Knowledge distillation has become a standard technique for compressing large language models into efficient student models,but existing methods often struggle to balance prediction accuracy with explanation quality.Re...Knowledge distillation has become a standard technique for compressing large language models into efficient student models,but existing methods often struggle to balance prediction accuracy with explanation quality.Recent approaches such as Distilling Step-by-Step(DSbS)introduce explanation supervision,yet they apply it in a uniform manner that may not fully exploit the different learning dynamics of prediction and explanation.In this work,we propose a task-structured curriculum learning(TSCL)framework that structures training into three sequential phases:(i)prediction-only,to establish stable feature representations;(ii)joint prediction-explanation,to align task outputs with rationale generation;and(iii)explanation-only,to refine the quality of rationales.This design provides a simple but effective modification to DSbS,requiring no architectural changes and adding negligible training cost.We justify the phase scheduling with ablation studies and convergence analysis,showing that an initial prediction-heavy stage followed by a balanced joint phase improves both stability and explanation alignment.Extensive experiments on five datasets(e-SNLI,ANLI,CommonsenseQA,SVAMP,and MedNLI)demonstrate that TSCL consistently outperforms strong baselines,achieving gains of+1.7-2.6 points in accuracy and 0.8-1.2 in ROUGE-L,corresponding to relative error reductions of up to 21%.Beyond lexical metrics,human evaluation and ERASERstyle faithfulness diagnostics confirm that TSCL produces more faithful and informative explanations.Comparative training curves further reveal faster convergence and lower variance across seeds.Efficiency analysis shows less than 3%overhead in wall-clock training time and no additional inference cost,making the approach practical for realworld deployment.This study demonstrates that a simple task-structured curriculum can significantly improve the effectiveness of knowledge distillation.By separating and sequencing objectives,TSCL achieves a better balance between accuracy,stability,and explanation quality.The framework generalizes across domains,including medical NLI,and offers a principled recipe for future applications in multimodal reasoning and reinforcement learning.展开更多
Improving plasma uniformity is a critical issue in the development of large-area radio-frequency(RF)inductively coupled plasma(ICP)sources.In this work,the effects of coil structure and electromagnetic shielding on th...Improving plasma uniformity is a critical issue in the development of large-area radio-frequency(RF)inductively coupled plasma(ICP)sources.In this work,the effects of coil structure and electromagnetic shielding on the spatial distribution and uniformity of the plasma are systematically investigated using a three-dimensional fluid model.The model integrates plasma and electromagnetic field modules to simulate the discharge characteristics of a large-area RF ICP source with dimensions of 100 cm×50 cm.The results reveal that the electron density distribution varies significantly with the coil structure.For the rotating and translating coil structures,the electron density is high at off-axis positions and low at the center.In contrast,the mirror coil structure exhibits a significantly higher electron density at the chamber center,resulting in a high-center and low-edge density distribution.Among the three configurations,the rotating coil structure provides the best plasma uniformity.The incorporation of electromagnetic shielding further improves plasma uniformity,particularly for the mirror coil structure.For the rotating and translating coil structures,the electron density exhibits a saddle-shaped distribution regardless of electromagnetic shielding.However,introducing electromagnetic shielding into the mirror coil structure reduces the electron density at the chamber center and decreases the non-uniformity degree by 18.4%.Overall,the mirror coil structure with electromagnetic shielding achieves the highest uniformity,with an exceptional plasma uniformity of 94%.This work offers valuable insights for the design of large-area ICP sources in advanced plasma processing systems.展开更多
Convective storms and lightning are among the most important weather phenomena that are challenging to forecast.In this study,a novel multi-task learning(MTL)encoder-decoder U-net neural network was developed to forec...Convective storms and lightning are among the most important weather phenomena that are challenging to forecast.In this study,a novel multi-task learning(MTL)encoder-decoder U-net neural network was developed to forecast convective storms and lightning with lead times for up to 90 min,using GOES-16 geostationary satellite infrared brightness temperatures(IRBTs),lightning flashes from Geostationary Lightning Mapper(GLM),and vertically integrated liquid(VIL)from Next Generation Weather Radar(NEXRAD).To cope with the heavily skewed distribution of lightning data,a spatiotemporal exponent-weighted loss function and log-transformed lightning normalization approach were developed.The effects of MTL,single-task learning(STL),and IRBTs as auxiliary input features on convection and lightning nowcasting were investigated.The results showed that normalizing the heavily skew-distributed lightning data along with a log-transformation dramatically outperforms the min-max normalization method for nowcasting an intense lightning event.The MTL model significantly outperformed the STL model for both lightning nowcasting and VIL nowcasting,particularly for intense lightning events.The MTL also helped delay the lightning forecast performance decay with the lead times.Furthermore,incorporating satellite IRBTs as auxiliary input features substantially improved lightning nowcasting,but produced little difference in VIL forecasting.Finally,the MTL model performed better for forecasting both lightning and the VIL of organized convective storms than for isolated cells.展开更多
