深度学习识别光网络单元故障的设计与应用

为解决依赖装维上门鉴别光网络单元故障带来的不便,可以从机器视觉入手实现自动化故障识别。近年,ImageNet挑战赛的成功推动了物体识别技术的跨越式发展,特别是基于卷积的深度学习技术在视觉识别方面已经达到人类水平,为光网络单元故障的自动识别提供了技术基础。文章对识别光网络单元的工作状态进行了研究,将设备工作状态分为7个场景,提出了利用手机APP采集图片识别故障的解决方案并投入了实际生产;重点阐述了深度学习模块的设计与实现,提出一种通过算法整合的方式综合运用物体检测和图像分类算法,分3阶段逐步求精,解决了图片过滤,光网络单元型号和状态识别等问题,实现了基于计算机视觉自动识别光网络单元故障。从数据上看产品的端到端准确率超过84%,识别速度达到10 FPS,月均提供服务超过1万人次,在减少用户等待的同时节约了人力资源。     

Effects of Cooling Rate and Component on the Microstructure and Mechanical Properties of Mg–Zn–Y Alloys

The microstructure and mechanical properties of Mg–6Zn–1Y and Mg–6Zn–3Y(wt%) alloys under different cooling rates were investigated. The results show that the second dendrite arm spacing(SDAS) of Mg–6Zn–1Y and Mg–6Zn–3Y is reduced by 32 and 30% with increasing cooling rates(Rc) from 10.2 to 23 K/s, which can be predicted using a empirical model of SDAS=68 R 0:45:45cand SDAS=73 R 0c, respectively. The compressive strength of both alloys increases with increasing the cooling rate, which is attributed to the increase of volume fraction(Vf) of secondary phases under high cooling rate. The interaction of the cooling rate and component with SDAS has been theoretically analyzed using interdependence theory.     

基于特征融合的中文简历解析方法研究

针对基于规则和统计的传统中文简历解析方法效率低、成本高、泛化能力差的缺点，提出一种基于特征融合的中文简历解析方法，即级联Word2Vec生成的词向量和用BLSTM（Bidirectional Long Short-Term Memory）建模字序列生成的词向量，然后再结合BLSTM和CRF（Conditional Random Fields）对中文简历进行解析（BLSTM-CRF）。为了提高中文简历解析的效率，级联包含字序列信息的词向量和用Word2Vec生成的词向量，融合成一个新的词向量表示；再由BLSTM强大的学习能力融合词的上下文信息，输出所有可能标签序列的分值给CRF层；再由CRF引入标签之间约束关系求解最优序列。利用梯度下降算法训练神经网络，使用预先训练的词向量和Dropout优化神经网络，最终完成对中文简历的解析工作。实验结果表明，所提的特征融合方法优于传统的简历解析方法。     

A dense and compact laser cladding layer with solid metallurgical bonding significantly improved corrosion resistance of Mg alloy for engineering applications

Although Mg alloy attracts great attention for engineering applications because of high specific strength and low density, low corrosion resistance limits its extensive use. In this study, Mg–Al–Zn–Mn alloy was treated via a laser cladding process to generate a dense and compact laser cladding layer with solid metallurgical bonding on the substrate for improving corrosion resistance, effectively hindering the corrosion pervasion into Mg alloy. The corrosion current density declined from 103 μA/cm² for Mg alloy to 13 μA/cm² for the laser cladding layer in NaCl aqueous solution. Moreover, the laser cladding layer was slightly corroded in comparison with Mg alloy in NaCl aqueous solution. Besides, the microhardness of the cladding layer reached a mean value of 170.5 HV, 3.1 times of Mg alloy (56.8 HV) due to the in situ formation of hardening intermetallic phases. Wear resistance of laser cladding layer was also obviously improved. These results demonstrated that the laser cladding layer obviously enhanced anticorrosion property of Mg alloy for engineering applications.     

Formation of a protective layer against corrosion on Mg alloy via alkali pretreatment followed by vanillic acid treatment

Although Mg alloy possesses high specific strength, low density, and good biocompatibility, poor corrosion resistance hinders its further applications. In the present study, an innovative protective layer against corrosion was prepared on the AZ31 Mg alloy via alkali pretreatment followed by vanillic acid treatment. The alkali pretreatment supplied –OH for the AZ31 Mg alloy surface to react with vanillic acid. The vanillic acid treatment played a crucial role in enhancing the corrosion resistance due to the excellent ability to act as a barrier and retard aqueous solution penetration, which effectively isolated the underlying Mg alloy from the corrosive environment. The corrosion current density of alkali and vanillic acid-treated Mg alloy (AZ31V) almost showed two orders of magnitude lower values in comparison with that of the AZ31 Mg alloy, and the corrosion potential of AZ31V Mg alloy increased from −1.41 to −1.25 V. The immersion tests proved that there was no occurrence of severe corrosion. Hence, the alkali pretreatment and vanillic acid treatment may represent a promising method to improve the corrosion resistance of Mg alloy.     

Automatic breast tumor detection in ABVS images based on convolutional neural network and superpixel patterns

Breast cancer is one of the most common female malignancies, as well as the second leading cause of mortality for women. Early detection and treatment can dramatically decrease the mortality rate. Recently, automated breast volume scanner (ABVS) has become one of the most frequently used diagnose methods for breast tumor screening because of its operator-independent and reproducible advantages. However, it is a challenging job to obtain the tumors’ accurate locations and shapes by reviewing hundreds of ABVS slices. In this paper, a novel computer-aided detection (CADe) system is developed to reduce clinicians’ reading time and improve the efficiency. The CADe system mainly contains three parts: tumor candidate acquisition, false-positive reduction and tumor segmentation. Firstly, a local phase-based approach is built to obtain breast tumor candidates for further recognition. Subsequently, a convolutional neural network (CNN) is applied to reduce false positives (FPs). The introduction of CNN can help to avoid complicated feature extraction as well as elevate the accuracy and efficiency. Finally, superpixel-based segmentation is used to outline the breast tumor. Here, superpixel-based local binary pattern (SLBP) is proposed to assist the segmentation, which improves the performance. The methods were evaluated on a clinical ABVS dataset whose abnormal cases were manually labeled by an experienced radiologist. The experiment results were mainly composed of two parts. At the FP reduction stage, the proposed CNN achieved 100% and 78.12% sensitivity with FPs/case of 2.16 and 0. At the segmentation stage, our SLBP obtained 82.34% true positive, 15.79% false positive and 83.59% Dice similarity. In summary, the proposed CADe system demonstrated promising potential to detect and outline breast tumors in ABVS images.