Abstract: The improved genetic algorithm is proposed to identify the defects of pile detected by low-strain integrity testing. Considering the premature phases and slow rate of convergence in evening of genetic algorithms, adaptive niche technology and adaptive probabilities of crossover and mutation are both introduced into the classical simple genetic algorithm, with one dimension theory of stress wave and finite difference method to identify the defects of pile. The study shows that this method is suitable for the automatic identification of defects of pile detected by low-strain integrity testing. It can also avoid premature phases and slow rate of convergence in evening that often happen in simple genetic algorithms. Finally, a good performance is achieved.

Key words: pile, defect, automatic identification, genetic algorithm, adaptive probabilities, adaptive niche technology