Recommendation system can greatly alleviate the "information overload" in the big data era. Existing recommendation methods, however, typically focus on predicting missing rating values via analyzing user-it...Recommendation system can greatly alleviate the "information overload" in the big data era. Existing recommendation methods, however, typically focus on predicting missing rating values via analyzing user-item dualistic relationship, which neglect an important fact that the latent interests of users can influence their rating behaviors. Moreover, traditional recommendation methods easily suffer from the high dimensional problem and cold-start problem. To address these challenges, in this paper, we propose a PBUED(PLSA-Based Uniform Euclidean Distance) scheme, which utilizes topic model and uniform Euclidean distance to recommend the suitable items for users. The solution first employs probabilistic latent semantic analysis(PLSA) to extract users' interests, users with different interests are divided into different subgroups. Then, the uniform Euclidean distance is adopted to compute the users' similarity in the same interest subset; finally, the missing rating values of data are predicted via aggregating similar neighbors' ratings. We evaluate PBUED on two datasets and experimental results show PBUED can lead to better predicting performance and ranking performance than other approaches.展开更多
In this paper, the method based on uniform design and neural network is proposed to model the complex system. In order to express the system characteristics all round, uniform design method is used to choose the model...In this paper, the method based on uniform design and neural network is proposed to model the complex system. In order to express the system characteristics all round, uniform design method is used to choose the modeling samples and obtain the overall information of the system;for the purpose of modeling the system or its characteristics, the artificial neural network is used to construct the model. Experiment indicates that this method can model the complex system effectively.展开更多
The cast preformed forming process(CPFP) is increasingly considered and applied in the metal forming industries due to its short process, low cost, and environmental friendliness, especially in the aerospace field. Ho...The cast preformed forming process(CPFP) is increasingly considered and applied in the metal forming industries due to its short process, low cost, and environmental friendliness, especially in the aerospace field. However, how to establish a unified model of a non-uniform as-cast billet depicting the flow stress and microstructure evolution behaviors during hot working is the key to microstructure prediction and parameter optimization of the CPFP. In this work, hot compression tests are performed using a non-uniform as-cast 42 CrMo billet at 1123–1423 K and 0.01–1sà1. The effect laws of the non-uniform state of the as-cast billet with different initial grain sizes on the flow stress and microstructure are revealed deeply. Based on experimental results, a unified model of flow stress and grain size evolutions is developed by the internal variable modeling method. Verified results show that the model can well describe the responses of the flow stress and microstructure to deformation conditions and initial grain sizes. To further evaluate its reliability, the unified model is applied to FE simulation of the cast preformed ring rolling process.The predictions of the rolling force and grain size indicate that it could well describe the flow stress and microstructure evolutions during the process.展开更多
Hot compression tests of Mg–11 Gd–4 Y–2 Zn–0.4 Zr alloy(GWZK114)were conducted at a deformation temperature range of 300–500°C and a strain rate range of 0.01–10.0 s-1.Based on systematic microstructure obs...Hot compression tests of Mg–11 Gd–4 Y–2 Zn–0.4 Zr alloy(GWZK114)were conducted at a deformation temperature range of 300–500°C and a strain rate range of 0.01–10.0 s-1.Based on systematic microstructure observation,it is confirmed that long period stacking ordered(LPSO)phase displays essential and evolving roles on the dynamic recrystallization(DRX)behavior.The results indicate that the plastic deformation is mainly coordinated by simultaneous exist of LPSO kinking of lamella 14 H-LPSO phase and DRX at 350–450℃,and DRX at 500℃.Further,it is found that the LPSO kinking induced during 350–450℃can delay the DRX.A phenomenological DRX model of GWZK114 alloy is established to be XDRX=1.exp[-0.5((ε-εc)/ε^*)0.91].Non-uniform distribution of plastic strain during compression was considered via finite element method and it ensures a good prediction of DRX fraction under a large plastic strain.Meanwhile,an enhanced DRX model,taking its formulation as XDRX={1.exp[-0.5((ε-εc)/ε*)0.91]}(T/(226.8)-1)n,n=3.82ε0.083,is proposed for the first time to capture the hindering effect of 14 H-LPSO kinking on DRX behavior.The predicted results of this enhanced DRX model agree well with the experimental cases,where 14 H-LPSO kinking is dominated or partially involved(300–450℃).Besides,a size model of DRX grains is also established and can depict the evolution of DRX grain size for all the investigated compression conditions with accounting for temperature rising at high strain rates(5 s^-1 and 10 s^-1).展开更多
For two-way contingency tables with ordered categories, the present paper gives a theorem that the independence model holds if and only if the logit uniform association model holds and equality of concordance and disc...For two-way contingency tables with ordered categories, the present paper gives a theorem that the independence model holds if and only if the logit uniform association model holds and equality of concordance and discordance for all pairs of adjacent rows and all dichotomous collapsing of the columns holds. Using the theorem, we analyze the cross-classification of duodenal ulcer patients according to operation and dumping severity.展开更多
