M$^2$AD: Multi-Sensor Multi-System Anomaly Detection through Global Scoring and Calibrated Thresholding

2025-04-21

AI Summary

Anomaly detection in industrial operating systems faces dual challenges: modeling cross-system dependencies and handling heterogeneity across multi-system, multi-sensor time-series data. This paper proposes the first unsupervised anomaly detection framework tailored for heterogeneous multi-system time series. It leverages deep time-series models to learn normal behavior and generate residuals; introduces a global scoring mechanism to capture inter-system dependencies while jointly calibrating thresholds via Gamma distribution fitting to unify sensor- and system-level heterogeneity; and employs Gaussian Mixture Models (GMM) for global residual aggregation. We theoretically prove that the dual mechanism—global scoring with statistical threshold calibration—jointly models dependency and heterogeneity, thereby transcending conventional single-system detection paradigms. Evaluated on multiple benchmarks, our method achieves an average 21% improvement in F1-score and has been successfully deployed on 130 warehouse robots at Amazon, significantly enhancing predictive maintenance accuracy.

Technology Category

• Intelligent Robots: Multi-Robot Systems
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Multiagent Systems: Multiagent Systems under Uncertainty

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

With the widespread availability of sensor data across industrial and operational systems, we frequently encounter heterogeneous time series from multiple systems. Anomaly detection is crucial for such systems to facilitate predictive maintenance. However, most existing anomaly detection methods are designed for either univariate or single-system multivariate data, making them insufficient for these complex scenarios. To address this, we introduce M$^2$AD, a framework for unsupervised anomaly detection in multivariate time series data from multiple systems. M$^2$AD employs deep models to capture expected behavior under normal conditions, using the residuals as indicators of potential anomalies. These residuals are then aggregated into a global anomaly score through a Gaussian Mixture Model and Gamma calibration. We theoretically demonstrate that this framework can effectively address heterogeneity and dependencies across sensors and systems. Empirically, M$^2$AD outperforms existing methods in extensive evaluations by 21% on average, and its effectiveness is demonstrated on a large-scale real-world case study on 130 assets in Amazon Fulfillment Centers. Our code and results are available at https://github.com/sarahmish/M2AD.

Problem

• Detects anomalies in heterogeneous multi-sensor time series data
• Addresses limitations of single-system anomaly detection methods
• Improves accuracy through global scoring and calibrated thresholding

Innovation

• Deep models capture normal behavior residuals
• Gaussian Mixture Model aggregates anomaly scores
• Gamma calibration handles sensor-system heterogeneity