A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
To perform the geometric and structural analyses of variable geometric trusses (VGT) simultaneously and to analyze the coupling behaviour between actuator responses and member forces, new fine-grained parallel algorithms have been proposed and implemented to an multiple instruction stream, multiple data stream (MIMD)-type parallel computer, nCUBE2. One processor is assigned to one truss node, and the integrated geometric/structural analysis is carried out by using a local rule and message passing with each other node. The proposed synchronous and asynchronous parallel algorithms are found to be effective, whereas the asynchronous one shows a better efficiency, although the simultaneity of the coordinate data sent from the neighbouring nodes is not assured in the asynchronous algorithm. Further, the asynchronous algorithm is efficient even in the case where a considerable unbalance of processor load exists. The proposed method is capable of treating the nonlinear interactions between actuator response and member force, and thus the integration of the geometric and structural analyses is performed. The method is also effective for indeterminate VGTs. The computation time of integrated analysis was about 3 h for a 30-node variable geometry truss using 30 processors. It is found that this type of local-rule-based approach is capable of treating any nonlinear behaviours, and parallel processing is inevitable.
To perform the geometric and structural analyses of variable geometric trusses (VGT) simultaneously and to analyze the coupling behaviour between actuator responses and member forces, new fine-grained parallel algorithms have been proposed and implemented to an multiple instruction stream, multiple data stream (MIMD)-type parallel computer, nCUBE2. One processor is assigned to one truss node, and the integrated geometric/structural analysis is carried out by using a local rule and message passing with each other node. The proposed synchronous and asynchronous parallel algorithms are found to be effective, whereas the asynchronous one shows a better efficiency, although the simultaneity of the coordinate data sent from the neighbouring nodes is not assured in the asynchronous algorithm. Further, the asynchronous algorithm is efficient even in the case where a considerable unbalance of processor load exists. The proposed method is capable of treating the nonlinear interactions between actuator response and member force, and thus the integration of the geometric and structural analyses is performed. The method is also effective for indeterminate VGTs. The computation time of integrated analysis was about 3 h for a 30-node variable geometry truss using 30 processors. It is found that this type of local-rule-based approach is capable of treating any nonlinear behaviours, and parallel processing is inevitable.
Parallel computing for analysis of variable geometry trusses
Parallelrechnerbetrieb zur Errechnung von Fachwerkträgern mit variabler Geometrie
AIAA Journal ; 34 ; 1468-1473
1996
6 Seiten, 13 Bilder, 15 Quellen
Article (Journal)
English
Fachwerkträger , geometrische Form , Variable , Algorithmus , Parallelbetrieb , Datenverarbeitungsanlage , Strukturanalyse , Stellantrieb , Schwingungsverhalten , Knotenpunkt , Bewegungsgleichung , Lösung (Ergebnis) , Simulationsmodell , Rechenzeit , Software , Iteration , Messwertverarbeitung , Konvergenz , Dehnung , Nichtlinearität
An Object Oriented Modeling Approach to Intelligent Variable Geometry Trusses
| British Library Online Contents | 1997
Trusses Optimisation Under Multiple Variable Load
| ASCE | 2000
Trusses Optimisation Under Multiple Variable Load
| British Library Conference Proceedings | 2000
New method of computing stresses in trusses
| Engineering Index Backfile | 1891