No abstract

Feature selection is one of key technologies for fault diagnosis. Especially for high dimensional data, Feature selection can not only find the feature subset with sufficient information, but also improve the classification accuracy and efficiency. In order to decrease the number of diagnosis parameter in fault diagnosis of Liquid-propellant Rocket Engine, the paper proposes one feature selection method based on improved particle swarm optimization, the method applies the quantum evolution thoughts to PSO. The particle is restricted in the range from -π/2 to 0, so the particle can correspond to the quantum angle. The parameter optimization function is designed. The improved algorithm can decrease the number of parameter in fault diagnosis of Liquid-propellant Rocket Engine from 25 to 6.

In order to solve the problem of knowledge acquisition in equipment fault diagnosis,the paper introduces a method based on improved particle swarm optimization. Firstly the paper transforms the nonlinear equations which describe the system into optimization problem with constriction. Since the equation is nonlinear and multidimensional ,standard particle swarm cant solve the problem due to the weakness of premature. So one improved particle swarm optimization is proposed. During the evolution, density evaluation, clone and mutation operator is proposed under the thought of immunity. The results of simulation show that the immune particle swarm optimization can simulate effectively and acquire the system knowledge.

In order to analyze the structural characteristics or calculate the support force of large-scale complex systems with spherical joints, an approximated method was raised simplifying the force of inner bodies to contact pressure with a hypothesis that the contact zones is ideally spherical. The contact pressure distribution is obtained and normal force-displacement relationship is simulated with finite element methods (FEMs). Finally, the goodness of fit is calculated with statistical hypothetical test theory treating the FEM results as the sample data.

No abstract