首页|改进的小波去噪算法与SSA预测模型在地铁监测中的应用

改进的小波去噪算法与SSA预测模型在地铁监测中的应用

扫码查看
为了有效去除地铁形变监测信号的噪声,提高形变监测精度,提出了一种以1/2 噪声幅度为阈值的小波去噪算法,对地铁形变实时监测信号进行去噪.结果表明,与硬阈值法、软阈值法和可调参数阈值函数法相比,1/2 阈值函数法去噪效果更好,平均SNR为24.627dB,与硬阈值法、软阈值法、可调参数阈值函数法相比分别提高了 3.15%、3.12%、1.07%;平均RMSE为 0.23mm,与硬阈值法、软阈值法、可调参数阈值函数法相比分别提高了 10.07%、6.47%、4.32%.采用奇异谱分析(Singular Spectrum Analysis,SSA)预测模型对监测数据进行了预测,预测精度,平均MAE为0.24mm、RMSE为0.26mm,就RMSE值而言,较BP神经网络提高了16.45%.
Application of improved wavelet denoising algorithm and SSA prediction model in subway monitoring
In order to effectively remove noise from subway deformation monitoring signals and improve deformation monito-ring accuracy,this paper designs a wavelet denoising algorithm with a threshold of 1/2 noise amplitude to denoise real-time subway displacement monitoring signals.The research results show that compared with hard threshold method,soft threshold method,and adjustable parameter threshold function method,the 1/2 threshold function method has the best denoising effect,with an average SNR of 24.627dB.Compared with hard threshold method,soft threshold method,and adjustable pa-rameter threshold function method,it has increased by 3.15%,3.12%,and 1.07%,respectively.The average RMSE is 0.23mm,which is10.07%,6.47%,and4.32%higher than the hard threshold method,soft threshold method,and adjustable parameter threshold function method,respectively.The singular spectrum analysis(SSA)prediction model was used to pre-dict the monitoring data.The prediction accuracy was as follows:the average MAE was 0.24mm,and the RMSE value was 0.26mm.In terms of RMSE value,it improved by 16.45%compared to the BP neural network.

waveletSSA prediction modelBP neural network

朱增洪、孔晓宇

展开 >

江西省建筑设计研究总院集团有限公司,江西南昌 330046

小波 SSA预测模型 BP神经网络

2023

南昌工程学院学报
南昌工程学院

南昌工程学院学报

影响因子:0.272
ISSN:1006-4869
年,卷(期):2023.42(6)
  • 8