Dynamic fuzzy reliability and safety assessment of passive safety systems in small modular reactors

Abstract

Small Modular Reactors (SMRs) have garnered significant interest in recent years, and they may play an essential role in the future of energy supply and power generation. One major advantage associated with SMRs is their improved safety, which is expected to be accomplished by integrating a wide range of novel design elements, such as Passive Safety Systems (PSS) and components. Passive safety systems rely on natural processes rather than active intervention, hence possessing some unique features compared to traditional safety systems. Safety analysis of SMR’s PSSs encounters challenges, including Limited Operating Experience (OPEX), limited data availability, and above all, dynamic behaviour of the system. Systematic analysis of the reliability and safety of PSSs is yet to be performed to understand and evaluate the reliability and safety of these systems. Probabilistic Safety Assessment (PSA) has played an important role in evaluating the reliability and safety of Nuclear Power Plant (NPPs) operations in the past decades. However, several limitations associated with the classical PSA, such as failing to capture the dynamic processes and timing sequences of events, make it difficult to be directly utilized to evaluate the safety of PSSs. These could potentially be addressed to a large extent by incorporating fuzzy logic and Artificial Neural Network (ANN) into the analysis, thus addressing some intrinsic properties associated with PSSs. This thesis proposes a dynamic fuzzy-PSA, fuzzy-FMEA (Failure Mode and Effects Analysis) and ANN-based fault tree analysis. The Passive Residual Heat Removal System (PRHRS) in the CAREM (Spanish: Central Argentina de Elementos Modulares)-25 small reactors under the Station Blackout (SBO) accident is analyzed to demonstrate the performance of the ANN-based FT analysis and dynamic fuzzy logic analysis. The effectiveness of the PRHRS in removing residual heat is evaluated using both methods. It is shown that engaging fuzzy logic into the PSA and Failure mode & Effects Analysis (FMEA) can handle uncertainty and imprecision in data and knowledge and improve classical risk assessment methods by implementing dynamics fuzzy operators and Pandora gates (Priority-AND and Priority-OR). On the other hand, ANN-based FT can analyze and handle a large number of data irrespective of the number of Basic Events (BEs), logical gates, and system complexity and identify patterns that would be difficult and time-consuming for classical FT analysis. Comparing the results of these two methods with existing research in the open literature shows that these models are valid and more efficient than conventional PSA methods. In the future work, ANN and fuzzy logic are expected to be linked together to enhance the capabilities of classical PSA when analyzing the reliability and safety of PSSs in SMRs.