top of page

Model-Based Safety & Dependability

Research theme 04 · Led by Dr David Parker

In short

Derive the safety analysis from the system model instead of writing it by hand. Describe how each component can fail, and the fault trees and FMEA tables are generated — and regenerated whenever the design changes.

Why it matters

This is the group’s oldest line of work, and the one with the longest industrial record. It starts from a simple observation: a hand-built fault tree is separate from the design it describes, so as soon as the design moves, the analysis is wrong, and nobody necessarily notices.

 

Model-based safety analysis removes that gap. The failure behaviour of each component is captured once, alongside the component. The system-level analysis is then built automatically from the architecture. Change the architecture and the analysis follows, so safety assessment can happen during design, not after it.

 

Once the analysis is automatic, optimisation becomes possible: searching architectures and redundancy choices for the best dependability within a cost or weight budget, instead of comparing the two or three options an engineer had time to draw.

Theme lead: Dr David Parker

Senior Lecturer · Director of Education, School of Digital and Physical Sciences

 

Works on model-based safety analysis, HiP-HOPS and design optimisation for dependability.

 

Also contributing: Prof. Yiannis Papadopoulos · Martin Walker

 

View profile →

HiP-HOPS

Hierarchically Performed Hazard Origin and Propagation Studies: the group’s model-based analysis method, which builds fault trees and FMEAs from component failure descriptions and an architecture.

Andromeda

A model-connected framework for safety assessment and assurance: it links safety cases to system models, safety assessment activities and argument patterns, so that much of the work of building an assurance argument can be automated.

Fault tree analysis and FMEA

The underlying techniques: how they combine, how they scale to large architectures, and how they stay true to the system as the design changes.

Optimisation

Automated exploration of design alternatives — redundancy, component choice, layout — against dependability and cost together, instead of treating safety as a check at the end.

Synthesis with AI

Where this theme meets theme 02: using learned and generative methods to propose the component failure descriptions and model pieces that model-based analysis depends on, still the most labour-intensive part of the method.

Open questions

• The annotation bottleneck: the analysis is only as good as the component failure models, and producing those is still mostly manual.

• How to extend this kind of analysis to components that learn, where there’s no list of failure modes to describe.

• How to analyse systems of systems whose members change at runtime, so the architecture isn’t fixed at design time.

 

Fault trees already out of date? DEIS tools, HiP-HOPS among them, are used by vehicle, aerospace and marine manufacturers to generate safety analyses directly from design models.

 

Talk to this theme →

​

Related themes:

02 · AI for Safety

06 · Safety and Assurance Tools

05 · Robots & Autonomous Systems

​

← All research themes

bottom of page