The human motion generation model can extract structural features from existing human motion capture data,and the generated data makes animated characters move.The 3D human motion capture sequences contain complex spa...The human motion generation model can extract structural features from existing human motion capture data,and the generated data makes animated characters move.The 3D human motion capture sequences contain complex spatial-temporal structures,and the deep learning model can fully describe the potential semantic structure of human motion.To improve the authenticity of the generated human motion sequences,we propose a multi-task motion generation model that consists of a discriminator and a generator.The discriminator classifies motion sequences into different styles according to their similarity to the mean spatial-temporal templates from motion sequences of 17 crucial human joints in three-freedom degrees.And target motion sequences are created with these styles by the generator.Unlike traditional related works,our model can handle multiple tasks,such as identifying styles and generating data.In addition,by extracting 17 crucial joints from 29 human joints,our model avoids data redundancy and improves the accuracy of model recognition.The experimental results show that the discriminator of the model can effectively recognize diversified movements,and the generated data can correctly fit the actual data.The combination of discriminator and generator solves the problem of low reuse rate of motion data,and the generated motion sequences are more suitable for actual movement.展开更多
Panel data combine cross-section data and time series data. If the cross-section is locations, there is a need to check the correlation among locations. ρ and λ are parameters in generalized spatial model to cover e...Panel data combine cross-section data and time series data. If the cross-section is locations, there is a need to check the correlation among locations. ρ and λ are parameters in generalized spatial model to cover effect of correlation between locations. Value of ρ or λ will influence the goodness of fit model, so it is important to make parameter estimation. The effect of another location is covered by making contiguity matrix until it gets spatial weighted matrix (W). There are some types of W—uniform W, binary W, kernel Gaussian W and some W from real case of economics condition or transportation condition from locations. This study is aimed to compare uniform W and kernel Gaussian W in spatial panel data model using RMSE value. The result of analysis showed that uniform weight had RMSE value less than kernel Gaussian model. Uniform W had stabil value for all the combinations.展开更多
Stability, boundedness and persistence are three important aspects for an ecological model. In this paper, a further analysis of a class of anaerobic digestion ecological models is performed. Based on the Liupunov Met...Stability, boundedness and persistence are three important aspects for an ecological model. In this paper, a further analysis of a class of anaerobic digestion ecological models is performed. Based on the Liupunov Method, the local stability of all equilibria in the system is got. According to the vector fields described by the system, the proof of the boundedness of the solution on the anaerobic digestion processes is completed in three steps. The method proposed in the discussion on the boundedness can be generalized to the similar problems. Results in this paper give information on how to run the ecological system well by adjusting the system parameters.展开更多
Cooperative multi-agent reinforcement learning(MARL)is a key technology for enabling cooperation in complex multi-agent systems.It has achieved remarkable progress in areas such as gaming,autonomous driving,and multi-...Cooperative multi-agent reinforcement learning(MARL)is a key technology for enabling cooperation in complex multi-agent systems.It has achieved remarkable progress in areas such as gaming,autonomous driving,and multi-robot control.Empowering cooperative MARL with multi-task decision-making capabilities is expected to further broaden its application scope.In multi-task scenarios,cooperative MARL algorithms need to address 3 types of multi-task problems:reward-related multi-task,arising from different reward functions;multi-domain multi-task,caused by differences in state and action spaces,state transition functions;and scalability-related multi-task,resulting from the dynamic variation in the number of agents.Most existing studies focus on scalability-related multitask problems.However,with the increasing integration between large language models(LLMs)and multi-agent systems,a growing number of LLM-based multi-agent systems have emerged,enabling more complex multi-task cooperation.This paper provides a comprehensive review of the latest advances in this field.By combining multi-task reinforcement learning with cooperative MARL,we categorize and analyze the 3 major types of multi-task problems under multi-agent settings,offering more fine-grained classifications and summarizing key insights for each.In addition,we summarize commonly used benchmarks and discuss future directions of research in this area,which hold promise for further enhancing the multi-task cooperation capabilities of multi-agent systems and expanding their practical applications in the real world.展开更多
In this paper, the influencing factors that affect few-mode and multi core optical fiber channel are analyzed in a comprehensive way. The theoretical modeling and computer simulation of the information channel are car...In this paper, the influencing factors that affect few-mode and multi core optical fiber channel are analyzed in a comprehensive way. The theoretical modeling and computer simulation of the information channel are carried out and then the modeling scheme of few-mode multicore optical fiber channel based on non-uniform mode field distribution is put forward. The proposed modeling scheme can not only exponentially increases the system capacity through fewmode multi-core optical fiber channel, but has better transmission performance compared to the channel of the same type to the uniform channel revealing from the simulation results.展开更多
