基于数据的机械论黑箱辨识方法研究  

Research on Black-Box Identification Based on DBM Method

作　　者：孔照丹[1] 罗文波[1] 何蕊[1] 

机构地区：[1]哈尔滨工业大学卫星技术研究所,黑龙江哈尔滨150080

出　　处：《系统仿真学报》2007年第11期2429-2433,共5页Journal of System Simulation

摘　　要：总结了基于数据的机械论(Data-based Mechanistic)黑箱辨识算法(简称DBM)在常规黑箱辨识算法中的基本框架和两类重要模型时变参数(Time Variant Parameter)模型(简称TVP)和状态依存参数(State Dependent Parameter)模型(简称SDP)的特点,并利用DBM方法和其他非线性辨识算法分别对时变参数方程和杜芬方程进行分析和比较,发现尽管方法本身具有拟合精度高,不存在过拟合现象,可以较精确的反应参数对于状态或者时间的依存关系等特点,但是SDP方法在分析复杂问题时存在搜索效率低,计算时间长的缺点。最后通过引入遗传算法对搜索方式进行改进,从而降低了所要搜索的规模,大大减少了计算量。The basic characteristic of the Data-based Mechanistic （DBM） approach in the context of Black-Box nonlinear identification was generalized. The characteristic of two important Time Variant Parameter （TVP） and State Dependent Parameter （SDP） models were presented and the efficiency of the DBM method was illustrated by two examples based on the analysis of Duffing Equations and lime-Variant Equation, in contrast with other methods. Finally, as the calculating number increased exponentially with the increase of the number of the states, which the SDP depends on, genetic algorithm was introduced to improve the search efficiency.

关 键 词：非线性系统 黑箱辨识 基于数据的机械论方法 遗传算法 

分 类 号：N945.1[自然科学总论—系统科学]