The numerical simulation and slope stability prediction are the focus of slope disaster research.Recently,machine learning models are commonly used in the slope stability prediction.However,these machine learning mode...The numerical simulation and slope stability prediction are the focus of slope disaster research.Recently,machine learning models are commonly used in the slope stability prediction.However,these machine learning models have some problems,such as poor nonlinear performance,local optimum and incomplete factors feature extraction.These issues can affect the accuracy of slope stability prediction.Therefore,a deep learning algorithm called Long short-term memory(LSTM)has been innovatively proposed to predict slope stability.Taking the Ganzhou City in China as the study area,the landslide inventory and their characteristics of geotechnical parameters,slope height and slope angle are analyzed.Based on these characteristics,typical soil slopes are constructed using the Geo-Studio software.Five control factors affecting slope stability,including slope height,slope angle,internal friction angle,cohesion and volumetric weight,are selected to form different slope and construct model input variables.Then,the limit equilibrium method is used to calculate the stability coefficients of these typical soil slopes under different control factors.Each slope stability coefficient and its corresponding control factors is a slope sample.As a result,a total of 2160 training samples and 450 testing samples are constructed.These sample sets are imported into LSTM for modelling and compared with the support vector machine(SVM),random forest(RF)and convo-lutional neural network(CNN).The results show that the LSTM overcomes the problem that the commonly used machine learning models have difficulty extracting global features.Furthermore,LSTM has a better prediction performance for slope stability compared to SVM,RF and CNN models.展开更多
As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge device...As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.展开更多
Long-term navigation ability based on consumer-level wearable inertial sensors plays an essential role towards various emerging fields, for instance, smart healthcare, emergency rescue, soldier positioning et al. The ...Long-term navigation ability based on consumer-level wearable inertial sensors plays an essential role towards various emerging fields, for instance, smart healthcare, emergency rescue, soldier positioning et al. The performance of existing long-term navigation algorithm is limited by the cumulative error of inertial sensors, disturbed local magnetic field, and complex motion modes of the pedestrian. This paper develops a robust data and physical model dual-driven based trajectory estimation(DPDD-TE) framework, which can be applied for long-term navigation tasks. A Bi-directional Long Short-Term Memory(Bi-LSTM) based quasi-static magnetic field(QSMF) detection algorithm is developed for extracting useful magnetic observation for heading calibration, and another Bi-LSTM is adopted for walking speed estimation by considering hybrid human motion information under a specific time period. In addition, a data and physical model dual-driven based multi-source fusion model is proposed to integrate basic INS mechanization and multi-level constraint and observations for maintaining accuracy under long-term navigation tasks, and enhanced by the magnetic and trajectory features assisted loop detection algorithm. Real-world experiments indicate that the proposed DPDD-TE outperforms than existing algorithms, and final estimated heading and positioning accuracy indexes reaches 5° and less than 2 m under the time period of 30 min, respectively.展开更多
Recent advancements in natural language processing have given rise to numerous pre-training language models in question-answering systems.However,with the constant evolution of algorithms,data,and computing power,the ...Recent advancements in natural language processing have given rise to numerous pre-training language models in question-answering systems.However,with the constant evolution of algorithms,data,and computing power,the increasing size and complexity of these models have led to increased training costs and reduced efficiency.This study aims to minimize the inference time of such models while maintaining computational performance.It also proposes a novel Distillation model for PAL-BERT(DPAL-BERT),specifically,employs knowledge distillation,using the PAL-BERT model as the teacher model to train two student models:DPAL-BERT-Bi and DPAL-BERTC.This research enhances the dataset through techniques such as masking,replacement,and n-gram sampling to optimize knowledge transfer.The experimental results showed that the distilled models greatly outperform models trained from scratch.In addition,although the distilled models exhibit a slight decrease in performance compared to PAL-BERT,they significantly reduce inference time to just 0.25%of the original.This demonstrates the effectiveness of the proposed approach in balancing model performance and efficiency.展开更多
The complexity of river-tide interaction poses a significant challenge in predicting discharge in tidal rivers.Long short-term memory(LSTM)networks excel in processing and predicting crucial events with extended inter...The complexity of river-tide interaction poses a significant challenge in predicting discharge in tidal rivers.Long short-term memory(LSTM)networks excel in processing and predicting crucial events with extended intervals and time delays in time series data.Additionally,the sequence-to-sequence(Seq2Seq)model,known for handling temporal relationships,adapting to variable-length sequences,effectively capturing historical information,and accommodating various influencing factors,emerges as a robust and flexible tool in discharge forecasting.In this study,we introduce the application of LSTM-based Seq2Seq models for the first time in forecasting the discharge of a tidal reach of the Changjiang River(Yangtze River)Estuary.This study focuses on discharge forecasting using three key input characteristics:flow velocity,water level,and discharge,which means the structure of multiple input and single output is adopted.The experiment used the discharge data of the whole year of 2020,of which the first 80%is used as the training set,and the last 20%is used as the test set.This means that the data covers different tidal cycles,which helps to test the forecasting effect of different models in different tidal cycles and different runoff.The experimental results indicate that the proposed models demonstrate advantages in long-term,mid-term,and short-term discharge forecasting.The Seq2Seq models improved by 6%-60%and 5%-20%of the relative standard deviation compared to the harmonic analysis models and improved back propagation neural network models in discharge prediction,respectively.In addition,the relative accuracy of the Seq2Seq model is 1%to 3%higher than that of the LSTM model.Analytical assessment of the prediction errors shows that the Seq2Seq models are insensitive to the forecast lead time and they can capture characteristic values such as maximum flood tide flow and maximum ebb tide flow in the tidal cycle well.This indicates the significance of the Seq2Seq models.展开更多
