Shenyang Institute of Automation,Chinese Academy Of Sciences
Not a member yet
23582 research outputs found
Sort by
Composition analysis of ceramic raw materials using laser-induced breakdown spectroscopy and autoencoder neural network
No full textIn the ceramic production process, the content of Si, Al, Mg, Fe, Ti and other elements in the ceramic raw materials has an important impact on the quality of the ceramic products. Exploring a method that can quickly and accurately analyze the content of key elements in ceramic raw materials is of great significance to improve the quality of ceramic products. In this work, laser-induced breakdown spectroscopy (LIBS) is used for rapid analysis of ceramic raw materials. The chemical element composition and content of ceramic raw materials are quite different, which leads to serious matrix effects. Building an artificial neural network model is an effective way to solve the complex matrix effects, but model training can easily lead to overfitting due to the high number of spectral features and the limited number of samples. In order to solve this problem, we propose a feature extraction method that combines the linear regression (LR) and the sparse and under-complete autoencoder (SUAC) neural network. This LR + SUAC method performs nonlinear feature extraction and dimension reduction on high-dimensional spectral data. The spectral data dimension is reduced from 8188 to 100 through the LR layer, and further reduced to 32 through the SUAC encoding layer. Further, a quantitative analysis model for the elemental composition of ceramic raw materials is established by the combination of LR + SUAC and Back Propagation Neural Network (BPNN). Since the input data dimension and redundant information are greatly reduced by LR + SUAC, the overfitting problem of BPNN is greatly reduced. Experiment results showed that the LR + SUAC + BPNN method obtained the best quantitative analysis performance compared with several other methods in the cross-validation process
Self-adaptive coding for spiking neural network
No full text脉冲神经网络(SNN)采用脉冲序列表征和传递信息，与传统人工神经网络相比，更具有生物可解释性。但典型SNN的特征提取能力受到其结构限制，对于图像数据等多分类任务的识别准确率不高，不能与卷积神经网络(CNN)相媲美。针对该问题，提出了一种新型的自适应编码脉冲神经网络(SCSNN)，将CNN的特征提取能力和SNN的生物可解释性结合起来，采用生物神经元动态脉冲触发特性构建网络结构，并设计了一种新的替代梯度反向传播方法直接训练网络参数。所提出的SCSNN网络分别在MNIST数据集和Fashion-MNIST数据集做了验证，取得较好的识别结果，在MNIST数据集上准确率达到了99.62%，在Fashion-MNIST数据集上准确率达到了93.52%，验证了本模型的有效性。</p
Nanosecond pulsed laser-assisted modified copper surface structure: Enhanced surface microhardness and microbial corrosion resistance
No full textMicrobiologically influenced corrosion (MIC) is an unavoidable problem in several industries. Copper (Cu) and its alloys are widely used engineering materials. However, MIC of Cu remains a persistent challenge to their performance and functional lifetime under aggressive environments. This study investigated nanosecond pulsed laser processing (LP), which may enhance the corrosion resistance of Cu. The microstructural evolution and corrosion behavior of LP-Cu in the presence of sulfate-reducing bacteria (SRB) were evaluated. Typical deformation-induced microstructural features of high-density dislocations were analyzed on the top surface of LP-Cu coupon. Electrochemical measurements suggested that LP-Cu coupons exhibited better corrosion resistance in SRB-inoculated solution compared with their original counterpart. The enhanced corrosion resistance by LP primarily resulted from the combined influences of compressive residual stress and work hardening in the surface. However, overlap percentage played a key role in improving corrosion resistance. LP produced optimal corrosion resistance at 50% overlap. Therefore, this study introduces a unique and an option for anticorrosion control in manufacturing processes and potentially implements it onto other materials to improve its microbial corrosion resistance through LP.</p
Tool-path continuity determination based on machine learning method
