Using computational time reversal method for localization of forming and propagating crack

0519747 - ÚT 2020 RIV CZ eng A - Abstrakt
Mračko, Michal - Kolman, Radek - Kober, Jan - Převorovský, Zdeněk - Plešek, Jiří
Using computational time reversal method for localization of forming and propagating crack.
NDT in Progress 2019. Praha: Ústav termomechaniky AV ČR, v. v. i., 2019 - (Převorovský, Z.). s. 124. ISBN 978-80-87012-72-7.
[NDT in Progress 2019 /10./. 07.10.2019-10.10.2019, Praha]
Grant CEP: GA MŠMT(CZ) EF15_003/0000493; GA ČR(CZ) GA17-22615S
Institucionální podpora: RVO:61388998
Klíčová slova: time reversal * explicit finite element analysis * elastic wave propagation * non-destructive testing * crack localization
Obor OECD: Materials engineering

The time reversal (TR) method has found its application in many fields concerning wave propagation. Our object of interest is the application in non-destructive testing (NDT). In NDT, this
method can be used for tracing the source of vibrations in solid bodies, the source being a crack or some other defect. The TR method uses a backward wave propagation for refocusing and
reconstruction of the original source. The TR process consists of two steps. In the first step – the Frontal task, a real body is loaded at the given place with the defined loading signal and an output is
recorded in a prescribed position of the body. In the second step – the Reverse task, this responding signal is reversed in time and loaded into the computational model so as to locate so called
scatterers (e.g. cracks). In computational TR method, both steps are performed numerically. Here we focus on localization of an initializing and a propagating crack in the prestressed finite element
(FE) model. We also study how the length of the computation (number of reflections of the elastic waves) influences the probability of localization of the crack. Special attention is paid to the way of
prescription of the loading signal. For numerical solution, we use the linear FE method, with the lumped mass matrix, a one-point Gauss integration rule and an hourglass control. For the direct integration in time the explicit central difference scheme is employed. This integration scheme is conditionally stable and reversible in time. We evaluate the quality of localization mainly by observing the total energy distribution at the end of the Reversal task. We compare results for several lengths of computation (between 1 000 and 50 000 time steps). The conclusions show that with increasing length of computation (more information loaded into the model) the probability of localization of the crack also increases (the energy refocuses in the location of the source of vibrations).
Trvalý link: http://hdl.handle.net/11104/0304940