Data‐driven approach to design fatigue‐resistant notched structures

Abstract

Fatigue is the most common failure mode for engineering materials in various industrial applications. Generally, designing new products and components based on typical deterministic fatigue design approaches is slow and expensive. Although numerous investigations have been carried on this topic over decades, there still lacks a robust and efficient method to design a super‐fatigue‐resistant and testless structure. In this investigation, a novel data‐driven approach based on the deep learning algorithm is applied on designing a fatigue‐resistant notched structure under different loading via numerical simulations and deep learning. The notches with various shapes are designed by employing the same material. Then finite element (FE) simulations are performed to obtain the mechanical behaviors of notched structures under different loading conditions. The deep learning algorithm is applied to predict mechanical behaviors of unknown notched structures based on input training set of simple notched geometries and relative lower computational cost. It is demonstrated that deep learning on the fatigue‐resistant structure is a promising approach of fatigue design. This work offers an alternative design strategy of fatigue‐resistant structure and cost‐effective solution to accelerate the design of engineering components under different loading conditions.

Finite element and generalized regression neural network modelling of multiple cracks growth under the influence of multiple crack parameters

“…In recent years, inspired by how human brain neurons connect and work to acquire an experience stipulated by the environment and then respond to the environment in a new manner, artificial neural networks (ANN) and NN have emerged as robust computational approaches. NN can also be considered as an information-based or data-driven computational approach (Kalayci et al, 2020;Strohmann et al, 2021;Wang et al, 2023). In an NN, complex relationships between input and output variables, which are often too complicated to handle by conventional closed-form models, can be revealed by the process of acquiring experience (learning process) through a set of training data that captures underlying patterns from the phenomena of interest (Haykin, 2009).…”

Using deep learning method to predict dimensionless values of stress intensity factors and T‐stress of edge notch disk bend (ENDB) specimen

“…Between the ML methods, deep learning (DL) approaches can identify complex and difficult patterns in data to generate accurate predictions. 19,20 In deep neural network (DNN) systems, the network trains by the database, and after performing a numerical process, the network produces the output data. 21,22 The DNN can precisely extract patterns, model nonlinear problems, and enable multi-layer neural network calculations.…”

“…The ML approach leads to an increase in the accuracy of predictions using artificial intelligence algorithms. Between the ML methods, deep learning (DL) approaches can identify complex and difficult patterns in data to generate accurate predictions 19,20 . In deep neural network (DNN) systems, the network trains by the database, and after performing a numerical process, the network produces the output data 21,22 .…”

Fatigue evolution prediction for fiber‐reinforced plastics based on frequency‐wavenumber wavefield of guided wave using deep‐learning model

“…Due to the complex fatigue behavior exhibited by FRPs, 20 addressing numerous engineering issues related to structural fatigue poses inherent challenges. However, in recent years, the rapid development of artificial intelligence (AI) technology has provided a promising avenue for handling these fatigue‐related problems, such as structural fatigue‐resistance design, 21 structural property evaluation, 22 and fatigue fracture recognition 23 . In order to characterize the fatigue state of composites, many feature extraction methods based on deep learning models can be considered, such as the deep neural network (DNN), 22 the convolutional neural network (CNN), 24 and the autoencoder (AE) 25 .…”