Corrosion degrades the performance of any civil structure. This work proposes a methodology based on the autocorrelation of vibration signals, a data treatment stage, and a processing stage based on one-dimension convolutional neural networks (1D-CNNs) to detect, locate, and quantify the corrosion damage. The autocorrelation method generally highlights or broadens relevant features into the vibration signals. The treatment stage splits, shuffles, and normalizes the autocorrelated data, then 1D-CNNs analyze data to compute a set of damage indicators. These indices represent the probability of damage in the structure for a particular location and a specific severity level. A truss-type bridge model is studied to test the proposed method, a truss-type bridge model located at the Autonomous University of Queretaro, Mexico, is studied. There are three levels of damage, i.e., incipient, moderate, severe, and healthy conditions. Obtained results demonstrate that the proposed method is a valuable tool since 100% of effectiveness is obtained. [ABSTRACT FROM AUTHOR]