关键词: 径向基函数, 循环子空间回归, 遗传算法, 混合规划, 脉冲萃取, 参数优化

Abstract: Radial basis function-cyclic subspace regression (RBF-CSR) approach is a rapid one-step modeling method, escaping the difficulty of ANN architecture design. ANN and multivariate analysis are integrated into this approach, but it is difficult to optimize the parameters of the model because the method is a mixed integer programming.A eugenic hybrid coding genetic algorithm（EHCGA）, which used different coding methods for different type of variables was proposed to give the optimal parameters of the RBF-CSR model. In EHCGA deterministic optimization method was added to standard GA based on eugenic strategy, which could greatly improve the convergence speed and accuracy of GA. RBF-CSR model optimized by EHCGA called RBF-CSR-EHCGA model was successfully applied to modeling recovery of caprolactam from waste water by using pulsed-sieve-plate extraction column.

Key words: 径向基函数, 循环子空间回归, 遗传算法, 混合规划, 脉冲萃取, 参数优化