Abstract: The sidelobe level and beamwidth would go up when elements in array antenna are failed. The genetic algorithm is improved by keeping the optimal individuals as the new initial population. This algorithm could speed up the convergence and preserve the best chromosome. In addition, the 26-element array antenna with -40dB side lobe level and 3 failed elements was corrected by the improved genetic algorithm based on advantage retention so as to restore the sidelobe level of the original pattern. The numerical results demonstrate that the algorithm can effectively reduce the deterioration of pattern.