Correlation Analysis And Improvement Of PSFs Of The Spar-H Method

The Standardized Plant Analysis of Risk-Human reliability analysis(SPAR-H)has become an internationally known and accepted human reliability analysis method for its simplicity,convenience and ease of use,and has also been widely used in the engineering of nuclear power plants(NPPs)in China.However,researchers and users have found two flaws in the PSFs system when using the SPAR-H method to analyze internal events at nuclear power plants.(1)Correlation among the contents of PSFs.Some PSF content overlaps;(2)Causal relationship: the causal relationship between PSFs was not considered when calculating the human error probabilities(HEPs).The defects in these two aspects increase the uncertainty of human reliability analysis and lead to the double counting of human error probability.In order to make the results of human reliability analysis more reliable,it is necessary to improve the PSFs system of SPAR-H method,that is,to solve the above two defects.This study is mainly divided into the following three steps to solve these problems.(1)To identify the overlapping PSFs.Each of the eight PSFs of SPAR-H method contains a number of sub-factors.These sub-factors serve as indicators to determine the PSFs and their levels involved in certain human error events.Overlap mainly lies among these sub-factors,which leads to different results gained by the experts when they assign PSFs and the responding levels.With the increase of human error events,such overlapping PSFs will show certain rules through the results of expert analysis.By studying the human error event/accident reports,the overlapping characteristics can be explored and the overlapping PSFs can be found.Based on this,89 human error related event reports were collected from 219 operating event reports of Chinese NPPs from 2007 to2017.Data mining techniques(Association Rule Analysis,Exploratory factor analysis,and Pearson Correlation analysis)were used to analyze statistical results in order to identify the overlapping PSFs.Results show that(1)there is a significant correlation among Complexity,Stress/stressor,Fitness for duty,and Available time;(2)there also exists correlation among Work process,Procedure,Ergonomics/HMI,and Experience/training.When Procedure,Ergonomics/HMI,or Experience/training have been assigned to an event,Work process will also be selected with a high probability.(2)To improve the problem PSFs.Firstly,On the basis of the overlapping PSFs identified in(1),literature review was used to further identify the problem PSFs,together with taking the digitized environment of the main control room of nuclear power plant into consideration.Then,the three aspects for improving the problem PSFs were proposed: a.Avoid the overlapping among PSFs;b.Increase the variability of PSFs;c.Consider the digitized context,wherein the literature analysis method and expert judgment method were applied.After that,the improved PSFs were used to analyze the human error events reports of digitized nuclear power plants.Kendall's W Test was used to test the results made by different experts on the same human error events,which was to verify whether the improved PSFs reduced the overlaps.Then,the results made by the experts were used to calculate HEPs of the corresponding events.Paired samples test was conducted to analyze the HEPs and historical data,which was to verify whether the improved PSFs are reasonable.The test results showed that(a)Kendall's W coefficients are all over 0.8 and significant,indicating that the results made by different experts were highly consistent,which shows that the overlaps of the improved PSFs were reduced.(b)Paired samples test shows that there is no significant difference between the HEPs calculated based on the modified PSFs and the historical data,indicating that the modified PSFs is reasonable.(3)To establish a quantitative causal model of the relationship among PSFs.By incorporating the model into the SPAR-H method,HEPs can be estimated more accurately in engineering applications.The completion of(1)and(2)laid a foundation for the study of the complex causal relationship among PSFs.In this study,the system dynamics(SD)method was used to establish the causal relationship model.Firstly,the causal relationship among the PSFs was described qualitatively using literature analysis and expert judgment method,and the relationship flow diagram was then established.Later,the functional relationship of each PSF was determined based on mutual information(MI)and analytic hierarchy process(AHP).The VENSIMPLE software was used to simulate the model.The SD model was validated from three aspects:(1)Three cases in the appendix of NUREG/CR-6883 were taken as simulation analysis objects,and the simulation results were compared with the results of the cases.(2)Verifying whether the simulation values of the range of HEPs lie in the interval [0,1];(3)Sensitivity analysis for the model.The results showed that:(1)Statistically,there was no significant difference of the HEPs between after and before the causal relationship among the PSFs considered,and the former was smaller than the latter;(2)The HEPs simulated by SD model is in the interval [0,1];(3)Sensitivity analysis results show that Procedures and Ergonomics/HMI are the two PSFs that significantly affect operators' performance.The above results indicate that the causality model of PSFs established in this study is reasonable.The completion of the above research is helpful to improve the PSFs system of the SPAR-H method,which contributes to the human reliability analysis.In terms of qualitative analysis,it can more clearly guide human factor experts to analyze the influencing factors in events and reduce the uncertainty of expert analysis.For quantification,after considering the causal relationship among PSFs,the calculation of HEPs may be closer to the real value,which reduces the conservativeness of the SPAR-H method.