Hareket denkleminin nümerik integrasyonunda çift hassasiyet tekniği uygulayarak yuvarlama hatalarının azaltılması


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Akdeniz Üniversitesi, Mühendislik Fakültesi, Bilgisayar Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2022

Tezin Dili: İngilizce

Öğrenci: BURÇAK YEŞİLIRMAK

Danışman: Murat Ak

Özet:

Calculation of numerical integration is realized with thousands of floating-point numbers and round-off errors accumulate as the number of steps increases in the integration (Goldberg 1991). The aim of this study is to reduce the round-off errors accumulated in the numerical integration of Newtonian equation of motion using pair-precision technique. Pair-precision technique is frequently used in studies of the stability of the Solar System for the purpose of investigating dynamical processes with more precise results (Quinn and Tremaine 1989; Quinlan 1993; Rein and Spiegel 2014). Through this technique a 105-bit precision integration was performed using 53-bit double precision variables. The obtained results were compared with the results of double precision test integration using two-body orbital constants and a gain of 8 digits has been achieved. Consequently local integration errors were reduced 108 times despite by a factor of 3.04 CPU time cost of the technique.