Damage assessment of prestressed concrete beams using artificial neural network (ANN) approach

An artificial neural network (ANN) based approach is presented for the assessment of damage in prestressed concrete beams from natural frequency measurements. The details of an experimental programme suitably designed and carried out to induce the desired extents of damages in the prestressed concrete beams and generate the training and test data for the ANN are presented. The analysis of the static and dynamic behavior of perfect and damaged prestressed concrete beams reveal that there exists a close relationship among the natural frequency, deflection, crack width, first crack load, ultimate load and degree of damage. Therefore, these parameters were mainly used as input data for training and testing the ANN. A feed forward ANN learning by back propagation algorithm implemented using MATLAB has been employed in this study. The main focus of this work has been to study the feasibility of using an ANN trained with only natural frequency data to assess the damage in prestressed concrete beams. This is explored by comparing the performance of an ANN trained only with natural frequency data with other ANNs trained with a mix of static and dynamic data. It has been demonstrated that an ANN trained only with dynamic data can assess the damage with less than 10% error, when the error is the difference between the actual damage in percent and predicted damage in percent. The shortcomings of this study have also been presented.