Hydrological models are developed to simulate river flows over a watershed for many practical applications in the field of water resource management. The present paper compares the performance of two recurrent neural ...Hydrological models are developed to simulate river flows over a watershed for many practical applications in the field of water resource management. The present paper compares the performance of two recurrent neural networks for rainfall-runoff modeling in the Zou River basin at Atchérigbé outlet. To this end, we used daily precipitation data over the period 1988-2010 as input of the models, such as the Long Short-Term Memory (LSTM) and Recurrent Gate Networks (GRU) to simulate river discharge in the study area. The investigated models give good results in calibration (R2 = 0.888, NSE = 0.886, and RMSE = 0.42 for LSTM;R2 = 0.9, NSE = 0.9 and RMSE = 0.397 for GRU) and in validation (R2 = 0.865, NSE = 0.851, and RMSE = 0.329 for LSTM;R2 = 0.9, NSE = 0.865 and RMSE = 0.301 for GRU). This good performance of LSTM and GRU models confirms the importance of models based on machine learning in modeling hydrological phenomena for better decision-making.展开更多
The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the ...The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the inception of financial markets.By examining historical transaction data,latent opportunities for profit can be uncovered,providing valuable insights for both institutional and individual investors to make more informed decisions.This study focuses on analyzing historical transaction data from four banks to predict closing price trends.Various models,including decision trees,random forests,and Long Short-Term Memory(LSTM)networks,are employed to forecast stock price movements.Historical stock transaction data serves as the input for training these models,which are then used to predict upward or downward stock price trends.The study’s empirical results indicate that these methods are effective to a degree in predicting stock price movements.The LSTM-based deep neural network model,in particular,demonstrates a commendable level of predictive accuracy.This conclusion is reached following a thorough evaluation of model performance,highlighting the potential of LSTM models in stock market forecasting.The findings offer significant implications for advancing financial forecasting approaches,thereby improving the decision-making capabilities of investors and financial institutions.展开更多
A Long Short-Term Memory(LSTM) Recurrent Neural Network(RNN) has driven tremendous improvements on an acoustic model based on Gaussian Mixture Model(GMM). However, these models based on a hybrid method require a force...A Long Short-Term Memory(LSTM) Recurrent Neural Network(RNN) has driven tremendous improvements on an acoustic model based on Gaussian Mixture Model(GMM). However, these models based on a hybrid method require a forced aligned Hidden Markov Model(HMM) state sequence obtained from the GMM-based acoustic model. Therefore, it requires a long computation time for training both the GMM-based acoustic model and a deep learning-based acoustic model. In order to solve this problem, an acoustic model using CTC algorithm is proposed. CTC algorithm does not require the GMM-based acoustic model because it does not use the forced aligned HMM state sequence. However, previous works on a LSTM RNN-based acoustic model using CTC used a small-scale training corpus. In this paper, the LSTM RNN-based acoustic model using CTC is trained on a large-scale training corpus and its performance is evaluated. The implemented acoustic model has a performance of 6.18% and 15.01% in terms of Word Error Rate(WER) for clean speech and noisy speech, respectively. This is similar to a performance of the acoustic model based on the hybrid method.展开更多
In view of the fact that the atmospheric motion is an irreversible process, a memory function which can recall the observation data in the past is introduced, moreover, a special concept of self-memorization of the at...In view of the fact that the atmospheric motion is an irreversible process, a memory function which can recall the observation data in the past is introduced, moreover, a special concept of self-memorization of the atmospheric motion is proposed, and a so-called self-memorization equation of the atmospheric motion has been derived. Based on the self-memorization principle, a numerical model for decadal forecast is established by means of the thermodynamic equation and the precipitation equation. The verification scores of the hindcasts of the model in the period from 1 to 12 years are much higher than that of monthly weather forecasts at present.展开更多
A macroscopic based multi-mechanism constitutive model is constructed in the framework of irreversible thermodynamics to describe the degeneration of shape memory effect occurring in the thermo-mechanical cyclic defor...A macroscopic based multi-mechanism constitutive model is constructed in the framework of irreversible thermodynamics to describe the degeneration of shape memory effect occurring in the thermo-mechanical cyclic deformation of NiTi shape memory alloys (SMAs). Three phases, austenite A, twinned martensite and detwinned martensite , as well as the phase transitions occurring between each pair of phases (, , , , and are considered in the proposed model. Meanwhile, two kinds of inelastic deformation mechanisms, martensite transformation-induced plasticity and reorientation-induced plasticity, are used to explain the degeneration of shape memory effects of NiTi SMAs. The evolution equations of internal variables are proposed by attributing the degeneration of shape memory effect to the interaction between the three phases (A, , and and plastic deformation. Finally, the capability of the proposed model is verified by comparing the predictions with the experimental results of NiTi SMAs. It is shown that the degeneration of shape memory effect and its dependence on the loading level can be reasonably described by the proposed model.展开更多
