论文部分内容阅读
为了解决常规优化方法存在的梯度计算困难、病态、非唯一解等问题,采用贝叶斯方法来实现随机模型修正,同时为了加快贝叶斯方法推断修正参数的后验概率密度的速度,将高斯过程替代模型技术和基于延缓拒绝自适应的Metropolis(DRAM)算法的先进马尔科夫链蒙特卡罗(MCMC)方法用于贝叶斯推断,最后,用一斜拉人行桥实例来验证基于贝叶斯推理的随机模型修正方法。结果表明:贝叶斯方法的修正结果与实测结果吻合较好,验证了基于贝叶斯推理的随机模型修正方法的可行性。
In order to solve the problems of gradient calculation, pathological and non-unique solutions to conventional optimization methods, the Bayesian method is used to modify the stochastic model. In order to speed up the estimation of the posterior probability density of the modified parameters by the Bayesian method, Proceedings of the National Academy of Sciences of the United States of America, Proceedings of the National Academy of Sciences of the United States of America, Proceedings of the National Academy of Sciences of the United States of America, Proceedings of the National Academy of Sciences of the United States of America, Proceedings of the National Academy of Sciences of the United States of America Stochastic model correction method of reasoning. The results show that the modified Bayesian method is in good agreement with the measured data, which verifies the feasibility of the stochastic model updating method based on Bayesian inference.