Reduction of computing time by genetic algorithm for the nondestructive technique to detect a crack in a large scale concrete structure

As described in this paper, we present a method for estimating the position of a crack in a large concrete structure using several accelerometers. An accelerometer array is installed in the concrete structure and a low-frequency vibration is produced with a small impulse hammer. A reflection wave is thereby generated from any crack that exists in the structure. Because the concrete structure is elastic, it has three wave-propagation modes: the surface-wave mode, the primary-wave mode, and the secondary-wave mode. It is difficult to estimate the position precisely because the power of thenecessaryprimary-wavemodeisweakerthanthatofsurface-wavemode. Toestimatethecrackpositionprecisely,we have proposed a method for eliminating the unwanted surface-wave and side-wall reflections, in which five parameters are used to estimate a single unwanted surface-wave or a side-wall reflection using the least mean squares technique. Because of the long time necessary to estimate even a single unwanted wave (a surface-wave or a side-wall reflection) however, the method is not practical if two or more waves mutually overlap: for instance, 10 parameters are necessary if two waves overlap. Therefore, we propose the use of genetic algorithms (GA). Results show that our method reduces the processing time remarkably. We were able to distinguish two waves reflected from the two close boundaries of a caisson.