CEEMD-GRU组合道路噪声预测模型  

Road noise prediction model with CEEMD GRU combination

作　　者：冯增喜[1] 崔巍[1] 何鑫 赵锦彤 孙欣 张茂强 杨芸芸 韦娜[2] FENG Zengxi;CUI Wei;HE Xin;ZHAO Jin-tong;SUN Xin;ZHANG Maoqiang;YANG Yunyun;WEI Na(School of Building Services Science and Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China;Academy of Arts,Xi'an University of Architecture and Technology,Xi'an 710055,China)

机构地区：[1]西安建筑科技大学建筑设备科学与工程学院,西安710055 [2]西安建筑科技大学艺术学院,西安710055

出　　处：《安全与环境学报》2023年第6期2128-2136,共9页Journal of Safety and Environment

基　　金：国家重点研发计划项目(2017YFC070410702)。

摘　　要：针对道路降噪问题,为降低主动噪声控制方法中滤波算法收敛性能要求,提出了一种基于互补集合经验模态分解(Complemementary Ensemble Empirical Mode Decomposition,CEEMD)与门控循环单元(Gated Recurrent Unit,GRU)组合的道路噪声预测模型(CEEMD-GRU)。首先,基于CEEMD算法将输入的噪声音频序列分解为多个本征模态函数分量和一个残差分量,以深度挖掘数据隐含的波动信息;其次,利用CEEMD分解的输入噪声序列各分量和输出噪声序列构建CEEMD-GRU神经网络噪声预测模型;最后,基于西安市某路段采集的噪声数据对该模型的有效性进行验证。结果表明:该模型EMA为0.0191,RMSE为0.0308,R^(2)为0.5892,预测声信号能够代替主动噪声控制中自适应控制器的实际初级声信号,为主动噪声控制的控制过程提供了更充分的响应时间。A road noise prediction model(CEEMD GRU)based on the combination of complementary ensemble empirical mode decomposition(CEEMD)and gated recurrent unit(GRU)is proposed to reduce the requirements for the convergence performance of the filtering algorithm in the traditional active noise control method,solve the secondary acoustic feedback in the control process,and provide more adequate preparation for the subsequent active noise control.Firstly,the model decomposes the input noise signal sequence into multiple eigenmode function components and a residual component based on the CEEMD algorithm to deeply mine the fluctuation information implicit in the data.Secondly,the multiple input noise sequence eigenmode function components decomposed by the CEEMD algorithm and the output noise sequence are used to construct the CEEMD GRU neural network noise prediction model.Finally,based on the noise data collected from a road section in Xi'an,the validity of the model is verified and analyzed,and the prediction effects of various models are compared.The experimental results show that the average absolute error EMA of the prediction results of the model is 0.0191,which is respectively 61.2%,55.0%,and 42.9%lower than that of the BP neural network model,the GRU neural network model and the EMD GRU combined prediction model.The root mean square error ERMS is 0.0308,which is respectively 41.4%,34.9%,and 23.4%lower than the BP neural network model.The R2 of the prediction results of the model is 0.5892,which is respectively 89.1%,47.9%,and 18.4%higher than that of the BP neural network model,the GRU neural network model,and the EMD GRU combined prediction model.Therefore,this model can show a better prediction effect than BP neural network model,GRU neural network model and EMD GRU combined prediction model,and the predicted acoustic signal can replace the actual primary acoustic signal of the adaptive controller in active noise control.The control execution process of the noise control provides a more sufficient response time and

关 键 词：环境工程学 噪声与振动控制 噪声预测 主动噪声控制 门控循环单元 互补集合经验模态分解 

分 类 号：X593[环境科学与工程—环境工程]