The invention belongs to the field of
casting defect detection, and discloses a
casting defect identification method based on a
convolutional neural network. The method comprises the steps that (a) X-
ray images of a plurality of defective castings are collected, the defect types of the castings are marked, each defect type is endowed with a
type number, and a
data set corresponding to the castingX-
ray images and the defect type numbers is established; (b) constructing and training a
convolutional neural network to obtain a prediction
network model for predicting the defect type, and correcting the prediction
network model until the prediction precision of the prediction
network model meets a prediction precision threshold, thereby obtaining a final prediction network model; and (c) for the X-
ray image of the to-be-detected
casting, framing out defects in the image, and predicting by adopting a final prediction network model to obtain a defect
type number of each defect so as to complete the recognition of the defect type of the to-be-detected casting. According to the invention, the defect identification efficiency and accuracy are improved, and
digital data support is provided for casting quality feedback.