ML22070A159
ML22070A159 | |
Person / Time | |
---|---|
Issue date: | 07/31/2018 |
From: | Chang Y, Jing Xing NRC/RES/DRA/HFRB |
To: | |
Xing, Jing - 301 415 2410 | |
References | |
Download: ML22070A159 (21) | |
Text
Use of IDHEAS to Generalize Human Performance Data for Estimation of Human Error Probabilities Jing Xing, Y. James Chang US Nuclear Regulatory Commission Presentation to AHFE, July, 2018 1
Whats next in human reliability analysis
- DATA, DATA, DATA
- Existing human error data - from various fields, in different formats, varying context and levels of details
- Data generalization and use for human reliability analysis -
the Integrated Human Event Analysis System (IDHEAS) has an inherent structure for generalizing and integrating human error data 2
Human error data: The ideal world and reality
- of errors (failure mode)
HEP (failure mode under specific context) =
- of Occurrence (under the context)
- Ideal world:
- The same task for a failure mode is repeated thousands of times with the same people under the identical context;
- Do this for all possible contexts Failure modes # Occurrence Context Variety Well-defined Known, sufficient Context clearly Sufficient data for all failure modes number of task defined and failure modes and occurrences repeated contexts 2 3
Human error data: The ideal world and reality
- of errors (failure mode)
HEP (failure mode under specific context) =
- of Occurrence (under the context)
- Reality:
X Failure modes unknown X Number of occurrences not reported X Context undocumented and/or unrepeated X Lack of variety - limited failure mode / context tested X Not talking to each other Type of human error data Failure modes # Occurrence Context Variety Statistical X X X Human error analysis X X Operational database Unrepeated Limited 2
Experimental X
Examples of statistical data
- Statistical study in 2016 - Medical errors are the third leading cause of death in the U.S., after heart disease and cancers, causing at least 250,000 deaths every year (Ref. 1)
- France - Nuclear Power plant replacement of the Dungeness B Data Processing System - The installation team completed 22,000 plant connections to the new system with a less than 2% error rate. (Ref. 3)
- X Occurrence of the tasks not reported
- X Failure modes unspecified
- X Context undocumented and unrepeated 5
Examples of human error analysis / root causal analysis
- Percent of error types (failure modes) - Airplane maintenance errors (Ref. 6)
Installation error - 44%
Approved data not followed - 28%
Servicing error - 12%
Poor troubleshooting standards - 0.7%
Poor maintenance practices - 9%
Poor inspection standards - 5%
Misinterpretation of approved data - 2%
- Percent of Airplane maintenance error contributing factors (Ref. 7)
Information Equipment Configuration Job/Task Training Individual Environmental Organization Supervision Communication
- Failure modes / contributing factors classified and ranked
- X Occurrence of the tasks not reported
- X Relation between failure modes / contributing factors unspecified 6
Examples of observed human error rates in operations (human performance databases)
- Error rates for nuclear power plant maintenance tasks (Ref. 4):
- 1/7 for transporting fuel assemblies with the fuel handling machine
- 1/48 for removing a ground connection from a switchgear cabinet
- 1/888 for reassembly of component elements
- Reported error rates in medical pharmacies (Ref. 5):
- 5% for failure to select ambiguously labeled control/package
- 2% for failed task related to values/units/scales/indicators
- 0.6% for procedural omission
- Human error rates reported for the failure modes
- X Relation of failure mode / contributing factors (maybe) unspecified 7
Example: Human error rates in experimental studies The effect of incomplete information on decision-making in simulated pilot de-icing (Ref.8)
Task: Make decision on de-icing in flight simulation under icing weather Failure mode: Incorrectly select or use information for decision-making Context: Incomplete or unreliable information (30%), time pressure Results: Providing additional accurate information improves handling of icing encounters. Performance drops below the baseline when inaccurate information (high uncertainty) is provided in the decision-aid.
