迭代法

The examples indicate that the algorithm has strong pertinence and high efficiency, hi the discrete variable optimization, the problem is solved in the relative quotient method On the base of analyzing the relative quotient method, the conjugate gradient direction is used to modify the old search directions, the iterative matrix of the algorithm, and the method to resolve the discrete variable optimization, the relative conjugate difference quotient algorithm is presented.

离散变量优化算法是从相对差商法开始的，在详细分析相对差上法的基础上，用共轭梯度方向修正原有的搜索方向，并对算法的迭代矩阵进行相关的修改，最终形成一种用于求解离散变量优化问题的RCDQ法。

Meanwhile the biconjugate gradient method instead of the conjugate gradient method is used to accelerate the iteration process.

同时用双共轭梯度法代替共轭梯度法来加速迭代过程。

Firstly, the approach formulates a cost functional to turn the inverse problem into a constrained minimization problem according to least squares criterion, then the resulting constrained minimization problem is transformed into an unconstrained minimization problem by using a penalty function technique, and then the closed Fr chet derivatives of the Lagrange function with respect to the properties are derived based on the calculus of variations, finally, one can solve the resulting problem by using any gradient-based algorithm and the finite-difference time-domain method.

该方法首先以最小二乘准则构造目标函数，将逆问题表示为约束最小化问题；接着应用罚函数法转化为无约束最小化问题；然后基于变分计算导出闭式的拉格朗日函数关于特征参数的Fr chet导数；最后借助梯度算法和时域有限差分法迭代反演德拜模型参数。

It provided a correct and reasonable mathematical model for the analysis of the ultimate bearing capacity and crankle stability of the thin-walled structure.

论文在比较了两种运动描述方法（完全拉格朗日法和更新拉格朗日法）异同点并分析了非线性几何关系及大变形下的弹性物理关系的基础上，基于增量虚功原理推导了几何非线性空间梁元单元切线刚度矩阵，采取分模块形式，分别表示了轴力、弯矩、双力矩对几何矩阵的贡献；通过球面显式荷载增量迭代式弧长法求解方法编制程序，很好地实现了几何非线性增量荷载的自动步长选择及全过程跟踪，为薄壁杆系结构的极限承载力分析和弯扭失稳分析，提供了较为精确合理的分析力学模型。

We presents an accurate F method based on corresponding point adjustment.The method adjusts coresponding points according to the fixedness of projective transformed cross ratio,then calculates F matrix accurately through linear and non-linear methods. When computing intrinsic parameter,A matrix,we simplify the step,and stress on the two important parameters of A.The result will be getten through solving Kruppa equation based on SVD decomposition.In order to compute extrinsic parameters,we use linear method to get initial R and t,then apply non-linear method to accurate them.

提出了基于匹配点调整的F求精方法，先根据摄影交比不见性对手工选择的匹配点进行调整，再用线性、非线性结合的方法求精F矩阵；在计算内部参数A中，进行了一定的简化，把重心放在A中重要的两个参数上，用SVD分解法计算Kruppa方程；在计算外部参数时，首先用线性法求解R、t，然后再用非线性法迭代求精。

In the self - calibration scheme , the thesis emphasizes the accuracy of camera intrinsic and extrinsic parameters . we presents an accurate f method based on corresponding point adjustment . the method adjusts coresponding points according to the fixedness of projective transformed cross ratio , then calculates f matrix accurately through linear and non - linear methods . when computing intrinsic parameter , a matrix , we simplify the step , and stress on the two important parameters of a . the result will be getten through solving kruppa equation based on svd decomposition . in order to compute extrinsic parameters , we use linear method to get initial r and t , then apply non - linear method to accurate them

提出了基于匹配点调整的f求精方法，先根据摄影交比不见性对手工选择的匹配点进行调整，再用线性、非线性结合的方法求精f矩阵；在计算内部参数a中，进行了一定的简化，把重心放在a中重要的两个参数上，用svd分解法计算kruppa方程；在计算外部参数时，首先用线性法求解r、 t ，然后再用非线性法迭代求精。

In contrast to conventional optimization algorithms whose iterates are computed and analyzed deterministically, randomized methods rely on stochastic processes and random number/vector generation as part of the algorithm and/or its analysis.

传统的最佳化演算法中迭代的计算和分析是确定的，与之相比，随机方法依靠随机过程和乱数字/向量的生成作为演算法和演算法分析的一部分。

Based on filtering theory, a residual life predict model is set up according to condition information, and a general formula of parameter estimation is deduced using maximum likelihood estimation, thus, when the life distributing of an equipment is known, the model can be rapidly educed by taking the distribution into the formula is educed, and its residual life can be predicted without complex iterative process is predicted.

利用滤波理论，建立了基于状态信息的剩余寿命预测通用模型，并推导出采用极大似然估计法下参数估计的通式，使得在已知寿命服从其他任何分布形式时直接代入通式便可快速地得出其模型，从而预测其剩余寿命，省去了复杂繁琐的迭代过程。

Simulation results show that the relative-effective error of bridge structure parameter identification is less than 10%, the application of Levenberg-Marquardt algorithm is feasible, its quantitative message can be put forward for bridge structure condition assessment.

对一连续粱的数值模拟计算表明，在Gauss-Newton法迭代发散的情况下，Levenberg—Marquardt法的识别结果相对误差在10％左右，Levenberg—Marquardt法基本能实现对真实结构参数的识别，为结构进一步的状态评估提供了结构模型最基本的量化信息。

In contrast to conventional optimization algorithms whose iterates are computed and analyzed deterministically, randomized methods rely on stochastic processes and random number/vector generation as part of the algorithm and/or its analysis.

传统的最佳化演算法中迭代的计算和分析具有确定性，与之相比，随机方法依靠随机过程和乱数字/向量的生成作为演算法和演算法分析的一部分。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。