频率法索力测定概率分析及参数识别

发布时间：2018-11-27 13:45

【摘要】：频率法因其经济高效、简便实用的特点在索力测量方法中占有重要地位。近几十年来，，国内外学者由频率和其他结构参数计算索力方面拟合了十几种公式，使频率法有很多计算选择；但是这些公式都是确定性分析，没有考虑频率和结构参数随机性对索力结果的影响，另外在多种公式中如何选择最优公式，也需要分析评价。在结构参数未知或模糊的情况下，国内外学者从反问题的角度对关键参数分析识别，但几乎都类似梯度法的方式从反向识别，对于索力测试这样一个非线性系统，这样容易使解陷入极值区间，造成解不稳定。针对以上不足，笔者主要做了一下几方面的工作：（1）在满足工程实际要求的前提下，以高精度为目标，设计一套能概率性确定索力结果和计算索力测量失效概率，对现存频率法索力公式进行精度和误差传递大小评价，对索力测定中未知参数进行正向识别的方法—神经网络-蒙特卡洛算法。该算法在拟合显式精度、计算索力标准差和失效概率方面优于响应面法，又比直接有限元法节省时间。（2）运用算法研究了多参数的随机性对长中短索索力结果的影响，频率、索长、抗弯刚度、边界条件、线质量变异系数如果都为0.1，则长索索力结果变异系数大于0.3，短索索力结果变异系数大于0.5。对常见公式从精度和误差传递大小方面进行了评价，在两端固结边界条件下，任伟新、王建飞、Hiroshi Zui、邵旭东四人公式的计算结果满足工程精度要求，应优先采用。（3）基于神经网络和蒙特卡洛结合法正分析识别频率法索力测定中的未知参数。利用多阶频率作为目标值，全局网格搜索最优解。索力及长度识别结果稳定并且满足工程要求精度。（4）以南盘江悬索桥索力测量为背景，将算法应用于工程实际，将遴选出的优秀公式计算索力，确定索力的可靠范围和计算测量失效概率，并且对未知参数进行识别，最重要是索力的识别。
[Abstract]:Frequency method plays an important role in cable force measurement because of its economical, efficient, simple and practical characteristics. In recent decades, scholars at home and abroad have fitted more than ten formulas for calculating cable force by frequency and other structural parameters, which makes the frequency method have many calculation options. However, these formulas are deterministic and do not consider the influence of frequency and structural parameters randomness on cable force results. In addition, how to choose the optimal formula in many formulas also needs to be analyzed and evaluated. In the case of unknown or fuzzy structural parameters, scholars at home and abroad analyze and identify the key parameters from the point of view of inverse problems, but almost all of them are similar to the gradient method from the reverse recognition, for a nonlinear system such as cable force testing. It is easy to make the solution fall into the extremum range and cause the solution to be unstable. In view of the above shortcomings, the author mainly does some work in the following aspects: (1) under the premise of meeting the practical requirements of engineering, a set of energy probability is designed to determine the cable force result and calculate the failure probability of cable force measurement. The accuracy and error transfer size of the existing Burkina Faso force formula are evaluated, and the neural network-Monte Carlo algorithm is used to identify the unknown parameters in the determination of cable force. The algorithm is superior to the response surface method in fitting the explicit precision, calculating the standard deviation of cable force and the failure probability, and saving time compared with the direct finite element method. (2) the effect of randomness of multiple parameters on the results of cable force is studied. The frequency, cable length, bending stiffness, boundary condition and coefficient of variation of line mass are all 0.1. The coefficient of variation of long cable force and short cable force is greater than 0.3 and 0.5 respectively. The common formulas are evaluated in terms of accuracy and error transfer. Under the condition of boundary consolidation at both ends, the calculation results of Ren Weixin and Wang Jianfei, Hiroshi Zui, Shaoxu Dongsiren formula meet the requirements of engineering precision and should be adopted first. (3) based on the neural network and Monte Carlo method, the unknown parameters in the determination of Burkina Faso force are analyzed. Using multi-order frequency as the target value, the global grid search for the optimal solution. The results of cable force and length identification are stable and meet the requirements of engineering accuracy. (4) based on the measurement of cable force of the suspension bridge of Nanpanjiang River, the algorithm is applied to the engineering practice, the excellent formula is selected to calculate the cable force, the reliable range of the cable force and the failure probability of the measurement are determined, and the unknown parameters are identified. The most important is the identification of cable force.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U448.25;U441

【参考文献】