Human Reliability Analysis Of Heavy Water Reactor Nuclear Power Plant

Nuclear power plant (NPP) is a social-technic complex, including technical equipments, human and organization&environment as well as their subsystems. Safety is critical to its existence and development. NPP accident would bring about great casualties and property losses as well as huge adverse social effects.NPP safety analysis is crucial to operation safety. Therefore, probabilistic safety assessment (PSA) was set forth and persistently improved. It adopts various kinds of safety assessment techniques and identifies possible consequences to calculate the probabilities of various casual factors to realize safety. In PSA, human reliability analysis (HRA) is a crucial factor contributing to the risk involved in the development of events and interaction between human and system. It affects the evolvement of event sequence and could have great impacts upon NPP safety.Heavy water reactor NPP system is complicated. It has more specified safety equipments and adopts hybrid (semi-digital) human-machine interface (HMI). Conventional HRA techniques become inappropriate to assess human reliability on the new HMI. To develop a new HRA method to assess Chinese heavy water reactor NPP becomes a main part of its PSA program. The dissertation presents a new HRA PABI+THERP model based on extensive research work and field studies. The main work includes:1) It studies NPP HMI development and its impact upon HRA. It discusses the main features of heavy water reactor NPP HIS. This part provides a foundation for the HRA model development.2) It makes a research work on how to incorporate human behaviors into a complicated NPP system and what behaviors worth consideration. The paper studies the major influencing factors, on the basis of which, HRA model is established. An HRA model is composed of two parts, diagnosis and action, which are influenced by performance shaping factors (PSF). This part of research work is the most fundamental theoretical work to the establishment of a heavy water reactor NPP HRA model.3) The paper studies the main features of post-accident operator behavior by experiments on a full-size simulator. In the experiment, operator’s diagnosis strategy was first studied and the diagnosis tasks were coded on the basis of cognitive complexity. On the full-size simulator in the reference heavy water reactor NPP, three crews simulated HTS (heat transport system) accident and its diagnosis tasks were observed and kinescoped. Similar tasks were grouped and their respective commission times were obtained by software.4) The paper sets forth a post-accident PABI (process analysis based on interface) diagnosis model. PABI model, on one hand, accounts time as an influence factor on diagnosis and on other hand analyzes human’s cognitive behavior process in a semi-digital main control room. Based on a careful study on the main attributes of diagnosis tasks, PABI diagnosis model employs Hidden Markov conditional transferring mechanism to dissect the diagnosis tasks into a consecutive process but at some points they are discrete while the corresponding interfaces and human factors are taken into account. The research takes HTA leakage as a case and calculates the error rates of the diagnosis tasks and the error rates in case of recovery. Errors of seven groups of diagnosis tasks considering recovery are calculated and event tree is used to make a combination to work out the human error rate of HTS leakage event.5) The paper conducts a research work on the integration of HRA model. It studies the possible action error modes in a heavy water reactor semi-digital MCR and has made an improvement on the error classification definition. It studies the time-interface issue involved in the combination of PABI+THERP and sets forth the main influencing factors and assumptions to facilitate and constitute the field analysis. The paper has presented a detailed event analysis and has conducted an analysis on all initiating events in PSA in the heavy-water reactor NPP. Finally three typical human factor events were selected to make a comparison with those of design PSA.