The objective of this paper is to model the size-dependent thermo-mechanical behaviors of a shape memory polymer (SMP) microbeam.Size-dependent constitutive equations,which can capture the size effect of the SMP,are p...The objective of this paper is to model the size-dependent thermo-mechanical behaviors of a shape memory polymer (SMP) microbeam.Size-dependent constitutive equations,which can capture the size effect of the SMP,are proposed based on the modified couple stress theory (MCST).The deformation energy expression of the SMP microbeam is obtained by employing the proposed size-dependent constitutive equation and Bernoulli-Euler beam theory.An SMP microbeam model,which includes the formulations of deflection,strain,curvature,stress and couple stress,is developed by using the principle of minimum potential energy and the separation of variables together.The sizedependent thermo-mechanical and shape memory behaviors of the SMP microbeam and the influence of the Poisson ratio are numerically investigated according to the developed SMP microbeam model.Results show that the size effects of the SMP microbeam are significant when the dimensionless height is small enough.However,they are too slight to be necessarily considered when the dimensionless height is large enough.The bending flexibility and stress level of the SMP microbeam rise with the increasing dimensionless height,while the couple stress level declines with the increasing dimensionless height.The larger the dimensionless height is,the more obvious the viscous property and shape memory effect of the SMP microbeam are.The Poisson ratio has obvious influence on the size-dependent behaviors of the SMP microbeam.The paper provides a theoretical basis and a quantitatively analyzing tool for the design and analysis of SMP micro-structures in the field of biological medicine,microelectronic devices and micro-electro-mechanical system (MEMS) self-assembling.展开更多
In order to propel the development of metal magnetic memory （MMM） technique in fatigue damage detection, the Jiles-Atherton model （J-A model） was modified to describe MMM mechanism in elastic stress stage. A serie...In order to propel the development of metal magnetic memory （MMM） technique in fatigue damage detection, the Jiles-Atherton model （J-A model） was modified to describe MMM mechanism in elastic stress stage. A series of rotating bending fatigue experiments were conducted to study the stress-magnetization relationship and verify the correctness of modified J-A model. In MMM detection, the magnetization of material irreversibly approaches to the local equilibrium state Mo instead of global equilibrium state M^n under cyclic stress, and the M0-a curves are loops around the Mar,-a curve. The modified J-A model is constructed by replacing M~ in J-A model with M0, and it can describe the magnetomechanical effect well at low external magnetic field. In the rotating bending fatigue experiments, the MMM field distribution in normal direction around cylinder specimen is similar to the stress distribution, and the calculation result of model coincides with experiment result after some necessary modifications. The MMM field variation with time at a certain point in fatigue process is divided into three stages with the variation of stable stress-stain hysteresis loop, and the calculation results of model can explain not only the three stages of MMM field changes, but also the different change laws when the applied magnetic field and initial magnetic field are different. The MMM field distribution in normal direction along specimen axis reflects stress concentration effect at artificial defect, and the magnetic signal fluctuates around the defect at late fatigue stage. The calculation results coincide with the initial MMM principle and can explain signal fluctuates around the defect. The modified J-A model can explain experiment results well, and it is fit for MMM field characterization.展开更多
Metal magnetic memory（MMM） testing has been widely used to detect welded joints. However, load levels, environmental magnetic field, and measurement noises make the MMM data dispersive and bring difficulty to quanti...Metal magnetic memory（MMM） testing has been widely used to detect welded joints. However, load levels, environmental magnetic field, and measurement noises make the MMM data dispersive and bring difficulty to quantitative evaluation. In order to promote the development of quantitative MMM reliability assessment, a new MMM model is presented for welded joints. Steel Q235 welded specimens are tested along the longitudinal and horizontal lines by TSC-2M-8 instrument in the tensile fatigue experiments. The X-ray testing is carried out synchronously to verify the MMM results. It is found that MMM testing can detect the hidden crack earlier than X-ray testing. Moreover, the MMM gradient vector sum K_（vs） is sensitive to the damage degree, especially at early and hidden damage stages. Considering the dispersion of MMM data, the K_（vs） statistical law is investigated, which shows that K_（vs） obeys Gaussian distribution. So K_（vs） is the suitable MMM parameter to establish reliability model of welded joints. At last, the original quantitative MMM reliability model is first presented based on the improved stress strength interference theory. It is shown that the reliability degree R gradually decreases with the decreasing of the residual life ratio T, and the maximal error between prediction reliability degree R_1 and verification reliability degree R_2 is 9.15%. This presented method provides a novel tool of reliability testing and evaluating in practical engineering for welded joints.展开更多
This article examines some general atmospheric circulation and climate models in the context of the notion of “memory”. Two kinds of memories are defined： statistical memory and deterministic memory. The former is ...This article examines some general atmospheric circulation and climate models in the context of the notion of “memory”. Two kinds of memories are defined： statistical memory and deterministic memory. The former is defined through the autocorrelation characteristic of the process if it is random （chaotic）, while for the latter, a special memory function is introduced. Three of the numerous existing models are selected as examples. For each of the models, asymptotic （at t →∞） expressions are derived. In this way, the transients are filtered out and that which remains concerns the final behaviour of the models.展开更多
