Applied Mechanics and Materials Vols. 644-650

Abstract: In the research of Delay Tolerant Network (DTN), DTN routing algorithm is a key research issue. The performance of a non-flooding routing algorithm is verified in our paper. The verified algorithm is an Adaptive Priority Routing Algorithm (APRA) which is based on fuzzy strategies. Firstly, we introduce the principle of APRA, then using Opportunistic Network Environment (ONE) -simulation software to compare the performance of Epidemic algorithm, Spray and Wait algorithm, PRoPHET algorithm and APRA. By comparing overhead of netword, rate of messages delivered and average dealy, it finds that the APRA performs better. At last, the weaknesses of this paper and further improvement are also discussed.

Abstract: On the basis of summarizing the concept filtering methods in the current Ontology learning, a method of domain concept filtering in the semantic level based on combination of word embedding and conventional statistics was presented, which can identify low-frequency words well, and as far as possible to ensure universality. Through experimental contrast, the proposed approach was proved to have a higher accuracy rate than the ways based on statistics.

Abstract: With reference to a certain type of flying drones with imitation airfoil design a seagull flat wings and on the basis of its wing tip winglet in this paper. Through to the numerical simulation of two wings, it is concluded that the bionic wing aerodynamic performance is superior to the conventional airfoil wing, after adding wing tip winglet bionic wings effectively reduced the downwash velocity, reduce the induced drag, makes the wing aerodynamic performance is improved. Provide theoretical reference for the design of the uav wing

Abstract: In this paper, we apply the Sugeno integral to text classification, which generalizes its usefulness as a classifier. In the classifier, the traditional weight method is replaced by fuzzy measure. Based on word co-occurrence frequency theory, which is improved in the paper, and based on 2-additive measure theory, this paper build a fuzzy measure model well. The paper creates the Sugeno integral text classifier in the last.

Abstract: Video segmentation algorithm based on background reconstruction, and the statistical method of cumulation obtain to complete occlusion of background information domain by, using video object in the current frame and the background image subtraction, successfully overcome the occlusion problem,and in a static background, more accurate segmentation results can be obtained for the moving target.

Abstract: This paper studied Lucene search engine technology in enterprise content management system and make it effectively expanded. It implemented four function modules in the architecture of full-text retrieval based on Lucene: Text analysis module, index module, query module and store module. The intelligent enterprise content management system is reflected in a perfect combination of intellectual technology which includes data retrieval and the modern open source frameworks. It intensified contents processing analyzing functions.

Abstract: The BP neural network as the traditional prediction method has certain advantages, but it has some drawbacks, Such as slow convergence and sensitive to the initial weights, etc. The PSO algorithm is introduced into the neural network training, using the particle swarm algorithm to optimize the neural network weights and threshold. Through the establishment of the particle swarm - BP neural network model for power load budget, it improves the accuracy and stability of the forecast.

Abstract: This paper proposes a new optimization algorithm for solving 0-1 knapsack problem includes greedy degree strategy and dynamic expectation efficiency strategy. The results of experiment show that GDEE algorithm can be used to solving 0-1 knapsack problem. In comparison with traditional dynamic expectation efficiency algorithm, this algorithm can get the best solution and improves convergence speed and saves the storage space.

Abstract: With further consideration of factors that affect the value of the indirect trust, and depth analysis by statistical principles, we proposed a new trust value estimation model. On basis of this model, the method of cluster nodes election has been studied to design a rout decision-making algorithm which combined trust value, density, distance between nodes and forwarding capabilities. Experimental results show that the trust value assessment model built on trust value calculation is very reliable with increasing cycle times, trust value gradually moving to 1. The improved algorithm has faster convergence than algorithms in [5] and [6]. In information transfer process, when the routing nodes are within a certain range, and the algorithm is also more stable. When the network topology changes, the algorithm can be re-routed quickly, indicating that the algorithm has a wide range of applications.

Abstract: The method provided in this paper can be according to the current population to readjust the weight, thus obtain toward the positive ideal point search pressure, finally converge to the optimal solution. This paper combines the feasible direction into genetic algorithm. This method can lead the individual to optimal solution region along feasible direction which approach the optimal solution sets. Through evaluating the degree of distance between chromosome and constrain, we introduce membership function into traditional GA and embed the information of infeasible solutions into fitness function. Propose a self-adapting evaluation function. This method can readjust the weights according to current group and then get the stress of searching to the ideal positive point. To a kind of fuzzy multi-objective optimization problem, propose a method of best satisfaction to transform the fuzzy models to clear ones and solve the models using GA based on interactive method. Then testify its validity though examples.