Anomaly Detection for Bivariate Signals

TL;DR

This paper introduces an empirical method for anomaly detection in multivariate time series using conditional quantiles, demonstrated on artificial data and aircraft engine monitoring.

Contribution

It proposes a simple, empirical approach based on conditional quantiles for detecting anomalies in multivariate signals, applicable to real-world industrial scenarios.

Findings

Effective in artificial data tests

Proven useful in aircraft engine monitoring

Provides confidence bounds for anomaly detection

Abstract

The anomaly detection problem for univariate or multivariate time series is a critical question in many practical applications as industrial processes control, biological measures, engine monitoring, supervision of all kinds of behavior. In this paper we propose a simple and empirical approach to detect anomalies in the behavior of multivariate time series. The approach is based on the empirical estimation of the conditional quantiles of the data, which provides upper and lower bounds for the confidence tubes. The method is tested on artificial data and its effectiveness has been proven in a real framework such as that of the monitoring of aircraft engines.