Advances in analysis of total uncertainties in a semi‐invasive residual stress measurement method

The ability to characterise residual stress distribution accurately and over different length scales, particularly deep into an engineering part, plays a significant role in assessing structural integrity. Two most commonly used techniques to measure residual stress fields deep into engineering components include neutron diffraction (ND) and deep‐hole drilling (DHD). As the measurements depend on several physical quantities, they are susceptible to error. The error or uncertainties may turn substantial and compromise the suitability of the results. Although noninvasive, the neutron diffraction technique is neither readily available nor portable and is limited to approximately 60‐mm‐thick specimen; errors associated with results become unacceptable at greater flight paths. Moreover, a mock‐up representing the engineering component is normally used in the ND technique. In contrast, the DHD technique is portable and measures residual stresses with high spatial resolution. An error propagation technique was applied to develop an error analysis procedure taking into consideration various stages of the DHD method and successfully applied to different DHD measurements. An essential feature comprising the effect of plasticity due to the creation of reference hole in the DHD procedure has not yet been taken into account in the error analysis procedure. This paper briefly describes how the uncertainties due to the creation of the initial reference hole can be determined. The effect of plasticity in the drilling procedure is quantified in this study. This error is combined with other sources of error and formulated to determine the total error. An incremental DHD technique was used to measure the complex triaxial residual stress field in an as‐welded circular disc, and the measured data were used to illustrate the total error using the error analysis method developed in the study.     

Physics-informed deep learning for one-dimensional consolidation

Neural networks with physical governing equations as constraints have recently created a new trend in machine learning research.In this context,a review of related research is first presented and discussed.The potential offered by such physics-informed deep learning models for computations in geomechanics is demonstrated by application to one-dimensional(1D)consolidation.The governing equation for 1D problems is applied as a constraint in the deep learning model.The deep learning model relies on automatic differentiation for applying the governing equation as a constraint,based on the mathematical approximations established by the neural network.The total loss is measured as a combination of the training loss(based on analytical and model predicted solutions)and the constraint loss(a requirement to satisfy the governing equation).Two classes of problems are considered:forward and inverse problems.The forward problems demonstrate the performance of a physically constrained neural network model in predicting solutions for 1D consolidation problems.Inverse problems show prediction of the coefficient of consolidation.Terzaghi’s problem,with varying boundary conditions,is used as a numerical example and the deep learning model shows a remarkable performance in both the forward and inverse problems.While the application demonstrated here is a simple 1D consolidation problem,such a deep learning model integrated with a physical law has significant implications for use in,such as,faster realtime numerical prediction for digital twins,numerical model reproducibility and constitutive model parameter optimization.     

多级深度还原法制备Ti6Al4V合金粉体的基础研究

目前Ti6Al4V合金粉体的生产方法主要有雾化法、机械合金法和氢化脱氢法,它们都以Kroll法生产海绵钛为基础。使用钛的氧化物作为原料的熔盐电解法和金属热还原法仍处于研究阶段。本文依据变价金属Ti和V的氧化物在还原过程中逐级还原的特性,提出使用金属氧化物(TiO₂, V₂O₅)作为原料的多级深度还原法制备Ti6Al4V合金粉体的新思路。首先计算了TiO₂-V₂O₅-Mg-Ca体系的吉布斯自由能变,结果表明先进行镁热自蔓延反应,后进行钙热深度还原反应制备Ti6Al4V合金粉体在热力学上具备可行性。然后通过实验进行了的验证。镁热自蔓延反应产物酸浸后除去MgO可得到氧含量为15.84%的多孔Ti-Al-V-O前驱体。配入金属Ca后进行深度脱氧可得到低氧Ti6Al4V合金粉体(Al: 6.2wt.%, V: 3.64wt.%, O: 0.24wt.%, Mg: 0.01wt.%, Ca: < 0.01wt.%)。     

基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

Adaptive investigation of the optical properties of polymer fibers from mixing noisy phase shifting microinterferograms using deep learning algorithms

