9 November 2024 in Applied Energy
16 March 2026 in IEEE Transactions on Automation Science and Engineering
Anomaly detection is critical for ensuring the reliability of industrial cyber-physical systems. Identifying anomalies based on data distribution is regarded as a promising approach. However, inherent noise in data collection and complex dependencies within the underlying structure can lead to class ambiguity. This ambiguity obscures the boundary between normal data and anomalies, thereby degrading the accuracy of distribution modeling. To address this issue, we shift the distribution modeling from the data space to a latent space to mitigate ambiguity and then propose a label free anomaly detection network, named ALDM. In ALDM, a contrastive-based methods is designed to facilitate the construction of a latent space, where the margin between normal data and anomalies has been expanded. Anomalies are then discerned through embeddings using a flow-based process. Recognizing the importance of distance metrics in contrastive-based methods, we propose an adaptive approach to obtain the optimal distance metric during network training instead of presetting a fixed formula. Furthermore, to accommodate anomalies of varying durations, we propose another event-wise performance index for evaluation. Extensive evaluations on three widely used benchmarks and a newly constructed dataset demonstrate that ALDM achieves state-of-the-art detection performance across both conventional metrics and our proposed index.
1 July 2026 in Journal of Computing in Civil Engineering
19 February 2025 in IEEE Transactions on Automatic Control
Recognizing complex events revealed by raw data is an increasingly crucial task that serves as one of the foundations for system monitoring and decision-making. Our goal is to accurately recognize the occurred complex events, that is, uniquely determine the occurred complex event sequence from the raw data. We abstract the outputs of data sources as a set of atomic events and then use an automaton to describe all atomic event sequences that can be generated by the given system. We represent a complex event as a set of atomic event sequences. For a given atomic event sequence and a complex event to be recognized, we introduce the notion of “partition” to stand for a possible single complex event sequence. By constructing an augmented automaton that includes all possible partitions, we derive a necessary and sufficient condition for the complex event recognition problem to be solvable. We then find an algorithm to check the condition. When the complex event recognition problem is solvable, any occurred complex event can be determined accurately and promptly online with existing methods like Aho–Corasick algorithm.