using bayesian networks to model accident causation in the uk railway industry william marsh risk...

15
Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of Computer Science Queen Mary, University of London George Bearfield Transport Safety and Reliability Atkins Rail, London

Upload: erik-mitchell

Post on 17-Dec-2015

218 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Using Bayesian Networks to Model Accident Causation in the UK Railway Industry

William Marsh Risk Assessment and Decision Analysis GroupDepartment of Computer ScienceQueen Mary, University of London

George Bearfield Transport Safety and ReliabilityAtkins Rail, London

Page 2: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Outline

Signals Passed at Danger (SPADs) Organisational Accidents Bayesian Networks Building a BN for SPADs Conclusions

Page 3: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

0

100

200

300

400

500

600

700

1997 1998 1999 2000 2001 2002

Num

ber

of S

PA

Ds

Signals Passed At Danger

Southall Ladbroke Grove

Page 4: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Signals Passed At Danger

‘Train has passed a stop signal without authority’

Incident on 27/3/03 at Southampton 360 yard overrun affected by low sunlight driver read adjacent signal signal is approached on a curve

wrong signal into the driver’s direct line of sight for a short time

Page 5: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Waterloo

Southampton

From: Railway SafetyAssessment of Railtrack’s Response to Improvement Notice I/RIS/991007/2 Covering the ‘Top 22’ Signals Passed Most Often at DangerHSE, 2002

Page 6: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Organisational Accidents

Operator errors have ‘organisational’ causes gradual relaxation of alertness pressure to increase efficiency

Currents acting within the Safety

Space

Increasing Resistance

Increasing Vulnerability

Page 7: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Organisational Causes of SPADs

Infrastructure: multi-SPAD signals Driver training and timetable pressure

‘Within the workforce there is a perception that emphasis on performance has affected attitudes to safety.’

Ladbroke Grove report

‘the industry is generally poor at identifying organisational issues that may underpin SPAD incidents …’

Page 8: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Bayesian NetworkMisinterpretation

Brakes not appliedSignal not located

Sighting obstruct.

Distraction

Late sighting

Read acrossSPAD

Read acrossat proceed

Phantom proceed

Late brakeapplication

Variable

Cause Table of ConditionalProbabilities

Page 9: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Organisational Model

Responsibilities of actors

Interactions between actors

DriverManagement

DriverTraining

Driver

Signal Route

Actors in the organisation (idea from Rasmussen’s AcciMap)

Page 10: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

BN Variables from Attributes

Actors and interactions can have attributes

DriverManagement

DriverTraining

Driver

Signal Route

qualitypressure

experiencealertness

visibilitycurve

traffic

routeknowledge

previoussignal

assessment

Page 11: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

SPAD Scenarios

Each SPAD scenario modelled as a BN events influences: attributes of driver, infrastructure,

… Scenario model merged

Read Across RA at Proceed No Brakes SPAD

Late Brakes

Infrastructure

Route KnowldgePressure

Alertness

Traffic

Page 12: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

SPAD Scenario

Event

Influence

Page 13: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Expert Judgement

Strength of probabilistic influences judged by experts Modify network structure Build probability tables

Aggregated data SPAD frequencies Used to validate judgements

Status Not yet completed!

Page 14: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Using the Causal Model

Assess frequency / risk Where are SPADs likely?

Monitor organisational changes Use audit results

Select interventions How can the frequency of SPADs be

reduced?

Page 15: Using Bayesian Networks to Model Accident Causation in the UK Railway Industry William Marsh Risk Assessment and Decision Analysis Group Department of

Summary

Integrated causal model of SPADs Organisational influences Event sequence

Bayesian networks Generalise other probabilistic modelling

Future challenges Use