差分进化算法在气动力参数辨识中的应用  

Application of Differential Evolution Algorithm for Aerodynamic Parameter Identification

作　　者：简兆圣 艾剑良 

机构地区：[1]复旦大学航空航天系,上海200433

出　　处：《复旦学报（自然科学版）》2017年第5期545-550,共6页Journal of Fudan University：Natural Science

摘　　要：差分进化算法是一种新兴的优化算法,与最小二乘法等梯度类算法相比,它能够进行全局寻优且对初值不敏感,具有广泛的应用前景.建立某型飞机刚体运动的6自由度非线性动力学模型,在叠加一定比例白噪声的情况下获得其仿真数据,使用差分进化算法辨识出该型飞机的纵向运动气动力参数,辨识结果与真实值较为吻合,证明该算法是可行的.多组试验表明:对于该型飞机的动力学模型和仿真数据,使用差分进化算法的辨识结果与使用最小二乘法、普通粒子群算法的辨识结果相比,具有更高的精度和更强的鲁棒性.Differential Evolution（ DE） algorithm is a new heuristic algorithm,compared to gradient-based methods including maximum likelihood estimation and least squares method,it does not require for the optimization problem to be differentiable,and has strong global search ability. A simulation for an airplane with given aerodynamic parameters as the true values is conducted,its input data and output data collected then are used to identify the aerodynamic parameters of the airplane. Among that the least squares method,basic Particle Swam Optimization（ PSO） algorithm and DE algorithm are applied. The identification results show that the mathematical model is correct, and aerodynamic parameters can be identified accurately. When added with different strength of white noise,the identification results are correct overall. In the meantime,the results show that,in this case,DE algorithm converges faster than basic PSO algorithm does and its results are more accurate than results with basic PSO algorithm.

关 键 词：气动力参数辨识 差分进化算法 粒子群算法 非线性系统辨识 

分 类 号：V249.4[航空宇航科学与技术—飞行器设计]