解等式约束优化问题的一个修正既约Hessian SQP方法

众所周知,既约Hessian方法是求解较大规模约束优化问题的一类有效方法,但已有的这类方法的全局收敛性分析需假定拉格朗日函数的既约Hessian矩阵的一致正定性.本文提出了一个修正的既约Hessian SQP方法,并且证明其在没有上面提及的假设条件下具有全局收敛性.     

一种求解TSP问题的多策略改进蚁群算法

蚁群算法是一种求解复杂组合优化问题的启发式仿生进化算法,并是求解TSP问题行之有效的一种随机算法.但此算法仍存在求解精度低、易陷入局部最优及求解效率低的问题,针对该问题提出一种多策略改进蚁群算法.采用最近邻法影响初始信息素的分布,达到降低算法初期较短路径上信息素浓度的目的,并在转移规则变异调整的基础上,结合路径的均值交叉进化策略,增强算法探索全局解空间和避免陷入局部最优的能力.然后,结合迭代和精英策略对信息素更新机制进行改进,进一步提高化算法的求解性能及求解效率,最后,对从TSPLIB数据库选出的8个实例进行求解并与其他算法进行对比,实验结果表明,改进算法在求解旅行商问题时的高效性,且具有较高的运算性能.     

Compressibility Effects in Turbulent Boundary Layers

Local cascade （LC） scheme and space-time correlations are used to study turbulent structures and their convection behaviour in the near-wall region of compressible boundaxy layers at Ma = 0.8 and 1.3. The convection velocities of fluctuating velocity components u （streamwise） and v （vertical） are investigated by statistically analysing scale-dependent ensembles of LC structures. The results suggest that u is convected with entropy perturbations while v with an isentropic process. An abnormal thin layer distinct from the conventional viscous sub-layer is discovered in the immediate vicinity of the wall （y^＋ ≤ 1） in supersonic flows. While in the region 1 〈 y^＋ 〈 30, streamwise streaks dominate velocity, density and temperature fluctuations, the abnormal thin layer is dominated by spanwise streaks in vertical velocity and density fluctuations, where pressure and density fluctuations are strongly correlated. The LC scheme is proven to be effective in studying the nature of supersonic flows and compressibility effects on wall-bounded motions.