【作者】 陈玮琪；

【导师】 王士同；

【作者基本信息】 江南大学 ， 轻工信息技术与工程， 2008， 博士

【摘要】 由于自由面和空泡的影响,物体穿越水面过程是一个强烈非定常流体力学问题,物体在这瞬变过程中将承受强烈的不定常流体动力载荷作用,这在工程设计上十分重要。但目前国内外对物体出水过程的水动力载荷还难以用试验手段或理论精确研究。本文设想利用现代辨识技术,从物体出水的弹道试验数据中并结合物体的动力学方程,对物体出水水动力系数进行建模。辨识技术在飞行器的线性气动参数辨识方面发展比较成熟,但对非线性气动参数的辨识还存在不少难点,而非线性流体动力系数的辨识与估计目前则还处于探索阶段。本文从物体出水水动力系数辨识的特点出发,在模型辨识、参数估计、试验设计和系统验证方面进行了相关研究工作。我们首先利用模糊系统研究了物体垂直出水非定常水动力系数的建模问题,然后利用粒子群优化算法PSO对水下航行体的多个定常水动力系数进行了模拟辨识,得到的结果与实际值相当吻合,证明了智能技术在水动力系数辨识中具有良好的工程应用价值和启发作用。但对于多维复杂的水动力系统,还需要进一步从非线性模型结构的优化、参数估计算法等方面进行改进。在模型辨识方面,我们专门研究了参数线性的非线性模型结构的选择技术及其算法。在Hilbert空间中提出并证明了一种新的双正交函数的前后向递推式,新的双正交函数可以给出任意函数在已知函数序列张成的子空间中的正交投影,特别是当张成子空间的函数序列扩大或缩小时,新的递推式可以快速给出在变化子空间中的新的正交投影,并且不需要求逆运算,也不需要Gram-Schmidt正交化技术。基于新递推式,本文提出了一个新的模型结构选择技术—逐步投影算法(SP),SP是一个双向选择技术,它首先通过前向选择技术获得初始模型项集合,然后通过后向选择技术进一步从初始模型项集合中筛选出最优的模型项,最后同时给出模型结构和参数估计。SP不仅可以用于系统辨识,还可以用于数据压缩、信号的稀疏逼近等方面。在参数估计方面,由于通过引入辨识准则,水动力系数辨识问题可转化为非线性约束优化问题,因此我们专门研究了适合水动力系数辨识的智能优化算法。一般意义上的智能优化算法,如PSO、DE等都是无约束优化技术,本身缺乏明确的处理约束条件的机制,当它们用于约束优化问题时需要增加其它处理约束条件的手段,如惩罚函数法,但这会带来新的问题。我们在PSO、DE的启发下提出了一个改进的群体智能优化算法—差分群体智能算法DS,DS在算法的机制中结合了约束条件的影响,具有一定的自适应处理约束条件的能力。通过对IEEE CEC06上的24个约束优化基准测试函数的大量试验,并和其它进化算法的比较,证明了DS对约束优化问题具有良好的快速优化能力和通用性,适合用于水动力系数辨识。在试验设计方面,由于6自由度物体出水水动力系统过于复杂,我们设计了约束模型单平面斜出水的水动力测量试验,并结合约束模型在出水过程中不同阶段的物理特点,分别对约束模型全湿流的定常水动力系数、约束模型出水的轴向和法向的非定常水动力系数分别进行了建模与辨识。在系统验证方面,我们利用辨识的水动力系数对相近试验条件下的其它出水试验的水动力进行了预测,预测结果与实际测量结果相当一致,由此验证了出水水动力系数模型的合理性。更多还原

