C. Krause, B. Uysal, M. Engler, C. Radek, M. Schaudig
{"title":"基于巴克豪森效应的机器学习技术在确定表面硬度中的应用","authors":"C. Krause, B. Uysal, M. Engler, C. Radek, M. Schaudig","doi":"10.1515/htm-2022-1029","DOIUrl":null,"url":null,"abstract":"Abstract Ensuring product and part quality impacts manufacturing productivity, efficiency and profitability. The goal of every manufacturing company is to quickly identify reduced quality in order to take appropriate measures to improve quality. The use of non-destructive testing methods such as Barkhausen noise in combination with artificial intelligence (AI), which immediately classifies the data, offers a way to implement the desired quality monitoring in a production line. In the present study, the measured data of the Barkhausen signal of surface hardened components with different degrees of tempering were analyzed. For this purpose, suitable AI models were developed and trained with the processed measurement data to generate prediction values for the surface hardness. Data preparation and further processing was carried out using the Spyder development environment with the Python programming language. The following models were applied, tested and optimized during the study: Support vector machine, random forest regression and an artificial neural network. The models were able to predict hardness levels with high accuracy after effective training. Overall, the neural network showed the best results. The applied procedures and methods are fast, non-destructive and provide results with acceptable measurement error, which allows their use in the production environment. Further improvements will be sought in the future, e. g. by applying a larger amount of training data, by changing the features used in the training and by increasing the measurement accuracy when capturing the Barkhausen signal.","PeriodicalId":44294,"journal":{"name":"HTM-Journal of Heat Treatment and Materials","volume":"6 1","pages":"409 - 424"},"PeriodicalIF":0.3000,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Application of Machine Learning Techniques to Determine Surface Hardness Based on the Barkhausen Effect\",\"authors\":\"C. Krause, B. Uysal, M. Engler, C. Radek, M. Schaudig\",\"doi\":\"10.1515/htm-2022-1029\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Abstract Ensuring product and part quality impacts manufacturing productivity, efficiency and profitability. The goal of every manufacturing company is to quickly identify reduced quality in order to take appropriate measures to improve quality. The use of non-destructive testing methods such as Barkhausen noise in combination with artificial intelligence (AI), which immediately classifies the data, offers a way to implement the desired quality monitoring in a production line. In the present study, the measured data of the Barkhausen signal of surface hardened components with different degrees of tempering were analyzed. For this purpose, suitable AI models were developed and trained with the processed measurement data to generate prediction values for the surface hardness. Data preparation and further processing was carried out using the Spyder development environment with the Python programming language. The following models were applied, tested and optimized during the study: Support vector machine, random forest regression and an artificial neural network. The models were able to predict hardness levels with high accuracy after effective training. Overall, the neural network showed the best results. The applied procedures and methods are fast, non-destructive and provide results with acceptable measurement error, which allows their use in the production environment. Further improvements will be sought in the future, e. g. by applying a larger amount of training data, by changing the features used in the training and by increasing the measurement accuracy when capturing the Barkhausen signal.\",\"PeriodicalId\":44294,\"journal\":{\"name\":\"HTM-Journal of Heat Treatment and Materials\",\"volume\":\"6 1\",\"pages\":\"409 - 424\"},\"PeriodicalIF\":0.3000,\"publicationDate\":\"2022-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"HTM-Journal of Heat Treatment and Materials\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1515/htm-2022-1029\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q4\",\"JCRName\":\"THERMODYNAMICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"HTM-Journal of Heat Treatment and Materials","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1515/htm-2022-1029","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"THERMODYNAMICS","Score":null,"Total":0}
Application of Machine Learning Techniques to Determine Surface Hardness Based on the Barkhausen Effect
Abstract Ensuring product and part quality impacts manufacturing productivity, efficiency and profitability. The goal of every manufacturing company is to quickly identify reduced quality in order to take appropriate measures to improve quality. The use of non-destructive testing methods such as Barkhausen noise in combination with artificial intelligence (AI), which immediately classifies the data, offers a way to implement the desired quality monitoring in a production line. In the present study, the measured data of the Barkhausen signal of surface hardened components with different degrees of tempering were analyzed. For this purpose, suitable AI models were developed and trained with the processed measurement data to generate prediction values for the surface hardness. Data preparation and further processing was carried out using the Spyder development environment with the Python programming language. The following models were applied, tested and optimized during the study: Support vector machine, random forest regression and an artificial neural network. The models were able to predict hardness levels with high accuracy after effective training. Overall, the neural network showed the best results. The applied procedures and methods are fast, non-destructive and provide results with acceptable measurement error, which allows their use in the production environment. Further improvements will be sought in the future, e. g. by applying a larger amount of training data, by changing the features used in the training and by increasing the measurement accuracy when capturing the Barkhausen signal.