Improved arithmetic optimization algorithm and its application to carbon fiber reinforced polymer-steel bond strength estimation

作     者：Shi, Xiaoling Yu, Xinping Esmaeili-Falak, Mahzad 

作者机构：Lyuliang Univ Dept Min Engn Luliang Peoples R China Zhejiang Guangsha Vocat & Tech Univ Construction Sch Intelligent Mfg Dongyang Peoples R China Islamic Azad Univ Dept Civil Engn North Tehran Branch Tehran Iran 

出 版 物：《COMPOSITE STRUCTURES》 (复合材料结构)

年 卷 期：2023年第306卷第1期

核心收录：

学科分类：08[工学] 0805[工学-材料科学与工程（可授工学、理学学位）] 080502[工学-材料学] 0801[工学-力学（可授工学、理学学位）] 

基　　金：Scientific and Technological Inno-vation Programs of Higher Education Institutions in Shanxi [2021L567] 

主　　题：Carbon fiber reinforced polymer-steel interface Bond strength Random forests Improved arithmetic optimization algorithm 

摘      要：In order to restore steel structures, bonding carbon fiber reinforced polymer (CFRP) laminates have been widely used. The bond strength (PU) between the CFRP and steel, along with the mechanical characteristics of the CFRP, is frequently crucial to the final strengthened effectiveness. However, the bond behavior at the CFRP-steel (CS) interface is incredibly complex, with several potential sources of failure, making it difficult to predict the PU and the stability of the CFRP-enhanced steel structure. In this specific instance, effective techniques were developed using a hybridized Random Forests (RF) methodology on collected CS single-shear experiment data to predict the PU of CS. The RF hyperparameters were tuned using the COOT optimizer (COOT), arithmetic optimization algorithm (AOA), and improved arithmetic optimization algorithm (IAOA). The IAOA was developed in this article by combining AOA with Aquila optimizer (AO) in order to overcome the shortage of it. When the training, testing, and data collection phases of each algorithm were executed in parallel, the results were uniformly excellent. The proposed IAOA -RF was the preferred approach, while other methods were also dependable in the prediction of CS interfacial PU, as determined by evaluating established designs using different aspects of analysis, such as different error criteria, the Taylor diagram, uncertainty analysis, scatter index analysis, and error distribution.