Advanced Materials Research Vols. 694-697

Abstract: An optimization design system for fir-tree root of turbine blade has been developed in this paper. In the system, a parametric model of the blade and rim was established based on the parametric design language APDL, and nonlinear contact method was used for analysis by ANSYS, meanwhile some optimization algorithms, such as Pattern Search Algorithm, Genetic Algorithm, Simulated Annealing Algorithm and Particle Swarm Optimization, were adopted to control the optimizing process. Five cases of manufacturing variation in contact surfaces between root and rim were taken into account, and the design objective was to minimize the maximum equivalent stress of root-rim by optimizing eight critical geometrical dimensions of the root and rim. As a result, the maximum equivalent stress of root-rim decreases markedly after the optimization in all cases. In consideration of both precision and computing time, particle swarm optimization is assessed as the best algorithm to solve structure optimization problem in this work. Corresponding to five different cases of manufacturing variation, the maximum equivalent stress of root and rim reduces by 7%, 8%; 27%, 24%; 27%, 22%; 25%, 19%; 10%, 14% using the Particle Swarm Optimization.

Abstract: This paper proposes an emergency obstacle avoidance control method based on driver steering intention recognition for steer-by-wire vehicle in order to solve the problem that the response rate and stability time are unsatisfactory. The paper focuses on the method to recognize driver steering intention, and builds a driver steering behavior model by using the multidimensional Gauss HMM theory, optimizes the model by using the Baum-Welch algorithm and conducts real-time verification on steering intention recognition by means of LabVIEW and driving simulator. The results indicate that the driver steering intention recognition method has higher recognition accuracy and can help to realize emergency obstacle avoidance control effectively for steer-by-wire vehicle.

Abstract: Abstract. A cross-regional multi-site inventory system with independent Poisson demand and continuous review (S-1,S) policy, in which there is bidirectional transshipment between the locations at the same area, and unidirectional transshipment between the locations at the different area. According to the M/G/S/S queue theory, birth and death process model and approximate calculation policy, we established inventory models respectively for the loss sales case and backorder case, and designed corresponding procedures to solve them. Finally, we verify the effectiveness of proposed models and methods by means of a lot of contrast experiments.

Abstract: The classical multi-class logistic regression classifier uses Newton method to optimize its loss function and suffers the expensive computations and the un-stable iteration process. In our work, we apply two state-of-art optimization techniques including conjugate gradient (CG) and BFGS to train multi-class logistic regression and compare them with Newton method on the classification accuracy of 20 datasets experimentally. The results show that CG and BFGS achieves better classification accuracy than the Newton method. Moreover, CG and BFGS have the lower time complexity, in contrast with Newton method. Finally, we also observe that CG and BFGS demonstrate similar performance.

Abstract: In the "call for paper" of 2013 IEEE Congress on Evolutionary Computation (CEC 2013), Special Session on "Differential Evolution: Past, Present and Future", "Experimental design and analysis of DE" is the third area. In this paper, we propose a rapid analysis approach based on Orthogonal Design (OD) and Analysis of Variance (ANOVA) for performance of DE. The analysis results can be the reliable basis of the principles guiding the creation of adapting rules in novel adaptive DE algorithms.

Abstract: This paper proposed a revised quantum-behaved particle swarm optimization algorithm utilizing comprehensive learning strategy to prevent the universal tendency of premature convergence, based on which introduced a novel data clustering algorithm as well. The optimal number of cluster could be automatically obtained by this novel clustering algorithm because a new special coding method for particles was used. Compared with another two dynamic clustering algorithms on five testing data sets, the proposed dynamic clustering algorithm based on the comprehensive learning strategy has the best performance and with the best potential application prospect.

Abstract: In daily life, lots of work need people maintain higher attention or vigilance. In the early study of vigilance, blink frequency, the impedance of skin, body temperature and blood pressure and other physiological signals was used to estimate the vigilance. EEG signal can more directly reflect the brain's activity than other physiological signals, and EEG signal have a higher time resolution. In this paper, ERP component was used to analyze the alert state, according to this study, the use of ERP component analysis can reflect the subjects’ alertness.

Abstract: In order to ensure the voltage stability of distribution network with doubly fed induction generator, it is more significant to determine the reactive power distribution of the grid reasonably except installation of on load tap changing transformer and shunt capacitor bank. However, the reactive power control capability of doubly fed induction generator is not fully implemented in the reactive power optimization of distribution network. In this paper, on the premise of utilizing wind energy maximally, the wind farm with doubly fed induction generator is considered as a continuous reactive power source to participate in the reactive power optimization. Then the constructed reactive power optimization problem is converted to a nonlinear mixed integer optimization problem which is solved by intelligent genetic algorithm. Finally, the IEEE 16 node system is chosen as an example for simulation, the computation results verify the effectiveness of the proposed algorithm.

Abstract: A nesting system based on minimum potential energy principle and intelligent optimization for ship part nesting problem was proposed. Discussing polygon judgment and separation, intersection test and collision problems of ship parts, a kind of polygon overlap detection method was put forward, and contacting process was analyzed by use of envelope rectangle intersection test algorithm; During analyzing ship part nesting process based on minimum potential energy principle and genetic algorithm fusion, basic physical meaning of nesting problem was explained from mechanics. Throng intelligent ship part nesting system verification, the algorithm is feasible, physical meaning is clear; it can realize ship part nesting.

Abstract: As many influence factors and technical problems should be taken into account in the scheme selection decision of multi-band camouflage screen design, a method based on hierarchical grey relational analysis combing analytic hierarchy process (AHP) and grey relational analysis (GRA) was developed. The hierarchy model was constructed firstly, and the judging matrixes were established through analytic hierarchy process. The weight value for each index which effects the evaluation was determined accordingly. The relational degree between an individual alternative scheme and the ideal alternative and the worst alternative are calculated by grey correlation degree, and the evaluation degree is calculated according the grey correlation degree and weight value. Based on the relational degree a quality order for the alternative schemes can be obtained. Experiment was carried out, which proved that the method mentioned in this paper had both strict mathematical theory basis and good practical utility. This method can be also used in scheme selection decision of other camouflage designs.