Mixed Gaussian process and state-space approach for fatigue crack growth prediction

S. Mohanty, R. Teale, Aditi Chattopadhyay, Pedro Peralta, C. Willhauck

Research output: Chapter in Book/Report/Conference proceedingConference contribution

15 Scopus citations

Abstract

This paper presents a hybrid model of fatigue crack growth in metallic alloys that combines a physics-based state-space model and a data-driven Gaussian Process (GP) and can be applied for variable-amplitude aircraft service loads. The physics driven part has a state variable formulation with underlying physics based on the crack closure concept used in the FASTRAN-11 model. The data driven part of the model uses a kernel based GP regression model, which is being considered as the equivalent of an infinite neuron neural network model. The state space part is linked to the data driven part via a constraint factor which accounts for the effects of the stress state on the plastic-zone size. Through simulations it is shown that the hybrid model gives better predictions of fatigue crack length and growth rate as compared to both the pure physics model and the pure data driven GP model.

Original languageEnglish (US)
Title of host publicationStructural Health Monitoring 2007
Subtitle of host publicationQuantification, Validation, and Implementation - Proceedings of the 6th International Workshop on Structural Health Monitoring, IWSHM 2007
EditorsFu-Kuo Chang
PublisherDEStech Publications
Pages1108-1115
Number of pages8
ISBN (Electronic)9781932078718
StatePublished - Jan 1 2007
Event6th International Workshop on Structural Health Monitoring: Quantification, Validation, and Implementation, IWSHM 2007 - Stanford, United States
Duration: Sep 11 2007Sep 13 2007

Publication series

NameStructural Health Monitoring 2007: Quantification, Validation, and Implementation - Proceedings of the 6th International Workshop on Structural Health Monitoring, IWSHM 2007
Volume2

Other

Other6th International Workshop on Structural Health Monitoring: Quantification, Validation, and Implementation, IWSHM 2007
CountryUnited States
CityStanford
Period9/11/079/13/07

ASJC Scopus subject areas

  • Computer Science Applications
  • Health Information Management

Fingerprint Dive into the research topics of 'Mixed Gaussian process and state-space approach for fatigue crack growth prediction'. Together they form a unique fingerprint.

  • Cite this

    Mohanty, S., Teale, R., Chattopadhyay, A., Peralta, P., & Willhauck, C. (2007). Mixed Gaussian process and state-space approach for fatigue crack growth prediction. In F-K. Chang (Ed.), Structural Health Monitoring 2007: Quantification, Validation, and Implementation - Proceedings of the 6th International Workshop on Structural Health Monitoring, IWSHM 2007 (pp. 1108-1115). (Structural Health Monitoring 2007: Quantification, Validation, and Implementation - Proceedings of the 6th International Workshop on Structural Health Monitoring, IWSHM 2007; Vol. 2). DEStech Publications.