In this article, an adaptive denoising method is suggested to accurate investigate the optical and structural features of polymeric fibers from noisy phase shifting microinterferograms. The mixed class of noise that may produce in the phase-shifting interferometric techniques is established. To our knowledge, this is an early study considered the mixing noises that may occur in microinterferograms. The suggested method utilized the convolution neural networks to detect the noise class and then denoising, it according to its class. Four convolution neural networks (Googlenet, VGG-19, Alexnet, and Alexnet–SVM) are refined to perform the automatic classification process for the noise class in the established data set. The network with the highest validation and testing accuracy of these networks is considered to apply the suggested method on realistic noisy microinterferograms for polymeric fibers, polypropylene and antimicrobial polyethylene terephthalate)/titanium dioxide, recoded using interference microscope. Also, the suggested method is applied on noisy microinterferograms include crazing and nanocomposite material. The demodulated phase maps and the three-dimensional birefringence profiles are calculated for tested fibers according to the suggested method. The obtained results are compared with the published data for these fibers and found to be in good agreements.     

瓷砖表面非常规尺寸瑕疵的检测

与普通目标检测任务不同, 瓷砖表面瑕疵检测的困难之处在于检测小尺寸和大长宽比等非常规尺寸的目标. 为了解决这两个问题, 本文提出了一种基于改进 Cascade R-CNN 的新型瓷砖表面瑕疵检测算法. 为了提高模型对小瑕疵的检测能力, 本文模型利用侧向连接结构进行上下层语义信息的融合, 使用可切换空洞率的空洞卷积来增加模型的感受野; 为了提高模型对于大长宽比瑕疵的检测能力, 本文模型在标准卷积上引入偏移域以更好提取目标特征信息. 此外, 本文模型调整 Cascade R-CNN 框架中预选锚框的大小和长宽比例. 实验结果表明, 在从瓷砖工厂收集的数据集上, 本文所提出算法的平均精度均值 (mean average precision, mAP) 达到了 73.5%, 比改进前的 Cascade R-CNN 模型提高了9.7%. 本文实验代码可从以下链接获取: https://github.com/mashibin/Ceramic-tile-defect-detection.     

低共熔溶剂用于CO2捕集的研究进展

低共熔溶剂(DESs)具有原料廉价易得、化学稳定性好、可设计、合成工艺简单、可循环使用和绿色环保等优点,在CO₂捕集领域受到广泛关注。重点综述了近年来DESs用于CO₂捕集的研究进展,总结了DESs捕集CO₂的能力,分析了DESs捕集CO₂的影响因素和DESs的循环使用性能,归纳了DESs捕集CO₂的机理(包括物理吸收、化学吸收和物理化学协同吸收),并总结了CO₂在DESs中的溶解度计算模型。分析发现,DESs捕集CO₂的影响因素中,温度、压力和水含量(质量分数)均对CO₂的捕集有影响,且DESs的结构是重要的影响因素;大部分DESs可循环使用;CO₂在DESs中溶解度计算模型的建立有效推动了DESs捕集CO₂的进一步发展。最后,指出了DESs捕集分离CO₂所面临的主要问题并对进一步的研究工作进行了分析讨论。     

Multimodal Attentive Fusion Network for audio-visual event recognition

Event classification is inherently sequential and multimodal. Therefore, deep neural models need to dynamically focus on the most relevant time window and/or modality of a video. In this study, we propose the Multimodal Attentive Fusion Network (MAFnet), an architecture that can dynamically fuse visual and audio information for event recognition. Inspired by prior studies in neuroscience, we couple both modalities at different levels of visual and audio paths. Furthermore, the network dynamically highlights a modality at a given time window relevant to classify events. Experimental results in AVE (Audio-Visual Event), UCF51, and Kinetics-Sounds datasets show that the approach can effectively improve the accuracy in audio-visual event classification. Code is available at: https://github.com/numediart/MAFnet     

炼油生化剩余污泥催化深度减量技术的工业应用

大幅减少炼油生化剩余污泥对降低企业污泥处理和处置费用、保护生态环境意义重大。考察了炼油生化剩余污泥催化深度减量SMR技术工业化的处理效果，运行结果表明：对含水率98.3%的炼油生化剩余污泥，经过SMR污泥减量装置的处理，污泥脱水性能得到显著改善，脱水残渣平均含水率57.1%、污泥量减少99.4%、挥发性悬浮物平均去除率78.3%、消解液BOD₅/COD比为0.31，达到了污泥深度减量化的目的。