Advanced Materials Research Vols. 765-767

Abstract: Path planning is an important research content of unmanned aircraft vehicle (UAV) technology, it involves many factors, and simulation is very difficult. Based on the knowledge of relevant mathematical knowledge and path planning, a UAV path planning simulation method is proposed, the method can obtain reconnaissance band width in reconnaissance area by reconnaissance instructions, and then UAV reconnaissance flight paths can be calculated combined with the shortest path principle, and UAV can complete reconnaissance missions based on path planning in the shortest possible time. This method is simple and practical, and it meets the needs of the UAV path planning simulation in the large-scale system simulation.

Abstract: Electrospinning technique is currently one of the most important methods for preparing nanofiber. According to the fact that the electric field at the receiver is relatively weak, the electric fields distribution in the typical electrospinning device was calculated and analyzed by ANSYS. It put forward the method of optimizing electric fields distribution by using an auxiliary electric field device. At the same boundary conditions, the electric fields distribution in the original device and improved binode device are compared. The simulation analysis provides an effective reference for the optimization design of electric fields distribution in the electrospinning device.

Abstract: This paper established the thermal Lattice Boltzmann model of fluid flow and heat transfer, which is based on double lattice Boltzmann distribution model [. The temperature distribution adopted the higher accuracy velocity model. Based on this thermal lattice Boltzmann model, this paper simulated forced convection of circular tube fluid. Comparing the simulation results with the traditional CFD calculation results, we could find that the thermal lattice Boltzmann method have unique advantages in effectiveness and flexibility than the traditional calculation method.

Abstract: Anti-surfaceship missile firepower allocation is a complex and important problem to submarine command. The method of partheno-genetic algorithm (PGA) is suit for firepower allocating. While keeping the virtues of traditional genetic algorithm, PGA overcomes its defects. Because of employing more simply operators, PGA decreases the complexity of calculation. Basing on the mathematics model of missile firepower allocation and building the steps of PGA, a case of firepower allocation to submarine by PGA was applied. Simulation results indicated that PGA is a simple, effective and fast algorithm. PGA can solve the submarines missile firepower allocation problem effectively and can be satisfied with the requirement of real-time and fast to submarine command.

Abstract: To establish an accurate model of aluminum electrolytic process, an novel variable selection strategy is proposed based on the false nearest neighbors (FNN) and randomization method (RM), which is abbreviated as FR. Firstly, the FNN is used to calculate the similarity measure of the respective variable; secondly, the RM is employed to test the significance level of each variable in turn; lastly, technical energy consumption model is established to verify the proposed method. The experimental results show that the method selects the best decision-making variables. Therefore, it provides a new method for the variable selection for complex industrial process.

Abstract: Recently, Particle Swarm Optimization (PSO) has received increasing interest from the optimization community due to its simplicity in implementation and its high speed computational capability. However, traditional PSO has premature convergence problem. To prevent the premature convergence of PSO, some modified algorithms is proposed, based on these modified algorithms a novel hybrid algorithm with random disturbance is proposed in this paper. The performance of the proposed algorithm is testified on a suite of benchmark functions. The simulation results show that this hybridized PSO has great capability of preventing premature convergence and faster computing speed.

Abstract: Aiming at the defect of the discernibility matrix formation, the three-process space-time function is analyzed, and the method is put forward which corresponding new elements dont involve the formation of discernibility matrix or the existed element is deleted, that is the method of discernibility matrix minimum formation. The algorithm of formation is also given in the method, with which, the number of discernibility matrix element is decreased, the space of data is saved and the speed of data mining is increased. The superiority of the method in time-space is verified during the experiment.

Abstract: An efficient algorithm for contact detection among many arbitrarily sized ellipsoids is developed. The numberical method is unable to efficiently deal with ellipsoid particles of greatly varying sizes. To overcome this challenge, We present an algebraic method to simulate polydisperse ellipsoid particles system. We offered several times speed-up compared to the numberical method. So that the problem of contact detection in polydisperse ellipsoid particles system essentially is solved. We will describe a generalization of the well-known linked cell list method and an improvement on the nearest neighbor list method.

Abstract: K-means algorithm has therefore become one of the methods widely used in cluster analysis. But the classification results of K-means algorithm depend on the initial cluster centers choice. We present a new neighborhood for PSO methods called the area of influence (AOI) and consider the combination of K-means has strong capacity of local searching and PSO has power global search ability. The improved PSO, i.e., improves the K-means local searching capacity, accelerates the convergence rate, and prevents the premature convergence effectively.

Abstract: in order to solve the problem which standard BP algorithm didnt have a good prediction accuracy for testing samples, a L-M Bayesian regularization algorithm was proposed by improved standard BP algorithm and applied to predict the resident consumption level of Cheng du. The experimental results show the L-M Bayesian regularization algorithm neural network has a higher accuracy, a more stable performance and a stronger generalization ability than another two improved algorithms in the same conditions and has a very good effect for the forecast of resident consumption level.