Geun-Seo Park, Min-Seok Hur, Tong-Seop Kim, Dong-Hyun Kim, Il-Young Jung
{"title":"基于CFD和人工神经网络的蜂窝迷宫密封性能预测","authors":"Geun-Seo Park, Min-Seok Hur, Tong-Seop Kim, Dong-Hyun Kim, Il-Young Jung","doi":"10.5293/kfma.2023.26.5.061","DOIUrl":null,"url":null,"abstract":"There is a growing interest in sealing technology, as the demand for gas turbine performance improvements increases. The honeycomb labyrinth seal is the most popular sealing technology, but it takes a lot of time to exactly predict its leakage performance considering various geometrical parameters and operating conditions. This study investigated a method to reduce computational costs involved in predicting the performance of honeycomb labyrinth seals using artificial neural networks(ANN) and computational fluid dynamics(CFD). Firstly, the central composite design, one of the design of experiment(DOE), was used to analyze the effects of various geometrical parameters on the leakage performance. The influences of geometric parameters were comparatively analyzed using a Pareto chart, and it was confirmed that clearance, tooth width, pitch, and honeycomb cell diameter were statistically significant(i.e. influential) parameters. Then, CFD simulation was performed using the combination of the selected geometric parameters and operating conditions, generating the database for the ANN to train. The high accuracy of the ANN’s prediction was confirmed by comparing its results with CFD simulations using mean squared error(MSE) and root mean squared error(RMSE). The MSE and RMSE values for the training data within the generated database were 1.475 × 10SUP-5/SUP and 0.003841, respectively. For the new unseen data, the MSE and RMSE values determined to be were 0.00021 and 0.01452 respectively.","PeriodicalId":491641,"journal":{"name":"한국유체기계학회 논문집","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Prediction of Honeycomb Labyrinth Seal Performance Using CFD and Artificial Neural Network\",\"authors\":\"Geun-Seo Park, Min-Seok Hur, Tong-Seop Kim, Dong-Hyun Kim, Il-Young Jung\",\"doi\":\"10.5293/kfma.2023.26.5.061\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"There is a growing interest in sealing technology, as the demand for gas turbine performance improvements increases. The honeycomb labyrinth seal is the most popular sealing technology, but it takes a lot of time to exactly predict its leakage performance considering various geometrical parameters and operating conditions. This study investigated a method to reduce computational costs involved in predicting the performance of honeycomb labyrinth seals using artificial neural networks(ANN) and computational fluid dynamics(CFD). Firstly, the central composite design, one of the design of experiment(DOE), was used to analyze the effects of various geometrical parameters on the leakage performance. The influences of geometric parameters were comparatively analyzed using a Pareto chart, and it was confirmed that clearance, tooth width, pitch, and honeycomb cell diameter were statistically significant(i.e. influential) parameters. Then, CFD simulation was performed using the combination of the selected geometric parameters and operating conditions, generating the database for the ANN to train. The high accuracy of the ANN’s prediction was confirmed by comparing its results with CFD simulations using mean squared error(MSE) and root mean squared error(RMSE). The MSE and RMSE values for the training data within the generated database were 1.475 × 10SUP-5/SUP and 0.003841, respectively. For the new unseen data, the MSE and RMSE values determined to be were 0.00021 and 0.01452 respectively.\",\"PeriodicalId\":491641,\"journal\":{\"name\":\"한국유체기계학회 논문집\",\"volume\":\"1 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-10-31\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"한국유체기계학회 논문집\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.5293/kfma.2023.26.5.061\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"한국유체기계학회 논문집","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.5293/kfma.2023.26.5.061","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Prediction of Honeycomb Labyrinth Seal Performance Using CFD and Artificial Neural Network
There is a growing interest in sealing technology, as the demand for gas turbine performance improvements increases. The honeycomb labyrinth seal is the most popular sealing technology, but it takes a lot of time to exactly predict its leakage performance considering various geometrical parameters and operating conditions. This study investigated a method to reduce computational costs involved in predicting the performance of honeycomb labyrinth seals using artificial neural networks(ANN) and computational fluid dynamics(CFD). Firstly, the central composite design, one of the design of experiment(DOE), was used to analyze the effects of various geometrical parameters on the leakage performance. The influences of geometric parameters were comparatively analyzed using a Pareto chart, and it was confirmed that clearance, tooth width, pitch, and honeycomb cell diameter were statistically significant(i.e. influential) parameters. Then, CFD simulation was performed using the combination of the selected geometric parameters and operating conditions, generating the database for the ANN to train. The high accuracy of the ANN’s prediction was confirmed by comparing its results with CFD simulations using mean squared error(MSE) and root mean squared error(RMSE). The MSE and RMSE values for the training data within the generated database were 1.475 × 10SUP-5/SUP and 0.003841, respectively. For the new unseen data, the MSE and RMSE values determined to be were 0.00021 and 0.01452 respectively.