Fluid Petri Nets and hybrid model-checking: a comparative case study

Marco Gribaudo, András Horváth, Andrea Bobbio, Enrico Tronci, Ester Ciancamerla, M. MinichinoReliability Engineering & System Safety 2003 · DOI

TL;DR

This paper proposes a unified approach to modeling and analyzing hybrid systems by leveraging Fluid Petri Nets (FPNs). They demonstrate how a single FPN model can be translated into formats suitable for various analysis tools, including hybrid automata, discrete models, and probabilistic models, enabling a comparative study of their capabilities.

What Problem Does This Paper Address?

Hybrid systems, which combine discrete control logic with continuous physical dynamics, are increasingly prevalent in critical applications like industrial control. Analyzing these systems presents a challenge: traditional formal methods excel at discrete aspects (reachability, safety), while stochastic methods are better suited for continuous aspects (performance, dependability). This paper aims to bridge this gap by proposing a unified modeling framework using Fluid Petri Nets (FPNs) that can be analyzed by diverse formal and stochastic tools, facilitating a comparative assessment of their strengths and weaknesses for hybrid system analysis.

Methods

The core of the paper is the demonstration of how a Fluid Petri Net (FPN) model of a hybrid system can be translated into different formalisms for analysis:

  • Hybrid Automata: The FPN model is converted into a hybrid automaton, a standard representation for hybrid systems, which is then analyzed using the HyTech model checker.
  • Discrete Models: The FPN is discretized to create a finite-state model suitable for the NuSMV model checker, allowing for exhaustive state-space exploration.
  • Probabilistic Models: The FPN is adapted into a Discrete-Time Markov Chain (DTMC) for analysis with the PRISM probabilistic model checker, enabling the evaluation of probabilistic properties.

The authors use a real-world temperature control system of a co-generative plant as a case study to illustrate and compare these conversion and analysis techniques.

Datasets

Not discussed in the paper. The paper uses a simulated "real-world" hybrid system (the ICARO co-generative plant's temperature control system) as a case study.

Key Insights and Ideas

  • Unified Modeling with FPNs: Fluid Petri Nets offer a flexible and powerful formalism to represent hybrid systems, accommodating both discrete and continuous dynamics within a single model.
  • Cross-Tool Compatibility: A single FPN model can serve as input for diverse analysis tools (HyTech, NuSMV, PRISM), enabling a comprehensive evaluation of a hybrid system from different perspectives (functional, discrete, probabilistic).
  • Trade-offs in Analysis Tools: Different tools have varying strengths and weaknesses. HyTech is powerful for linear hybrid automata but struggles with complex continuous dynamics. NuSMV is effective for finite-state systems but requires discretization. PRISM excels at probabilistic analysis but has limited arithmetic support.
  • Model Complexity vs. Analysis Power: There's a fundamental trade-off between the complexity of the hybrid system model and the tractability of its analysis. More detailed models often require more sophisticated (and computationally expensive) analysis techniques.
  • Importance of Simulation and Model Checking: Simulation is useful for exploring typical behaviors and finding obvious errors, while model checking is crucial for exhaustive verification and finding low-probability errors, especially in safety-critical systems.

Key Results

The paper successfully demonstrates the conversion of an FPN model of a temperature control system into hybrid automata, discrete, and probabilistic models.

  • HyTech: Analysis of the hybrid automaton model using HyTech was possible but required significant manual effort to approximate continuous dynamics with linear constraints, leading to a large number of locations. The system was verified for certain user demand ranges.
  • NuSMV: The discretized model allowed for exhaustive exploration, confirming that under specific user demand ranges, system requirements were met. Counterexamples were generated when requirements were violated.
  • PRISM: The probabilistic model allowed for the analysis of system behavior under probabilistic user demand variations. Results showed that increased probability of user demand changes led to a higher probability of requirement violations, particularly for the secondary temperature.

Limitations and Future Work

  • HyTech's Limitations: HyTech's restriction to linear hybrid automata and its difficulty with complex continuous dynamics necessitated model approximation, leading to state-space explosion and manual effort.
  • PRISM's Arithmetic Support: The limited arithmetic capabilities of PRISM required adjustments to the model compared to NuSMV.
  • Scalability: The scalability of the technique is limited by the exponential complexity of the FPN-to-hybrid automaton translation and the analysis techniques themselves.
  • Future Work: The authors suggest exploring heterogeneous modeling that combines stochastic and deterministic timing, as well as discrete and continuous variables, within a unified framework. They also mention ongoing development of tools to support such integrated approaches.

Summary updated

BibTeX
@inproceedings{gribaudo2003fluid,
  author = {Marco Gribaudo and András Horváth and Andrea Bobbio and Enrico Tronci and Ester Ciancamerla and M. Minichino},
  title = {{Fluid Petri Nets and hybrid model-checking: a comparative case study}},
  booktitle = {Reliability Engineering \& System Safety},
  year = {2003},
  doi = {10.1016/s0951-8320(03)00089-9},
}
Figure