The mineralization process of microbial-induced calcium carbonate precipitation(MICP)is influenced by many factors,and the uniformity of the calcium carbonate precipitation has become the main focus and challenge for ...The mineralization process of microbial-induced calcium carbonate precipitation(MICP)is influenced by many factors,and the uniformity of the calcium carbonate precipitation has become the main focus and challenge for MICP technology.In this study,the uniformity of the saturated calcareous sand treated with MICP was in-vestigated through one-dimensional calcareous sand column tests and model tests.The coefficient of variation was employed in one-dimensional sand column tests to investigate the impact of injection rate,cementation solution concentration,and number of injection cycles on the uniformity of the MICP treatment.Additionally,model tests were conducted to investigate the impact of injection pressure and methods on the treatment range and uniformity under three-dimensional seepage conditions.Test results demonstrate that the reinforcement strength and uniformity are significantly influenced by the injection rate of the cementation solution,with a rate of 3 mL/min,yielding a favorable treatment effect.Excessive concentration of the cementation solution can lead to significant non-uniformity and a reduction in the compressive strength of MICP-treated samples.Conversely,excessively low concentrations may result in decreased bonding efficiency.Among the four considered con-centrations,0.5 mol/L and 1 mol/L exhibit superior reinforcing effects.The morphological development of calcareous sandy foundation reinforcement is associated with the spatial distribution pattern of the bacterial solution,exhibiting a relatively larger reinforcement area in proximity to the lower region of the model and a gradually decreasing range towards the upper part.Under three-dimensional seepage conditions,in addition to the non-uniform radial cementation along the injection pipe,there is also vertical heterogeneity of cementation along the length of the injection pipe due to gravitational effects,resulting in preferential deposition of calcium carbonate at the lower section,The application of injection pressure and a double-pipe circulation injection method can mitigate the accumulation of bacterial solution and cementation solution at the bottom,thereby improving the reinforcement range and uniformity.展开更多
Long-duration energy storage has become critical for renewable energy integration.While redox flow batteries,especially vanadium-based systems,are scaling up in capacity,their performance at the stack level remains in...Long-duration energy storage has become critical for renewable energy integration.While redox flow batteries,especially vanadium-based systems,are scaling up in capacity,their performance at the stack level remains insufficiently optimized,demanding more profound mechanistic studies and engineering refinements.To address the difficulties in resolving the flow inhomogeneity at the stack scale,this study establishes a multi-physics field coupling model and analyzes the pressure distributions,flow rate differences,active substance concentration,and electrochemical characteristics.The results show that the uneven cell pressure distribution is a key factor affecting the consistency of the system performance,and the increase in the flow rate improves the reactant homogeneity,with both the average concentration and the uniformity factor increasing with the flow rate.In contrast,high current densities lead to an increased imbalance between electrochemical depletion and reactant replenishment,resulting in a significant decrease in reactant concentration in the under-ribs region.In addition,a higher flow rate can expand the high-current-density region where the stack operates efficiently.This study provides a theoretical basis for optimizing the design of the stack components.展开更多
Scour around bridge pier foundations is a complex phenomenon that can threaten structural stability.Accurate prediction of scour depth around compound piers remains challenging for bridge engineers.This study investig...Scour around bridge pier foundations is a complex phenomenon that can threaten structural stability.Accurate prediction of scour depth around compound piers remains challenging for bridge engineers.This study investigated the effect of foundation elevation on scour around compound piers and developed reliable scour depth prediction models for economical foundation design.Experiments were conducted under clear-water conditions using two circular piers:(1)a uniform pier(with a diameter of D)and(2)a compound pier consisting of a uniform pier resting on a circular foundation(with a foundation diameter(D_(f))of 2D)positioned at various elevations(Z)relative to the channel bed.Results showed that foundation elevation significantly affected scour depth.Foundations at or below the bed(Z/D≥0)reduced scour,while those projecting into the flow field(Z/D<0)increased scour.The optimal foundation elevation was found to be 0.1D below the bed level,yielding a 57%reduction in scour depth compared to the uniform pier due to its shielding effect against downflow and horseshoe vortices.In addition,regression,artificial neural network(ANN),and M5 model tree models were developed using experimental data from this and previous studies.The M5 model outperformed the traditional HEC-18 equation,regression,and ANN models,with a coefficient of determination greater than 0.85.Sensitivity analysis indicated that flow depth,foundation elevation,and diameter significantly influenced scour depth prediction,whereas sediment size had a lesser impact.展开更多
The received signal of the polarization sensitive array is proved to have trilinear model characteristics. The blind parallel factor（PARAFAC） signal detection algorithm for the polarization sensitive array is propos...The received signal of the polarization sensitive array is proved to have trilinear model characteristics. The blind parallel factor（PARAFAC） signal detection algorithm for the polarization sensitive array is proposed. The trilinear alternating least square （TALS） algorithm is used to abtain the source matrix, and then the matrix is judged. Simulation results show that the bit error rate （BER） of the detection algorithm is close to that of the non-blind decorrelating method and the algorithm works well under the array error condition. BER difference between the non-blind method and this algorithm is less than 2 dB under a high SNR. The algorithm is blind and robust. The channel fading, the direction of arrive（DOA） imformation and the polarization information are needless in the algorithm.展开更多
