engleski

Reverse Engineering with Shape Optimization using Workflow-Based Computation and Distributed Computing

Shape optimization is numerically a very demanding process, where several traditionally separate numerical processes meet together in a heterogeneous multidisciplinary environment. Typical elements involved are (1) the optimizer itself that implements the selected search strategy, (2) the parameterization procedure that provides the bidirectional mapping interface between the 3D geometric shape of the object and a limited number of parameters that control the parametric model of the geometry, (3) the simulation program that provides the results of the simulated interaction between the object and its environment (such as finite element analysis, FEA), and (4) control procedures that monitor the overall workflow and invoke mutual application calls (process flows and data flows) and handle data mining as well as acceptance or rejection of individual instances of candidate designs which are numerically generated and analyzed. In this paper, the set-up of this complex process, combined with evolutionary optimization, is developed with the objectives of numerical flexibility, computational efficiency in terms of combining sequential and parallel execution of individual applications, and low-cost implementation within an unmanaged computer cluster based on the Windows Communication Foundation, WCF.

Shape optimization; distributed numerical workflow; parallelization; WCF cluster

nije evidentirano