移动智能网的业务控制点对IN/Internet互通的支持

首先说明了业务控制点在IN/Internet互通的总体结构中的地位,然后讨论了原有移动智能网业务控制点的功能结构,在此基础上提出了对业务控制点的接口、智能网应用协议和SIB（业务独立构件）这三个方面的增强方案,对于类似的工作的实施有一定的参考价值。     

LOA-RPL: Novel Energy-Efficient Routing Protocol for the Internet of Things Using Lion Optimization Algorithm to Maximize Network Lifetime

Energy conservation is a significant task in the Internet of Things (IoT) because IoT involves highly resource-constrained devices. Clustering is an effective technique for saving energy by reducing duplicate data. In a clustering protocol, the selection of a cluster head (CH) plays a key role in prolonging the lifetime of a network. However, most cluster-based protocols, including routing protocols for low-power and lossy networks (RPLs), have used fuzzy logic and probabilistic approaches to select the CH node. Consequently, early battery depletion is produced near the sink. To overcome this issue, a lion optimization algorithm (LOA) for selecting CH in RPL is proposed in this study. LOA-RPL comprises three processes: cluster formation, CH selection, and route establishment. A cluster is formed using the Euclidean distance. CH selection is performed using LOA. Route establishment is implemented using residual energy information. An extensive simulation is conducted in the network simulator ns-3 on various parameters, such as network lifetime, power consumption, packet delivery ratio (PDR), and throughput. The performance of LOA-RPL is also compared with those of RPL, fuzzy rule-based energy-efficient clustering and immune-inspired routing (FEEC-IIR), and the routing scheme for IoT that uses shuffled frog-leaping optimization algorithm (RISA-RPL). The performance evaluation metrics used in this study are network lifetime, power consumption, PDR, and throughput. The proposed LOA-RPL increases network lifetime by 20% and PDR by 5%–10% compared with RPL, FEEC-IIR, and RISA-RPL. LOA-RPL is also highly energy-efficient compared with other similar routing protocols.     

基于自适应粒子群算法和支持向量机的控制图模式识别

针对目前多品种、复杂化的生产趋势,提出了一种基于自适应变异的粒子群算法（AMPSO）和支持向量机（SVM）的控制图失效模式识别的方法。利用SVM小样本学习能力,设计一对一的SVM多分类器进行控制图模式识别,并利用AMPSO算法优化SVM核函数的参数。通过对10种控制图模式(6种基本模式和4种混合模式)的20维特征仿真数据对该方法进行检验,并通过与BP、SVM、PSO SVM识别方法的对比分析。仿真试验表明该方法有效提高了控制图模式的识别精度,达到9814%,而BP仅有75%,为控制图在线实时识别提供了一种可行的途径。      

基于随机森林的统计控制图模式识别研究

引入随机森林方法进行统计控制图模式识别的研究。提取了控制图的统计特征和形状特征,设计了5种不同的特征组合方法,利用蒙特卡洛仿真方法产生训练数据集和测试数据集,选取了常用的3种模式识别方法（支持向量机方法、人工神经网络方法、决策树方法）进行对比。实验结果表明,随机森林方法相比其他3种分类器方法,在分类准确率和消耗时间两个维度上都有明显优势,可以应用于统计过程控制图模式识别。     

新能源功率预测算法优化研究

以风能和太阳能为代表的新能源具有随机性、间歇性和波动性,对新能源发电功率进行预测是有效解决以上问题的途径。在确定性预测中充分考虑风电出力和预测模型特性,提出分段支持向量机（piecewise support vector machine,PSVM）和神经网络（neural network,NN）预测算法;充分考虑天气特征对光伏出力的影响,提出基于气象特性分析的光伏出力预测算法。通过若干风电场的算例分析,证明了上述几种预测模型的实用性,为功率预测的可靠性分析提供支持。     

Chemiresistive Sensor Arrays for Gas/Volatile Organic Compounds Monitoring: A Review

Gas detection and monitoring are essential due to their direct impact on human health, environment, and ecosystem. Chemiresistive sensors are one of the most used classes of sensors for monitoring and measurement of gases thanks to their ease of fabrication, customizability, mechanical flexibility, and fast response time. While chemiresistive sensors can offer good sensitivity and selectivity to a particular gas in a controlled environment with known interferences, they may not be able to differentiate between various gases having similar physiochemical properties under uncontrolled conditions. To address this shortcoming of chemiresistive gas sensors, sensor arrays have been the subject of recent studies. Gas sensor arrays are a group of individual gas sensors that are arranged to simultaneously detect and differentiate multiple cross-reactive gases. In this regard, various sensor array technologies have been developed to differentiate a given set of gases using multivariate algorithms. This review provides an insight into the different algorithms that are used to extract the data from the sensor arrays, highlighting the fabrication techniques used for developing the sensor array prototypes, and different applications in which these arrays are used.     

Machine learning aided nanoindentation: A review of the current state and future perspectives

The solution of instrumented indentation inverse problems by physically-based models still represents a complex challenge yet to be solved in metallurgy and materials science. In recent years, Machine Learning (ML) tools have emerged as a feasible and more efficient alternative to extract complex microstructure-property correlations from instrumented indentation data in advanced materials. On this basis, the main objective of this review article is to summarize the extent to which different ML tools have been recently employed in the analysis of both numerical and experimental data obtained by instrumented indentation testing, either using spherical or sharp indenters, particularly by nanoindentation. Also, the impact of using ML could have in better understanding the microstructure-mechanical properties-performance relationships of a wide range of materials tested at this length scale has been addressed.The analysis of the recent literature indicates that a combination of advanced nanomechanical/microstructural characterization with finite element simulation and different ML algorithms constitutes a powerful tool to bring ground-breaking innovation in materials science. These research means can be employed not only for extracting mechanical properties of both homogeneous and heterogeneous materials at multiple length scales, but also could assist in understanding how these properties change with the compositional and microstructural in-service modifications. Furthermore, they can be used for design and synthesis of novel multi-phase materials.