文摘Knowledge distillation has become a standard technique for compressing large language models into efficient student models,but existing methods often struggle to balance prediction accuracy with explanation quality.Recent approaches such as Distilling Step-by-Step(DSbS)introduce explanation supervision,yet they apply it in a uniform manner that may not fully exploit the different learning dynamics of prediction and explanation.In this work,we propose a task-structured curriculum learning(TSCL)framework that structures training into three sequential phases:(i)prediction-only,to establish stable feature representations;(ii)joint prediction-explanation,to align task outputs with rationale generation;and(iii)explanation-only,to refine the quality of rationales.This design provides a simple but effective modification to DSbS,requiring no architectural changes and adding negligible training cost.We justify the phase scheduling with ablation studies and convergence analysis,showing that an initial prediction-heavy stage followed by a balanced joint phase improves both stability and explanation alignment.Extensive experiments on five datasets(e-SNLI,ANLI,CommonsenseQA,SVAMP,and MedNLI)demonstrate that TSCL consistently outperforms strong baselines,achieving gains of+1.7-2.6 points in accuracy and 0.8-1.2 in ROUGE-L,corresponding to relative error reductions of up to 21%.Beyond lexical metrics,human evaluation and ERASERstyle faithfulness diagnostics confirm that TSCL produces more faithful and informative explanations.Comparative training curves further reveal faster convergence and lower variance across seeds.Efficiency analysis shows less than 3%overhead in wall-clock training time and no additional inference cost,making the approach practical for realworld deployment.This study demonstrates that a simple task-structured curriculum can significantly improve the effectiveness of knowledge distillation.By separating and sequencing objectives,TSCL achieves a better balance between accuracy,stability,and explanation quality.The framework generalizes across domains,including medical NLI,and offers a principled recipe for future applications in multimodal reasoning and reinforcement learning.
基金supported by the National Natural Science Foundation of China(Grant Nos.12075049 and 11935005)。
文摘Improving plasma uniformity is a critical issue in the development of large-area radio-frequency(RF)inductively coupled plasma(ICP)sources.In this work,the effects of coil structure and electromagnetic shielding on the spatial distribution and uniformity of the plasma are systematically investigated using a three-dimensional fluid model.The model integrates plasma and electromagnetic field modules to simulate the discharge characteristics of a large-area RF ICP source with dimensions of 100 cm×50 cm.The results reveal that the electron density distribution varies significantly with the coil structure.For the rotating and translating coil structures,the electron density is high at off-axis positions and low at the center.In contrast,the mirror coil structure exhibits a significantly higher electron density at the chamber center,resulting in a high-center and low-edge density distribution.Among the three configurations,the rotating coil structure provides the best plasma uniformity.The incorporation of electromagnetic shielding further improves plasma uniformity,particularly for the mirror coil structure.For the rotating and translating coil structures,the electron density exhibits a saddle-shaped distribution regardless of electromagnetic shielding.However,introducing electromagnetic shielding into the mirror coil structure reduces the electron density at the chamber center and decreases the non-uniformity degree by 18.4%.Overall,the mirror coil structure with electromagnetic shielding achieves the highest uniformity,with an exceptional plasma uniformity of 94%.This work offers valuable insights for the design of large-area ICP sources in advanced plasma processing systems.
基金supported by the Science and Technology Grant No.520120210003,Jibei Electric Power Company of the State Grid Corporation of China。
文摘Convective storms and lightning are among the most important weather phenomena that are challenging to forecast.In this study,a novel multi-task learning(MTL)encoder-decoder U-net neural network was developed to forecast convective storms and lightning with lead times for up to 90 min,using GOES-16 geostationary satellite infrared brightness temperatures(IRBTs),lightning flashes from Geostationary Lightning Mapper(GLM),and vertically integrated liquid(VIL)from Next Generation Weather Radar(NEXRAD).To cope with the heavily skewed distribution of lightning data,a spatiotemporal exponent-weighted loss function and log-transformed lightning normalization approach were developed.The effects of MTL,single-task learning(STL),and IRBTs as auxiliary input features on convection and lightning nowcasting were investigated.The results showed that normalizing the heavily skew-distributed lightning data along with a log-transformation dramatically outperforms the min-max normalization method for nowcasting an intense lightning event.The MTL model significantly outperformed the STL model for both lightning nowcasting and VIL nowcasting,particularly for intense lightning events.The MTL also helped delay the lightning forecast performance decay with the lead times.Furthermore,incorporating satellite IRBTs as auxiliary input features substantially improved lightning nowcasting,but produced little difference in VIL forecasting.Finally,the MTL model performed better for forecasting both lightning and the VIL of organized convective storms than for isolated cells.
