{"title":"Scalable Domain Decomposition for Large-Scale Multibody-3D Finite Element Structures","authors":"Ravi T. Lumba, A. Datta","doi":"10.2514/6.2021-1078","DOIUrl":null,"url":null,"abstract":"A mesh partitioner was developed for large-scale 3D rotor structural dynamic models containing higher-order 3D finite elements and multibody joints. A parallel iterative substructuring solver was built to operate on the partitioned data structures. The algorithmic scalability of the solver was demonstrated on three sample problems of progressively increasing complexity; an elementary uniform beam, a real rotor blade - the NASA TRAM (1/4 scale V-22 model blade), and the full TRAM blade and hub assembly containing four flexible parts connected using six joints. The key conclusion is that a multibody blade and hub assembly can be partitioned and solved in a parallel and scalable manner. Rotor structures of up to 6.6 million degrees of freedom are solved on up to 2048 processors with detailed studies of scalability and efficiency. The uniform cantilevered beam is able to achieve a speedup of 1000 times, while the realistic rotor is able to achieve a speedup of over an order of magnitude, reducing the computational time to a matter of seconds. The use of the specialized mesh partitioner, with a robust corner node selection strategy, is key to minimizing computational time.","PeriodicalId":165313,"journal":{"name":"AIAA Scitech 2021 Forum","volume":"52 6 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"AIAA Scitech 2021 Forum","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.2514/6.2021-1078","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1
Abstract
A mesh partitioner was developed for large-scale 3D rotor structural dynamic models containing higher-order 3D finite elements and multibody joints. A parallel iterative substructuring solver was built to operate on the partitioned data structures. The algorithmic scalability of the solver was demonstrated on three sample problems of progressively increasing complexity; an elementary uniform beam, a real rotor blade - the NASA TRAM (1/4 scale V-22 model blade), and the full TRAM blade and hub assembly containing four flexible parts connected using six joints. The key conclusion is that a multibody blade and hub assembly can be partitioned and solved in a parallel and scalable manner. Rotor structures of up to 6.6 million degrees of freedom are solved on up to 2048 processors with detailed studies of scalability and efficiency. The uniform cantilevered beam is able to achieve a speedup of 1000 times, while the realistic rotor is able to achieve a speedup of over an order of magnitude, reducing the computational time to a matter of seconds. The use of the specialized mesh partitioner, with a robust corner node selection strategy, is key to minimizing computational time.