For the traditional target localization algorithms of frequency diverse array (FDA), there are some problems such as angle and distance coupling in single-frequency receiving FDA mode, large amount of calculation, and weak adaptability. This paper introduces a good learning and predictive method of target localization by using BP neural network on FDA, and FDA-IPSO-BP neural network algorithm is formed. The improved particle swarm optimization (IPSO) algorithm with nonlinear weights is developed to optimize the neural network weights and biases to prevent BP neural network from easily falling into local minimum points. In addition, the decoupling of angle and distance with single frequency increment is well solved. The simulation experiments show that the proposed algorithm has better target localization effect and convergence speed, compared with FDA-BP and FDA-MUSIC algorithms.

The current research of product variant design is focused on how to make it effective and feasible， but it is of no any methodology to get the optimized result of the designing. Based on the special research of the parametric variant design and genetic algorithms, a new optimal method of parametric and variant design is presented. By analyzing mechanical product parameters and their relevance, we build the master cast of the mechanical product. In the genetic algorithm module, the engineering requirements are expressed by mathematic model. The parameters are optimized through global searching, and the objective function and performanceconstraint function are built in genetic algorithm. After a global optimal convergence, the optimal solution will be found. The solution of product variant design is achieved after importing the optimal solution into the master cast. Moreover, the procedure based on genetic algorithm is realized, which aimed at a type of high speed shaft. The experimental result demonstrates that the proposed method is efficient and feasible.