基金supported in part by the National High‐tech R&D Program of China (863 Program) under Grant No. 2013AA102301technological project of Henan province (162102210214)
文摘Recommendation system can greatly alleviate the "information overload" in the big data era. Existing recommendation methods, however, typically focus on predicting missing rating values via analyzing user-item dualistic relationship, which neglect an important fact that the latent interests of users can influence their rating behaviors. Moreover, traditional recommendation methods easily suffer from the high dimensional problem and cold-start problem. To address these challenges, in this paper, we propose a PBUED(PLSA-Based Uniform Euclidean Distance) scheme, which utilizes topic model and uniform Euclidean distance to recommend the suitable items for users. The solution first employs probabilistic latent semantic analysis(PLSA) to extract users' interests, users with different interests are divided into different subgroups. Then, the uniform Euclidean distance is adopted to compute the users' similarity in the same interest subset; finally, the missing rating values of data are predicted via aggregating similar neighbors' ratings. We evaluate PBUED on two datasets and experimental results show PBUED can lead to better predicting performance and ranking performance than other approaches.
文摘In this paper, the method based on uniform design and neural network is proposed to model the complex system. In order to express the system characteristics all round, uniform design method is used to choose the modeling samples and obtain the overall information of the system;for the purpose of modeling the system or its characteristics, the artificial neural network is used to construct the model. Experiment indicates that this method can model the complex system effectively.
基金supported by the National Natural Science Foundation of China (No’s. 51575448 and 51135007)
文摘The cast preformed forming process(CPFP) is increasingly considered and applied in the metal forming industries due to its short process, low cost, and environmental friendliness, especially in the aerospace field. However, how to establish a unified model of a non-uniform as-cast billet depicting the flow stress and microstructure evolution behaviors during hot working is the key to microstructure prediction and parameter optimization of the CPFP. In this work, hot compression tests are performed using a non-uniform as-cast 42 CrMo billet at 1123–1423 K and 0.01–1sà1. The effect laws of the non-uniform state of the as-cast billet with different initial grain sizes on the flow stress and microstructure are revealed deeply. Based on experimental results, a unified model of flow stress and grain size evolutions is developed by the internal variable modeling method. Verified results show that the model can well describe the responses of the flow stress and microstructure to deformation conditions and initial grain sizes. To further evaluate its reliability, the unified model is applied to FE simulation of the cast preformed ring rolling process.The predictions of the rolling force and grain size indicate that it could well describe the flow stress and microstructure evolutions during the process.
文摘Hot compression tests of Mg–11 Gd–4 Y–2 Zn–0.4 Zr alloy(GWZK114)were conducted at a deformation temperature range of 300–500°C and a strain rate range of 0.01–10.0 s-1.Based on systematic microstructure observation,it is confirmed that long period stacking ordered(LPSO)phase displays essential and evolving roles on the dynamic recrystallization(DRX)behavior.The results indicate that the plastic deformation is mainly coordinated by simultaneous exist of LPSO kinking of lamella 14 H-LPSO phase and DRX at 350–450℃,and DRX at 500℃.Further,it is found that the LPSO kinking induced during 350–450℃can delay the DRX.A phenomenological DRX model of GWZK114 alloy is established to be XDRX=1.exp[-0.5((ε-εc)/ε^*)0.91].Non-uniform distribution of plastic strain during compression was considered via finite element method and it ensures a good prediction of DRX fraction under a large plastic strain.Meanwhile,an enhanced DRX model,taking its formulation as XDRX={1.exp[-0.5((ε-εc)/ε*)0.91]}(T/(226.8)-1)n,n=3.82ε0.083,is proposed for the first time to capture the hindering effect of 14 H-LPSO kinking on DRX behavior.The predicted results of this enhanced DRX model agree well with the experimental cases,where 14 H-LPSO kinking is dominated or partially involved(300–450℃).Besides,a size model of DRX grains is also established and can depict the evolution of DRX grain size for all the investigated compression conditions with accounting for temperature rising at high strain rates(5 s^-1 and 10 s^-1).
文摘For two-way contingency tables with ordered categories, the present paper gives a theorem that the independence model holds if and only if the logit uniform association model holds and equality of concordance and discordance for all pairs of adjacent rows and all dichotomous collapsing of the columns holds. Using the theorem, we analyze the cross-classification of duodenal ulcer patients according to operation and dumping severity.
