【摘要】 目的生成空间多点的顺风向风速时程,比较不同阶AR模型的模拟效果,评价由BIC准则判定的AR模型最优阶数的合理性.方法根据工程中普遍采用的顺风向脉动风速谱,利用MATLAB软件编制空间多点风速时程的自回归线性模型程序,并嵌入BIC准则以判定多维AR模型的最优阶数,使用不同阶模型对风场进行模拟并比较各阶模型的模拟效果.结果通过对西堠门大桥风场的模拟发现,BIC准则判定得到的AR模型最优阶数为2阶,同时2阶模型取得了不错的模拟效果;不同阶模型的模拟效果有较大差异,1阶模型离散性较大,6阶、8阶模型模拟精度稍好,4阶模型模拟效果最好.结论 BIC准则判定得到的AR模型最优阶数偏低,但根据其最优阶数进行的模拟也可取得不错的模拟效果;模型阶数对模拟效果有很大影响,阶数过低将导致结果离散性较大,阶数过高并不一定取得更高的模拟精度,选择合理的模型阶数既可取得最优的模拟效果又可节省计算时间.更多还原

【Abstract】 Wind velocity time-history is simulated using multidimensional auto regressive model,where velocities are correlated in accordance with the Kaimal power spectrum.In order to determine the optimal order of the model,BIC criterion is utilized.The multidimensional AR model is programmed by MATLAB in which BIC criterion is embedded.Simulation performed on the wind field of Xihoumen bridge shows that the optimal order of AR model is 2,and the 2nd order model performs effectively by comparing the simulated power spectrum with the target one.Models of different orders have different performance:1st order model has larger discreteness than others;6th and 8th order models perform better than 1st and 2nd order models in low frequency region;4th order model performs best among all models.These results indicate that the optimal order determined by BIC criterion is lower than the actual one.Different order models can lead to much different results.When order is too low,the discreteness will be high.When order is too high,it can′t obtain higher accuracy but just waste time.An appropriate order of AR model is important to get the best result.更多还原