Federated Learning for Multivariate Time Series Anomaly Detection in Industrial Automation

2026-05-26

AI Summary

This study addresses the absence of benchmark datasets tailored to discrete industrial automation cycles—characterized by multivariate time series with both scale and accurate labels—for federated learning–based anomaly detection. To bridge this gap, the authors introduce a novel dataset that captures the cyclic dynamics inherent in such industrial processes and conduct a systematic evaluation of multiple federated anomaly detection methods on this dataset alongside established public benchmarks. This work not only establishes the first high-quality benchmark for this specific domain but also critically examines the applicability and limitations of current federated approaches in cyclic industrial settings, thereby providing a solid foundation and clear directions for future research.

Abstract

Federated learning (FL) has broadened the horizon for multivariate time series anomaly detection (MTSAD). However, benchmarking such anomaly detection methods within FL paradigm poses data-centric challenges. The existing datasets do not counteract these challenges since they do not simultaneously provide sufficient scale, accurate labels, and freedom from common flaws. In addition, the role of cyclic process behavior, which is common in discrete industrial automation, remains underexplored for MTSAD for the current state of research. This paper aims to shed more light on the literature and address these gaps by introducing a dataset designed with cyclic dynamics arising from the repetitive nature of discrete automation processes and evaluates selected MTSAD methods on both the proposed dataset and a public benchmark dataset.

Problem

• Federated Learning
• Multivariate Time Series Anomaly Detection
• Industrial Automation
• Cyclic Process
• Benchmark Dataset

Innovation

• Federated Learning
• Multivariate Time Series Anomaly Detection
• Industrial Automation
• Cyclic Dynamics
• Benchmark Dataset