An integrated framework for functional model-based safety assessment of process systems using Cloud-Bayesian network
提出一种结合功能建模与云贝叶斯网络的模型安全评估框架，通过多层流建模生成因果树并利用改进云模型处理专家判断，实现故障传播的定量风险推理，在海上油气平台Minox系统中验证，辅助决策者提升风险管理。
• A novel model-based safety assessment framework was proposed combining functional modeling and Cloud-Bayesian Network. • Functional model-based hazard analysis enables system-level fault diagnosis and prognosis. • Cause tree generated by Multilevel Flow Modelling applies to Bayesian Network structure design. • An improved Cloud Model is used to process experts’ judgments, facilitating uncertainty reasoning of failure propagation. • Cloud-Bayesian Network provides quantitative risk reasoning for failure scenarios and hazard analysis, supporting decision-making. The increasing complexity of modern industrial systems makes it challenging to fully consider the complex interactions between components while also representing dynamic system behaviors. This poses challenges for traditional risk assessment methods, which are often labor-intensive and time-consuming. Model-Based Safety Assessment (MBSA) emerges as a promising solution, offering concrete knowledge representation and consistent reasoning. MBSA not only processes large system data volumes and manages complexity through structured modeling but also automates the error-prone manual safety analysis process, enhancing efficiency and reliability. Multilevel Flow Modelling (MFM), a model-based method belonging to symbolic AI with cognitive capabilities, provides a functional modeling framework to represent complex industrial processes. It captures mass, energy, and control information, enabling effective reasoning about failure propagation and system behavior. To quantify the qualitative reasoning conducted by MFM-based hazard analysis, Bayesian Network (BN) is introduced to enable a more comprehensive utilization of MFM for safety assessments. In cases of insufficient failure scenario data, subjective information with uncertainties from experts remains valuable. The improved Cloud model is proposed to process expert judgments, addressing cognitive fuzziness and stochasticity issues. This paper proposes a model-based safety assessment framework that enhances the MFM-based hazard analysis by probabilistic risk reasoning using Cloud-Bayesian Network. This framework facilitates critical hazard and failure propagation analysis while also enabling scenario verification through downstream effect analysis caused by varying degrees of deviation. The framework is demonstrated in the Minox system of an offshore oil & gas platform, providing critical insights for designing effective countermeasures and aiding decision-makers in enhancing risk management.