Stress intensity factor estimation for unbalanced rotating cracked shafts by artificial neural networks

Abstract

The knowledge of the stress intensity factor (SIF) values along a crack front is essential to calculate the crack growth rate and the remaining life of a mechanical component. In the case of a rotating shaft, usually it presents disalignments, which modify the SIF data with regard to a balanced one. This paper presents the use of an artificial neural network (ANN) for estimating the SIF at the crack front in an unbalanced shaft under rotating bending, previously, a quasi‐static numerical (finite element) model, which simulates a rotating shaft, has been developed to create the training cases for the ANN. The obtained results allow to study the influence of the unbalance of rotating shafts in the crack breathing mechanism and will allow to predict the influence of this behaviour on the values of the SIF and in the propagation of cracks.

Neural approach to estimate the stress intensity factor of semi‐elliptical cracks in rotating cracked shafts in bending

“…[35][36][37][38][39][40] Artificial neural networks are computational systems inspired by biological neural networks formed by different layers: an input layer, one or more hidden layers, and an output layer. [35][36][37][38][39][40] Artificial neural networks are computational systems inspired by biological neural networks formed by different layers: an input layer, one or more hidden layers, and an output layer.…”

“…In the last 2 decades, many researchers have used artificial neural networks (ANNs) to solve a great variety of complex engineering problems, as pattern recognition, optimization, or prediction. [35][36][37][38][39][40] Artificial neural networks are computational systems inspired by biological neural networks formed by different layers: an input layer, one or more hidden layers, and an output layer. The layers involve different nodes that are connected by weights.…”

Artificial neural network approach to predict the fracture parameters of the size effect model for concrete

“…This ability of ANN makes a simulation more robust and fault‐tolerant than that from a classical regression analysis, while maintaining the same efficiency of computation. Consequently, there has been a great upsurge in utilizing this useful tool to predict mechanical properties of materials . In particular, Shabani and Mazahery presented an integrated approach that combines ANN, GA and FEM to find an optimal solidification procedure to produce 2024 Al‐alloy with minimum wear and maximum strength.…”

“…Consequently, there has been a great upsurge in utilizing this useful tool to predict mechanical properties of materials. [11][12][13][14][15][16] In particular, Shabani and Mazahery 11 presented an integrated approach that combines ANN, GA and FEM to find an optimal solidification procedure to produce 2024 Al-alloy with minimum wear and maximum strength. The FEM-GA-ANN results were consistent with experimental measurements.…”

“…They also investigated the effect of different ANN training algorithms on the predicted results of the mechanical behaviour of an A356 composite reinforced with B 4 C particulates and showed that Levenberg-Marquardt learning algorithms gave the best results. 12 Muñoz-Abella et al 16 employed ANN in estimating the SIF values along a crack front with respect to the unbalanced and rotating cracked shafts in consideration of the breathing mechanism for any crack depth. However, currently, the application of ANN is rarely adopted in concrete fracture mechanics because of the complexity of concrete fracture behaviour, which is quite different from metal or composite behaviour.…”