Sequence-Oriented Diagnosis of Discrete-Event Systems
Authors: Gianfranco Lamperti, Stefano Trerotola, Marina Zanella, Xiangfu Zhao
JAIR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evidence suggests that, among these techniques, only lazy diagnosis may be viable in non-trivial application domains. No experimental results about sequence-oriented diagnosis of DESs can be found in previous works, let alone a comparative empirical analysis as illustrated in this paper. |
| Researcher Affiliation | Academia | Gianfranco Lamperti EMAIL Stefano Trerotola EMAIL Marina Zanella EMAIL Department of Information Engineering, University of Brescia Via Branze 38, 25123 Brescia, Italy Xiangfu Zhao EMAIL School of Computer and Control Engineering, Yantai University 30, Qingquan RD, Laishan District, Yantai 264005, China |
| Pseudocode | Yes | this paper is the first to provide the pseudocode of an algorithm that computes a fault space (cf. Definition 8) |
| Open Source Code | Yes | The software system is open source and available at github.com/Stefano-BS/Explanatory-Diagnosis (see the README.md file for usage details). |
| Open Datasets | No | The sample application described in this section is inspired by a small real-world example, illustrated by Boselli et al. (2014), which is relevant to the Italian Labour Market dataset. Our aim is to show that in this domain, as well as in similar ones, sequence-oriented diagnosis is more convenient than set-oriented diagnosis is. |
| Dataset Splits | No | In order to speed up the experimentation, the software system allows for the automatic generation of a DES based on ten parameters, namely: 1. The number m of components included in the DES. 2. The number of states included in the behavioral model of each component: this number is used as the mean value of a normal distribution from which the actual number is extracted based on the Box-Muller transform (Box & Muller, 1958). |
| Hardware Specification | Yes | Experiments were run on a computer with CPU Intel Xeon Gold 6140 (1-7 cores available) and 128 GB of working memory. |
| Software Dependencies | No | The software engines were implemented in the C programming language, mainly because of its flexibility and the ability to access the memory directly, even at byte level, by allocating, reallocating, and deallocating space explicitly and dynamically with the support of pointers, which is impossible in other higher-level languages, such as Java. This modus operandi allows the programmer to have full control on memory, which is essential for a software system that is supposed to construct and manipulate highly dynamical data structures under stringent time constraints. The relevant data structures generated by the software are shown to the user by exploiting the Graphviz package along with the dot engine, which allow for a graphic representation of a DES, a (possibly O-constrained) space, a (possibly partial) explainer, and a trace. |
| Experiment Setup | Yes | In order to speed up the experimentation, the software system allows for the automatic generation of a DES based on ten parameters, namely: 1. The number m of components included in the DES. 2. The number of states included in the behavioral model of each component: this number is used as the mean value of a normal distribution from which the actual number is extracted based on the Box-Muller transform (Box & Muller, 1958). 3. The internal connection degree of components, expressed as a percentage of n n 1, where n is the actual number of states of the component, indicating implicitly the number of transitions included in the behavioral model. 4. The external connection degree of components, expressed as a percentage of m m 1, indicating implicitly the number of links between components. 5. The observability degree, indicating the percentage of observable transitions. 6. The abnormality degree, indicating the percentage of faulty transitions. 7. The number of (distinct) observable labels. 8. The number of (distinct) fault labels. 9. The number of (distinct) events. 10. The probability of generation or consumption of events. |