A novel metamodeling approach for probabilistic LCF estimation of turbine disk. (February 2021)

• Record Type: Journal Article
• Authors: Song, Lu-Kai; Bai, Guang-Chen; Li, Xue-Qin
• Is Part Of: Engineering failure analysis. Volume 120(2021)
• Publication Date: 2021-02
• Subjects: Probabilistic LCF -- Turbine disk -- Distributed-coordinated strategy -- Neural network -- Metamodel; System failures (Engineering) -- Periodicals; Fracture mechanics -- Periodicals; Reliability (Engineering) -- Periodicals; Pannes -- Périodiques; Rupture, Mécanique de la -- Périodiques; Fiabilité -- Périodiques; Fracture mechanics; Reliability (Engineering); System failures (Engineering); Periodicals; Electronic journals; 620.112
• DOI: 10.1016/j.engfailanal.2020.105074
• Languages: English
• ISSNs: 1350-6307

Abstract

Highlights: A metamodeling approach (DCNNM) is proposed for probabilistic LCF estimation for the first time. We built the DCNNM by integrating distributed-coordinated strategy and neural network model. Probabilistic LCF estimation theory is established with the developed DCNNM approach. Probabilistic LCF estimation of turbine disk is completed, the service life is obtained. The developed metamodeling approach is validated to possess high efficiency and accuracy. Abstract: To improve the computing efficiency and accuracy of probabilistic low cycle fatigue (LCF) estimation for turbine disk, a distributed-coordinated neural network metamodel (DCNNM) is developed. By integrating the proposed neural network metamodel and distributed-coordinated strategy, the mathematical model of DCNNM is studied. The probabilistic LCF estimation theory is introduced in respect of the presented DCNNM. Moreover, the probabilistic LCF estimation for turbine disk is regarded as one case to evaluate the proposed method with respect to various randomness such as material variability, load variation and model randomness. We obtain the distributional traits, reliability degree and sensitivity degree of LCF failure cycle, which provides an effective guidance for the turbine disk life control. By comparing the direct Monte Carlo simulation, support vector regression (SVR), neural network metamodel (NNM), distributed-coordinated SVR (DCSVR) and DCNNM, we observe that the proposed DCNNM approach possesses high efficiency and accuracy for probabilistic LCF estimation of turbine disk. The present effort offers a useful insight for estimating LCF failure from a probabilistic perspective.