Search

Appearance defect detection of products is a demanding procedure in the manufacturing process. Existing appearance defect inspection mainly relies on manual visual inspection, which is neither efficient nor accurate enough to ensure the manufacturing quality. Thus, automatic defect inspection has become an urgent demand. A critical problem hindering extensive applications of automatic defect inspection is that there are only limited defective samples to develop classification algorithms, leading to inadequate accuracy or robustness to meet industrial requirements. This paper proposed a knowledge reuse strategy to train convolutional neural network (CNN) models to improve defect inspection accuracy and robustness. By introducing model-based transfer learning and data augmentation, the knowledge from other vision tasks is transferred to industrial defect inspection tasks, resulting in high accuracy with limited training samples. Experimental results on an injection molding product showed that the detection accuracy was improved to about 99% when only 200 images per category were available. In comparison, conventional CNN models and the support vector machine method could achieve an average accuracy of only about 88.70% and 86.90%, respectively. The proposed method was also robust enough in detecting complicated defects which had many diversified appearances. The visualization method also proved that the performance improvement of the proposed method was because the model accurately extracted the discriminative features of the defective regions in the input images. The proposed method is meaningful for automatic defect detection in the manufacturing process.

In this paper, the thermal behavior and process optimization for polyamide 6 during the selective laser sintering process were systematically investigated via numerical simulations and experimental testing. The effects of laser power and scanning speed were studied, and the simulation results showed that the increasing laser power and decreasing scanning speed led to the increase of the maximum temperature and the three-dimensional size of molten pool. Then, the single-layer experiments under different process parameters were carried out to verify the accuracy of the model. In addition, the energy density was applied to evaluate the combined effect of laser power and scanning speed on the thermal behavior. A process map for parameter optimization was obtained based on the predicted results, and the optimized parameters were in the region that the powder can be remelt resulting in an enough depth of molten pool, while the maximum temperature was below the decomposition temperature. Meanwhile, the specimen selected in the remelting region showed the highest tensile strength.

In this brief, we propose a novel nonparametric supervised linear dimension reduction (SLDR) algorithm that extracts the features by maximizing the pairwise separation probability. The separation probability, as a new class separability measure, describes the generalization accuracy when we use the obtained features to train a linear classifier. Obtaining high-quality features, the proposed method avoids the overlaps between classes that are close to each other in the input space and improves the subsequent classification performance. Experiments on benchmark data sets show the superiority of the proposed algorithm over some other state-of-the-art SLDR methods.

An unprecedented tandem photocatalysis protocol consists of two distinct photocatalytic cycles. This method allows for rapid and efficient construction of multisubstituted benzimidazole derivatives through an intramolecular cyclization and deprotection sequence from diamines under very mild conditions. Two birds with one light: An unprecedented tandem photocatalysis, which consists of two distinct photocatalytic cycles, has been successfully developed to allow for a rapid and efficient construction of multisubstituted benzimidazole derivatives through an intramolecular cyclization/deprotection sequence from diamines under very mild conditions.

Spin-orbit coupling is predicted to have a dramatic effect on thermal properties of a two-component atomic Bose gas. We show that in three spatial dimensions it lowers the critical temperature of condensation and enhances thermal depletion of the condensate fraction. In two dimensions we show that spin-orbit coupling destroys superfluidity at any finite temperature, modifying dramatically the cerebrated Berezinskii-Kosterlitz-Thouless scenario. We explain this by the increase of the number of low-energy states induced by spin-orbit coupling, enhancing the role of quantum fluctuations. [ABSTRACT FROM AUTHOR]

Numerical simulation of flow field and flow noise was carried out on SUBOFF with Lighthill acoustical analogy and large eddy simulation method. The results agreed well with the numerical and experimental results obtained from published literature. The validation of the computational method of flow noise used in this paper was carried out. The geometric model was established for the unmanned underwater vehicle with full appendages and the flow field’s computational region discretization was carried out by using structural grid. Numerical simulation of flow noise for unmanned underwater vehicle with full appendages was taken. The numerical results show that the angle of attack has little effect on the flow noise level of the vehicle while the angle ranged from 0° to 4°. Compared with the “＋”-type rudders, the flow noise level of the underwater vehicle with “X”-type rudders is about 3dB lower.

This paper proposes a novel constructive training algorithm for cascade neural networks. By reformulating the cascade neural network as a linear-in-the-parameters model, we use the orthogonal least squares (OLS) method to derive a novel objective function for training new hidden units. With this objective function, the sum of squared errors (SSE) of the network can be maximally reduced after each new hidden unit is added, thus leading to a network with less hidden units and better generalization performance. Furthermore, the proposed algorithm considers both the input weights training and output weights training in an integrated framework, which greatly simplifies the training of output weights. The effectiveness of the proposed algorithm is demonstrated by simulation results. [ABSTRACT FROM PUBLISHER]