% error Accurate and Accurate and Inaccurate additional incomplete additional information information information
% Stall 18.1 30 89
% recovery 26.7 63.8 75
- Failure modes, error rates, and specific context reported
- Quantitative impact of specific context factors reported
- X Not generalized for more complex context with multiple factors 8
Whats next in human reliability analysis
- DATA, DATA, DATA
- Existing human error data - from various fields, in different formats, varying context and levels of details
- Data generalization and use for human reliability analysis -
the Integrated Human Event Analysis System (IDHEAS) has an inherent structure for generalizing and integrating human error data 9
Generalizing human error data to inform human error probability estimation HEP = f(states of performance influencing factors)
Data source 1 Data source 2 Tasks Context Tasks Context Failure PIFs Failure PIFs modes modes A generic, adaptable set of failure modes and PIFs 10
Demonstration of IDHEAS-G cognitive failure modes Failure of Failures of cognitive Behaviorally observable macrocognitive process failure modes function D1- Fail to establish D3-1 Primary acceptance-criteria information is not Failure of Detection available D2 - Fail to attend to D3-2 Key alarm or alert Failure of sources of information not attended to Understanding D3-3 Key information D Fail to perceive the information not perceived Failure of D3-4 Information Decisionmaking D4- Fail to verify and misperceived (e.g.,
modify detection failing to discriminate Failure of Action Execution signals, reading errors)
D5- Fail to retain or communicate D3-5 Parameters Failure of Teamwork Information incorrectly monitored 11 10
Demonstration of IDHEAS-G PIF structure Systems and Personnel / team / Task /
Context environment organization situation
- Unfamiliar scenario
- Environmental factors - Procedures
- Multitasking,
- System opacity - Training Interruption, and PIF - Information - Work process distraction
- Tools and parts - Organization - Cognitive
- HSI factors complexity
- Teamwork factors - Mental fatigue and stress
- Alarm not salient - Teamwork - Physical demands PIF
- Mode confusion infrastructure attributes
- Key Information - Distributed teams masking
- Communication
- Ambiguity of equipment Indicators
- Communication protocol 11
Generalizing human error data to IDHEAS-G cognitive failure modes (CFMs) and PIFs Information Task complexity PIF attribute Cognitive function Training HSI - Cognitive failure modes
- CFM1
- CFM 2
- CFM3 Procedures Multitasking 13
Evaluate data - PIF effects on human errors Error factor (EF) = Error rate at a poor state of the PIF / error rate at the nominal state PIF - Multitasking, Distraction and interruption Ref Context and task Error rates and impact factor (EF)
Ref .8 Experiment on dual task: Airplane Error rate in detecting icing cue alone vs. dual-task:
pilots detecting de-icing cue and 2.8% vs 21% missing cue EF= 7.2 responding to air traffic control 5% vs 20% missing changes EF= 4 information 1% vs 37% wrong diagnosis EF= 37 Ref. 9 Effect of interruption on target Accuracy for no interruption vs interruption detection Simple Spatial .726 (.21) .803 (.11)
Complex Spatial . 549 (.254) .441 (.273)
EF(weak interruption on detection) =1.1 for simple task EF(weak interruption on detection) =0.9 for complex task Ref. 10 Driving simulation with cell phone Missing dangerous targets:
conversation 2.5% without cell phone distraction 7% with cell phone distraction EF(persistent distraction) = 2.8 Ref. 11 Experiment on performing error rate =0.15 for no interruption, sequences of action steps 0.3 for 2.8s interruption, EF(interruption) = 2 0.45 for 4.4s interruption, , EF(longer interruption) = 3 Ref. 12 The effect of interruption on driving 4% for no interruption and and fighting in military weapon 8% with interruption EF(interruption) =2 14 system
Interpret and represent human error data PIF - Multitasking, Distraction and interruption Low impact Moderate High impact PIF state - Distraction impact - Intermingled