A thermoviscoelastic modeling approach is developed to predict the recovery behaviors of the thermally activated amorphous shape memory polymers(SMPs)based on the generalized finite deformation viscoelasticity theory....A thermoviscoelastic modeling approach is developed to predict the recovery behaviors of the thermally activated amorphous shape memory polymers(SMPs)based on the generalized finite deformation viscoelasticity theory.In this paper,a series of moduli and relaxation times of the generalized Maxwell model is estimated from the stress relaxation master curve by using the nonlinear regression(NLREG)method.Assuming that the amorphous SMPs are approximately incompressible isotropic elastomers in the rubbery state,the hyperelastic response of the materials is well modeled with a hyperelastic model in Ogden form.In addition,the Williams-Landel-Ferry(WLF)equation is used to describe the horizontal shift factor obtained with time-temperature superposition principle(TTSP).The finite element simulations show good agreement with the experimental thermomechanical behaviors.Moreover,the possibility of developing a temperature-responsive intravascular stent with the SMP studied here is investigated in terms of its thermomechanical property.Therefore,it can be concluded that the model has good prediction capabilities for the recovery behaviors of amorphous SMPs.展开更多
The hysteresis characteristic is the major deficiency in the positioning control of magnetic shape memory alloy actuator. A Prandtl-Ishlinskii model was developed to characterize the hysteresis of magnetic shape memor...The hysteresis characteristic is the major deficiency in the positioning control of magnetic shape memory alloy actuator. A Prandtl-Ishlinskii model was developed to characterize the hysteresis of magnetic shape memory alloy actuator. Based on the proposed Prandtl-Ishlinskii model, the inverse Prandtl-Ishlinskii model was established as a feedforward controller to compensate the hysteresis of the magnetic shape memory alloy actuator. For further improving of the positioning precision of the magnetic shape memory alloy actuator, a hybrid control method with hysteresis nonlinear model in feedforward loop was proposed. The control method is separated into two parts: a feedforward loop with inverse Prandtl-Ishlinskii model and a feedback loop with neural network controller. To validate the validity of the proposed control method, a series of simulations and experiments were researched. The simulation and experimental results demonstrate that the maximum error rate of open loop controller based on inverse PI model is 1.72%, the maximum error rate of the hybrid controller based on inverse PI model is 1.37%.展开更多
Existing experimental results have shown that four types of physical mechanisms, namely, martensite transformation, martensite reorientation, magnetic domain wall motion and magnetization vector rotation, can be activ...Existing experimental results have shown that four types of physical mechanisms, namely, martensite transformation, martensite reorientation, magnetic domain wall motion and magnetization vector rotation, can be activated during the magneto-mechanical deformation of NiMnGa ferromagnetic shape memory alloy (FSMA) single crystals. In this work, based on irreversible thermodynamics, a three-dimensional (3D) single crystal constitutive model is constructed by considering the aforementioned four mechanisms simultaneously. Three types of internal variables, i.e., the volume fraction of each martensite variant, the volume fraction of magnetic domain in each variant and the deviation angle between the magnetization vector, and easy axis are introduced to characterize the magneto-mechanical state of the single crystals. The thermodynamic driving force of each mechanism and the thermodynamic constraints on the constitutive model are obtained from Clausius's dissipative inequality and constructed Gibbs free energy. Then, thermodynamically consistent kinetic equations for the four mechanisms are proposed, respectively. Finally, the ability of the proposed model to describe the magneto-mechanical deformation of NiMnGa FSMA single crystals is verified by comparing the predictions with corresponding experimental results. It is shown that the proposed model can quantitatively capture the main experimental phenomena. Further, the proposed model is used to predict the deformations of the single crystals under the non-proportional mechanical loading conditions.展开更多
基金funded by the National Natural Science Foundation of China (41807285)。
文摘The numerical simulation and slope stability prediction are the focus of slope disaster research.Recently,machine learning models are commonly used in the slope stability prediction.However,these machine learning models have some problems,such as poor nonlinear performance,local optimum and incomplete factors feature extraction.These issues can affect the accuracy of slope stability prediction.Therefore,a deep learning algorithm called Long short-term memory(LSTM)has been innovatively proposed to predict slope stability.Taking the Ganzhou City in China as the study area,the landslide inventory and their characteristics of geotechnical parameters,slope height and slope angle are analyzed.Based on these characteristics,typical soil slopes are constructed using the Geo-Studio software.Five control factors affecting slope stability,including slope height,slope angle,internal friction angle,cohesion and volumetric weight,are selected to form different slope and construct model input variables.Then,the limit equilibrium method is used to calculate the stability coefficients of these typical soil slopes under different control factors.Each slope stability coefficient and its corresponding control factors is a slope sample.As a result,a total of 2160 training samples and 450 testing samples are constructed.These sample sets are imported into LSTM for modelling and compared with the support vector machine(SVM),random forest(RF)and convo-lutional neural network(CNN).The results show that the LSTM overcomes the problem that the commonly used machine learning models have difficulty extracting global features.Furthermore,LSTM has a better prediction performance for slope stability compared to SVM,RF and CNN models.
