PSAM 16 - Paper Overview

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Lead Author: Sascha Schmidt

Model-Based Reliability Engineering of Automotive Drivetrain Architectures With Multi-Trajectory Simulation
Key words: Reliability Analysis Methods and Tools; Dynamic Reliability and Safety, Simulation, Example application (safety-critical automotive systems) The architecture of complex systems is often decided in an early stage of the design process. This leads to risky outcomes, as there is not yet much information available about the planned system. Non-functional properties such as reliability and safety are crucial for critical systems, yet the effect of low-level design decisions in modules on the overall system behavior is often unclear because of their emergent nature. Formal models and performability evaluation algorithms in the field of model-based systems design are useful tools to improve this situation [1-3]. In reliability and safety, the classic models such as fault trees and reliability block diagrams allow static systems, but fall short in the description of dynamic processes. Such a behavior is important to cover in the model for systems including dynamic fault tolerance, or if there is a significant influence of the underlying timed behavior. Dynamic models in the reliability and safety context have to support discrete events, states, probabilistic choices and stochastic activities. Markov chains and variants of stochastic Petri nets are used for reliability engineering of dynamic systems in international standards and in the literature [1, 3, 4, 5]. There are mainly two types of algorithms to compute performability measures of interest from such models: numerical analysis and simulation. Numerical analysis methods are (mainly) restricted to Markovian models and manageable reachability graph sizes. Simulation be applied to any model, but will lead to intractably long runs in reliability evaluations because of the computational effort to generate enough failure states to achieve statistical confidence in the estimated results. This problem is known as rare-event simulation, and there are several approaches described in the literature (mostly variants of importance sampling [6] and splitting [7, 8]). However, they only achieve the theoretically possible speedup if the models are simple and symmetric in the case of sampling, or if a heuristic guiding the simulation is known a-priori for splitting [9]. A more recently developed algorithm tries to overcome this by integrating elements of numerical analysis with simulation [10]. It aims at retaining the advantages of both approaches: For models with manageable reachability graph size, it works similarly to a numerical analysis; while in the case of larger state spaces, it will work more like a splitting simulation to speed up rare-event problems. There is, however, no switching: the method allows seamless adaptations “in between” the underlying algorithms of simulation (which follows exactly one system state trajectory) and numerical analysis (which covers all possible trajectories). The paper will show how the reliability and safety of dynamic systems can be efficiently evaluated by this method. It will use stochastic Petri nets for the modeling part and show selected use cases from the areas of safety-critical embedded control systems in the automotive field. A prototype implementation of the algorithm in our software tool TimeNET [11] is used to derive numerical values, showing typical design trade-offs as the result. References [1] K. Trivedi, A. Bobbio, Reliability and Availability Engineering: Modeling, Analysis, and Applications, Cambridge University Press 2017. [2] J. Faulin, A. A. Juan, S. Martorell, J.-E. Ramirez-Marquez, Eds., Simulation methods for reliability and availability of complex systems, Springer 2010. [3] A. Zimmermann, Stochastic Discrete Event Systems — Modeling, Evaluation, Applications, Springer, 2007. [4] Analysis techniques for dependability — Petri net techniques, IEC 62551:2012, IEC Norm DIN EN 00 338, Sep. 2013. [5] Application of Markov techniques, IEC 61165:2006 Ed. 2.0, IEC Norm DIN EN 00 338, May 2006. [6] P. W. Glynn, D. L. Iglehart, Importance sampling for stochastic simulations, Management Science, vol. 35, no. 11, (Nov.) 1989. [7] P. Glasserman, P. Heidelberger, P. Shahabuddin, T. Zajic, Multilevel splitting for estimating rare event probabilities, Operations Research, vol. 47, pp. 585–600, 1999. [8] M. Villen-Altamirano and J. Villen-Altamirano, On the efficiency of RESTART for multidimensional systems, ACM Transactions on Modeling and Computer Simulation, vol. 16, no. 3, pp. 251-279, Jul. 2006. [9] M. J. Garvels, J.-K. C. Van Ommeren, and D. P. Kroese, On the importance function in splitting simulation, European Transactions on Telecommunications, vol. 13, no. 4, pp. 363-371, 2002. [10] A. Zimmermann and T. Hotz, Integrating simulation and numerical analysis in the evaluation of generalized stochastic Petri nets, ACM Transactions on Modeling and Computer Simulation (TOMACS), vol. 29, no. 4, 2019. [11] A. Zimmermann, Modelling and Performance Evaluation with TimeNET 4.4, Proc. Quantitative Evaluation of Systems (QEST 2017) 14th Int. Conf., LNCS 10503, (Sep.) 2017, pp. 300-303.