【Abstract】 Due to the impact of free-surface and cavitation bubbles,the moving process of water-exit body is a very strong unsteady hydrodynamics problem.The body will endure very strong unsteady hydrodynamic forces during the transient process;this is of great importance in engineering design.However,it has been thus far,both at home and abroad,a rather difficult task to investigate the unsteady hydrodynamic forces due to water-exit body.In this dissertation,by using modern identification techniques,we would like to contribute to such investigations by focusing on mathematical modeling and parameter identification of the above mentioned hydrodynamic system, based on related experimental data and the corresponding hydrodynamics equations.Whilst identification techniques in the field of linear aerodynamics parameter identification have been considerably developed,some difficult problems remain untackled on identification of nolinear aerodynamics parameter.However,the identification and estimation of nonlinear hydrodynamics parameters is still in its initial stage.Starting with the characteristics of parameters of hydrodynamics due to water-exit body,we conduct investigations on model identification,parameter estimation,experimental design,and system verification.First of all,we study the parameter modeling problem of unsteady hydrodynamics due to vertical water-exit body based on fuzzy system.Next,we use Particle Swarm Optimization(PSO) algorithm to investigate the simulated identification of the several parameters of the unsteady hydrodynamic forces due to a submerged body. The results we obtained are considerably consistent with the real values,this being a strong evidence for the practical value and inspiriting effects of intelligence techniques in hydrodynamics parameter identification. Nevertheless,for multi-dimensional complex hydrodynamic systems,further improvements need to be done on several issues such as optimization of nonlinear model structure and parameter estimation algorithms.On the aspect of model identification,we specially study the selection techniques for nonlinear model structure with linear parameters and the corresponding algorithms.We propose a new set of bi-orthogonal functions in Hilbert space,derive the corresponding forward and backward recursive equations,and prove that the proposed bi-orthogonal functions can generate the orthogonal projection of an arbitrary function onto the subspace spanned by a given set of functions.It is appropriate to stress that the proposed recursive method avoids the computation of inverse operations,does not need the Gram-Schmidt orthogonalization technique,and allows us to update the whole set of bi-orthogonal function when the dimension of the subspace is enlarged or decreased. Based on the proposed recursive equations,we further propose a new model structure selection technique,termed Stepwise Projection(SP) algorithm.SP is a bi-direction selection technique,which firstly selects an initial set of model items through forward selection technique,and then further selects the optimal items from the initial set by use of backward selection technique,and finally yields model structure and parameter estimations simultaneously. SP is applicable not only to system identification,but to a broad range of signal processing problems such as data compression and sparse approximation of signals.On the aspect of parameter estimation,due to introduce of identification rules,the hydrodynamics parameter identification problems can be converted to nonlinear constrained optimization problems.Therefore,we investigate specialized intelligent optimization algorithms suitable for hydrodynamics parameter identification. The general intelligent optimization algorithms,taking PSO and Differential Evolution(DE) for instances,are unconstrained optimization techniques and thereby lack an implicit constraint-handling mechanism.Therefore, they need extra technique(say,penalty function method) to handle constraints when they are used for soling constrained optimization problems.But this would result in some new difficulties.Inspired by the philosophy of PSO,we propose an improved swarm intelligence algorithm,called Differential Swarm(DS),which incorporates into its algorithmic mechanism the impact of constraints and has a certain ability to self-adaptively handle constraints.The performance of DS is experimented on the benchmark set of 24 functions for CEC06 and compared with other evolutionary algorithms.Experimental results show that DS performs considerably well in terms of optimization ability,simplicity,convergence rate,and generality,and is suitable for hydrodynamics parameter identification. On the aspect of experimental design,since the hydrodynamic forces due to water-exit body of six-degree freedom is over complicated,we devise relevant experiments to measure the hydrodynamic forces acting on constrained model exiting water obliquely.Moreover,by considering the physical characteristics of such constrained model at different stages of water-exit process,we conduct research on modeling and identification of steady hydrodynamics parameters of no cavity flow,and of the axial and normal unsteady hydrodynamics parameters.On the aspect of system verification,we predict the hydrodynamic forces of other water-exit experiments under similar condition by using the hydrodynamics parameters obtained through identification techniques.It has been shown that the predicted results are highly consistent with the experimental results,thus verifying the correctness of the hydrodynamics parameter model of water-exit body.更多还原