No abstract

Machine vision plays an increasingly important role in industrial product quality detection. During processing, scratches, dents and other defects are inevitable on the surface of a smooth part. Although surface defects do not affect the overall performance of the product, their existence is unacceptable when a perfect product is required. The surface defect detection method based on machine vision and deep convolutional neural networks overcomes, to a certain extent, the problem of low detection efficiency, high false detection and missing detection rates in the traditional detection method. In this paper, a multistream semantic segmentation neural network is proposed to identify defects on smooth parts. Taking a seatbelt buckle as an example, the scratch and crush defects on the surface are classified. The network takes DeepLabV3+ as the framework and three types of image stream as the input of the network. In the backbone feature extraction network, the Xception structure is improved to MobilenetV2 and the convolutional block attention module (CBAM) is introduced into the decoding network, which improves the operational efficiency and accuracy. Compared with other classical networks, this network demonstrates good performance in the image dataset of the seatbelt buckle and realises fast and accurate semantic segmentation and classification of surface defects. The evaluation results of the network model have been significantly improved.

To study the mechanism of workpiece warpage under air-bending forming, a method combined experiment with numerical simulation is proposed in this paper. Based on Hill's yielding criterion and plane strain condition, 3D ABAQUS finite element models, used to form the semiellipse-shaped workpiece with super length and large opening of sheet metal, are established. The multistep incremental air-bending forming and nonuniform springback processes are simulated. It is pointed out that a large bending force and bed deformation of press brake can lead to warpage defect of workpiece. A technique used reverse deflection compensation for bed deformation of press brake, and female die with a large opening is found to eliminate warpage defect of workpiece. Manifested by the experiment for incremental air-bending forming of this workpiece, the proposed method yields satisfactory performance in improving forming accuracy and eliminating warpage defect of a large workpiece.

To conduct the ultrasonic weld inspection of polyethylene pipes, it is necessary to use low-frequency transducers due to the high sound energy attenuation of polyethylene. However, one of the challenges in this process is that the blind zone of the ultrasonic transducer may cover a part of the workpiece being tested. This leads to a situation where if a defect appears near the surface of the workpiece, its signal will be buried by the blind zone signal. This hinders the early identification of defects, which is not favorable in such a scenario. To address this issue, we propose a new approach to detect and locate the near-surface defects. We begin by performing a synchro-squeezing transform on the original A-scan signal to obtain an accurate time-frequency distribution. While successful in detecting and localizing near-surface defects, the method alone fails to identify the specific type of defect directly: a limitation shared with other signal processing methods. Thus, an effective and lightweight defect identification model was established that combines depth-wise separable convolution and an attention mechanism. Finally, the performance of the proposed model was compared and visually analyzed with other models. This paper successfully achieves the detection, localization, and identification of near-surface defects through the synchro-squeezing transform and the defect identification model. The results show that our model can identify both general and near-surface defects with an accuracy of 99.50% while having a model size of only 1.14 MB.