No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Morphology identification of dendrites of laser cladding layer based on deep learning
... Laser powder bed fusion (LPBF) technology, as an essential category of multiple additive manufacturing technologies, is able to produce polymeric composite materials economically and with obvious flexibility [1,2]. This technique is widely used in aerospace [3,4], automotive manufacturing [5,6], optical components [7][8][9], and other industrial fields. The technique of LPBF uses a high-energy intensity laser beam to selectively melt the powder, which can fabricate more complex and superior structures due to its smaller laser spot and thinner layer thickness [10][11][12]. ...
Polymeric composites such as Poly-ether-ether-ketone (PEEK)/carbon fiber (CF) have been widely utilized due to outstanding performances such as high specific strength and specific modulus. The PEEK/CF components via powder bed fusion additive manufacturing usually show brittle fracture behaviors induced by their poor interfacial affinity and inner voids. These defects are strongly associated with powder packing quality upon deposition. The particle dynamic model has been widely employed to study the interactions of particle motions. Powder property, bulk material property, and interfacial features of composite powders are key factors in the particle dynamic model. In this work, an efficient and systematic material evaluation is developed for composite powders to investigate their deposition mechanism. The discrete element method is utilized to simulate the dynamic behaviors of PEEK/CF composite powders. The powder properties, bulk material properties, and interfacial features of powders are calibrated and justified by experimental measurement, numerical simulation, and design of experiments. The particle dynamic model can explain the powder flow behaviors and interactions. The experimental and simulation AOR results show a maximal deviation of 4.89%. It reveals that the addition of short CF particles can assist the flow of PEEK powders and improve the packing quality of the composite powders. The results show an experimental improvement of 31.3% and 55.2% for PEEK/CF_30wt% and PEEK/CF_50wt%, with a simulated improvement of 27.4% and 50.2% for corresponding composite powders.
Laser cladding technology uses a high-power laser beam to melt the substrate and metal powder at high temperature to form a molten pool. Relying on the spontaneous cooling of the molten pool, a metal cladding coating is formed on the substrate to strengthen the surface properties of the substrate metal. However, the typical defects such as cracks are easy to occur during the cladding process, which greatly affects the performance and quality of the cladded layer. This paper proposes a method for the state identification of cladding and the crack detection in the laser cladding process. By monitoring the acoustic emission signal during the laser cladding process, the current cladding state such as the status of laser power, scanning speed, and powder feed rate, and the occurrence of cracks are identified. By collecting the acoustic emission signal, the method first performs the data preprocessing for signal feature components according to the characteristic parameters of the signal maximum peak value and the energy of the emission signal samples, and then a deep learning neural network is applied to extract the feature vectors based on the two major characteristics of the signal. Finally, the current cladding states are recognized and the generation of cracks are detected based on the extracted feature vector and the identification through the neural network.
Modern deep neural network training is typically based on mini-batch stochastic gradient optimization. While the use of large mini-batches increases the available computational parallelism, small batch training has been shown to provide improved generalization performance and allows a significantly smaller memory footprint, which might also be exploited to improve machine throughput. In this paper, we review common assumptions on learning rate scaling and training duration, as a basis for an experimental comparison of test performance for different mini-batch sizes. We adopt a learning rate that corresponds to a constant average weight update per gradient calculation (i.e., per unit cost of computation), and point out that this results in a variance of the weight updates that increases linearly with the mini-batch size m. The collected experimental results for the CIFAR-10, CIFAR-100 and ImageNet datasets show that increasing the mini-batch size progressively reduces the range of learning rates that provide stable convergence and acceptable test performance. On the other hand, small mini-batch sizes provide more up-to-date gradient calculations, which yields more stable and reliable training. The best performance has been consistently obtained for mini-batch sizes between and , which contrasts with recent work advocating the use of mini-batch sizes in the thousands.
A method of vehicle detection at night based on RetinaNet and optimized loss functions
- Zhang Bingli
- Haoran
- Shang
ZHANG Bingli, QIN Haoran, JIANG Shang, et al. A
method of vehicle detection at night based on RetinaNet
and optimized loss functions[J].
2DMBV image segmentation based on genetic algorithm with wavelet analysis
- Zhang Haijun
- Wang Chunguang
- Gaixia
ZHANG Haijun, WANG Chunguang, ZHAI Gaixia.
2DMBV image segmentation based on genetic algorithm
with wavelet analysis[J]. Techniques of Automation and
Applications，2008，27(4)：15-18.
Microstructure and properties of iron-based alloys coatings prepared by high-speed laser cladding
- Xu Yifei
- Sun Yaoning
- Wang Guojian
XU Yifei, SUN Yaoning, WANG Guojian, et al. Microstructure and properties of iron-based alloys coatings prepared by high-speed laser cladding[J].
Semantic segmentation algorithm based on improved MobileNetV2
- Meng Lu
- Xu Lei
- Jiayang Guo
MENG Lu, XU Lei, GUO Jiayang. Semantic segmentation algorithm based on improved MobileNetV2[J].
Reconstruction for Cherenkov-excited luminescence scanned tomography based on unet network
- Zhang Wenqian
- Jinchao
- Li Zhe
ZHANG Wenqian, FENG Jinchao, LI Zhe, et al. Reconstruction for Cherenkov-excited luminescence scanned
tomography based on unet network[J].
- Park J Woo S
- Lee J Y
WOO S, PARK J, LEE J Y, et al. CBAM: convolutional
block attention module[C]//Computer Vision -ECCV,
Berlin, German: Springer, 2018, 2018: 3-19.