- Interruption - Secondary task multitasking Macrocognitive - Prolonged - Concurrently function interruption multitasking Detection EF( weak EF(persistent EF(dual-task) = [5, interruption) = distraction)=2.8 7.5]
[0.9, 1.1]
Understanding EF(intermingled)=37 Decisionmaking EF(interruption on simple decision) =
1.6 EF(interruption on complex decision)
= 1.7 Action Execution EF(2.8s) = 2 HEP (interruption)
EF(4.4s)=3 =2 EF(interruption)=2 Teamwork Undetermined EF(interruption)=2 15
Integrating the data to inform PIF quantification Example PIF - Multitasking, interruption, and distraction Detection Understanding (diagnosis)
Effect on HEP Effect on HEP Performance influencing factor 15
Evaluate data - PIF effects on human errors PIF - Teamwork factors ID Context and task Error rate Nuclear waste handling facility Check-off sheet, low dependence 1E-1 maintenance and operation Check-off sheet, medium dependence 3E-1 Supervisor verification error Check-off sheet, high dependence and stress 5E-1 EF(independent checking) = 5 for high dependence EF(independent checking) = 3 for medium dependence Failure to restore from testing Two persons, operator check 5E-3 Single person, operator check 1E-2 Single person, no check 3E-2 EF(no team verification) = 2 Failure to restore following Two persons, operator check 3E-3 maintenance Single person, operator check 5E-3 Single person, no check 5E-2 EF(no team verification) = 1.7 Experiment of vigilance dual task - Paired team, low target presentation speed 19%
detecting targets (responding to Single person, low target presentation speed 29%
visual alarms) and completing Paired team, high target presentation speed 28%
jigsaw puzzle. Single person, high target presentation speed 38%
EF(team detection) = 1.5, 1.3 for low and high complexity 17
Evaluate Data - PIF effects on human errors PIF - Information completeness and Correctness ID Context and task Error rate 04 Expert judgment of HEPs for NPP HEP (information obviously incorrect) = 3E-2 internal at-power event IHEP (information not obviously incorrect) =8E-2E-1 Information misleading HEP(No information misleading) = 1E-3 EF = 30 for Information obviously incorrect EF=80 for Information not obviously incorrect 40 Experimental study on supporting Error rate - Percentage of early buffet:
decision making and action Accurate information 7.87%
selection under Accurate information but not timely) 20.56%
time pressure and information 30% inaccurate information 73.63.%
uncertainty in pilots de-icing simulation Error rate - Percentage of stall:
Accurate information 18%
Accurate information not timey 30%
(30%) inaccurate information 89%
EF = 1.5, 2.5 for accurate but not-timely or not-organized information EF= 5, 9 for 30% inaccurate information 18
Conclusions
- Human error data are available, not perfect, but can be used to inform quantification of human error reliabilities
- We preliminarily generalized the data to inform the quantification of performance influencing factors on human error probabilities 19
References
- 1. Makary MA, Daniel M (2016). Medical error-the third leading cause of death in the US. BMJ. 353:i2139
- 2. The National Motor Vehicle Crash Causation Survey (2015). Critical Reasons for Crashes Investigated in the National Motor Vehicle Crash Causation Survey. DOT HS 812 115
- 3. N. N. Chokshi ; J. P. Bailey ; A. Johnson ; D. Quenot ; J. F. Le Gall (2010). Integration testing of safety-related systems: Lessons learnt from Dungeness B DPS replacement project, 5th IET International Conference on System Safety 2010
- 4. Civil Aviation Authority (2015). Aircraft Maintenance Incident Analysis, CAP 1367.
- 5. Hobbs A, Williamson A (2003). Associations between errors and contributing factors in aircraft maintenance. Hum Factors. 45(2):186-201.
- 6. Preischl W, Hellmich M (2013). Human error probabilities from operational experience of German nuclear power plants Reliability Engineering and System Safety, 109:150-159
- 7. Rovira E, McGarry K, Parasuraman R (2007). Effects of Imperfect Automation on Decision Making in a Simulated Command and Control Task. Hum Factors 49(1):76-87 20
Thank you!
Jing.xing@nrc.gov 21