Leveraging Large Language Models for the Generation of Stochastic Models in System Availability and Reliability Analysis
Paulo R. M. Maciel1, Erick B. Nascimento1 and Jamilson R. Dantas1
1 Centro de Informática – Cin, Universidade Federal de Pernambuco, Recife-PE, , ,
DOI : https://doi.org/10.48545/advance2025-fullpapers-1_3
Abstract
This paper investigates the ability of large language models (LLMs) such as ChatGPT, Gemini, Claude, and Mistral to generate and expand Stochastic Petri Net (SPN) models specifically for system availability and reliability analysis. The study evaluates how LLMs can assist in automating the expansion of formal models in these domains, following a structured methodology: (a) modeling a basic availability and reliability system in Mercury-script; (b) providing detailed prompts to guide the extension of the model; (c) obtaining the expanded models from LLMs; (d) validating the generated models in Mercury for syntactic correctness, semantic accuracy, and dependability metrics; and (e) conducting a comparative analysis of the results across the different LLMs. The findings reveal the strengths and limitations of LLM-generated models regarding availability and reliability, providing insights into the potential of AI-assisted formal modeling for improving the efficiency and accuracy of system dependability evaluations.
Keywords
Large language models; Stochastic Petri Net; Mercury-script language; Availability and Reliability analysis