Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quantifying the value of positive transfer: An experimental case study

Aidan J. Hughes, Giulia Delo, Jack Poole, Nikolaos Dervilis, Keith Worden

TL;DR

The paper tackles data scarcity in structural health monitoring by quantifying the value of transferring information across a population of similar structures. It extends the EVIT framework to an experimental GARTEUR population, using a Dirichlet-regression-based, likelihood-guided mapping from structural similarity to predicted post-transfer performance, and demonstrates that EVIT monotonically increases with similarity, yielding positive transfer across all source domains. The approach combines normal-condition alignment, similarity-weighted distance metrics, and probabilistic, vector-valued regression to forecast target-domain outcomes and inform transfer strategies. Practically, the work provides a decision-theoretic basis for selecting source domains to optimize O&M decisions in PBSHM and highlights directions for enhancing transfer-learning efficacy in data-scarce settings.

Abstract

In traditional approaches to structural health monitoring, challenges often arise associated with the availability of labelled data. Population-based structural health monitoring seeks to overcomes these challenges by leveraging data/information from similar structures via technologies such as transfer learning. The current paper demonstrate a methodology for quantifying the value of information transfer in the context of operation and maintenance decision-making. This demonstration, based on a population of laboratory-scale aircraft models, highlights the steps required to evaluate the expected value of information transfer including similarity assessment and prediction of transfer efficacy. Once evaluated for a given population, the value of information transfer can be used to optimise transfer-learning strategies for newly-acquired target domains.

Quantifying the value of positive transfer: An experimental case study

TL;DR

The paper tackles data scarcity in structural health monitoring by quantifying the value of transferring information across a population of similar structures. It extends the EVIT framework to an experimental GARTEUR population, using a Dirichlet-regression-based, likelihood-guided mapping from structural similarity to predicted post-transfer performance, and demonstrates that EVIT monotonically increases with similarity, yielding positive transfer across all source domains. The approach combines normal-condition alignment, similarity-weighted distance metrics, and probabilistic, vector-valued regression to forecast target-domain outcomes and inform transfer strategies. Practically, the work provides a decision-theoretic basis for selecting source domains to optimize O&M decisions in PBSHM and highlights directions for enhancing transfer-learning efficacy in data-scarce settings.

Abstract

In traditional approaches to structural health monitoring, challenges often arise associated with the availability of labelled data. Population-based structural health monitoring seeks to overcomes these challenges by leveraging data/information from similar structures via technologies such as transfer learning. The current paper demonstrate a methodology for quantifying the value of information transfer in the context of operation and maintenance decision-making. This demonstration, based on a population of laboratory-scale aircraft models, highlights the steps required to evaluate the expected value of information transfer including similarity assessment and prediction of transfer efficacy. Once evaluated for a given population, the value of information transfer can be used to optimise transfer-learning strategies for newly-acquired target domains.
Paper Structure (11 sections, 7 equations, 2 figures, 3 tables)

This paper contains 11 sections, 7 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background Theory
  3. Transfer Learning for PBSHM
  4. Value of Information Transfer
  5. Case Study: GARTEUR Population
  6. Classification Task
  7. Transfer Task
  8. Similarity Assessment
  9. Predicting Transfer Efficacy
  10. Results
  11. Conclusions

Figures (2)

  • Figure 1: Learned probabilistic functions $p(q_k|\varsigma)$ for (a) the true prediction rate, (b) the false-positive prediction rate, (c) the false-negative prediction rate, and (d) the false-damage prediction rate. Blue shaded regions indicate the 90% confidence intervals (CI) computed using samples taken from the Dirichlet distributions.
  • Figure 2: The expected value of information transfer as a function of structural similarity.