Abstract: Traffic noise was composed of time series, which was collected at certain section in urban road. Autocorrelation function was used to calculate the delay time. When the autocorrelation coefficient attenuated to 0.4, the delay time to reconstruct phase space was 12. The training set included two groups of noise samples was employed to calculate the correlation dimension. When embedding dimension was 6, the correlation dimension kept invariant. Therefore, the embedding dimension of system was 3 and traffic noise samples were of chaotic characteristics. Based on the reconstructed phase space, state points and near points, the fitting parameters were calculated. The two groups of traffic noise samples were predicted respectively by chaotic models and then compared the results with those got by the training set. The results show that chaotic models are suitable to predict the traffic noise of urban road in a short time. The average relative errors of two groups of noise samples were 8.56% and 9.33% individually.

Key words: traffic engineering, urban road, traffic noise, chaotic characteristics, phase space reconstruction

摘要： 通过采集城市道路特征断面的交通噪声，形成时间序列，采用自相关函数法计算延迟时间，当相关系数衰减为0.4时，计算得到相空间重构的延迟时间为12。利用两组噪声样本的训练 集计算关联维数，当嵌入维数为6时，关联维数保持不变，确定系统的嵌入维数为3，从而判定两组交通噪声样本存在混沌特征。根据重构的相空间、状态点及临近点计算拟合参数，分别对两 组噪声样本进行混沌预测，并与检验集比对，结果表明：混沌模型可以用于短时城市道路交通噪声预测，两组噪声样本的预测平均相对误差分别为8.56%和9.33%。

关键词: 交通工程, 城市道路, 交通噪声, 混沌特征, 相空间重构