31
2 proving Scenario Analysis for H y: 2012 - 2014

Improving Scenario Analysis for HRA

Embed Size (px)

Citation preview

Page 1: Improving Scenario Analysis for HRA

2

Improving Scenario Analysis for HRAStudy: 2012 - 2014

Page 2: Improving Scenario Analysis for HRA

3

Study overview

ObjectiveTo investigate how scenario analysis is performed by

HRA analysts in real world settings, and why it is sometimes considered difficult

GoalTo develop a practical handbook for analysts,

documenting a recommended approach, good practices and guidance on how to overcome typical challenges

Page 3: Improving Scenario Analysis for HRA

4

Motivation for the study

Background International HRA Empirical study & US Domestic

study both identified that variability in human error predictions between analyst teams were, in part, due to deficiencies in qualitative scenario analysis methods and differences in approach

Follow-up interviews indicated that lack of guidance may be a primary cause of the differences in how HRA teams performed qualitative scenario analysis

Page 4: Improving Scenario Analysis for HRA

5

Key study activities

Interviews 2 rounds of interviews

Workshops 2 workshops

Initial interviews, 2013 Participants from the international HRA

empirical study or the US domestic study 9 people interviewed About half had minimal experience of HRA,

half perform HRA regularly

Detailed interviews, 2014 Additional participants from HRA community

in US & Europe 8 people interviewed Almost all perform HRA regularly as part of

their job

Halden workshop, May 2013 Attended by HRA practitioners, researchers

and regulators Presented a generic approach to HRA Agreed on what constitutes a qualitative

“scenario analysis”

EPRI Charlotte workshop, Oct 2014 Attended by 15 HRA practitioners from the US Presented the first draft practical handbook Collected additional data regarding

challenges, good practices & guidance needs

Page 5: Improving Scenario Analysis for HRA

6

Main findings from the study

Page 6: Improving Scenario Analysis for HRA

7

Similar approaches to HRA Most people who perform

HRA regularly adopt a similar approach to qualitative analysis…

… despite the fact that there is very little guidance available on how to perform qualitative analysis

… regardless of which HRA quantification method they use

Scenario definition

Qualitative data collection

Qualitative data analysis

Human error quantification

Page 7: Improving Scenario Analysis for HRA

9

Focus of this study The Improving Scenario

Analysis for HRA study focused on the first two steps:• Scenario definition• Qualitative data collection

Qualitative data analysis is investigated in a separate study by the Halden Reactor Project from 2015 – 2017.

Scenario definition

Qualitative data collection

Qualitative data analysis

Page 8: Improving Scenario Analysis for HRA

10

Use of the simulator

There is a difference of opinion amongst practitioners about the value/reality of collecting data from a training simulator

Cannot get a real understanding of how the scenario might unfold without seeing it in the simulator

However, observation of a training exercise in the simulator does not allow the analyst to see how the operator might fail

Simulator schedules are often fully booked and so the analyst might not be able to observe relevant scenarios within the timescales of the analysis

Page 9: Improving Scenario Analysis for HRA

11

Lack of practical guidance

Existing HRA guidance does not tell the analyst how to collect the necessary information for scenario analysis

The guidance on how to do scenario analysis is improving – for example, NUREG-1921 (Fire PRA)

But this still does not the more practical aspects of how to collect qualitative data

Page 10: Improving Scenario Analysis for HRA

12

Importance of operator input

Analysts must go to the site whenever possible to talk to operations personnel

This is essential to ensure that the HRA reflects the “as operated” plant, rather than “as built”

It is also essential to understand the contextual factors that will influence how the operators will response in that scenario

HRA cannot be a desktop exercise

Page 11: Improving Scenario Analysis for HRA

13

What type of information should analysts collect for HRA?

What it is really like to perform tasks• How difficult or complex the tasks might be, why, and what

impact this has on task performance• What kind of limitations or constraints (physical,

organizational, cultural) might exist and how these influence task performance

How operators would know that something is wrong• Whether the scenario event would be obvious to them• Whether and how they would be able to diagnose the event• Whether and how they would know what to do

Page 12: Improving Scenario Analysis for HRA

14

Types of information…

Which plant systems and interfaces operators would use and how they would use them• Whether they are easy to use and interpret• Whether they are reliable and trustworthy

The challenges that they could experience• Inadequate procedures• Lack of training• Poor quality of HMIs• The likelihood of incorrect mindset or group think

Page 13: Improving Scenario Analysis for HRA

15

Types of information…

The presence and effect of other Performance Shaping Factors (PSFs)• Time pressure• Stress• Workload

Operators can also• Confirm assumptions made by the analyst• Provide information on areas of uncertainty

Page 14: Improving Scenario Analysis for HRA

16

Challenges with collecting information

Page 15: Improving Scenario Analysis for HRA

17

Key challenges

Availability of knowledgeable and experienced operators

Access to where the operators are located

Availability of simulator time to run analysis scenarios

Unsurprisingly, the most knowledgeable and experienced operators are often the busiest

Analysts may have to go to the plant in the evening or at weekends to secure interviews

For control-room based analyses, it can be difficult (or impossible) to gain access to the control room for interviews or observations

This is particularly the case for contractors

Equally, simulator training schedules are defined well in advance and are extremely busy

If the analyst is allowed to observe an exercise, it will probably not be the analysis scenario

Page 16: Improving Scenario Analysis for HRA

18

Key challenges

Risk of obtaining incorrect or contradictory information

Reluctance to talk about potential human errors

Perceived credibility of the scenario being analyzed

Highly unlikely that operators will deliberately give incorrect information, but the might not know that they are wrong

Operators may contradict one another

Human error can be a sensitive subject Operators may be reluctant to admit they could

do something wrong Mindset of “that would never happen here”

If no previous experience of the analysis scenario, it might be difficult for operators to perceive that it is credible – e.g. beyond design basis events, Fukushima-type events, etc.