文摘The human motion generation model can extract structural features from existing human motion capture data,and the generated data makes animated characters move.The 3D human motion capture sequences contain complex spatial-temporal structures,and the deep learning model can fully describe the potential semantic structure of human motion.To improve the authenticity of the generated human motion sequences,we propose a multi-task motion generation model that consists of a discriminator and a generator.The discriminator classifies motion sequences into different styles according to their similarity to the mean spatial-temporal templates from motion sequences of 17 crucial human joints in three-freedom degrees.And target motion sequences are created with these styles by the generator.Unlike traditional related works,our model can handle multiple tasks,such as identifying styles and generating data.In addition,by extracting 17 crucial joints from 29 human joints,our model avoids data redundancy and improves the accuracy of model recognition.The experimental results show that the discriminator of the model can effectively recognize diversified movements,and the generated data can correctly fit the actual data.The combination of discriminator and generator solves the problem of low reuse rate of motion data,and the generated motion sequences are more suitable for actual movement.
文摘Panel data combine cross-section data and time series data. If the cross-section is locations, there is a need to check the correlation among locations. ρ and λ are parameters in generalized spatial model to cover effect of correlation between locations. Value of ρ or λ will influence the goodness of fit model, so it is important to make parameter estimation. The effect of another location is covered by making contiguity matrix until it gets spatial weighted matrix (W). There are some types of W—uniform W, binary W, kernel Gaussian W and some W from real case of economics condition or transportation condition from locations. This study is aimed to compare uniform W and kernel Gaussian W in spatial panel data model using RMSE value. The result of analysis showed that uniform weight had RMSE value less than kernel Gaussian model. Uniform W had stabil value for all the combinations.
基金Supported by the National Natural Science Foundation of China (No.60372012) and NSF of Chongqing (No.0831)
文摘Stability, boundedness and persistence are three important aspects for an ecological model. In this paper, a further analysis of a class of anaerobic digestion ecological models is performed. Based on the Liupunov Method, the local stability of all equilibria in the system is got. According to the vector fields described by the system, the proof of the boundedness of the solution on the anaerobic digestion processes is completed in three steps. The method proposed in the discussion on the boundedness can be generalized to the similar problems. Results in this paper give information on how to run the ecological system well by adjusting the system parameters.
基金The National Natural Science Foundation of China(62136008,62293541)The Beijing Natural Science Foundation(4232056)The Beijing Nova Program(20240484514).
文摘Cooperative multi-agent reinforcement learning(MARL)is a key technology for enabling cooperation in complex multi-agent systems.It has achieved remarkable progress in areas such as gaming,autonomous driving,and multi-robot control.Empowering cooperative MARL with multi-task decision-making capabilities is expected to further broaden its application scope.In multi-task scenarios,cooperative MARL algorithms need to address 3 types of multi-task problems:reward-related multi-task,arising from different reward functions;multi-domain multi-task,caused by differences in state and action spaces,state transition functions;and scalability-related multi-task,resulting from the dynamic variation in the number of agents.Most existing studies focus on scalability-related multitask problems.However,with the increasing integration between large language models(LLMs)and multi-agent systems,a growing number of LLM-based multi-agent systems have emerged,enabling more complex multi-task cooperation.This paper provides a comprehensive review of the latest advances in this field.By combining multi-task reinforcement learning with cooperative MARL,we categorize and analyze the 3 major types of multi-task problems under multi-agent settings,offering more fine-grained classifications and summarizing key insights for each.In addition,we summarize commonly used benchmarks and discuss future directions of research in this area,which hold promise for further enhancing the multi-task cooperation capabilities of multi-agent systems and expanding their practical applications in the real world.
基金supports from National High Technology 863 Program of China(No.2013AA013403,2015AA015501,2015AA015502,2015AA015504)National NSFC(No.61425022/61522501/61307086/61475024/61275158/61201151/61275074/61372109)+4 种基金Beijing Nova Program(No.Z141101001814048)Beijing Excellent Ph.D.Thesis Guidance Foundation(No.20121001302)the Universities Ph.D.Special Research Funds(No.20120005110003/20120005120007)Fund of State Key Laboratory of IPOC(BUPT)P.R.China
文摘In this paper, the influencing factors that affect few-mode and multi core optical fiber channel are analyzed in a comprehensive way. The theoretical modeling and computer simulation of the information channel are carried out and then the modeling scheme of few-mode multicore optical fiber channel based on non-uniform mode field distribution is put forward. The proposed modeling scheme can not only exponentially increases the system capacity through fewmode multi-core optical fiber channel, but has better transmission performance compared to the channel of the same type to the uniform channel revealing from the simulation results.
基金support of Natural Science Foundation of China(Grant No.52108324,No.52008207,and No.52108298)for conducting this study.
