{"title":"基于 DBO-ELM 的 C/SiC 复合材料有序砂轮磨损状态识别研究","authors":"Ye Guo , Bing Chen , Hongyu Zeng , Guangye Qing , Bing Guo","doi":"10.1016/j.wear.2024.205529","DOIUrl":null,"url":null,"abstract":"<div><p>In the grinding process of 2.5D C/SiC composites by the ordered grinding wheel, the signals are more complicated due to the special structure of the material and the ordered abrasive clusters, which makes it difficult to identify the wear state of the grinding wheel. To solve this problem, this study employs the analysis methods of direct and indirect. The acoustic emission signals, grinding force and grinding temperature signals throughout the entire lifespan of the grinding wheel were collected, and the topography of the grinding wheel was photographed. The difference in wear behavior between the ordered grinding wheel and the traditional disordered grinding wheel was compared and analyzed. The corresponding relationship between the ordered grinding wheel wear behavior and various signals was studied in depth. The results showed that the initial wear occurred during the 1st-80th grinding process, the stable wear occurred during the 81st-224th grinding process, and the serious wear occurred during the 225th-350th grinding process. Additionally, this research further extracted the key features of various signals, and used the Pearson correlation coefficient to identify the features highly related to grinding wheel wear. Subsequently, LDA was used to reduce the dimension of individual signal types. By comparing the recognition effects of different signal type combinations, the optimal combination was determined. Finally, this research proposed a DBO-ELM model for the recognition and classification of ordered grinding wheel wear state. Compared to four common machine learning models, namely KNN, SVM, ELM, and BP, the proposed DBO-ELM model demonstrated a classification accuracy of 94.86 %, which was increased by 25.57 %, 16 %, 16 % and 7.86 % respectively. This demonstrates that the DBO-ELM model proposed in this research has certain advantages and potential in the wear state recognition of abrasive clusters ordered grinding wheel.</p></div>","PeriodicalId":23970,"journal":{"name":"Wear","volume":null,"pages":null},"PeriodicalIF":5.3000,"publicationDate":"2024-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Research on wear state identification of ordered grinding wheel for C/SiC composites based on DBO-ELM\",\"authors\":\"Ye Guo , Bing Chen , Hongyu Zeng , Guangye Qing , Bing Guo\",\"doi\":\"10.1016/j.wear.2024.205529\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><p>In the grinding process of 2.5D C/SiC composites by the ordered grinding wheel, the signals are more complicated due to the special structure of the material and the ordered abrasive clusters, which makes it difficult to identify the wear state of the grinding wheel. To solve this problem, this study employs the analysis methods of direct and indirect. The acoustic emission signals, grinding force and grinding temperature signals throughout the entire lifespan of the grinding wheel were collected, and the topography of the grinding wheel was photographed. The difference in wear behavior between the ordered grinding wheel and the traditional disordered grinding wheel was compared and analyzed. The corresponding relationship between the ordered grinding wheel wear behavior and various signals was studied in depth. The results showed that the initial wear occurred during the 1st-80th grinding process, the stable wear occurred during the 81st-224th grinding process, and the serious wear occurred during the 225th-350th grinding process. Additionally, this research further extracted the key features of various signals, and used the Pearson correlation coefficient to identify the features highly related to grinding wheel wear. Subsequently, LDA was used to reduce the dimension of individual signal types. By comparing the recognition effects of different signal type combinations, the optimal combination was determined. Finally, this research proposed a DBO-ELM model for the recognition and classification of ordered grinding wheel wear state. Compared to four common machine learning models, namely KNN, SVM, ELM, and BP, the proposed DBO-ELM model demonstrated a classification accuracy of 94.86 %, which was increased by 25.57 %, 16 %, 16 % and 7.86 % respectively. This demonstrates that the DBO-ELM model proposed in this research has certain advantages and potential in the wear state recognition of abrasive clusters ordered grinding wheel.</p></div>\",\"PeriodicalId\":23970,\"journal\":{\"name\":\"Wear\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":5.3000,\"publicationDate\":\"2024-07-31\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Wear\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0043164824002941\",\"RegionNum\":1,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ENGINEERING, MECHANICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Wear","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0043164824002941","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, MECHANICAL","Score":null,"Total":0}
Research on wear state identification of ordered grinding wheel for C/SiC composites based on DBO-ELM
In the grinding process of 2.5D C/SiC composites by the ordered grinding wheel, the signals are more complicated due to the special structure of the material and the ordered abrasive clusters, which makes it difficult to identify the wear state of the grinding wheel. To solve this problem, this study employs the analysis methods of direct and indirect. The acoustic emission signals, grinding force and grinding temperature signals throughout the entire lifespan of the grinding wheel were collected, and the topography of the grinding wheel was photographed. The difference in wear behavior between the ordered grinding wheel and the traditional disordered grinding wheel was compared and analyzed. The corresponding relationship between the ordered grinding wheel wear behavior and various signals was studied in depth. The results showed that the initial wear occurred during the 1st-80th grinding process, the stable wear occurred during the 81st-224th grinding process, and the serious wear occurred during the 225th-350th grinding process. Additionally, this research further extracted the key features of various signals, and used the Pearson correlation coefficient to identify the features highly related to grinding wheel wear. Subsequently, LDA was used to reduce the dimension of individual signal types. By comparing the recognition effects of different signal type combinations, the optimal combination was determined. Finally, this research proposed a DBO-ELM model for the recognition and classification of ordered grinding wheel wear state. Compared to four common machine learning models, namely KNN, SVM, ELM, and BP, the proposed DBO-ELM model demonstrated a classification accuracy of 94.86 %, which was increased by 25.57 %, 16 %, 16 % and 7.86 % respectively. This demonstrates that the DBO-ELM model proposed in this research has certain advantages and potential in the wear state recognition of abrasive clusters ordered grinding wheel.
期刊介绍:
Wear journal is dedicated to the advancement of basic and applied knowledge concerning the nature of wear of materials. Broadly, topics of interest range from development of fundamental understanding of the mechanisms of wear to innovative solutions to practical engineering problems. Authors of experimental studies are expected to comment on the repeatability of the data, and whenever possible, conduct multiple measurements under similar testing conditions. Further, Wear embraces the highest standards of professional ethics, and the detection of matching content, either in written or graphical form, from other publications by the current authors or by others, may result in rejection.