Page 17: Improving Scenario Analysis for HRA

19

Challenges from the analyst

Risk of being biased about how operators work

The International HRA Empirical study identified that analysts have a tendency to consider that operators perform a single “initial diagnosis” at the start of the event only

Subsequent observations of operating crews in the simulator showed that operators continuously make decisions whilst working through procedures and executing tasks

Some of the analysts failed to identify important PSFs, resulting in underestimations of HEPs.

(It should be noted that the analysts did not have an opportunity to interview operators or simulator trainers in this study importance of getting operator input!)

Detect Diagnose Decide

Act

Event

Basic behavioral model

Page 18: Improving Scenario Analysis for HRA

20

Challenges from the analyst

Risk of over-reliance on a single information source

??

??

??

In the US HRA Empirical study, the analysts were allowed to interview simulator trainers as part of the HRA, but there were differences in how the analysts used the information from the trainers in their analyses

Two teams appeared optimistic in their timing analysis because the trainer underestimated the time needed to respond to an HFE, and they did not validate this estimation with anyone else

A third team was also told by the trainer that the operators would succeed within the time, but they then asked the trainer to talk through the action in detail, and identified some hidden error traps that would prevent the crew from succeeding

The subsequent empirical results showed that the crew did not manage to perform the action in time

Page 19: Improving Scenario Analysis for HRA

21

The handbook of good practices

Page 20: Improving Scenario Analysis for HRA

22

Aim of the handbook

To provide universal, method-independent guidance that can be tailored to fit the specific context of the scenario being analyzed

Difficult to write a handbook to fit every situation and HRA method

Handbook cannot (and does not intend to) replace the role of HRA training and mentoring

Can be seen as a knowledge transfer exercise – this is the first step towards documenting what experienced analysts know about HRA – tips & tricks, etc.

Page 21: Improving Scenario Analysis for HRA

23

Recommended approach

Page 22: Improving Scenario Analysis for HRA

24

Practical guidance

Page 23: Improving Scenario Analysis for HRA

25

Good practice - be prepared

Contact the site early in the HRA to secure time with knowledgeable and experienced operators• Be realistic about how much time you will need

with each person or group• Be flexible in case the person or group is not

available on the day

Do your homework before you go to the site to avoid wasting time with basic questions

Page 24: Improving Scenario Analysis for HRA

26

Good practice – interview technique

Ask probing questions to ensure enough detailed information is collected for analysis later in the HRA

Use examples to stimulate discussion of human error and to reinforce the credibility of the analysis scenario / HFEs

Use non-accusatory language when talking about human error E.g. Rather than asking what the operator could do wrong themselves,

ask what could go wrong that would prevent the operator from doing their job

Don’t use PRA-speak – put the “human” back into HRA!

Page 25: Improving Scenario Analysis for HRA

27

Good practice – talk to lots of people

Interview multiple operators and/or groups of operators (crews) to get consensus on the information received and to minimize bias

Don’t interview only operating personnel – supervisors, shift managers, simulator trainers, process experts, systems experts etc. also have valuable insights and information and often have an operations background

Page 26: Improving Scenario Analysis for HRA

28

Good practice – follow up

Be prepared to do a follow up visit to collect additional information if needed, to check details of the analysis, to confirm assumptions, etc.• If it is not possible to do a follow-up visit, then be

prepared to follow up by telephone or email

Send a follow up email after your visit thanking the participants for their time – this will benefit you in the long run!

Page 27: Improving Scenario Analysis for HRA

29

Improving HRA PracticesStudy: 2015 - 2017

Page 28: Improving Scenario Analysis for HRA

30

Motivation HRA analysts also report difficulties when performing

qualitative data analysis, e.g.

How far to decompose tasks in a task analysis How to identify credible human errors / Human Failure Events How to link task analysis and human error identification to fault

or event tree analysis (and ultimately to the Probabilistic Safety/Reliability Analysis)

The 2015-2017 Improving HRA Practices topic continues investigating the HRA process, with a focus on qualitative data analysis

Page 29: Improving Scenario Analysis for HRA

1. Problem definition

2. Qualitative data collection

3. Task analysis

4. Human error analysis

5. Representation

7. Quantification

8. Impact assessment

10. Quality assurance

11. Documentation

Is screening required?

HEP acceptable?

Factors influencing

performance and error causes or

mechanisms

9. Error reduction

6. ScreeningInsignificant errors not studied further

Yes

No

Yes

No

Improving performance

Error avoidance

Generic HRA process Improving Scenario Analysis project

(2012 – 2014)

Improving HRA practices project

(2015 – 2017)

HWR-1120

Page 30: Improving Scenario Analysis for HRA

Objectives To identify the different qualitative data analysis techniques

that are used in HRA, and the difficulties or challenges that analysts experience when using these

To identify qualitative data analysis techniques that are used in simulator experiments (e.g. at HAMMLAB) and investigate whether these could be used for HRA

To further develop the practical handbook to include a recommended approach and guidance for HRA analysts performing qualitative data analysis

Page 31: Improving Scenario Analysis for HRA

34

Thank you!Email: [email protected]