文摘The mineralization process of microbial-induced calcium carbonate precipitation(MICP)is influenced by many factors,and the uniformity of the calcium carbonate precipitation has become the main focus and challenge for MICP technology.In this study,the uniformity of the saturated calcareous sand treated with MICP was in-vestigated through one-dimensional calcareous sand column tests and model tests.The coefficient of variation was employed in one-dimensional sand column tests to investigate the impact of injection rate,cementation solution concentration,and number of injection cycles on the uniformity of the MICP treatment.Additionally,model tests were conducted to investigate the impact of injection pressure and methods on the treatment range and uniformity under three-dimensional seepage conditions.Test results demonstrate that the reinforcement strength and uniformity are significantly influenced by the injection rate of the cementation solution,with a rate of 3 mL/min,yielding a favorable treatment effect.Excessive concentration of the cementation solution can lead to significant non-uniformity and a reduction in the compressive strength of MICP-treated samples.Conversely,excessively low concentrations may result in decreased bonding efficiency.Among the four considered con-centrations,0.5 mol/L and 1 mol/L exhibit superior reinforcing effects.The morphological development of calcareous sandy foundation reinforcement is associated with the spatial distribution pattern of the bacterial solution,exhibiting a relatively larger reinforcement area in proximity to the lower region of the model and a gradually decreasing range towards the upper part.Under three-dimensional seepage conditions,in addition to the non-uniform radial cementation along the injection pipe,there is also vertical heterogeneity of cementation along the length of the injection pipe due to gravitational effects,resulting in preferential deposition of calcium carbonate at the lower section,The application of injection pressure and a double-pipe circulation injection method can mitigate the accumulation of bacterial solution and cementation solution at the bottom,thereby improving the reinforcement range and uniformity.
基金supported by National Natural Science Foundation of China(No.524B2078,12426307,51906203)Guangdong Major Project of Basic and Applied Basic Research(2023B0303000002)+6 种基金Guangdong Basic and Applied Basic Research Foundation(2023B1515120005)Natural Science Foundation of Shenzhen(JCYJ20241202125327036,JCYJ20240813100103005)Shenzhen Engineering Research Center of Redox Flow Battery for Energy Storage(XMHT20230208003)Research Project on Medium-and Long-Duration Flow Battery Energy Storage Technology(2024KJTW0015)China Association for Science and Technology(OR2308010)High level of special funds(G03034K001)supported by the Center for Computational Science and Engineering at the Southern University of Science and Technology.
文摘Long-duration energy storage has become critical for renewable energy integration.While redox flow batteries,especially vanadium-based systems,are scaling up in capacity,their performance at the stack level remains insufficiently optimized,demanding more profound mechanistic studies and engineering refinements.To address the difficulties in resolving the flow inhomogeneity at the stack scale,this study establishes a multi-physics field coupling model and analyzes the pressure distributions,flow rate differences,active substance concentration,and electrochemical characteristics.The results show that the uneven cell pressure distribution is a key factor affecting the consistency of the system performance,and the increase in the flow rate improves the reactant homogeneity,with both the average concentration and the uniformity factor increasing with the flow rate.In contrast,high current densities lead to an increased imbalance between electrochemical depletion and reactant replenishment,resulting in a significant decrease in reactant concentration in the under-ribs region.In addition,a higher flow rate can expand the high-current-density region where the stack operates efficiently.This study provides a theoretical basis for optimizing the design of the stack components.
文摘Scour around bridge pier foundations is a complex phenomenon that can threaten structural stability.Accurate prediction of scour depth around compound piers remains challenging for bridge engineers.This study investigated the effect of foundation elevation on scour around compound piers and developed reliable scour depth prediction models for economical foundation design.Experiments were conducted under clear-water conditions using two circular piers:(1)a uniform pier(with a diameter of D)and(2)a compound pier consisting of a uniform pier resting on a circular foundation(with a foundation diameter(D_(f))of 2D)positioned at various elevations(Z)relative to the channel bed.Results showed that foundation elevation significantly affected scour depth.Foundations at or below the bed(Z/D≥0)reduced scour,while those projecting into the flow field(Z/D<0)increased scour.The optimal foundation elevation was found to be 0.1D below the bed level,yielding a 57%reduction in scour depth compared to the uniform pier due to its shielding effect against downflow and horseshoe vortices.In addition,regression,artificial neural network(ANN),and M5 model tree models were developed using experimental data from this and previous studies.The M5 model outperformed the traditional HEC-18 equation,regression,and ANN models,with a coefficient of determination greater than 0.85.Sensitivity analysis indicated that flow depth,foundation elevation,and diameter significantly influenced scour depth prediction,whereas sediment size had a lesser impact.
文摘The received signal of the polarization sensitive array is proved to have trilinear model characteristics. The blind parallel factor（PARAFAC） signal detection algorithm for the polarization sensitive array is proposed. The trilinear alternating least square （TALS） algorithm is used to abtain the source matrix, and then the matrix is judged. Simulation results show that the bit error rate （BER） of the detection algorithm is close to that of the non-blind decorrelating method and the algorithm works well under the array error condition. BER difference between the non-blind method and this algorithm is less than 2 dB under a high SNR. The algorithm is blind and robust. The channel fading, the direction of arrive（DOA） imformation and the polarization information are needless in the algorithm.
