An online-offline prognosis model for fatigue life prediction under biaxial cyclic loading with overloads

Guoyi Li, Siddhant Datta, Aditi Chattopadhyay, Nagaraja Iyyer, Nam Phan

Research output: Contribution to journalArticlepeer-review

10 Scopus citations


This paper presents a robust online-offline model for the prediction of crack propagation under complex in-phase biaxial fatigue loading in the presence of overloads of different magnitudes. The online prognosis model comprises a combination of finite element analysis and data-driven regression to predict the crack propagation under constant loading, while the offline model is trained using experimental data to inform the post-overload crack growth retardation behavior to the online model. The developed methodology is validated by conducting biaxial fatigue experiments using aluminum AA7075-T651 alloy cruciform specimens. A close correlation is observed between the experimental results and model predictions. The results show that the model successfully predicts the crack retardation behavior under the influence of overloads with different magnitudes occurring at different stages of fatigue crack growth. Error analysis is conducted to investigate the sensitivities of the number of training points and crack increments to the prediction accuracy. In addition, the error propagation with respect to the crack length is studied, which provides constructive suggestions for further model improvement.

Original languageEnglish (US)
Pages (from-to)1175-1190
Number of pages16
JournalFatigue and Fracture of Engineering Materials and Structures
Issue number5
StatePublished - May 2019


  • Gaussian process machine learning
  • biaxial fatigue
  • crack propagation
  • online-offline model
  • overload
  • prognosis

ASJC Scopus subject areas

  • General Materials Science
  • Mechanics of Materials
  • Mechanical Engineering


Dive into the research topics of 'An online-offline prognosis model for fatigue life prediction under biaxial cyclic loading with overloads'. Together they form a unique fingerprint.

Cite this