48 Chapter 2. Automatic Inference of Fault Tree Models via Multi-Objective Evolutionary Algorithms Initialisation Failure dataset (Section 2.4.1) + Initial parameters (Section 2.5.1) Yes No Convergence criteria Inferred FT Parent fault trees Step 2 Step 1 Extraction of MCS from the failure dataset (MD) (Optional) Step 1.2 (Section 2.5.2) (i) Disconnect (ii) Connect (iii) Swap (iv) Create (v) Delete (vi) Cross-over (vii) Change gate Step 3 Genetic operators No Yes Step 5 Multi-objective evolutionary algorithm Computation of metrics (Section 2.5.5) (Section 2.3) (Section 2.5.6) NSGA-II + C-D (Section 2.3.1 and 2.3.2) Step 4 Single AND-gate Single OR-gate ... ... (Section 2.5.3) (Section 2.5.4) Do you consider MCSs in the MOEA? Converged? Figure 2.4: General process of the FT-MOEA algorithm to infer FTs from a failure dataset. - Max. generations with unchanged best candidate (uc): if after uc number of generations the best individual (i.e., the FT with the smallest size, and smallest error(s) within the best Pareto set) remains unchanged, then we assume the process has converged and is therefore terminated. - Max. number of generations (ng): Terminates the optimisation process if the number of generations exceeds ng and none of the other convergence criteria is met. 2.6.2 Step 1.2 - Extraction of MCSs from the failure dataset (optional step) Minimal Cut Sets (MCSs) are minimal combinations of component failures leading to system failure, encoding the system’s failure modes. Including this information in optimisation can enhance algorithm e"ciency. However, using MCSs in the optimisation process is optional. MCSs should be considered only if the failure dataset is noise-free and the expected FT is not overly complex (see Section 2.7.4 for more on FTs complexity). Otherwise,
RkJQdWJsaXNoZXIy MjY0ODMw