Frequency analysis of cracked beams using machine learning | ||
| مدل سازی در مهندسی | ||
| Articles in Press, Accepted Manuscript, Available Online from 14 September 2025 | ||
| Document Type: Mechanics article | ||
| DOI: 10.22075/jme.2025.36673.2800 | ||
| Authors | ||
| مسعود میر* ; Rasta Razeghian | ||
| Department of Mechanical Engineering, Faculty of advanced technologies, Qochan University of Technology, Quchan, Iran | ||
| Receive Date: 21 January 2025, Revise Date: 12 June 2025, Accept Date: 17 June 2025 | ||
| Abstract | ||
| The presence of cracks in a beam changes the dynamic characteristics of the beam. Therefore, to assess the condition of the beam, its natural frequencies must be examined. In this study, using a numerical solution based on the Rayleigh method, the natural frequencies of a beam with two cracks are calculated based on the depth and location of the cracks. Next, using the Python programming language, the aforementioned mathematical relationship is entered into this program to solve this relationship sequentially for different inputs by creating iterative loops. The goal of this is to produce a dataset that can be used to train machine learning algorithms such as random forest regression, gradient boosting regression, multilayer perceptron, and decision tree regression to predict the natural frequency. The key innovation in this study is the use of a network search method to determine the optimal amount of data for each algorithm, which increases accuracy and introduces a new criterion for comparison called "required data volume". The study found that increasing the size of the dataset generally increases the prediction accuracy of the algorithms. In addition, algorithms that predict a single output have higher accuracy compared to those that predict multiple outputs. The study demonstrates the effective use of machine learning algorithms for predicting natural frequencies. The gradient boosting regression algorithm with an accuracy of 84.10% and the random forest regression algorithm with an accuracy of 83.73% emerged as the superior methods for predicting beam frequencies. | ||
| Keywords | ||
| Machine learning; Cracked beam; Natural frequency; Python; Multi layer perceptron; Decision tree regression | ||
| References | ||
|
| ||
|
Statistics Article View: 310 |
||