The present study is aimed at developing approaches to detecting and classifying defects in the blade apparatus of turbomachines and assessing the manifestation of defects on the characteristics of the installation. The relevance of the work is connected with the use of digital technologies in the problem of defectoscopy of the blade row apparatus. In the course of the study, a YOLOv8s convolutional neural network model was prepared and trained on two data sets: with artificially visualized images of defects and photographs of blades after operation. Based on the analysis of the results of training and testing of the YOLOv8s model on a database with artificial images of defects in the blade row, an accuracy of 97.8% was achieved. On the prepared real data set, YOLOv8s has an average accuracy of mAP0.5 = 0.84. The paper also describes an approach to assessing the effect of blade defects on the characteristics of turbomachines based on the results of numerical experiments and the use of machine learning methods to predict the development of defects or evaluate changes in the characteristics of turbomachines with defects. The study shows the promise of using a digital approach to detecting and evaluating the impact of blade defects on the performance of turbomachines.