A new methodology for multi-step-ahead forecasting was proposed herein which combined the wavelet transform(WT), artificial neural network(ANN) and forecasting strategies based on the changing characteristics of avail...A new methodology for multi-step-ahead forecasting was proposed herein which combined the wavelet transform(WT), artificial neural network(ANN) and forecasting strategies based on the changing characteristics of available parking spaces(APS). First, several APS time series were decomposed and reconstituted by the wavelet transform. Then, using an artificial neural network, the following five strategies for multi-step-ahead time series forecasting were used to forecast the reconstructed time series: recursive strategy, direct strategy, multi-input multi-output(MIMO) strategy, DIRMO strategy(a combination of the direct and MIMO strategies), and newly proposed recursive multi-input multi-output(RECMO) strategy which is a combination of the recursive and MIMO strategies. Finally, integrating the predicted results with the reconstructed time series produced the final forecasted available parking spaces. Three findings appear to be consistently supported by the experimental results. First, applying the wavelet transform to multi-step ahead available parking spaces forecasting can effectively improve the forecasting accuracy. Second, the forecasting resulted from the DIRMO and RECMO strategies is more accurate than that of the other strategies. Finally, the RECMO strategy requires less model training time than the DIRMO strategy and consumes the least amount of training time among five forecasting strategies.展开更多
深度学习(deep learning,DL)应用于推动智能故障诊断的发展,带来了显著的性能提升。然而,现有方法大多无法捕捉机械设备的时间信息和全局特征,无法收集到足够的故障信息。同时,由于运行环境复杂恶劣,单源故障诊断方法难以稳定、广泛地...深度学习(deep learning,DL)应用于推动智能故障诊断的发展,带来了显著的性能提升。然而,现有方法大多无法捕捉机械设备的时间信息和全局特征,无法收集到足够的故障信息。同时,由于运行环境复杂恶劣,单源故障诊断方法难以稳定、广泛地提取故障特征。因此,提出一种基于多源信息融合(multi-source information fusion,MSIF)的连续小波和TransXNet三相异步电机故障诊断新方法,通过提取和整合丰富的特征来提高诊断性能稳定性。首先,搭建三相异步电机故障实验平台,使用加速度传感器与电流传感器采集电机多种工况下的振动信号与电流信号,获得多源信号;其次,提出一种新的轻量级混合网络模块:双动态令牌混合器(dual dynamic token mixer,D-Mixer),其可以动态地利用全局和局部信息,同时注入大的感受野和强大的归纳偏差,而不牺牲输入依赖性,提高了特征提取能力。提出多尺度前馈网络(multi-scale feed-forward network,MS-FFN),在前馈网络中进行多尺度特征聚合。通过交替使用D-Mixer和MS-FFN,构建一种新型的混合CNN-Transformer网络:TransXNet;然后,利用连续小波变换将多源信号进行时频变换,提出数据级融合策略获得多源信息图,将多源信息图输入到TransXNet中进行特征分割以及聚合以完成特征提取,以训练并验证所提出的TransXNet有效性;最后,使用多源测试样本来验证所提出方法的诊断性能。结果表明,基于TransXNet强大的特征提取能力,识别准确率达到100%。通过对比调整兰德指数、归一化互信息、F1分数和准确率4个评价指标以及抗噪性分析,得出所提方法优于目前故障诊断领域最先进的方法(state of the art,SOTA),在故障诊断领域具有很好的前景。展开更多
Sonar images have complex background, low contrast, and deteriorative edges; these characteristics make it difficult for researchers to dispose the sonar objects. The multi-resolution analysis represents the signals i...Sonar images have complex background, low contrast, and deteriorative edges; these characteristics make it difficult for researchers to dispose the sonar objects. The multi-resolution analysis represents the signals in different scales efficiently, which is widely used in image processing. Wavelets are successful in disposing point discontinuities in one dimension, but not in two dimensions. The finite Ridgelet transform （FRIT） deals efficiently with the singularity in high dimension. It presents three improved denoising approaches, which are based on FRIT and used in the sonar image disposal technique. By experiment and comparison with traditional methods, these approaches not only suppress the artifacts, but also obtain good effect in edge keeping and SNR of the sonar image denoising.展开更多
