Search results

With the rapid growth of the domestic lending industry, assessing whether the borrower of each loan is at risk of default is a pressing issue for financial institutions. Although there are some models that can handle such problems well, there are still some shortcomings in some aspects. The purpose of this paper is to improve the accuracy of credit assessment models.

In this paper, three different stages are used to improve the classification performance of LSTM, so that financial institutions can more accurately identify borrowers at risk of default. The first approach is to use the K-Means-SMOTE algorithm to eliminate the imbalance within the class. In the second step, ResNet is used for feature extraction, and then two-layer LSTM is used for learning to strengthen the ability of neural networks to mine and utilize deep information. Finally, the model performance is improved by using the IDWPSO algorithm for optimization when debugging the neural network.

On two unbalanced datasets (category ratios of 700:1 and 3:1 respectively), the multi-stage improved model was compared with ten other models using accuracy, precision, specificity, recall, G-measure, F-measure and the nonparametric Wilcoxon test. It was demonstrated that the multi-stage improved model showed a more significant advantage in evaluating the imbalanced credit dataset.

In this paper, the parameters of the ResNet-LSTM hybrid neural network, which can fully mine and utilize the deep information, are tuned by an innovative intelligent optimization algorithm to strengthen the classification performance of the model.

The purpose of this study is to solve the problem of longer dead-time in the rear bridge leg switches and lower efficiency in the Four-Switch Buck-Boost LLC Resonant Converter.

The paper adopts time-domain analysis to derive the time-domain expression for optimal dead time, analyzing the conditions for achieving soft switching of the transistors. It further explores the relationship between the dead time of the bridge arm switching transistors and the input/output of the converter under different operating conditions. Specifically, the dead time of the upper bridge arm transistors increases with the converter input voltage and decreases with the output current. In contrast, the dead time of the lower bridge arm transistors is independent of the converter output current and decreases with increasing converter input voltage.

By simulating and constructing a 500 W experimental prototype, experimental results indicate that designing the dead time of the switch according to the optimal dead time proposed in this paper significantly improves efficiency when the converter operates from heavy load to full load. When the transformer takes minimum input, maximum input and intermediate bus voltage inputs respectively, its peak efficiency is increased by 0.6%, 1.7% and 1.1%, respectively, compared to the traditional four-switch Buck–Boost LLC resonant converter.

Experimental validation confirms the correctness of the optimal dead time design and analyzes the impact of different operating conditions of the converter on the dead time. This is of significant importance for the rational design of switch dead times and the enhancement of converter efficiency.

The purpose of this paper is to understand the consumption and demand of Chinese citizens for public digital culture, and make suggestions for government-supported public digital culture providers.

Through a questionnaire survey, this study investigates the provision of public digital cultural services (PDCS) from the perspective of consumption and demands.

The results indicate: the Chinese populace as a whole had low expenses on digital cultural services, and had not effectively utilized them to support their own development; significant disparities exist between demographics, particularly between urban and rural residents; the populace were strongly interested in participation in public digital culture, but the services had low actual utilization rates; and the services had been unable to meet the users’ quality-related demands.

The first study to approach the provision of PDCS from the side of consumption and user demand.

Although building information modeling (BIM) has brought competitive advantages and many new jobs, the BIM-related job market is still confusing in China, which will undermine the adoption of BIM. This paper aims to show what kinds of BIM-related jobs are there in China, what employers require and whether all BIM engineers are the same kind.

A text mining approach, structural topic model, was used to process the job descriptions of 1,221 BIM-related online job advertisements in China, followed by a cluster analysis based on it.

First, 10 topics of requirements with the impact of experience and educational background to them were found, namely, rendering software, international project, design, management, personal quality, experience, modeling, relation and certificate. Then, six types were clustered, namely, BIM modeler, BIM application engineer, BIM consultant, BIM manager, BIM developer and BIM designer. Finally, different kinds of BIM engineers proved this title was an expediency leading to confusion.

This paper can provide a clear and insightful look into the confusing and unheeded BIM-related job market in China and might help to cope with the abuse of job titles. It could also benefit both employers and candidates in their recruitment for better matching.

Limited by the types of sensors, the state information available for musculoskeletal robots with highly redundant, nonlinear muscles is often incomplete, which makes the control face a bottleneck problem. The aim of this paper is to design a method to improve the motion performance of musculoskeletal robots in partially observable scenarios, and to leverage the ontology knowledge to enhance the algorithm’s adaptability to musculoskeletal robots that have undergone changes.

