Generally, the on-line identification of holes or cracks in a structure is a pressing task of non-destructive identification. In this paper, the method used is different from that generally studied previously: The detectors of the inverse problem are the static strains simply measured by strain gauges, and the system of on-line identification is accomplished through an artificial neural network (ANN). It is more and more feasible and accurate to on-line measure the static strains by applying highly developed smart materials. To express the complex relationship between the strains and the parameters of holes or cracks, a network of ANN with two hidden layers is designed. Not only the size but also the location and orientation of a hole/crack in a composite plate can be identified on-line. The weights and thresholds in the networks can be updated based upon the well-trained values if new training data are added. Consequently, the training time will be saved. To perform the optimal learning efficiency and a ccuracy, many numerical results are provided in this paper.
All Science Journal Classification (ASJC) codes