BowtieRisk.jl

BowtieRisk.jl provides a Julia framework for bowtie risk modeling with an event-chain view, escalation factors, and basic dependency handling. It is designed to support structured hazard analysis and the assignment of probabilities similar to tools like RiskyProject.

This is a new project scaffold with a small, explicit core model. It focuses on clear data structures, transparent assumptions, and simple calculations that can be extended for domain-specific needs.

Core Concepts

• Hazard: the source of potential harm.
• Threats: initiating causes that may trigger a top event.
• Top Event: the moment control is lost (center of the bowtie).
• Consequences: outcomes following the top event.
• Barriers: preventive (left side) or mitigative (right side) controls.
• Escalation factors: conditions that reduce barrier effectiveness.
• Dependencies: shared-cause failures across barrier groups.
• Simulation: Monte Carlo evaluation with barrier distributions.
• Reporting: Markdown and CSV outputs for sensitivity data.
• Templates: built-in starter models for common scenarios.
• Schema: JSON schema for UI integrations.
• Event Chain: ordered events with probabilities and barriers.

Probability Model (Baseline)

This package assumes independent threats and independent barriers by default. You can switch to a dependency-aware model for shared-cause failures. Under the independent assumptions:

• Threat residual = p(threat) * Π(1 - barrier_effectiveness)
• Top event probability = 1 - Π(1 - threat_residual)
• Consequence probability = p(top_event) * Π(1 - barrier_effectiveness)
• Risk score = probability * severity

These formulas are intentionally simple and transparent so they can be replaced with richer methods later.

Quick Start

using BowtieRisk

hazard = Hazard(:LossOfContainment, "Loss of containment from vessel")

threats = [
    Threat(:Overpressure, 0.02, "Pressure exceeds design"),
    Threat(:Corrosion, 0.01, "Wall thinning"),
]

preventive = [
    Barrier(:ReliefValve, 0.7, :preventive, "Relieves excess pressure", 0.1, :none),
    Barrier(:Inspection, 0.5, :preventive, "Detects corrosion", 0.0, :none),
]

consequences = [
    Consequence(:Release, 0.6, "Release to atmosphere"),
    Consequence(:Injury, 0.8, "Personnel injury"),
]

mitigative = [
    Barrier(:GasDetection, 0.6, :mitigative, "Detects release", 0.0, :shared_power),
    Barrier(:Evacuation, 0.5, :mitigative, "Evacuate area", 0.0, :shared_power),
]

model = BowtieModel(
    hazard,
    TopEvent(:ContainmentLost, "Containment is lost"),
    [ThreatPath(threats[1], [preventive[1]], EscalationFactor[]),
     ThreatPath(threats[2], [preventive[2]], EscalationFactor[])],
    [ConsequencePath(consequences[1], [mitigative[1]], EscalationFactor[]),
     ConsequencePath(consequences[2], [mitigative[1], mitigative[2]], EscalationFactor[])],
    ProbabilityModel(:independent),
)

summary = evaluate(model)
println(summary.top_event_probability)

Diagramming Support

BowtieRisk.jl includes helpers that export Mermaid or GraphViz diagram specs. These are design aids for arranging bowtie diagrams.

spec = to_mermaid(model)
println(spec)

dot = to_graphviz(model)
println(dot)

write_model_json("bowtie.json", model)

dist = Dict(
    :ReliefValve => BarrierDistribution(:beta, (2.0, 5.0, 0.0)),
    :GasDetection => BarrierDistribution(:triangular, (0.2, 0.5, 0.9)),
)

sim = simulate(model; samples=500, barrier_dists=dist)
tornado = sensitivity_tornado(model; delta=0.1)

write_report_markdown("report.md", model; tornado_data=tornado)
write_tornado_csv("tornado.csv", tornado)

model = template_model(:process_safety)
write_schema_json("bowtie.schema.json")

Development

julia --project=. -e 'using Pkg; Pkg.instantiate()'
julia --project=. -e 'using Pkg; Pkg.test()'

API Snapshot

Hazard, Threat, TopEvent, Consequence, Barrier, EscalationFactor
ProbabilityModel, ThreatPath, ConsequencePath, BowtieModel
Event, EventChain, chain_probability
BarrierDistribution, SimulationResult
evaluate, simulate, sensitivity_tornado
to_mermaid, to_graphviz
write_model_json, read_model_json
report_markdown, write_report_markdown, write_tornado_csv