Work

A data-driven approach for predicting stress intensity factors of a single-edge cracked plate with a random polygon-shaped void

Year: 2025

Type: article

Abstract: This study represents a data-driven framework for predicting mode I (KI) and mode II (KII) Stress Intensity Factors (SIFs) in single-edge cracked plates with central polygon-shaped voids. Finite element simulations were conducted in Abaqus software to generate a dataset by varying key parameters, including the polygon’s number of vertices, angle, average radius, and crack length. Two machine learning models were employed to analyze the dataset created by the finite element method: Group Method of Data Handling (GMDH) networks and an Artificial Neural Network (ANN). The GMDH networks were optimized using the least squares method and the Root Mean Squared Error (RMSE) criteria, while the ANN, designed as a feedforward fully connected network, was trained with the backpropagation algorithm and the gradient descent optimization technique using TensorFlow and Keras libraries. The ANN demonstrated exceptional accuracy, with a R2 value exceeding 0.99 for KI predictions and 0.98 for KII, significantly outperforming GMDH models, particularly in capturing the nonlinear behavior of KII.

Source: Mechanical Engineering Advances

DOI: 10.59400/mea3338

Authors: Mehrad Zargar Ershadi, Saeid Nickabadi, Majid Askari Sayar, Alireza Alidoust, Reza Ansari

Institutions: University of Guilan, Imam Khomeini Marine Science University, Isfahan University of Technology

Language: English

Topic: Fatigue and fracture mechanics

Subfield: Mechanics of Materials

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: diamond

OpenAlex ID: W7127044022

FWCI:

Cites: 29

Cited by:

Related to: 10