Deep learning for computer vision has achieved remarkable results based on massive, diverse, and well-annotated training sets. However, it is difficult to collect defect datasets that cover all possible features, especially for small, weak defects. Therefore, in this article, a defect image generation method with controllable defect regions and strength is proposed. Regarded as image inpainting that uses a generative adversarial network, generated defect regions are controlled by using defect masks. Moreover, the defect direction vector is constructed in the latent variable space based on the feature continuity between defects and nondefects to control the defect strength, which enables a one-to-many correspondence between defect masks and images. Moreover, a defect attention loss is also designed to force the generation model to focus on the defect regions. Experimentally, our method yields generated images of better quality and diversity and thus significantly improves defect segmentation performance (an intersection over union of 63.20% and 61.86% on the Kolektor surface-defect and the metal hook defect datasets, respectively), especially for small, weak defects.

In industrial quality inspection, large amounts of data on the desired product appearance are available at the time of training, while significantly few defective samples are available. In this study, we proposed new memory-augmented adversarial autoencoders to detect and localize defects in real-time using defect-free samples alone for model training. This research was conducted by reconstructing images using an adversarial autoencoder and detection results from the Fréchet Markov distance (FMD). A threshold was determined based on the statistical characteristics of defect-free samples in the training set. Innovatively, we introduced a memory module and redesigned the reconstruction loss function to avoid the situation where the reconstruction ability is too strong or too poor, which lead to missing detection of defects. Then we proposed FMD, which can accurately measure the distance between the distribution of test samples and positive samples. Moreover, the statistics-based threshold determination method is used to meet different industrial needs. The accuracy, robustness, and computational overhead of the proposed MAA were evaluated using three datasets obtained from the production line and two benchmark datasets. The results indicated the effectiveness and ability of the proposed method to adapt to the real-time nature of industrial production.

Surface defect segmentation is a critical task in industrial quality control. Existing neural network architectures often face challenges in providing both real-time performance and high accuracy, limiting their practical applicability in time-sensitive, resource-constrained industrial setting. To bridge this gap, we introduce A-Net, an A-shape lightweight neural network specifically designed for real-time surface defect segmentation. Initially, A-Net introduces a pioneering A-shaped architecture tailored to efficiently handle both low-level details and high-level semantic information. Secondly, a series of lightweight feature extraction blocks are designed, explicitly engineered to meet the stringent demands of industrial defect segmentation. Finally, rigorous evaluations across multiple industry-standard benchmarks demonstrate A-Net’s exceptional efficiency and high performance. Compared to the well-estabilished U-Net, A-Net achieves comparable or superior intersection over union (IoU) scores with gains of −0.21%, −0.3%, +4.7%, and +5.94% on NEU-seg, DAGM-seg, MCSD-seg, and MT dataset, respectively. Remarkably, A-Net does so with only 0.39M parameters, a 98.8% reduction, and 0.44G floating point operations (FLOPs), a 99% decrease in computational load. Besides, A-Net shows extremely fast inference speed on edge device without GPU because of its low FLOPs. A-Net contributes to the development of effective and efficient defect segmentation networks, suitable for real-world industrial applications with limited resources.

In the tool orientation planning for five-axis sculptured surface machining, the geometrical constraints are usually considered. Actually, the effect of nongeometrical constraints on tool orientation planning is also important. This paper studied one nongeometrical constraint which was cutting force induced static deflection under different tool orientations, and proposed a cutter deflection model based on that. In the study of the cutting force, the undeformed chip thickness in filleted end milling was modeled by geometrical analysis and coordinate transformation of points at the cutting edge. In study of static flexibility of multi-axis machine, static flexibility of the entire machining system was taken into consideration. The multi-axis machining system was divided into the transmission axes-handle (AH) end and the cutting tool end. The equivalent shank method was developed to calculate the static flexibility of the AH end. In this method, static flexibility anisotropy of the AH end was considered, and the equivalent lengths of the AH end were obtained from calibration experiments. In cutter deflection modeling, force manipulability ellipsoid (FME) was applied to analyze the static flexibility of the AH end in arbitrary directions. Based on the synthetic static flexibility and average cutting force, cutter deflections were derived and estimated through developing program realization. The predicted results were compared with the experimental data obtained by machining 300 M steel curved surface workpiece, and a good agreement was shown, which indicated the effectiveness of the cutter deflection model. Additional experiments of machining flat workpiece were performed, and the relationship of cutter deflections and tool orientations were revealed directly. This work could be further employed to optimize tool orientations for suppressing the surface errors due to cutter deflections and achieving higher machining accuracy.

Surface defect detection based on computer vision remains a challenging task due to the uneven illumination, low contrast and miscellaneous patterns of defects. Current methods usually present undesirable detection accuracy and lack adaptability for the various scenes. In the paper, the novel uneven illumination surface defects inspection (UISDI) method is proposed to address these issues. First, the multi-scale saliency detection (MSSD) method is proposed to construct a coarse defect map and obtain the corresponding background regions. Second, a novel background similarity prior-based intrinsic image decomposition model (BSIID) is applied to divide the defect image into a non-defective shading layer and a defective reflectance layer. An accelerated optimization solution is proposed to solve the minimization problem of the intrinsic image decomposition model. Last, the enhanced defect image is obtained by filtering the reflectance image and is then utilized to accurately segment the defect region from the coarse defect map. The experiments conducted using four real-world defect datasets demonstrate that the proposed method outperforms state-of-the-art methods.