论文部分内容阅读
针对马尔可夫链蒙特卡罗(MCMC)模型修正方法在待修正参数维数较高时不易收敛和计算效率低下的问题,建立了融合自适应算法和相关向量机的快速模型修正方法。基于广义无偏见先验分布,推导了待修正参数的后验分布;在标准MCMC方法的基础上,引入延缓拒绝算法以提高新样本接受概率;引入自适应算法以自主调整建议分布的带宽。通过相关向量机建立待修正参数与有限元模型理论计算值之间的回归模型,以提高模型修正的计算效率。数值模拟和试验结构的模型修正结果表明,该方法的收敛速度较快,计算效率优于传统的一阶优化模型修正方法,为解决不确定性模型修正中的计算效率提供了一种新手段。
In order to solve the problem that Markov chain Monte Carlo (MCMC) model modification method is difficult to converge and the computational efficiency is low when the dimension of the parameter to be modified is high, a fast model correction method based on fusion adaptive algorithm and correlation vector machine is established. Based on the generalized unbiased prior distribution, the posterior distribution of the parameters to be modified was deduced. Based on the standard MCMC method, a delay rejection algorithm was introduced to increase the probability of accepting new samples, and an adaptive algorithm was introduced to adjust the bandwidth of the proposed distribution. The regression model between the parameter to be modified and the theoretical calculation value of the finite element model is established by a correlation vector machine to improve the calculation efficiency of the model correction. The results of numerical simulation and model modification of the experimental structure show that the convergence rate of this method is faster and the computational efficiency is better than the traditional first-order optimization model. It provides a new method to solve the computational efficiency in the uncertainty model revision.