No full textComputer-aided manufacturing (CAM) software outputs machining data by encoding a tool-path into a series of G-codes which are composed of various lengths of line segments. The discontinuities of these line segments may cause inefficiency for computer numerical control (CNC) system. To achieve high-speed continuous motions, corner smoothing algorithms based on look-ahead methods are widely used. However, it is difficult to meet smoothing trajectories in real-time requirements. Based on machine learning method, in this paper, a support vector machine (SVM) system is presented for directly outputting classification results of the various geometric continuities at the transition corners. The feature values used for generating continuity classification model are extracted from sampling paths of the previous publication work: the machining parameters, length, fairness criteria, the root mean square (RMS) contour errors, and dominant stage type of movement of each sampling path are calculated. The acceleration/deceleration (ACC/DEC) feedrate planning scheme is used to determine the feedrate at the transition corners. Simulations and experiments show that the proposed algorithm can realize accurately and efficiently continuity classification in real-time requirements under the conditions of machining accuracy.</p
Deep learning with laser-induced breakdown spectroscopy (LIBS) for the classification of rocks based on elemental imaging
No full textIn geological research, the identification and classification of rock lithology plays an important role in many fields such as resource exploration, earth evolution and paleontology research. Laser-induced breakdown spectroscopy (LIBS), which is capable of real-time, on-situ, micro-destructive determination of the elemental composition of any substance (solid, liquid, or gas), has been developed as a technology for ‘geochemical fingerprinting’ in a variety of geochemical applications. However, for rock samples with coarse grains, the bulk analysis based on the average spectrum is insufficient. This study proposes a new method for identifying multiple types of rocks, which utilizes the rapid multi-element compositional imaging capability of LIBS, and combines with the deep learning theory. The LIBS-based images characterizing the spatial distribution of elements on rock surface were achieved firstly, and then were classified by the Inception-v3 network combined with the transfer learning method. In addition, to solve the problem of the small scale of the image dataset obtained in the laboratory, specific data augmentation methods such as cutting-recombining and filtering were proposed. Moreover, the superior of this method was verified by the three classification experiments of shale, gneiss and granite.</p
Infrared and visible image fusion using improved generative adversarial networks
No full text红外与可见光图像融合技术能够同时提供红外图像的热辐射信息和可见光图像的纹理细节信息，在智能监控、目标探测和跟踪等领域具有广泛的应用。两种图像基于不同的成像原理，如何融合各自图像的优点并保证图像不失真是融合技术的关键，传统融合算法只是叠加图像信息而忽略了图像的语义信息。针对此问题，提出一种改进的生成对抗网络，生成器设计了局部细节特征和全局语义特征两路分支捕获源图像的细节和语义信息；在判别器中引入谱归一化模块，解决传统生成对抗网络不易训练的问题，加速网络收敛；引入了感知损失，保持融合图像与源图像的结构相似性，进一步提升了融合精度。实验结果表明，本文方法在主观评价与客观指标上均优于其他代表性方法，对比基于全变分模型方法，平均梯度和空间频率分别提升了55.84%和49.95%。</p
Research on Edge Cloud Resource Pricing Mechanism Based on Stackelberg Game
No full text移动边缘计算（mobile edge computing，MEC）支持终端设备将任务或应用程序卸载到边缘云服务器处理，边缘云服务器处理外来任务会消耗本地资源，为激励边缘云提供资源服务，构建向终端设备收费以奖励边缘云的资源定价机制尤为重要。现有的定价机制依赖中间商的静态定价，费用高且终端任务处理不及时，难以实现边缘云计算资源的有效利用。针对上述问题，本文提出一种基于Stackelberg博弈的边缘云资源定价机制。首先，针对资源定价时终端设备因资金不足而导致的本地任务搁置问题，提出包含贷款和激励的辅助机制，实现终端设备任务的及时处理；其次，提出影响资源定价的4种价格导向因素，制定了一致性与弹性两种定价方案，提高定价的准确性和效率，并为动态定价做准备；然后，为了使终端设备与边缘云直接进行动态定价，构建基于斯坦克伯格（Stackelberg）博弈的资源定价机制模型，将资源需求与定价问题转化为边缘云收益最大与终端设备支付成本最小问题；最后，通过改进的强化学习State Action Reward State Action（SARSA）算法得到资源需求及定价的最优策略。实验表明，本文提出的定价机制在边缘云收益最大化方面优于其他定价算法12%以上，同时弹性定价方案下边缘云的收益优于一致性定价方案的24%。</p
Codesign of Architecture, Control, and Scheduling of Modular Cyber-Physical Production Systems for Design Space Exploration
