Optimization-Assisted Event Tree Analysis

Standard ETA describes what the risk is given the current system configuration: it maps scenarios and calculates their probabilities. Optimization-assisted ETA adds a prescriptive layer that searches over possible design or resource allocation decisions to find the configuration that achieves the best risk or cost outcome. ETA answers 'how risky is this system?'; OA-ETA answers 'how should we invest to make it safer?'

The choice depends on problem structure. If decision variables are continuous and constraints are linear, mathematical programming (LP/MILP) is preferred for its guaranteed optimality. When the decision space is discrete and non-linear — as is common with barrier selection — genetic algorithms, particle swarm optimization, and simulated annealing are frequently used. Some studies use multi-objective evolutionary algorithms (e.g., NSGA-II) to generate Pareto-optimal risk-cost trade-off curves.

Uncertainty can be addressed through sensitivity analysis (varying probabilities over their plausible ranges and checking whether the optimal solution changes), interval arithmetic, or Monte Carlo sampling integrated into the optimization loop. If uncertainty is large, robust optimization formulations — which optimize performance under worst-case or expected-case uncertainty — are preferable to point-estimate optimization.

Only to a limited extent. The optimization layer requires numerical probability values to compute an objective function. If only qualitative likelihood categories (high/medium/low) are available, you can assign representative numerical values to each category, but results should be interpreted cautiously as ordinal rankings rather than precise risk estimates. For primarily qualitative data, a standard FMEA with priority scoring is usually more appropriate.

Fault tree analysis (FTA) and event tree analysis are complementary techniques often used together in probabilistic risk assessment. FTA determines the probability that a specific barrier fails (working top-down from failure to causes); ETA uses those barrier failure probabilities as inputs (working forward from an initiating event to consequences). Optimization can be applied to either individually or to the combined FTA-ETA model, with the optimizer adjusting component reliabilities across both trees simultaneously.