Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Explainable Condition Monitoring via Probabilistic Anomaly Detection Applied to Helicopter Transmissions

Aurelio Raffa Ugolini, Jessica Leoni, Valentina Breschi, Damiano Paniccia, Francesco Aldo Tucci, Luigi Capone, Mara Tanelli

TL;DR

This work proposes to focus on learning the probability distribution of healthy observations only, and detect Anomalies at runtime, via the definition of probabilistic measures of deviation from nominality, which allow to detect and anticipate faults.

Abstract

We present a novel Explainable methodology for Condition Monitoring, relying on healthy data only. Since faults are rare events, we propose to focus on learning the probability distribution of healthy observations only, and detect Anomalies at runtime. This objective is achieved via the definition of probabilistic measures of deviation from nominality, which allow to detect and anticipate faults. The Bayesian perspective underpinning our approach allows us to perform Uncertainty Quantification to inform decisions. At the same time, we provide descriptive tools to enhance the interpretability of the results, supporting the deployment of the proposed strategy also in safety-critical applications. The methodology is validated experimentally on two use cases: a publicly available benchmark for Predictive Maintenance, and a real-world Helicopter Transmission dataset collected over multiple years. In both applications, the method achieves competitive detection performance with respect to state-of-the-art anomaly detection methods.

Explainable Condition Monitoring via Probabilistic Anomaly Detection Applied to Helicopter Transmissions

TL;DR

This work proposes to focus on learning the probability distribution of healthy observations only, and detect Anomalies at runtime, via the definition of probabilistic measures of deviation from nominality, which allow to detect and anticipate faults.

Abstract

We present a novel Explainable methodology for Condition Monitoring, relying on healthy data only. Since faults are rare events, we propose to focus on learning the probability distribution of healthy observations only, and detect Anomalies at runtime. This objective is achieved via the definition of probabilistic measures of deviation from nominality, which allow to detect and anticipate faults. The Bayesian perspective underpinning our approach allows us to perform Uncertainty Quantification to inform decisions. At the same time, we provide descriptive tools to enhance the interpretability of the results, supporting the deployment of the proposed strategy also in safety-critical applications. The methodology is validated experimentally on two use cases: a publicly available benchmark for Predictive Maintenance, and a real-world Helicopter Transmission dataset collected over multiple years. In both applications, the method achieves competitive detection performance with respect to state-of-the-art anomaly detection methods.
Paper Structure (27 sections, 6 theorems, 24 equations, 10 figures, 11 tables, 1 algorithm)

This paper contains 27 sections, 6 theorems, 24 equations, 10 figures, 11 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Our take on the problem
  3. Machine Fault Detection
  4. Vibration Condition Monitoring (VCM) of Helicopter Transmissions
  5. Outline
  6. Setting and Goal
  7. Learning Conditional Densities with CoCoAFusE
  8. Parameterization of the gates
  9. Affine Gaussian Base Experts
  10. Choice of Prior Distributions
  11. Building Measures of Anomaly
  12. Selecting, Inspecting and Assessing CoCoAFusE Models
  13. Hyperparameter Tuning and Selection
  14. Making CoCoAFusE Transparent to End Users
  15. Assessing Fault Detection Capabilities
  16. ...and 12 more sections

Key Result

Proposition 1

Let $Z$ be a continuous random variable with full support on $\mathbb{R}$ with cumulative distribution function $F_Z:\mathbb{R}\to(0, 1)$, i.e., $Z \sim F_Z$. Then, $U = F_Z(Z)$ has uniform distribution on $(0, 1)$.

Figures (10)

  • Figure 1: Illustrative example of Failure Detection-related concepts.
  • Figure 2: Predictive Inference for the Vibration Index of the Azure Train Dataset.
  • Figure 3: Predictive Inference Results for the Pressure (First Row), Rotate (Second Row), Vibration (Third Row), and Volt (Last Row) Indices of the Azure Test Dataset.
  • Figure 4: Anomaly Scores and Alerts on Pooled Indices of the Azure Dataset (Test Set).
  • Figure 5: Predictive Inference for the ENHKUR Health Index of First Helicopter on the Train (top), Validation (middle), and Test (bottom) Sets.
  • ...and 5 more figures

Theorems & Definitions (19)

  • Remark 1: Gate's priors and identifiability issues
  • Proposition 1: The Probability Transform
  • Corollary 1
  • Lemma 1
  • Remark 2
  • Lemma 2: 1979barrowSplineNotationApplied
  • Remark 3
  • Remark 4: Reducing the Computational Complexity of \ref{['eq:weighted-sum-cdf']}
  • Lemma 3
  • Corollary 2
  • ...and 9 more