基金supported by the National Natural Science Foundation of China(U21A20519)。
文摘As a large amount of data is increasingly generated from edge devices,such as smart homes,mobile phones,and wearable devices,it becomes crucial for many applications to deploy machine learning modes across edge devices.The execution speed of the deployed model is a key element to ensure service quality.Considering a highly heterogeneous edge deployment scenario,deep learning compiling is a novel approach that aims to solve this problem.It defines models using certain DSLs and generates efficient code implementations on different hardware devices.However,there are still two aspects that are not yet thoroughly investigated yet.The first is the optimization of memory-intensive operations,and the second problem is the heterogeneity of the deployment target.To that end,in this work,we propose a system solution that optimizes memory-intensive operation,optimizes the subgraph distribution,and enables the compiling and deployment of DNN models on multiple targets.The evaluation results show the performance of our proposed system.
文摘Long-term navigation ability based on consumer-level wearable inertial sensors plays an essential role towards various emerging fields, for instance, smart healthcare, emergency rescue, soldier positioning et al. The performance of existing long-term navigation algorithm is limited by the cumulative error of inertial sensors, disturbed local magnetic field, and complex motion modes of the pedestrian. This paper develops a robust data and physical model dual-driven based trajectory estimation(DPDD-TE) framework, which can be applied for long-term navigation tasks. A Bi-directional Long Short-Term Memory(Bi-LSTM) based quasi-static magnetic field(QSMF) detection algorithm is developed for extracting useful magnetic observation for heading calibration, and another Bi-LSTM is adopted for walking speed estimation by considering hybrid human motion information under a specific time period. In addition, a data and physical model dual-driven based multi-source fusion model is proposed to integrate basic INS mechanization and multi-level constraint and observations for maintaining accuracy under long-term navigation tasks, and enhanced by the magnetic and trajectory features assisted loop detection algorithm. Real-world experiments indicate that the proposed DPDD-TE outperforms than existing algorithms, and final estimated heading and positioning accuracy indexes reaches 5° and less than 2 m under the time period of 30 min, respectively.
基金supported by Sichuan Science and Technology Program(2023YFSY0026,2023YFH0004).
文摘Recent advancements in natural language processing have given rise to numerous pre-training language models in question-answering systems.However,with the constant evolution of algorithms,data,and computing power,the increasing size and complexity of these models have led to increased training costs and reduced efficiency.This study aims to minimize the inference time of such models while maintaining computational performance.It also proposes a novel Distillation model for PAL-BERT(DPAL-BERT),specifically,employs knowledge distillation,using the PAL-BERT model as the teacher model to train two student models:DPAL-BERT-Bi and DPAL-BERTC.This research enhances the dataset through techniques such as masking,replacement,and n-gram sampling to optimize knowledge transfer.The experimental results showed that the distilled models greatly outperform models trained from scratch.In addition,although the distilled models exhibit a slight decrease in performance compared to PAL-BERT,they significantly reduce inference time to just 0.25%of the original.This demonstrates the effectiveness of the proposed approach in balancing model performance and efficiency.
基金The National Natural Science Foundation of China under contract Nos 42266006 and 41806114the Jiangxi Provincial Natural Science Foundation under contract Nos 20232BAB204089 and 20202ACBL214019.
