Deep neural networks have been widely applied to bearing fault diagnosis systems and achieved impressive success recently.To address the problem that the insufficient fault feature extraction ability of traditional fa...Deep neural networks have been widely applied to bearing fault diagnosis systems and achieved impressive success recently.To address the problem that the insufficient fault feature extraction ability of traditional fault diagnosis methods results in poor diagnosis effect under variable load and noise interference scenarios,a rolling bearing fault diagnosis model combining Multi-Scale Convolutional Neural Network(MSCNN)and Long Short-Term Memory(LSTM)fused with attention mechanism is proposed.To adaptively extract the essential spatial feature information of various sizes,the model creates a multi-scale feature extraction module using the convolutional neural network(CNN)learning process.The learning capacity of LSTM for time information sequence is then used to extract the vibration signal’s temporal feature information.Two parallel large and small convolutional kernels teach the system spatial local features.LSTM gathers temporal global features to thoroughly and painstakingly mine the vibration signal’s characteristics,thus enhancing model generalization.Lastly,bearing fault diagnosis is accomplished by using the SoftMax classifier.The experiment outcomes demonstrate that the model can derive fault properties entirely from the initial vibration signal.It can retain good diagnostic accuracy under variable load and noise interference and has strong generalization compared to other fault diagnosis models.展开更多
Space-time video super-resolution(STVSR)serves the purpose to reconstruct high-resolution high-frame-rate videos from their low-resolution low-frame-rate counterparts.Recent approaches utilize end-to-end deep learning...Space-time video super-resolution(STVSR)serves the purpose to reconstruct high-resolution high-frame-rate videos from their low-resolution low-frame-rate counterparts.Recent approaches utilize end-to-end deep learning models to achieve STVSR.They first interpolate intermediate frame features between given frames,then perform local and global refinement among the feature sequence,and finally increase the spatial resolutions of these features.However,in the most important feature interpolation phase,they only capture spatial-temporal information from the most adjacent frame features,ignoring modelling long-term spatial-temporal correlations between multiple neighbouring frames to restore variable-speed object movements and maintain long-term motion continuity.In this paper,we propose a novel long-term temporal feature aggregation network(LTFA-Net)for STVSR.Specifically,we design a long-term mixture of experts(LTMoE)module for feature interpolation.LTMoE contains multiple experts to extract mutual and complementary spatial-temporal information from multiple consecutive adjacent frame features,which are then combined with different weights to obtain interpolation results using several gating nets.Next,we perform local and global feature refinement using the Locally-temporal Feature Comparison(LFC)module and bidirectional deformable ConvLSTM layer,respectively.Experimental results on two standard benchmarks,Adobe240 and GoPro,indicate the effectiveness and superiority of our approach over state of the art.展开更多
针对长时目标跟踪所面临的目标被遮挡、出视野等常常会导致跟踪漂移或丢失的问题,基于MDNet提出一种深度长时目标跟踪算法(long-term object tracking based on MDNet,LT-MDNet)。首先,引入了一种改进的收缩损失函数,以解决模型训练时...针对长时目标跟踪所面临的目标被遮挡、出视野等常常会导致跟踪漂移或丢失的问题,基于MDNet提出一种深度长时目标跟踪算法(long-term object tracking based on MDNet,LT-MDNet)。首先,引入了一种改进的收缩损失函数,以解决模型训练时正负样本不均衡的问题;其次,设计了一种高置信度保留样本池,对在线跟踪时的每一帧的有效并且置信度最高结果进行保留,并在池满时替换最低置信度的保留样本;最后,在模型检测到跟踪失败或连续跟踪帧数达到特定阈值时,利用保留样本池进行在线训练更新模型,从而使模型在应对长时跟踪时保持鲁棒和高效。实验结果表明,LT-MDNet在跟踪精度和成功率上都展现了极强的竞争力,并且在目标被遮挡、出视野等情况下保持了优越的跟踪性能和可靠性。展开更多
To promote urban sustainability and resilience,there is an increasing demand for actionable science that links science and decision making based on social-ecological knowledge.Approaches,frameworks,and practices for s...To promote urban sustainability and resilience,there is an increasing demand for actionable science that links science and decision making based on social-ecological knowledge.Approaches,frameworks,and practices for such actionable science are needed and have only begun to emerge.We propose that approaches based on the co-design and co-production of knowledge can play an essential role to meet this demand.Although the antecedents for approaches to the co-design and co-production of knowledge are decades old,the integration of science and practice to advance urban sustainability and resilience that we present is different in several ways.These differences include the disciplines needed,diversity and number of actors involved,and the technological infrastructures that facilitate local-to global connections.In this article,we discuss how the new requirements and possibilities for co-design,co-production,and practical use of social-ecological research can be used as an ecology for the city to promote urban sustainability and resilience.While new technologies are part of the solution,traditional approaches also remain important.Using our urban experiences with long-term,place-based research from several U.S.Long-Term Ecological Research sites and U.S.Department of Agriculture,Forest Service Urban Field Stations,we describe a dynamic framework for linking research with decisions.We posit that this framework,coupled with a user-defined,theory-based approach to science,is instrumental to advance both practice and science.Ultimately,cities are ideal places for integrating basic science and decision making,facilitating flows of information through networks,and developing sustainable and resilient solutions and futures.展开更多
文摘Deep neural networks have been widely applied to bearing fault diagnosis systems and achieved impressive success recently.To address the problem that the insufficient fault feature extraction ability of traditional fault diagnosis methods results in poor diagnosis effect under variable load and noise interference scenarios,a rolling bearing fault diagnosis model combining Multi-Scale Convolutional Neural Network(MSCNN)and Long Short-Term Memory(LSTM)fused with attention mechanism is proposed.To adaptively extract the essential spatial feature information of various sizes,the model creates a multi-scale feature extraction module using the convolutional neural network(CNN)learning process.The learning capacity of LSTM for time information sequence is then used to extract the vibration signal’s temporal feature information.Two parallel large and small convolutional kernels teach the system spatial local features.LSTM gathers temporal global features to thoroughly and painstakingly mine the vibration signal’s characteristics,thus enhancing model generalization.Lastly,bearing fault diagnosis is accomplished by using the SoftMax classifier.The experiment outcomes demonstrate that the model can derive fault properties entirely from the initial vibration signal.It can retain good diagnostic accuracy under variable load and noise interference and has strong generalization compared to other fault diagnosis models.
