ALGORITHM OF SOLUTIONS FOR MIXEDNONLINEAR VARIATIONAL-LIKE INEQUALITIES IN REFLEXIVE BANACH SPACE

In this paper, the author studies a class of mixed nonlinear variational-like inequalities in reflexive Banach space. By applying a minimax inequality obtained by the author, some existence uniqueness theorems of solutions for the mixed nonlinear variational-like inequalities are proved. Next, by applying the auxiliary problem technique, rite author suggests an innovative iterative algorithm to compute the approximate solutions of the mixed nonlinear variational-like inequalities. Finally, the convergence criteria is also discussed.

Progressive failure processes of reinforced slopes based on general particle dynamic method

In order to resolve grid distortions in finite element method(FEM), the meshless numerical method which is called general particle dynamics(GPD) was presented to simulate the large deformation and failure of geomaterials. The Mohr-Coulomb strength criterion was implemented into the code to describe the elasto-brittle behaviours of geomaterials while the solid-structure(reinforcing pile) interaction was simulated as an elasto-brittle material. The Weibull statistical approach was applied to describing the heterogeneity of geomaterials. As an application of general particle dynamics to slopes, the interaction between the slopes and the reinforcing pile was modelled. The contact between the geomaterials and the reinforcing pile was modelled by using the coupling condition associated with a Lennard-Jones repulsive force. The safety factor, corresponding to the minimum shear strength reduction factor "R", was obtained, and the slip surface of the slope was determined. The numerical results are in good agreement with those obtained from limit equilibrium method and finite element method. It indicates that the proposed geomaterial-structure interaction algorithm works well in the GPD framework.

ON THE COUPLED VIBRATION OF AN IDEAL FLUID WITH A LINEAR ELASTIC STRUCTURE

The purpose of this paper is to analyse theoretically and numerically the coupled vibration of an ideal fluid with a linear elastic structure, It is proved in the paper that the natural frequencies of the coupled vibration do exist and are all real positive. The paper presents an efficient method to transform a coupled fluid-structure system to the structure with added mass and the vibrational analysis of the former is replaced by the latter in vacuum only. Numerical solution is outlined for the transformed problem and a compact jrequeney equation is derived in which fluid variables do not appear. This simplifies the analysis significantly. A convergent proof has been given to guarantee the reliability of the solution. The paper also offers a general algorithm combined with Ritz method, boundary element method, and finite element method to analyse the transformed problem. Based on this algorithm, one can apply a known structural analysing program, with a little modification, to solve many different kinds of fluid-structure coupling problems. Some numerical results are given to show the efficiency of the algorithm.

Target recognition algorithm for passive sonar system with high generalization ability

A new algorithm based on a Supervised Self-Organizing neural network for the pas sive sonar target recognition was proposed. Because of the incompleteness of the passive sonar exemplar set, the algorithm introduced a Multi-Activation-function structure and Supervised Self-Organizing competitive learning algorithm into the classic feed-forward neural networks,and obviously improved the generalization ability in target recognition. Besides, it can effi ciently reduce the learning time and avoid the local optimum. The recognition experiments of realistic passive sonar signals show that this new algorithm has good generalization ability and high recognition rate

AN SQP ALGORITHM WITH NONMONOTONE LINE SEARCHFOR GENERAL NONLINEAR CONSTRAINED OPTIMIZATION PROBLEM

In this paper, an SQP type algorithm with a new nonmonotone line search technique for general constrained optimization problems is presented. The new algorithm does not have to solve the second order correction subproblems for each iterations, but still can circumvent the so-called Maratos effect. The algorithm's global convergence and superlinear convergent rate have been proved. In addition, we can prove that, after a few iterations, correction subproblems need not be solved, so computation amount of the algorithm will be decreased much more. Numerical experiments show that the new algorithm is effective.