Novi, učinkoviti iteracijski postupak proračuna konstrukcija - poopćenje suvremenih postupaka

Opis projekta
The idea of the project is to develop a fast iterative method for structural analysis which is, in the case of equal memory usage, faster and for larger memory usage, much faster than modern direct and iterative solvers. The procedure that we propose has additional importance, because the majority of modern iterative methods can be represented as a special case of our procedure. Therefore, it can be considered as a generalization of most modern iterative methods. In brief, it is actually an iterative procedure in which at each iteration step, discrete Ritz method is applied. At every step, coordinate vectors which form a subspace are generated, within which local energy minimum is sought, thus decreasing the total energy of the system, which therefore converges to the required minimum. The number of coordinate vectors (subspace dimension) is not limited, but aims to be small - much smaller than the number of unknowns. How to generate a good subspace - coordinate vectors is the central problem the project needs to address. Basically, with the method proposed it is possible to combine the good properties of several iterative methods simultaneously. Furthermore, every new vector generation procedure opens the possibility to fasten our method. Additionally, the property of conjugacy which underlies some iterative procedures and which is valid only in linear problems, in our case is not absolutely necessary. Therefore, the method can be successfully applied in nonlinear calculations in which the conjugacy is not even defined. This fact is really important, because in the case of nonlinear practical applications, iterative methods are used exclusively. Finally, applying our original method based on integer arithmetic, the exact solution of appropriate, practical, realistic (not just simple, benchmark) examples can be determined, so we can evaluate the convergence, stability and accuracy of any numerical procedure, including the method proposed.

Akronim
YODA

Šifra projekta
IP-2014-09-2899

Poziv
2014-09

Financijer
HRZZ - Hrvatska zaklada za znanost

Detalji financiranja
 IP - Istraživački projekti

Ukupan iznos financiranja projekta
591.599,00 HRK

Trajanje projekta
01-09-2015 - 09-01-2019

Ustanova - uloga
Građevinski fakultet, Zagreb - (591.599,00 HRK) [01-09-2015 - 09-01-2019]

Osoba - Ustanova - uloga [pocetak - zavrsetak]
Damir Lazarević (MBZ: 217624, CROSBI: 16287) - nema - Voditelj projekta [01-09-2015 - 09-01-2019]