基金Project(51561135003)supported by the International Cooperation and Exchange of the National Natural Science Foundation of ChinaProject(51338003)supported by the Key Project of National Natural Science Foundation of China
文摘A new methodology for multi-step-ahead forecasting was proposed herein which combined the wavelet transform(WT), artificial neural network(ANN) and forecasting strategies based on the changing characteristics of available parking spaces(APS). First, several APS time series were decomposed and reconstituted by the wavelet transform. Then, using an artificial neural network, the following five strategies for multi-step-ahead time series forecasting were used to forecast the reconstructed time series: recursive strategy, direct strategy, multi-input multi-output(MIMO) strategy, DIRMO strategy(a combination of the direct and MIMO strategies), and newly proposed recursive multi-input multi-output(RECMO) strategy which is a combination of the recursive and MIMO strategies. Finally, integrating the predicted results with the reconstructed time series produced the final forecasted available parking spaces. Three findings appear to be consistently supported by the experimental results. First, applying the wavelet transform to multi-step ahead available parking spaces forecasting can effectively improve the forecasting accuracy. Second, the forecasting resulted from the DIRMO and RECMO strategies is more accurate than that of the other strategies. Finally, the RECMO strategy requires less model training time than the DIRMO strategy and consumes the least amount of training time among five forecasting strategies.
文摘深度学习(deep learning,DL)应用于推动智能故障诊断的发展,带来了显著的性能提升。然而,现有方法大多无法捕捉机械设备的时间信息和全局特征,无法收集到足够的故障信息。同时,由于运行环境复杂恶劣,单源故障诊断方法难以稳定、广泛地提取故障特征。因此,提出一种基于多源信息融合(multi-source information fusion,MSIF)的连续小波和TransXNet三相异步电机故障诊断新方法,通过提取和整合丰富的特征来提高诊断性能稳定性。首先,搭建三相异步电机故障实验平台,使用加速度传感器与电流传感器采集电机多种工况下的振动信号与电流信号,获得多源信号;其次,提出一种新的轻量级混合网络模块:双动态令牌混合器(dual dynamic token mixer,D-Mixer),其可以动态地利用全局和局部信息,同时注入大的感受野和强大的归纳偏差,而不牺牲输入依赖性,提高了特征提取能力。提出多尺度前馈网络(multi-scale feed-forward network,MS-FFN),在前馈网络中进行多尺度特征聚合。通过交替使用D-Mixer和MS-FFN,构建一种新型的混合CNN-Transformer网络:TransXNet;然后,利用连续小波变换将多源信号进行时频变换,提出数据级融合策略获得多源信息图,将多源信息图输入到TransXNet中进行特征分割以及聚合以完成特征提取,以训练并验证所提出的TransXNet有效性;最后,使用多源测试样本来验证所提出方法的诊断性能。结果表明,基于TransXNet强大的特征提取能力,识别准确率达到100%。通过对比调整兰德指数、归一化互信息、F1分数和准确率4个评价指标以及抗噪性分析,得出所提方法优于目前故障诊断领域最先进的方法(state of the art,SOTA),在故障诊断领域具有很好的前景。
基金This project was supported by the National Natural Science Foundation of China (60672034)the Research Fund for the Doctoral Program of Higher Education(20060217021)the Natural Science Foundation of Heilongjiang Province of China (ZJG0606-01)
文摘Sonar images have complex background, low contrast, and deteriorative edges; these characteristics make it difficult for researchers to dispose the sonar objects. The multi-resolution analysis represents the signals in different scales efficiently, which is widely used in image processing. Wavelets are successful in disposing point discontinuities in one dimension, but not in two dimensions. The finite Ridgelet transform （FRIT） deals efficiently with the singularity in high dimension. It presents three improved denoising approaches, which are based on FRIT and used in the sonar image disposal technique. By experiment and comparison with traditional methods, these approaches not only suppress the artifacts, but also obtain good effect in edge keeping and SNR of the sonar image denoising.
