Fault Trees from Data: Efficient Learning with an Evolutionary Algorithm

TL;DR

This paper introduces an evolutionary algorithm that automatically learns fault tree models from observational data, improving accuracy and efficiency in reliability modeling for complex cyber-physical systems.

Contribution

The paper presents a novel evolutionary algorithm for learning fault trees from data, outperforming existing methods and applicable to both synthetic and industrial datasets.

Findings

Outperforms existing fault tree learning methods

Provides near-optimal fault tree models

Effective on both synthetic and real industrial data

Abstract

Cyber-physical systems come with increasingly complex architectures and failure modes, which complicates the task of obtaining accurate system reliability models. At the same time, with the emergence of the (industrial) Internet-of-Things, systems are more and more often being monitored via advanced sensor systems. These sensors produce large amounts of data about the components' failure behaviour, and can, therefore, be fruitfully exploited to learn reliability models automatically. This paper presents an effective algorithm for learning a prominent class of reliability models, namely fault trees, from observational data. Our algorithm is evolutionary in nature; i.e., is an iterative, population-based, randomized search method among fault-tree structures that are increasingly more consistent with the observational data. We have evaluated our method on a large number of case studies, both on synthetic data, and industrial data. Our experiments show that our algorithm outperforms other methods and provides near-optimal results.