銲接件振動應力消除模擬並以類神經網路進行預估;Simulation of Vibration Stress Relief of Weldments and Prediction with Artificial Neural Network

NCU Institutional Repository > 工學院 > 機械工程研究所 > 博碩士論文 > Item 987654321/93770

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/93770

题名:	銲接件振動應力消除模擬並以類神經網路進行預估;Simulation of Vibration Stress Relief of Weldments and Prediction with Artificial Neural Network
作者:	李柏泓;Li, Bo-Hong
贡献者:	機械工程學系
关键词:	振動應力消除;沙博什模型;有限元素模擬;類神經網路;Vibration stress relief;Chaboche model;Simulation using finite element software;Artificial neural network
日期:	2023-07-20
上传时间:	2024-09-19 17:36:08 (UTC+8)
出版者:	國立中央大學
摘要:	本研究期望提供一個類神經網路模型可以快速預估不同工況的殘留應力在經過振動應力消除後的應力值，透過此類神經網路模型可以只需直接輸入所需少數特徵就可以快速獲得應力消除的量，以立即預估振動應力消除的效果。如何消除殘留應力一直都是金屬加工的重要課題之一，隨著技術發展，振動應力消除法(VSR)逐漸受到關注，該技術藉由疊加工件內部的殘留應力與外部施予的負載來產生塑性變形，並由此達到應力重新分布，降低材料內部應力的目的，然而其應力消除量不好量測。故本研究透過進行銲接有限元素模擬產生殘留應力，再對其施予振動應力消除，即反覆施力效果模擬，並收集有限元素模擬結果的網格節點資料形成數據庫，以該數據庫訓練類神經網路，期望能訓練出準確度高的類神經網路，期能直接以此對振動應力消除後的殘留應力進行預估。;This research aims to provide an artificial neural network model that can efficiently estimate the residual stress under different operating conditions of vibration stress relief. By utilizing, it is expected to bypass the time-consuming finite element simulations. Through the direct input of the necessary features to the artificial neural network model, it is feasible to expeditiously acquire stress values after applying the vibratory force. Elimination of residual stress has always been one of the important issues in metal processing. With the development of innovative technology, Vibration Stress Relief (VSR) has gradually attracted attention. Vibration stress relief achieves because of the plastic deformation due to superimposing the residual stress and the applied cyclic load. However, there is no systematic way to estimate the stress reduction effect. This study, therefore, tried to numerically study the stress reduction resulting from the application of a cyclical load. Next the numerical data obtained from the finite element software, ANSYS, were collected. Artificial neural network model was build using these numerical data. The differences between the numerical data and the predicted data of the model were less than 5 % in Mean Absolute Percentage Error (MAPE).
显示于类别:	[機械工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	15	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....