Safety and Risk Evaluation - The SERENE Method

ESPRIT 4 Project 22187, Fourth Framework Programme

The SERENE (SafEty and Risk Evaluation using bayesian NEts) method is concerned with the functional safety of complex systems, particularly programmable electronic systems which fall within the scope of draft IEC 61508 and similar standards. Functional safety concerns the ability of a system to carry out the actions necessary to achieve or maintain a safe state (IEC 1995, adapted). In a complex system the demonstration of functional safety must take account of both random and systematic failures. Systematic failures include those that result from design errors. All complex systems are potentially subject to systematic failures, but this difficulty applies most of all to software, for which systematic failures are the only form of failure.

The SERENE Project

The SERENE project produced a decision support method, with a supporting tool, for quantifying safety of complex systems using Bayesian Networks.

The method helps safety analysts and engineers combine diverse forms of evidence together in order to predict safety or risk at any stage in the development life-cycle. A number of key benefits are provided to users of the SERENE method:

• rationally combine different sources and types of evidence in a single model
• identify weaknesses in the safety argument such that it can be improved
• identify weaknesses in products and processes to aid process improvement
• specify degrees of confidence associated with predictions
• provide a sound basis for rational discussion and negotiation about the systems development and deployment

Why Bayesian Networks is Used for Safety and Risk Evaluation

Bayesian Networks (BNs) form the core technology in the SERENE method. These allow the specification of risk models that represent the key factors and their inter-relationships (qualitative model) with probability values and distributions estimated via expert judgement or from data (quantitative model). With the SERENE tool you can build large-scale risk BN models quickly and efficiently. The SERENE tool helps you to:

• use idioms to build up a safety argument from repeating patterns
• draw cause-effect BN graphs using an intuitive visual editor
• specify probability tables manually or automatically using deterministic and statistical functions
• apply object oriented Bayesian nets (OOBNs) to organise and structure your analysis
• execute the network using fast evidence propagation algorithms
• perform what-if and sensitivity analysis on the results
• export the results in HTML format for inclusion in reports

The method has now been applied in a number of case studies, as part of the project itself, and in a number of commercial projects. Indeed the SERENE method has also been used for general risk and reliability assessment in addition to predicting systems safety.

Project Details:

Duration: 30 months
Start date: 1996-06-01
End date: 1998-11-30
Project cost: 1.8644 million Euro

Coordinator:
William MARSH ERA Technology UNITED KINGDOM

Partners:
ERA Technology UNITED KINGDOM
HUGIN EXPERT A/S DENMARK
CITY UNIVERSITY UNITED KINGDOM
Pgcc Technologie FRANCE
Electricite de France Direction des Etudes et Recherches FRANCE
TECHNISCHER ÜBERWACHUNGS-VEREIN NORD E.V. GERMANY