Implementation of a structured mean time to first failure computation scheme for static fault trees based on the Tensor Train format. Application of the TT-based approach for Generalized Stochastic Petri-Nets is in progress.
Usage: The CLIMain class provides a command-line interface to the tool. Use the calc command for MTFF calculation. Use the -h option to see the help page with all the available options. Example for computing the MTFF (-m 1, because we need the first moment of the failure time) of the dft described by galileofile.dft, using the AMEn-ALS solver with enrichment 4 and dampening factor 1e-5:
stoatt-sft calc -f galileofile.dft -m 1 -s AMEn-ALS --enrichment 4 --damp 1e-5
The core module depends on the delta decision diagram library. It is not open-source yet, so it is included as a jar file dependency.
The gspn module depends on another library, turnout-petrinet, which is not open-source yet either. This library is not included yet, but as the modules are separate gradle projects, the SFT tool can be compiled without any problems.
Run the :stoatt-sft:distTar
or :stoatt-sft:distZip
gradle task to create a tar or ZIP distribution,
including an executable and all dependencies. The resulting artifact can be found in the
stoatt-sft/build/distributions
directory.
The implementation is based on the algorithm described in the paper D. Szekeres, K. Marussy, I. Majzik: Tensor-Based Reliability Analysis of Complex Static Fault Trees.
The paper is available at the EDCC 2021 conference website.
@inproceedings{szekeres2021tensor,
title={Tensor-Based Reliability Analysis of Complex Static Fault Trees},
author={Szekeres, D\'aniel and Marussy, Krist\'of and Majzik, Istv\'an},
booktitle={European Dependable Computing Conference},
pages={33--40},
year={2021},
organization={IEEE},
ISBN="978-1-6654-3671-7"
}
The project uses the Apache Public License v2.0.