No full textDesign space exploration (DSE) of cyber-physical production systems (CPPS) is a search problem in the space of potential compositional configurations. Current design methodologies follow the separated design paradigm in which the architecture, control, and scheduling are separately designed. Optimization of each part considers only the corresponding goals of interest and overlooks other aspects by adopting gross assumptions, which makes it difficult to determine the global optimal solution for a given system. To address this problem, in this article, we propose a codesign method that considers the design spaces of architecture, control, and scheduling as monolithic, mixed discrete-continuous spaces. We formulate DSE as an optimization problem and propose a generic iterative algorithm schema involving simulation in the loop to solve the abovementioned new problem. To illustrate the effectiveness of the proposed method, a real single-stage reducer assembly production system is considered. The design results demonstrate that our method provides better solutions than does the separated design method
A hierarchical conditional random field-based attention mechanism approach for gastric histopathology image classification
No full textIn the Gastric Histopathology Image Classification (GHIC) tasks, which are usually weakly supervised learning missions, there is inevitably redundant information in the images. Therefore, designing networks that can focus on distinguishing features has become a popular research topic. In this paper, to accomplish the tasks of GHIC superiorly and assist pathologists in clinical diagnosis, an intelligent Hierarchical Conditional Random Field based Attention Mechanism (HCRF-AM) model is proposed. The HCRF-AM model consists of an Attention Mechanism (AM) module and an Image Classification (IC) module. In the AM module, an HCRF model is built to extract attention regions. In the IC module, a Convolutional Neural Network (CNN) model is trained with the attention regions selected, and then an algorithm called Classification Probability-based Ensemble Learning is applied to obtain the image-level results from the patch-level output of the CNN. In the experiment, a classification specificity of 96.67% is achieved on a gastric histopathology dataset with 700 images. Our HCRF-AM model demonstrates high classification performance and shows its effectiveness and future potential in the GHIC field. In addition, the AM module and transfer learning technique allow the network to generalize well to other types of image data except histopathology images, and we obtain 95.5% and 95.8% accuracies on IG02 and Oxford-IIIT Pet Datasets
IL-MCAM: An interactive learning and multi-channel attention mechanism-based weakly supervised colorectal histopathology image classification approach
No full textIn recent years, colorectal cancer has become one of the most significant diseases that endanger human health. Deep learning methods are increasingly important for the classification of colorectal histopathology images. However, existing approaches focus more on end-to-end automatic classification using computers rather than human-computer interaction. In this paper, we propose an IL-MCAM framework. It is based on attention mechanisms and interactive learning. The proposed IL-MCAM framework includes two stages: automatic learning (AL) and interactivity learning (IL). In the AL stage, a multi-channel attention mechanism model containing three different attention mechanism channels and convolutional neural networks is used to extract multi-channel features for classification. In the IL stage, the proposed IL-MCAM framework continuously adds misclassified images to the training set in an interactive approach, which improves the classification ability of the MCAM model. We carried out a comparison experiment on our dataset and an extended experiment on the HE-NCT-CRC-100K dataset to verify the performance of the proposed IL-MCAM framework, achieving classification accuracies of 98.98% and 99.77%, respectively. In addition, we conducted an ablation experiment and an interchangeability experiment to verify the ability and interchangeability of the three channels. The experimental results show that the proposed IL-MCAM framework has excellent performance in the colorectal histopathological image classification tasks.</p