A memory and attention-based reinforcement learning method is proposed for musculoskeletal robots with prior knowledge of muscle synergies. First, to deal with partially observed states available to musculoskeletal robots, a memory and attention-based network architecture is proposed for inferring more sufficient and intrinsic states. Second, inspired by muscle synergy hypothesis in neuroscience, prior knowledge of a musculoskeletal robot’s muscle synergies is embedded in network structure and reward shaping.

Based on systematic validation, it is found that the proposed method demonstrates superiority over the traditional twin delayed deep deterministic policy gradients (TD3) algorithm. A musculoskeletal robot with highly redundant, nonlinear muscles is adopted to implement goal-directed tasks. In the case of 21-dimensional states, the learning efficiency and accuracy are significantly improved compared with the traditional TD3 algorithm; in the case of 13-dimensional states without velocities and information from the end effector, the traditional TD3 is unable to complete the reaching tasks, while the proposed method breaks through this bottleneck problem.

In this paper, a novel memory and attention-based reinforcement learning method with prior knowledge of muscle synergies is proposed for musculoskeletal robots to deal with partially observable scenarios. Compared with the existing methods, the proposed method effectively improves the performance. Furthermore, this paper promotes the fusion of neuroscience and robotics.

The purpose of this paper is to consider three factors, namely, intra-week demand fluctuations, interrelationship between the number of robots and order scheduling and conflicting objectives (i.e. cost minimization and customer satisfaction maximization), to optimize the robot logistics system.

The number of robots and the sequence of delivery orders are first optimized using the heuristic algorithm NSGACoDEM, which is designed using genetic algorithm and composite difference evolution. The superiority of this method is then confirmed by a case study of a four-star grade hotel in South Korea and several comparative experiments.

Two performance metrics reveal the superior performance of the proposed approach compared to other baseline approaches. Results of comparative experiments found that the consideration of three influencing factors in the operation design of a robot logistic system can effectively balance cost and customer satisfaction over the course of a week in hotel operation and optimize robot scheduling flexibility.

The results of this study reveal that numerous factors (e.g. intra-week demand fluctuations) can optimize the performance efficiency of robots. The proposed algorithm can be used by hotels to overcome the influence of intra-week demand fluctuations on robot scheduling flexibility effectively and thereby enhance work efficiency.

The design of a novel algorithm in this study entails enhancing the current robot logistics system. This algorithm can successfully manage cost and customer satisfaction during off-seasons and peak seasons in the hotel industry while offering diversified schemes to various types of hotels.

In regarding to operational efficiency and safety improvements, multiple tower crane service scheduling problem is one of the main problems related to tower crane operation but receives limited attention. The current work presents an optimization model for scheduling multiple tower cranes' service with overlapping areas while achieving collision-free between cranes.

The cooperative coevolutionary genetic algorithm (CCGA) was proposed to solve this model. Considering the possible types of cross-tasks, through effectively allocating overlapping area tasks to each crane and then prioritizing the assigned tasks for each crane, the makespan of tower cranes was minimized and the crane collision avoidance was achieved by only allowing one crane entering the overlapping area at one time. A case study of the mega project Daxing International Airport has been investigated to evaluate the performance of the proposed algorithm.

The computational results showed that the CCGA algorithm outperforms two compared algorithms in terms of the optimal makespan and the CPU time. Also, the convergence of CCGA was discussed and compared, which was better than that of traditional genetic algorithm (TGA) for small-sized set (50 tasks) and was almost the same as TGA for large-sized sets.

This paper can provide new perspectives on multiple tower crane service sequencing problem. The proposed model and algorithm can be applied directly to enhance the operational efficiency of tower cranes on construction site.

This study conducts a comparative analysis of the diversification effects of China's national carbon market (CEA) and the EU ETS Phase IV (EUA) within major commodity markets.

The study employs the TVP-VAR extension of the spillover index framework to scrutinize the information spillovers among the energy, agriculture, metal, and carbon markets. Subsequently, the study explores practical applications of these findings, emphasizing how investors can harness insights from information spillovers to refine their investment strategies.

