The Application of Convolutional Neural Networks and Fuzzy Logic in Control and Automation Systems for Recognizing and Assessing the Danger Level of Hot Cracks in the Production of Cast Products

The use of computer technologies applying combined systems of convolutional neural networks and fuzzy logic in the analysis of casting images for detecting and classifying hot cracks represents an innovative approach in foundry production, possessing significant potential for solving quality and safety issues of products. Trained on large and diverse sets of casting images, these systems are capable of identifying hot cracks, recognizing their unique characteristics such as shape, size, and orientation. This is achieved through targeted study of patterns inherent in hot cracks and their comparison with an extensive database of images. The presence of noise, artifacts, or blurred crack boundaries in images can complicate the task of classification and analysis. Fuzzy logic is employed to allow the system to more flexibly interpret data, assessing the degree of uncertainty and ambivalence in the information.