Applied Mechanics and Materials Vols. 433-435

Abstract: Constructing the similarity matrix is the key step for spectral clustering, and its goal is to model the local neighborhood relationships between the data points. In order to evaluate the influence of similarity matrix on performance of the different spectral clustering algorithms and find the rules on how to construct an appropriate similarity matrix, a system empirical study was carried out. In the study, six recently proposed spectral clustering algorithms were selected as evaluation object, and normalized mutual information, F-measures and Rand Index were used as evaluation metrics. Then experiments were carried out on eight synthetic datasets and eleven real word datasets respectively. The experimental results show that with multiple metrics the results are more comprehensive and confident, and the comprehensive performance of locality spectral clustering algorithm is better than other five algorithms on synthetic datasets and real word datasets.

Abstract: Time-varying data widely exists anywhere in the objective world, and whose diverse distribution as time is a hybrid stochastic process. Whereas, determining how to achieve knowledge from time-varying database is still an important content of data-mining. In fact, existing techniques of data-mining are difficult to deal with this problem. The paper deeply studies and analyses finite dimension distribution function families of stochastic process and gives out existent theorem of time-varying data classification, proposes a new more effective technique called time-varying datasets classification approach based on slide-window neural networks, and gives out smoothing algorithm and convergent condition with solving this problem as well as simulation examples. The result shows the proposed method is a very effective time-varying data classification method.

Abstract: For the problem that threat assessment often has some uncertainty and the correlation exist among threat factors, a technique based on intuitionistic fuzzy sets Choquet integral is proposed with intuitionistic fuzzy sets and fuzzy integral being introduced into information fusion area. First, threat estimators based on different factors are constructed with intuitionistic fuzzy sets theory. The uncertainty of each estimator is represented with membership function and non-membership function. Then, the significances of the estimators are modeled with fuzzy measures. Subsequently, threat assessment results are obtained using Choquet integral. Finally, the proposed method is validated through the air combat threat assessment of 20 typical targets.

Abstract: To improve the accuracy of fault diagnosis, this paper reply to biological immune theory, candidate detectors have been created in non-self-space, according to biological mutation mechanism and the clone selection theory, to mature candidate detector and to reduce training time. Fault samples have classified in non-self-space, in order to reduce unclear boundaries of fault samples. The use of adaptive threshold are adapted to in the diagnosis process, to reduce the diagnosis of black holes arising, the error rate and missed rate reduce, therefore the accuracy of fault diagnosis improve. Fault test and simulation results show that the effectiveness of the method.

Abstract: Localization of sensor nodes is essential for wireless sensor network when it is applied to the special applications.This paper proposed a rang-free localization algorithm based on SVM. In this algorithm, multi-class SVMs are applied. So to improve the performance, a fast SVM algorithm is proposed in this paper.Finally, the experimental results demonstrate the algorithm proposed has small localization error, and it is robust and stable.

Abstract: Taking the lithium iron phosphate power battery as the research object, through analysis on characteristics of the battery, this paper chooses the improved second-order RC model as the model of battery whose complexity is moderate and it can better reflect the battery dynamic and static characteristics. Then by pulse discharge experiments and with improved recursive least squares algorithm to identify model parameters online, and puts forward up the adaptive kalman filtering algorithm to estimate battery SOC. The results show that the adaptive kalman filter algorithm can effectively improve battery SOC estimation precision.

Abstract: In Inertia confinement fusion (ICF) physical experiments, target positioning accuracy directly affects the success of target hitting. The proposed positioning method firstly used template matching to extract the target features in image, then calculated the target’s spatial coordinate and rotation matrix by integrating the feature values from three CCDs. We used a PI Hexapods Micro-robot to adjust the target to a desired position. The experiment results show the PI Hexapods Micro-robot is confirmed to be able to adjust the target in a desired position, which verifies the practicality of the proposed positioning algorithm to be used in the real ICF physical experiment.

Abstract: Many methods were presented to define fuzzy entropy to measure fuzzy degree of a fuzzy set and a variety of fuzzy entropy formulae were derived and constructed from the definitions of fuzzy entropy. In this paper, a new definition of fuzzy entropy is presented based on a simple order relation and computation formulae of fuzzy entropy is given. Then, the unique representations of fuzzy entropy are given by applying several set of reasonable conditions to fuzzy entropy.

Abstract: Biological cluster behavioral models rarely discuss about group cohesion issues at present, which makes biological colony susceptible to external environmental factors and leads to the group readily partitioning into multiple small groups. In order to enhance group cohesion and improve colony effect, the paper takes artificial fish as the research object and proposes a new model on artificial fish cluster behavior. This model is improved from standard self-propelled particles model, utilizing topological distance to ensure surrounding neighbors and to restrain the cognitive level of artificial fish. Meanwhile, Sphere method based on bounding box is adopted to test collision and proposes a hybrid collision processing scheme that only applies to cluster behavior, aimed at avoiding the penetration phenomenon of artificial fish. Finally, Experiments have been conducted. Results show that this model can simulate the cluster behavior of artificial fish precisely and make the group cohesion stronger. The proposed method is a feasible solution to the cluster model problem of cohesion.

Abstract: For fault forecast and information exchange problems in multi-device health management, the paper proposed a multi-device and multi-parameter fault forecast technology based on Internet of Things, and improved fault prediction algorithm based on Elman neural network feedback. Therefore, it can implement approximating nonlinear functions by arbitrary accuracy, and use the feedback reference to historical data. Thereby, it can provide early detection, isolation, management and forecasting for component failure warning, the initial issuance of the fault condition and ancillary component failure and other states multi-device health management. It also improves self-learning and adaptive capacity of the system, and effectively improves the robustness of the fault prediction system.