Model-Based Reliability Engineering of Automotive Drivetrain Architectures With Multi-Trajectory Simulation

Key words: Reliability Analysis Methods and Tools; Dynamic Reliability and Safety, Simulation, Example application (safety-critical automotive systems) The architecture of complex systems is often decided in an early stage of the design process. This leads to risky outcomes, as there is not yet much information available about the planned system. Non-functional properties such as reliability and safety are crucial for critical systems, yet the effect of low-level design decisions in modules on the overall system behavior is often unclear because of their emergent nature. Formal models and performability evaluation algorithms in the field of model-based systems design are useful tools to improve this situation [1-3]. In reliability and safety, the classic models such as fault trees and reliability block diagrams allow static systems, but fall short in the description of dynamic processes. Such a behavior is important to cover in the model for systems including dynamic fault tolerance, or if there is a significant influence of the underlying timed behavior. Dynamic models in the reliability and safety context have to support discrete events, states, probabilistic choices and stochastic activities. Markov chains and variants of stochastic Petri nets are used for reliability engineering of dynamic systems in international standards and in the literature [1, 3, 4, 5]. There are mainly two types of algorithms to compute performability measures of interest from such models: numerical analysis and simulation. Numerical analysis methods are (mainly) restricted to Markovian models and manageable reachability graph sizes. Simulation be applied to any model, but will lead to intractably long runs in reliability evaluations because of the computational effort to generate enough failure states to achieve statistical confidence in the estimated results. This problem is known as rare-event simulation, and there are several approaches described in the literature (mostly variants of importance sampling [6] and splitting [7, 8]). However, they only achieve the theoretically possible speedup if the models are simple and symmetric in the case of sampling, or if a heuristic guiding the simulation is known a-priori for splitting [9]. A more recently developed algorithm tries to overcome this by integrating elements of numerical analysis with simulation [10]. It aims at retaining the advantages of both approaches: For models with manageable reachability graph size, it works similarly to a numerical analysis; while in the case of larger state spaces, it will work more like a splitting simulation to speed up rare-event problems. There is, however, no switching: the method allows seamless adaptations “in between” the underlying algorithms of simulation (which follows exactly one system state trajectory) and numerical analysis (which covers all possible trajectories). The paper will show how the reliability and safety of dynamic systems can be efficiently evaluated by this method. It will use stochastic Petri nets for the modeling part and show selected use cases from the areas of safety-critical embedded control systems in the automotive field. A prototype implementation of the algorithm in our software tool TimeNET [11] is used to derive numerical values, showing typical design trade-offs as the result. References [1] K. Trivedi, A. Bobbio, Reliability and Availability Engineering: Modeling, Analysis, and Applications, Cambridge University Press 2017. [2] J. Faulin, A. A. Juan, S. Martorell, J.-E. Ramirez-Marquez, Eds., Simulation methods for reliability and availability of complex systems, Springer 2010. [3] A. Zimmermann, Stochastic Discrete Event Systems — Modeling, Evaluation, Applications, Springer, 2007. [4] Analysis techniques for dependability — Petri net techniques, IEC 62551:2012, IEC Norm DIN EN 00 338, Sep. 2013. [5] Application of Markov techniques, IEC 61165:2006 Ed. 2.0, IEC Norm DIN EN 00 338, May 2006. [6] P. W. Glynn, D. L. Iglehart, Importance sampling for stochastic simulations, Management Science, vol. 35, no. 11, (Nov.) 1989. [7] P. Glasserman, P. Heidelberger, P. Shahabuddin, T. Zajic, Multilevel splitting for estimating rare event probabilities, Operations Research, vol. 47, pp. 585–600, 1999. [8] M. Villen-Altamirano and J. Villen-Altamirano, On the efficiency of RESTART for multidimensional systems, ACM Transactions on Modeling and Computer Simulation, vol. 16, no. 3, pp. 251-279, Jul. 2006. [9] M. J. Garvels, J.-K. C. Van Ommeren, and D. P. Kroese, On the importance function in splitting simulation, European Transactions on Telecommunications, vol. 13, no. 4, pp. 363-371, 2002. [10] A. Zimmermann and T. Hotz, Integrating simulation and numerical analysis in the evaluation of generalized stochastic Petri nets, ACM Transactions on Modeling and Computer Simulation (TOMACS), vol. 29, no. 4, 2019. [11] A. Zimmermann, Modelling and Performance Evaluation with TimeNET 4.4, Proc. Quantitative Evaluation of Systems (QEST 2017) 14th Int. Conf., LNCS 10503, (Sep.) 2017, pp. 300-303.

Paper AR216 Preview

Author and Presentation Info

Lead Author Name: Sascha Schmidt (sascha.schmidt@tu-ilmenau.de)

Bio: Sascha Schmidt studied computer science at the Technical University of Ilmenau. He received his Master of Science degree in 2021 and has since been working as a research associate at the Department of Systems- and Software Engineering at the Technical University of Ilmenau. In the German BMWK project AnRox on "Fail-safe and efficient electric drive system for robot cabs" he is working on a methodology for reliability evaluation using stochastic Petri net models.

Country: Germany
Company: Technische Universität Ilmenau
Job Title: Research Assistant

PSAM 16 Conference Paper Overview

Welcome to the PSAM 16 Conference paper and speaker overview page.

Paper AR216 Preview

Author and Presentation Info

Download paper AR216.

Download the presentation PowerPoint file.