178 Chapter 8. Discussion With SymLearn, we can handle larger problems and thus improve scalability. Data and implementations are available at zenodo.org/record/5571811. 3. The FT-MOEA-CM extension (Chapter 4) expands the multi-objective optimisation function by incorporating metrics computed from the confusion matrix. With FT-MOEA-CM, we improved robustness by consistently achieving global optima for larger problems. Data and implementations are available at https://gitlab.utwente.nl/fmt/fault-trees/ft-moea. Our findings suggest that using MOEAs for the inference of FT models generally has a positive impact in terms of robustness, scalability, and convergence speed. Why are algorithms for FT inference important? In the digital era, vast amounts of data are collected, o!ering opportunities to manage complex systems and processes better. Graph learning algorithms capture the underlying intricate relationships within data by modelling essential connections between vertices through edges while adhering to the properties of the graph model. These models are valuable for human interpretation, as they translate complex data relationships into more comprehensible forms. Moreover, their rich mathematical properties make them important tools for managing complex systems and processes. In terms of FTs, this is particularly relevant, as FT models are applicable in many industries. E!ective algorithms for learning FT models from data o!er benefits such as accounting for updated FT structures as changes occur and generating more e"cient structures. For example, our COVID-19 case study originally presented an FT structure of 33 elements, and after applying our algorithms, an equivalent structure with 13 elements was revealed. Risk asset managers may find that smaller and more compact equivalent FT could aid in redefining the meaning of intermediate events. What are the implications of using MOEAs in the inference of FTs? Inspired by the FT-EAalgorithm proposed in Linard, Bucur, and Stoelinga, 2019, we observed that while FT-EA introduces a novel approach and opens up an interesting research direction, there are areas in need of improvement. One of these is the consistency of the FT structure, which involves obtaining the same (or equivalent) graph model from identical failure datasets. This consistency was not always achieved with FT-EA, especially in larger problems. Why is this consistency important? A consistent FT model reveals patterns that aid in understanding the relationships between basic and intermediate events. In other words, it automatically uncovers the complex logical connections related to how failures propagate through the system. This is particularly valuable for
RkJQdWJsaXNoZXIy MjY0ODMw