Machine learning‐based efficient stress intensity factor calculation for aeroengine disk probabilistic risk assessment under polynomial stress fields

Abstract

In probabilistic failure risk assessment, the accuracy and efficiency of the stress intensity factor calculation are important. The universal weight function method has been widely adopted for efficiency, but this method still has some debatable parts. For accurate and efficient stress intensity factor prediction, two approaches for machine learning techniques are specially designed. Three tests are conducted for the first approach where Gaussian process regression, tree-structure models, and artificial neural network are evaluated and compared for the ability of interpolation and extrapolation. Results show that the artificial neural network and extremely randomized trees perform better.Hybrid models in the second approach are also proposed, and results show that the accuracy of SIF calculation can be improved by 5-35% compared with the weight function. Real aeroengine disks are adopted, and the errors of machine learning methods are less than 20% in the disk life calculation.

Estimation of stress intensity factor for surface cracks in the firtree groove structure of turbine disk using pool-based active learning with Gaussian Process Regression

“…Recently, K-solutions combined with ML have become a topic of growing interest (Keprate et al, 2017;Liu et al, 2020;Muñoz-Abella et al, 2015;Xu et al, 2021). With the capacity to solve complex input-output problems of an underlying system, ML is especially useful in the calculation of K. These solutions proceed in three steps (in this paper, the method that follows these steps is called the traditional ML method): (1) calculating the limit number of K data as the training data set; (2) training an ML model using training the data set and (3) predicting values of K using the trained ML model.…”

Prediction of rock tensile-shear fracture propagation and crack interaction

“…The current stress intensity factor (SIF) prediction models include BP-GA [44]; BP [45]; GPR, DT, RF, ET, GBDT, ANN [46]; CNN [47]; gray prediction model [48]; LSSVM [49]; DT, RF, ET, FCNN [50]; SVM [51]; and ANFIS-GA [52]. These methods are practical and reliable for SIF pre diction.…”

Application of machine learning methods in multiaxial fatigue life prediction

“…Sometimes ML methods are used to predict the fatigue life of machines 31–33 . In addition to the commonly used ML methods such as DNN, RF, or support vector machine (SVM), the applicability of other methods is sometimes tested 22,23,27,34,35 …”

“…Silva et al 22 combined an extremely randomized trees (ERT) algorithm with a finite element method to predict the fatigue life of different adhesively bonded joints. Xu et al 23 used different ML methods to estimate the stress intensity factor for an aero‐engine disk, employing Gaussian process regression, decision tree (DT), random forest (RF), extremely randomized trees, gradient boosting, and ANN, as well as an ML‐based hybrid model. Burghardt et al 24 successfully estimated the stress–strain response at notches under uniaxial and multiaxial proportional loadings using ANNs.…”