文摘The complexity of river-tide interaction poses a significant challenge in predicting discharge in tidal rivers.Long short-term memory(LSTM)networks excel in processing and predicting crucial events with extended intervals and time delays in time series data.Additionally,the sequence-to-sequence(Seq2Seq)model,known for handling temporal relationships,adapting to variable-length sequences,effectively capturing historical information,and accommodating various influencing factors,emerges as a robust and flexible tool in discharge forecasting.In this study,we introduce the application of LSTM-based Seq2Seq models for the first time in forecasting the discharge of a tidal reach of the Changjiang River(Yangtze River)Estuary.This study focuses on discharge forecasting using three key input characteristics:flow velocity,water level,and discharge,which means the structure of multiple input and single output is adopted.The experiment used the discharge data of the whole year of 2020,of which the first 80%is used as the training set,and the last 20%is used as the test set.This means that the data covers different tidal cycles,which helps to test the forecasting effect of different models in different tidal cycles and different runoff.The experimental results indicate that the proposed models demonstrate advantages in long-term,mid-term,and short-term discharge forecasting.The Seq2Seq models improved by 6%-60%and 5%-20%of the relative standard deviation compared to the harmonic analysis models and improved back propagation neural network models in discharge prediction,respectively.In addition,the relative accuracy of the Seq2Seq model is 1%to 3%higher than that of the LSTM model.Analytical assessment of the prediction errors shows that the Seq2Seq models are insensitive to the forecast lead time and they can capture characteristic values such as maximum flood tide flow and maximum ebb tide flow in the tidal cycle well.This indicates the significance of the Seq2Seq models.
文摘Hydrological models are developed to simulate river flows over a watershed for many practical applications in the field of water resource management. The present paper compares the performance of two recurrent neural networks for rainfall-runoff modeling in the Zou River basin at Atchérigbé outlet. To this end, we used daily precipitation data over the period 1988-2010 as input of the models, such as the Long Short-Term Memory (LSTM) and Recurrent Gate Networks (GRU) to simulate river discharge in the study area. The investigated models give good results in calibration (R2 = 0.888, NSE = 0.886, and RMSE = 0.42 for LSTM;R2 = 0.9, NSE = 0.9 and RMSE = 0.397 for GRU) and in validation (R2 = 0.865, NSE = 0.851, and RMSE = 0.329 for LSTM;R2 = 0.9, NSE = 0.865 and RMSE = 0.301 for GRU). This good performance of LSTM and GRU models confirms the importance of models based on machine learning in modeling hydrological phenomena for better decision-making.
文摘The stock market is a vital component of the broader financial system,with its dynamics closely linked to economic growth.The challenges associated with analyzing and forecasting stock prices have persisted since the inception of financial markets.By examining historical transaction data,latent opportunities for profit can be uncovered,providing valuable insights for both institutional and individual investors to make more informed decisions.This study focuses on analyzing historical transaction data from four banks to predict closing price trends.Various models,including decision trees,random forests,and Long Short-Term Memory(LSTM)networks,are employed to forecast stock price movements.Historical stock transaction data serves as the input for training these models,which are then used to predict upward or downward stock price trends.The study’s empirical results indicate that these methods are effective to a degree in predicting stock price movements.The LSTM-based deep neural network model,in particular,demonstrates a commendable level of predictive accuracy.This conclusion is reached following a thorough evaluation of model performance,highlighting the potential of LSTM models in stock market forecasting.The findings offer significant implications for advancing financial forecasting approaches,thereby improving the decision-making capabilities of investors and financial institutions.
基金supported by the Ministry of Trade,Industry & Energy(MOTIE,Korea) under Industrial Technology Innovation Program (No.10063424,'development of distant speech recognition and multi-task dialog processing technologies for in-door conversational robots')
文摘A Long Short-Term Memory(LSTM) Recurrent Neural Network(RNN) has driven tremendous improvements on an acoustic model based on Gaussian Mixture Model(GMM). However, these models based on a hybrid method require a forced aligned Hidden Markov Model(HMM) state sequence obtained from the GMM-based acoustic model. Therefore, it requires a long computation time for training both the GMM-based acoustic model and a deep learning-based acoustic model. In order to solve this problem, an acoustic model using CTC algorithm is proposed. CTC algorithm does not require the GMM-based acoustic model because it does not use the forced aligned HMM state sequence. However, previous works on a LSTM RNN-based acoustic model using CTC used a small-scale training corpus. In this paper, the LSTM RNN-based acoustic model using CTC is trained on a large-scale training corpus and its performance is evaluated. The implemented acoustic model has a performance of 6.18% and 15.01% in terms of Word Error Rate(WER) for clean speech and noisy speech, respectively. This is similar to a performance of the acoustic model based on the hybrid method.
基金The study was supported by the National Natural Science Foundation of China under GrantNo.49875025 and National Key Program fo
文摘In view of the fact that the atmospheric motion is an irreversible process, a memory function which can recall the observation data in the past is introduced, moreover, a special concept of self-memorization of the atmospheric motion is proposed, and a so-called self-memorization equation of the atmospheric motion has been derived. Based on the self-memorization principle, a numerical model for decadal forecast is established by means of the thermodynamic equation and the precipitation equation. The verification scores of the hindcasts of the model in the period from 1 to 12 years are much higher than that of monthly weather forecasts at present.
