Correlation Relationship of Performance Shaping Factors (PSFs) for Human Reliability Analysis

At TMI-2, operators permitted thousands of gallons of water to escape from the reactor plant before realizing that the coolant pumps were behaving abnormally. The coolant pumps were then turned off, which in turn led to the destruction of the reactor itself as cooling was completely lost within the core. Human also plays a role in many aspects of complex systems e.g. in design and manufacture of hardware, interface between human and system and also in maintaining such systems as well as for coping with unusual events that place the NPP system at a risk. This is why human reliability analysis (HRA) - an aspect of risk assessments which systematically identifies and analyzes the causes and consequences of human decisions and actions - is important in nuclear power plant operations. It either upgrades or degrades human performance; therefore it has an impact on the possibility of error. These PSFs can be used in various HRA methods to estimate Human Error Probabilities (HEPs). There are many current HRA methods who propose sets of PSFs for normal operation mode of NPP. Some of these PSFs in the sets have some degree of dependency and overlap. Overlapping PSFs introduce error in HEP evaluations due to the fact that some elements are counted more than once in data; this skews the relationship amongst PSF and masks the way that the elements interact to affect performance. This study uses a causal model that represents dependencies and relationships amongst PSFs for HEP evaluation during normal NPP operational states. The model is built taking into consideration the dependencies among PSFs and thus eliminating overlap. The use of an interdependent model of PSFs is expected to produce more accurate HEPs compared to other current methods. PSF sets produced in this study can be further used as nodes (variables) and directed arcs (causal influence between nodes) in HEP evaluation methods such as Bayesian belief (BN) networks. This study was done to estimate the relationships between PSFs using correlation analysis and identify patterns in the PSFs using Principal Factor Analysis (PFA). The study is specifically based on Operational Performance Information Systems (OPIS) database. This study was conducted to determine causal relationships between PSFs and also find sets of PSFs (error forcing context) which contribute more to human error probabilities. These goals were achieved using correlation and principal factor analysis