy II},
      day = {7},
      month = {November},
      year = {2013},
      note = {Presented at the <a
                href="http://ptolemy.org/conferences/13" >10th
                Biennial Ptolemy Miniconference</a>, Berkeley.
  },
      abstract = {Fault modeling has been a primary focus of the
                well-established fields of circuit design and
                fault-tolerant control system research. For the
                emerging research field of cyber-physical systems,
                this modeling aspect plays an even more
                fundamental role due to the increased complexity
                in the networked-real time system. As concurrency
                and distributed behavior become ubiquitous in
                system design, faults due to networks, latency and
                cyber-physical interactions gain importance and
                detecting faults in pre-deployment stages becomes
                essential. In the domain of modeling and
                simulation, behavioral modeling often focuses on
                implementing the correct behavior that takes no
                account of the faults that might occur in system
                execution. However, numerous fault scenarios exist
                in data content, communication, analog and digital
                component failures, all of which can drive system
                behavior beyond the model context. Cyber-Physical
                System models can be refined continuously to
                address more possible behavior under different
                fault conditions, however, at the expense of
                increasing model complexity beyond usefulness.
                Aspect-oriented modeling is a means to represent
                different concerns related to system design as
                separate entities that are evaluated in
                concurrency with a functional representation. The
                separation of concerns in this sense has several
                benefits in fault modeling such as enabling
                modular and flexible representations for fault
                aspects while maintaining model simplicity and the
                models that define of correct operation. In
                Ptolemy, an aspect-oriented modeling formalism is
                developed to inject families of deterministic,
                nondeterministic and stochastic faults and to
                apply anomaly detection techniques to monitor
                anomalies in system behavior. The Modal model
                domain has been extended to enable probability
                expressions at transitions, and in turn, to
                represent Markov chain and Hidden Markov Model
                behavior. On the anomaly detection side, a
                component library has been developed for Hidden
                Markov and Mixture Model parameter estimation and
                classification.},
      URL = {http://chess.eecs.berkeley.edu/pubs/1038.html}
  }
  

Posted by Barb Hoversten on 21 Nov 2013.
Groups: ptolemy
For additional information, see the Publications FAQ or contact webmaster at chess eecs berkeley edu.