基金Financial supports by the National Natural Science Foundation of China (Grant 11532010)the project for Sichuan Provincial Youth Science and Technology Innovation Team, China (Grant 2013TD0004)
文摘A macroscopic based multi-mechanism constitutive model is constructed in the framework of irreversible thermodynamics to describe the degeneration of shape memory effect occurring in the thermo-mechanical cyclic deformation of NiTi shape memory alloys (SMAs). Three phases, austenite A, twinned martensite and detwinned martensite , as well as the phase transitions occurring between each pair of phases (, , , , and are considered in the proposed model. Meanwhile, two kinds of inelastic deformation mechanisms, martensite transformation-induced plasticity and reorientation-induced plasticity, are used to explain the degeneration of shape memory effects of NiTi SMAs. The evolution equations of internal variables are proposed by attributing the degeneration of shape memory effect to the interaction between the three phases (A, , and and plastic deformation. Finally, the capability of the proposed model is verified by comparing the predictions with the experimental results of NiTi SMAs. It is shown that the degeneration of shape memory effect and its dependence on the loading level can be reasonably described by the proposed model.
基金Project supported by the National Key Research and Development Program of China(No.2017YFC0307604)the Talent Foundation of China University of Petroleum(No.Y1215042)the Graduate Innovation Program of China University of Petroleum(East China)(No.YCX2019084)
文摘The objective of this paper is to model the size-dependent thermo-mechanical behaviors of a shape memory polymer (SMP) microbeam.Size-dependent constitutive equations,which can capture the size effect of the SMP,are proposed based on the modified couple stress theory (MCST).The deformation energy expression of the SMP microbeam is obtained by employing the proposed size-dependent constitutive equation and Bernoulli-Euler beam theory.An SMP microbeam model,which includes the formulations of deflection,strain,curvature,stress and couple stress,is developed by using the principle of minimum potential energy and the separation of variables together.The sizedependent thermo-mechanical and shape memory behaviors of the SMP microbeam and the influence of the Poisson ratio are numerically investigated according to the developed SMP microbeam model.Results show that the size effects of the SMP microbeam are significant when the dimensionless height is small enough.However,they are too slight to be necessarily considered when the dimensionless height is large enough.The bending flexibility and stress level of the SMP microbeam rise with the increasing dimensionless height,while the couple stress level declines with the increasing dimensionless height.The larger the dimensionless height is,the more obvious the viscous property and shape memory effect of the SMP microbeam are.The Poisson ratio has obvious influence on the size-dependent behaviors of the SMP microbeam.The paper provides a theoretical basis and a quantitatively analyzing tool for the design and analysis of SMP micro-structures in the field of biological medicine,microelectronic devices and micro-electro-mechanical system (MEMS) self-assembling.
基金Projects(11072056, 10772061) supported by the National Natural Science Foundation of ChinaProject(A200907) supported by the Natural Science Foundation of Heilongjiang Province,ChinaProject(20092322120001) supported by the PhD Programs Foundations of Ministry of Education of China
文摘In order to propel the development of metal magnetic memory （MMM） technique in fatigue damage detection, the Jiles-Atherton model （J-A model） was modified to describe MMM mechanism in elastic stress stage. A series of rotating bending fatigue experiments were conducted to study the stress-magnetization relationship and verify the correctness of modified J-A model. In MMM detection, the magnetization of material irreversibly approaches to the local equilibrium state Mo instead of global equilibrium state M^n under cyclic stress, and the M0-a curves are loops around the Mar,-a curve. The modified J-A model is constructed by replacing M~ in J-A model with M0, and it can describe the magnetomechanical effect well at low external magnetic field. In the rotating bending fatigue experiments, the MMM field distribution in normal direction around cylinder specimen is similar to the stress distribution, and the calculation result of model coincides with experiment result after some necessary modifications. The MMM field variation with time at a certain point in fatigue process is divided into three stages with the variation of stable stress-stain hysteresis loop, and the calculation results of model can explain not only the three stages of MMM field changes, but also the different change laws when the applied magnetic field and initial magnetic field are different. The MMM field distribution in normal direction along specimen axis reflects stress concentration effect at artificial defect, and the magnetic signal fluctuates around the defect at late fatigue stage. The calculation results coincide with the initial MMM principle and can explain signal fluctuates around the defect. The modified J-A model can explain experiment results well, and it is fit for MMM field characterization.
基金Supported by National Natural Science Foundation of China(Grant Nos.11272084,11472076)PetroChina Innovation Foundation(Grant No.2015D-5006-0602)Postdoctoral Science Research Developmental Foundation of Chinese Heilongjiang Province(Grant No.LBH-Q13035)
文摘Metal magnetic memory（MMM） testing has been widely used to detect welded joints. However, load levels, environmental magnetic field, and measurement noises make the MMM data dispersive and bring difficulty to quantitative evaluation. In order to promote the development of quantitative MMM reliability assessment, a new MMM model is presented for welded joints. Steel Q235 welded specimens are tested along the longitudinal and horizontal lines by TSC-2M-8 instrument in the tensile fatigue experiments. The X-ray testing is carried out synchronously to verify the MMM results. It is found that MMM testing can detect the hidden crack earlier than X-ray testing. Moreover, the MMM gradient vector sum K_（vs） is sensitive to the damage degree, especially at early and hidden damage stages. Considering the dispersion of MMM data, the K_（vs） statistical law is investigated, which shows that K_（vs） obeys Gaussian distribution. So K_（vs） is the suitable MMM parameter to establish reliability model of welded joints. At last, the original quantitative MMM reliability model is first presented based on the improved stress strength interference theory. It is shown that the reliability degree R gradually decreases with the decreasing of the residual life ratio T, and the maximal error between prediction reliability degree R_1 and verification reliability degree R_2 is 9.15%. This presented method provides a novel tool of reliability testing and evaluating in practical engineering for welded joints.
