If the elements in an ultrasound array are damaged, the beamforming performance and imaging quality (contrast and resolution) will be significantly reduced due to the distorted beam pattern with for example, a significant increase of sidelobe level. Defective elements can be detected through either in-direct impedance measurement or direct monitoring of the acoustic output using fibre Bragg grating (FBG) sensors. In this paper, we demonstrate that via recalculating the weights (apodization) of the remaining functional elements, a new pattern that is close to the original can be formed, thus the array failure is corrected. Since a damaged array can be considered as a non-uniformly spaced array, analytic methods are generally unable to tackle this kind of problem. A numerical approach based on genetic algorithms (GA) to reconfigure the apodization distribution of remaining elements is presented. This technique provides a cost-effective alternative to hardware replacement, and potentially produces ‘smart ultrasonic arrays’ with capability of self-restoration from failures throughout the operating lifetime when incorporating with a detection procedure.