Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semi-Supervised Health Index Monitoring with Feature Generation and Fusion

Gaëtan Frusque, Ismail Nejjar, Majid Nabavi, Olga Fink

TL;DR

This work employs the deep semisupervised anomaly detection (DeepSAD) embeddings to tackle the challenge of extracting features associated with the system's health state and proposes an alternating projection algorithm with isotonic constraints to transform the embedding into a normalized HI with an increasing trend.

Abstract

The Health Index (HI) is crucial for evaluating system health and is important for tasks like anomaly detection and Remaining Useful Life (RUL) prediction of safety-critical systems. Real-time, meticulous monitoring of system conditions is essential, especially in manufacturing high-quality and safety-critical components such as spray coatings. However, acquiring accurate health status information (HI labels) in real scenarios can be difficult or costly because it requires continuous, precise measurements that fully capture the system's health. As a result, using datasets from systems run-to-failure, which provide limited HI labels only at the healthy and end-of-life phases, becomes a practical approach. We employ Deep Semi-supervised Anomaly Detection (DeepSAD) embeddings to tackle the challenge of extracting features associated with the system's health state. Additionally, we introduce a diversity loss to further enrich the DeepSAD embeddings. We also propose applying an alternating projection algorithm with isotonic constraints to transform the embedding into a normalized HI with an increasing trend. Validation on the PHME2010 milling dataset, a recognized benchmark with ground truth HIs, confirms the efficacy of our proposed HI estimations. Our methodology is further applied to monitor the wear states of thermal spray coatings using high-frequency voltage. These contributions facilitate more accessible and reliable HI estimation, particularly in scenarios where obtaining ground truth HI labels is impossible.

Semi-Supervised Health Index Monitoring with Feature Generation and Fusion

TL;DR

This work employs the deep semisupervised anomaly detection (DeepSAD) embeddings to tackle the challenge of extracting features associated with the system's health state and proposes an alternating projection algorithm with isotonic constraints to transform the embedding into a normalized HI with an increasing trend.

Abstract

The Health Index (HI) is crucial for evaluating system health and is important for tasks like anomaly detection and Remaining Useful Life (RUL) prediction of safety-critical systems. Real-time, meticulous monitoring of system conditions is essential, especially in manufacturing high-quality and safety-critical components such as spray coatings. However, acquiring accurate health status information (HI labels) in real scenarios can be difficult or costly because it requires continuous, precise measurements that fully capture the system's health. As a result, using datasets from systems run-to-failure, which provide limited HI labels only at the healthy and end-of-life phases, becomes a practical approach. We employ Deep Semi-supervised Anomaly Detection (DeepSAD) embeddings to tackle the challenge of extracting features associated with the system's health state. Additionally, we introduce a diversity loss to further enrich the DeepSAD embeddings. We also propose applying an alternating projection algorithm with isotonic constraints to transform the embedding into a normalized HI with an increasing trend. Validation on the PHME2010 milling dataset, a recognized benchmark with ground truth HIs, confirms the efficacy of our proposed HI estimations. Our methodology is further applied to monitor the wear states of thermal spray coatings using high-frequency voltage. These contributions facilitate more accessible and reliable HI estimation, particularly in scenarios where obtaining ground truth HI labels is impossible.
Paper Structure (31 sections, 13 equations, 8 figures, 10 tables, 1 algorithm)

This paper contains 31 sections, 13 equations, 8 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. HI generation using Embedding Diversified DeepSAD
  4. DeepSAD
  5. Generating embedding with more diversity
  6. Feature fusion using an Alternating Projection Algorithm with isotonic contraints
  7. Proposed feature fusion methodology
  8. Algorithm
  9. Training and real-time HI construction
  10. Application on a benchmark dataset: The PHME 2010 milling wear datasets
  11. Dataset
  12. Input features and metrics
  13. Metrics using the ground truth HI
  14. Standard unsupervised HI quality criteria
  15. Performance of the APAIC merging algorithm
  16. ...and 16 more sections

Figures (8)

  • Figure 1: Diversity function apply to each eigenvalues of $\mathbf{C}$.
  • Figure 2: Schematic illustration of the alternating projection algorithm for a 2-D space. $E_I$ represents the space of perfect health indicators, while $E_Y$ denotes the space generated by the dataset.
  • Figure 3: (a) PHME 2010 data acquisition experimental platform - (b) Ground truth HI for the three complete lifecycles dataset
  • Figure 4: This visualization illustrates the evolution of the APAIC merging algorithm over 1000 iterations on a validation dataset. The algorithm's progression is depicted through a series of curves, transitioning from blue to green every 10 iterations. The initial state is represented in black, while the final result is highlighted in red.
  • Figure 5: Comparison between the estimated (solid line) and ground truth (dotted) for the following methods (a) APAIC Feature (b) SVDD tax2004support (c) Pseudo-label nieves2022semi (d) MOO moradi2023intelligent (e) DeepSAD ruff2019deep (f) RADS (g) 2DS (h) RA2DS
  • ...and 3 more figures