文摘Space-time video super-resolution(STVSR)serves the purpose to reconstruct high-resolution high-frame-rate videos from their low-resolution low-frame-rate counterparts.Recent approaches utilize end-to-end deep learning models to achieve STVSR.They first interpolate intermediate frame features between given frames,then perform local and global refinement among the feature sequence,and finally increase the spatial resolutions of these features.However,in the most important feature interpolation phase,they only capture spatial-temporal information from the most adjacent frame features,ignoring modelling long-term spatial-temporal correlations between multiple neighbouring frames to restore variable-speed object movements and maintain long-term motion continuity.In this paper,we propose a novel long-term temporal feature aggregation network(LTFA-Net)for STVSR.Specifically,we design a long-term mixture of experts(LTMoE)module for feature interpolation.LTMoE contains multiple experts to extract mutual and complementary spatial-temporal information from multiple consecutive adjacent frame features,which are then combined with different weights to obtain interpolation results using several gating nets.Next,we perform local and global feature refinement using the Locally-temporal Feature Comparison(LFC)module and bidirectional deformable ConvLSTM layer,respectively.Experimental results on two standard benchmarks,Adobe240 and GoPro,indicate the effectiveness and superiority of our approach over state of the art.
文摘针对长时目标跟踪所面临的目标被遮挡、出视野等常常会导致跟踪漂移或丢失的问题,基于MDNet提出一种深度长时目标跟踪算法(long-term object tracking based on MDNet,LT-MDNet)。首先,引入了一种改进的收缩损失函数,以解决模型训练时正负样本不均衡的问题;其次,设计了一种高置信度保留样本池,对在线跟踪时的每一帧的有效并且置信度最高结果进行保留,并在池满时替换最低置信度的保留样本;最后,在模型检测到跟踪失败或连续跟踪帧数达到特定阈值时,利用保留样本池进行在线训练更新模型,从而使模型在应对长时跟踪时保持鲁棒和高效。实验结果表明,LT-MDNet在跟踪精度和成功率上都展现了极强的竞争力,并且在目标被遮挡、出视野等情况下保持了优越的跟踪性能和可靠性。
基金support from the Urban Sustainability Research Coordination Network(National Science Foundation Grant No.1140070)Childers received support from the Central Arizona-Phoenix Long-Term Ecological Research Program(National Science Foundation Grant No.DEB-1027188)Grove received support from the Baltimore Ecosystem Study Long-Term Ecological Research Program(National Science Foundation Grant No.DEB-1027188).
文摘To promote urban sustainability and resilience,there is an increasing demand for actionable science that links science and decision making based on social-ecological knowledge.Approaches,frameworks,and practices for such actionable science are needed and have only begun to emerge.We propose that approaches based on the co-design and co-production of knowledge can play an essential role to meet this demand.Although the antecedents for approaches to the co-design and co-production of knowledge are decades old,the integration of science and practice to advance urban sustainability and resilience that we present is different in several ways.These differences include the disciplines needed,diversity and number of actors involved,and the technological infrastructures that facilitate local-to global connections.In this article,we discuss how the new requirements and possibilities for co-design,co-production,and practical use of social-ecological research can be used as an ecology for the city to promote urban sustainability and resilience.While new technologies are part of the solution,traditional approaches also remain important.Using our urban experiences with long-term,place-based research from several U.S.Long-Term Ecological Research sites and U.S.Department of Agriculture,Forest Service Urban Field Stations,we describe a dynamic framework for linking research with decisions.We posit that this framework,coupled with a user-defined,theory-based approach to science,is instrumental to advance both practice and science.Ultimately,cities are ideal places for integrating basic science and decision making,facilitating flows of information through networks,and developing sustainable and resilient solutions and futures.
