Multiaxial fatigue prediction and uncertainty quantification based on back propagation neural network and Gaussian process regression

Gao, Jingjing; Wang, Jun; Zhou, Xu; Wang, Cunjun; Song, Yan

doi:10.1016/j.ijfatigue.2022.107361

Cited by 23 publications

(2 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…29 The surrogate modeling technique-which employs simplified mathematical models as substitutes for high-fidelity models-has been widely applied in research on uncertainty in multiaxial fatigue life. [30][31][32][33] Jing et al 34 considered the uncertainties of material and load and utilized a back propagation neural network model to replace FEA in evaluations of the fatigue reliability of a cylinder head. Gao et al 35 introduced a probabilistic method based on a substructure Kriging surrogate model to predict combined high and low cycle fatigue life of gas turbine blades.…”

Section: Introductionmentioning

confidence: 99%

“…The key to solving this problem is to establish a surrogate model that can effectively replace the expensive‐to‐evaluate FEA model 29 . The surrogate modeling technique—which employs simplified mathematical models as substitutes for high‐fidelity models—has been widely applied in research on uncertainty in multiaxial fatigue life 30–33 . Jing et al 34 considered the uncertainties of material and load and utilized a back propagation neural network model to replace FEA in evaluations of the fatigue reliability of a cylinder head.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A probabilistic life prediction framework for vibration fatigue of turbine blade based on refined polynomial chaos expansion

Fang,

Liu,

et al. 2024

Fatigue Fract Eng Mat Struct

View full text Add to dashboard Cite

Vibration fatigue is a typical form of multiaxial fatigue. In this study, initial vibration fatigue tests were conducted on rocket turbine blades, and significant dispersion in fatigue life was found. Consistent with this finding, a novel probabilistic life prediction framework was proposed. This framework integrates sensitivity analysis and sequential sampling technology and introduces polynomial chaos expansion as a computationally efficient alternative to finite element analysis. And a continuous mechanics‐based damage evolution model was employed to examine the vibration fatigue life of turbine blades. The findings validate the effectiveness of the framework, as no significant difference was found between the experimental results and simulated predictions at the 95% confidence level. Furthermore, comparison with the Monte Carlo simulation indicated that this framework achieves comparable prediction accuracy, while significantly reducing the required samples by 2 orders of magnitude, which effectively addresses the fatigue problem of small sample data. This framework enables rapid and accurate multiaxial fatigue probabilistic life prediction, which holds important implications for the reliability design of reusable spacecraft.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A probabilistic life prediction framework for vibration fatigue of turbine blade based on refined polynomial chaos expansion

Fang,

Liu,

et al. 2024

Fatigue Fract Eng Mat Struct

View full text Add to dashboard Cite

show abstract

A novel interpretable hybrid model for multi-step ahead dissolved oxygen forecasting in the Mississippi River basin

Ali,

Mohammadi Ghaleni,

Moghaddasi

et al. 2024

Stoch Environ Res Risk Assess

View full text Add to dashboard Cite

Impact of the interface on the fatigue life of steel-based explosively welded heterostructured plates

Derda

Karolczuk

Robak

et al. 2023

Archiv.Civ.Mech.Eng

View full text Add to dashboard Cite

Melted zones, microcracks, shear bands, and elastic incompatibility of explosively welded materials are features that may initialize cracks at the interface and reduce fatigue strength. This study aims to determine the effect of interfacial defect-like structures on the fatigue strength of explosively welded corrosion-resistant plates. Cyclic axial loading was applied to seven distinct layer-by-layer compositions of Ti Gr 1, Zr 700 alloys, and carbon steels. The interfacial wave height as a metric of potential fatigue life influencing factors along with measured strain amplitude was applied as the input quantities for the Machine Learning based model, i.e. the Gaussian process for regression (GPR). This is a novel and successful application of GPR to estimate the effect of interfacial wave height on the fatigue life of explosively welded plates. For the first time, the effect of the interface feature on fatigue life was estimated quantitatively. The Digital Image Correlation technique was applied to measure the field of cyclic strain for the purpose of verifying if a single strain amplitude is representative of a heterostructured plate. It was found that interfacial wave height is an important feature and its increase by 100 µm reduces the fatigue life of analysed plates by 36%. Additionally, to validate the applicability of explosively welded plates to engineering structures under cyclic loading, the experimental fatigue lives were compared with the design curve of the American Society of Mechanical Engineers (ASME) code.

show abstract

Multiaxial fatigue prediction and uncertainty quantification based on back propagation neural network and Gaussian process regression

Cited by 23 publications

References 47 publications

A probabilistic life prediction framework for vibration fatigue of turbine blade based on refined polynomial chaos expansion

A probabilistic life prediction framework for vibration fatigue of turbine blade based on refined polynomial chaos expansion

A novel interpretable hybrid model for multi-step ahead dissolved oxygen forecasting in the Mississippi River basin

Impact of the interface on the fatigue life of steel-based explosively welded heterostructured plates

Contact Info

Product

Resources

About