An on-line continuous updating Gaussian mixture model for damage monitoring under time-varying structural boundary condition

Abstract : Damage monitoring under time-varying structural boundary condition is one of the most difficult tasks in piezoelectric transducers (PZTs) and Lamb wave based SHM methods for engineering applications. Because the structural boundary changes such as variations in the tightness of bolts between structures can lead to false monitoring result even the structure is in health state. This paper proposes a Lamb wave based on-line continuous updating Gaussian Mixture Model (GMM) to study the problem. Based on the baseline GMM constructed by features of Lamb wave signals in structural health state, an on-line continuous updating GMM is studied to learn the dynamic changes of Lamb wave monitoring signals without any prior knowledge of damage patterns. The KullbackÐLeibler (KL) divergence is used as a degradation index to estimate the structural damage by measuring the difference between the baseline GMM and the on-line GMM. The proposed method is validated on an aircraft steel beam. The validation results show that the method is effective for bolt hole crack growth monitoring under the time-varying changes in the tightness degree of the bolts.
Liste complète des métadonnées

Cited literature [20 references]  Display  Hide  Download

https://hal.inria.fr/hal-01020317
Contributor : Anne Jaigu <>
Submitted on : Tuesday, July 8, 2014 - 9:51:23 AM
Last modification on : Sunday, July 29, 2018 - 8:46:01 AM
Document(s) archivé(s) le : Wednesday, October 8, 2014 - 11:22:12 AM

File

0086.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01020317, version 1

Collections

Citation

Lei Qiu, Shenfang Yuan, Qiao Bao, Tianxiang Huang. An on-line continuous updating Gaussian mixture model for damage monitoring under time-varying structural boundary condition. EWSHM - 7th European Workshop on Structural Health Monitoring, IFFSTTAR, Inria, Université de Nantes, Jul 2014, Nantes, France. ⟨hal-01020317⟩

Share

Metrics

Record views

252

Files downloads

345