Identifying damage to structures is extremely important, especially in the field of engineering maintenance, and experimental modal analysis is a powerful tool for detecting damage in the field of vibration. The current study presents a method for reliable damage detection in steel structures such as cantilever beams through the use of practical data for typical analysis, and different excitation techniques were used, such as the impact hammer test and shaker test. Later, pattern recognition artificial neural networks were used to predict and classify damage in terms of location and depth in the cantilever beam by using model analysis data represented by natural frequencies and differences in the deformation of mode shape. Through the results, it was shown that the artificial neural network method of pattern recognition is a promising method for predicting damage to structures, where the accuracy of classification ratio is 100% and 98% for two models of classification (damage depth and damage location). It also provides an alternative to current methods of diagnosing and discovering damage with high speed and accuracy. It provides a new perspective for designing pattern recognition schemes in ANN in the field of automation control.