文摘This article examines some general atmospheric circulation and climate models in the context of the notion of “memory”. Two kinds of memories are defined： statistical memory and deterministic memory. The former is defined through the autocorrelation characteristic of the process if it is random （chaotic）, while for the latter, a special memory function is introduced. Three of the numerous existing models are selected as examples. For each of the models, asymptotic （at t →∞） expressions are derived. In this way, the transients are filtered out and that which remains concerns the final behaviour of the models.
基金supported by the Natural Science Foundation of Jiangsu Province of China (No. BK20170759)the National Natural Science Foundation of China (No. 11572153)+3 种基金Jiangsu Government Scholarship for Overseas Studiesa project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD)Outstanding Scientific and Technological Innovation Team in Colleges and Universities of Jiangsu Provincethe Doctor Special Foundation and the Research Fund of Nanjing Institute of Technology (Nos. ZKJ201603, YKJ201312)
文摘A thermoviscoelastic modeling approach is developed to predict the recovery behaviors of the thermally activated amorphous shape memory polymers(SMPs)based on the generalized finite deformation viscoelasticity theory.In this paper,a series of moduli and relaxation times of the generalized Maxwell model is estimated from the stress relaxation master curve by using the nonlinear regression(NLREG)method.Assuming that the amorphous SMPs are approximately incompressible isotropic elastomers in the rubbery state,the hyperelastic response of the materials is well modeled with a hyperelastic model in Ogden form.In addition,the Williams-Landel-Ferry(WLF)equation is used to describe the horizontal shift factor obtained with time-temperature superposition principle(TTSP).The finite element simulations show good agreement with the experimental thermomechanical behaviors.Moreover,the possibility of developing a temperature-responsive intravascular stent with the SMP studied here is investigated in terms of its thermomechanical property.Therefore,it can be concluded that the model has good prediction capabilities for the recovery behaviors of amorphous SMPs.
基金Project(51105170) supported by the National Natural Science Foundation of ChinaProject supported by the Scientific Research Foundation for the Returned Overseas Chinese Scholars,Ministry of Education,China
文摘The hysteresis characteristic is the major deficiency in the positioning control of magnetic shape memory alloy actuator. A Prandtl-Ishlinskii model was developed to characterize the hysteresis of magnetic shape memory alloy actuator. Based on the proposed Prandtl-Ishlinskii model, the inverse Prandtl-Ishlinskii model was established as a feedforward controller to compensate the hysteresis of the magnetic shape memory alloy actuator. For further improving of the positioning precision of the magnetic shape memory alloy actuator, a hybrid control method with hysteresis nonlinear model in feedforward loop was proposed. The control method is separated into two parts: a feedforward loop with inverse Prandtl-Ishlinskii model and a feedback loop with neural network controller. To validate the validity of the proposed control method, a series of simulations and experiments were researched. The simulation and experimental results demonstrate that the maximum error rate of open loop controller based on inverse PI model is 1.72%, the maximum error rate of the hybrid controller based on inverse PI model is 1.37%.
基金the National Natural Science Foundation of China (Grant 11602203)Young Elite Scientist Sponsorship Program by the China Association for Science and Technology (Grant 2016QNRC001)Fundamental Research Funds for the Central Universities (Grant 2682018CX43).
文摘Existing experimental results have shown that four types of physical mechanisms, namely, martensite transformation, martensite reorientation, magnetic domain wall motion and magnetization vector rotation, can be activated during the magneto-mechanical deformation of NiMnGa ferromagnetic shape memory alloy (FSMA) single crystals. In this work, based on irreversible thermodynamics, a three-dimensional (3D) single crystal constitutive model is constructed by considering the aforementioned four mechanisms simultaneously. Three types of internal variables, i.e., the volume fraction of each martensite variant, the volume fraction of magnetic domain in each variant and the deviation angle between the magnetization vector, and easy axis are introduced to characterize the magneto-mechanical state of the single crystals. The thermodynamic driving force of each mechanism and the thermodynamic constraints on the constitutive model are obtained from Clausius's dissipative inequality and constructed Gibbs free energy. Then, thermodynamically consistent kinetic equations for the four mechanisms are proposed, respectively. Finally, the ability of the proposed model to describe the magneto-mechanical deformation of NiMnGa FSMA single crystals is verified by comparing the predictions with corresponding experimental results. It is shown that the proposed model can quantitatively capture the main experimental phenomena. Further, the proposed model is used to predict the deformations of the single crystals under the non-proportional mechanical loading conditions.
