Applied Mechanics and Materials Vols. 571-572

Abstract: With the prevalence of credit system, the stipulation of “academic warning” is written into the teaching management constitution by more colleges and universities. However, the present research in this stipulation is only limited to the simulation of multivariate normal distribution. This paper aims to improve the current setting of academic warning through Monte Carlo simulation of multivariate Copula functions, and to calculate more reasonable academic warning credit line. The result demonstrates that the accuracy is significantly improved, therefore, this approach can provide a new train of thought and universal method for colleges and universities to set specific standards.

Abstract: Silkworm egg incubation is a very complicated process and is hard to establish the mathematical model. In order to improve the temperature and humidity of silkworm incubation chamber control method, the paper deal with the method of iteration learning control looking for expected input. Based on the summary of human experience, designing learning rule is given. The simulation result shows iterative learning control has ideal tracking control.

Abstract: This paper investigates a new method of fault diagnoses and reliability analysis based on inverse fuzzy model method. The proposed method employs inverse fuzzy model solving method to estimate the component state based on the measured system performance and relationship about component state and system performance which is constructed by expert. The method is proved to be effective in fault diagnoses.

Abstract: Community detection in complex network has been an active research area in data mining and machine learning. This paper proposed a community detection method based on multi-objective evolutionary algorithm, named CDMOEA, which tries to find the Pareto front by maximize two objectives, community score and community fitness. Fast and Elitist Multi-objective Genetic Algorithm is used to attained a set of optimal solutions, and then use Modularity function to choose the best one from them. The locus based adjacency representation is used to realize genetic representation, which ensures the effective connections of the nodes in the network during the process of population Initialization and other genetic operator. Uniform crossover is introduced to ensure population’s diversity. We compared it with some popular community detection algorithms in computer generated network and real world networks. Experiment results show that it is more efficient in community detection.

Abstract: This paper proposes an indoor positioning system based on visible light communication technology and intelligent terminals which equipped with image sensors, electronic compasses and gyroscope sensors. Each LED source is modulated with unique codes which represent its absolute location. The receiving terminal gets the LED’s absolute location through the optical link and capture the scene image with image sensor simultaneously. The electronic compass and the gyroscope sensor measure the terminal’s yaw angle, pitch angle and roll angle in real time, together with the scene image, relative position of the terminal and the LED light source can be calculated. Combine the absolute location of the LED light source with the relative position, the system is able to estimate the terminal’s position within the accuracy of about 2 centimeter. Position accuracy can be improved by using high-precision camera and attitude sensors.

Abstract: This paper presents a hybrid algorithm combining immune genetic algorithm (IGA) with simulated annealing (SA) to overcome the shortages of both the two algorithms respectively. SA is introduced to solve the problem of IGA in fault diagnosis, unable to reach whole convergence and etc. by designing a new kind of self-adaption strategy of genetic parameters. Finally, the Schaffer function is introduced to show the optimization ability of this proposed IGA-SA algorithm.

Abstract: Since the 1950s, with the great development of computer technology and bionics, particle swarm optimization (PSO) was raised. The particle swarm optimization mimics the nature biological group behaviors, and has the following advantages compared to classic optimization algorithms: it is a global optimization process and doesn’t depend on the initial state; it can be applied widely without prior knowledge on the optimization problems; the ideas and the implements of PSO are quite simple, the steps are standardization, and it’s very convenient to integrate it with other algorithms; PSO is based on the swarm intelligence theory, and it has very good potential parallelism. Particle swarm optimization has a feature that fitness value is used to exchange information in the population, and guides the population to close the optimal solution. Therefore, a mount of fitness should be calculated in swarm intelligence optimization algorithms in order to find the optimal solution or an approximate one. However, when the calculation of the fitness is quite complex, the time cost of this kind of algorithms will be too large. What’s more, the fitness of optimization problems in the real world is often difficult to calculate. Addressing this problem,Efficient Particle Swarm Optimization Algorithm Based on Affinity Propagation (EAPSO) is proposed in this paper.

Abstract: The adjoint assimilation technique is used to invert the prescribed initial field in the Bohai Sea. Based on this technique, the practical performances of the limited-memory BFGS (L-BFGS) method, the Regularization method, and the Gradient Descent (GD) method are investigated computationally through a series experiments. Experimental results demonstrate that the prescribed initial field can be successfully estimated by these three methods. Inversion result with the Regularization method is better than that with the L-BFGS method, although errors of observations are higher. Though higher simulation errors than L-BFGS and Regularization method, the difference between the prescribed distribution and inversion result is the lowest, indicating that inversion result with the traditional GD method is the best.

Abstract: According to the characteristics of fault types of the transformer ,RBF neural network is used to diagnose transformer fault. The paper regards six gases as inputs of the neural network and establishes RBF neural network model which can diagnose six transformer faults: low temperature overheat, medium temperature overheat, high temperature overheat, low energy discharge, high energy discharge and partial discharge . The Matlab simulation studies show that transformer fault diagnosis model based on RBF neural network diagnosis for failure beyond the traditional three-ratio method. The rate of the transformer fault diagnosis accuracy reaches 91.67% which is also much higher than the traditional three ratio method.

Abstract: Speech feature extraction is discussed. Mel frequency cepstral coefficients (MFCC) and perceptual linear prediction coefficient (PLP) method is analyzed. These two types of features are extracted in Lhasa large vocabulary continuous speech recognition system. Then the recognition results are compared.