First, the CEA provide ample opportunities for portfolio diversification between the energy, agriculture, and metal markets, a desirable feature that the EUA does not possess. Second, a portfolio comprising exclusively energy and carbon assets often exhibits the highest Sharpe ratio. Nevertheless, the inclusion of agricultural and metal commodities in a carbon-oriented portfolio may potentially compromise its performance. Finally, our results underscore the pronounced advantage of minimum spillover portfolios; particularly those that designed minimize net pairwise volatility spillover, in the context of China's national carbon market.

This study addresses the previously unexplored intersection of information spillovers and portfolio diversification in major commodity markets, with an emphasis on the role of CEA.

This study focuses on the CFD numerical simulation and analysis of the vortex stacking problem at the top of the impeller of a high-speed fuel pump, mainly using LCS and entropy production theory to visualize the vortex at the top of the impeller as well as quantitatively analyzing the energy loss caused by the vortex at the top of the impeller. By combining the two methods, the two are well verified with each other that the stacking problem of the vortex at the top of the impeller and the location of the energy loss caused by the vortex are consistent with the vortex location. Such a method can reveal the problem of vortex buildup at the top of the lobe well, and provide a novel guidance idea for improving the performance of high-speed fuel pumps.

Based on CFD numerical simulation and analysis, this study mainly uses LCS and entropy production theory to visualize the top vortex of the impeller. Through the combination of the two methods, the accumulation problem of the top vortex of the impeller and the location of the energy loss caused by the vortex can be well revealed.

(1) The CFD numerical simulation analysis of the high-speed fuel pump is carried out, and the test is conducted to verify the numerical simulation results. The inlet and outlet pressure difference? P is used as the validation index, and the error analysis shows that the error between numerical simulation and test results is within 10%, which meets our requirements. Therefore, we carry out the next analysis with the help of CFD numerical simulation. By analyzing the full working condition simulation, its inlet and outlet differential pressure? P and efficiency? Are evaluated. It is found that its differential pressure decreases with the flow rate and its efficiency reaches its maximum at Qv = 9.87 L/s with a maximum efficiency of 78.32%. (2) We used the LCS in the analysis of vortices at the top of the impeller blades of a high-speed fuel pump. One of the metrics used to describe the LCS in fluid dynamics is the FTLE. The high FTLE region represents the region with the highest and fastest particle trajectory stretching velocity in the fluid flow. We performed a cross-sectional analysis of the FTLE field on the different height surfaces of the impeller on 25% Plane, 50% Plane, and 75% Plane, respectively. And a quarter turn of the rotor rotation was analyzed as a cycle divided into 8 moments. It is found that on 25% Plane, the vortex at the top of the lobe is not obvious, but there are high FTLE values on the shroud surface. On 50% Plane, the lobe top vortex is relatively obvious and the number of vortices is three. The vortex pattern remains stable with the rotating motion of the rotor. At 75% Plane, the lobe top vortex is more visible and its number of vortices increases to about 5 and the vortex morphology is relatively stable. The FTLE ridges visualize the vortex profile. This is a good guide for fluid dynamics analysis. (3) At the same time, we use the entropy production theory to quantitatively analyze the energy loss, and define the entropy production rate Ep. Through the entropy production analysis of the impeller shroud surface and the suction surface of the pressure surface of the blades at eight moments, we find that the areas of high energy loss are mainly concentrated in the leading and trailing edges of the blades as well as in the shroud surface close to the leading edge of the blades, and the value of the entropy production rate is up to 106 W/m3/K. The areas of high energy loss in the leading edge of the blades as well as the trailing edge show a curved arc, and the energy loss is decreasing as it moves away from the shroud surface and closer to the hub surface. The high energy loss areas at the leading and trailing edges of the blades are curved, and the energy loss decreases as they move away from the shroud surface and closer to the hub surface. The energy loss at the pressure surface of the blade is relatively small, about 5 × 105 W/m3/K, which is mainly concentrated near the leading edge of the blade near the shroud surface and the trailing edge of the blade near the hub surface. Such energy loss corresponds to the vortex LCS at the top of the impeller, and the two mirror each other.

This study focuses on the CFD numerical simulation and analysis of the vortex stacking problem at the top of the impeller of a high-speed fuel pump, mainly using LCS and entropy production theory to visualize the vortex at the top of the impeller as well as quantitatively analyzing the energy loss caused by the vortex at the top of the impeller. By combining the two methods, the two are well verified with each other that the stacking problem of the vortex at the top of the impeller and the location of the energy loss caused by the vortex are consistent with the vortex location. Such a method can reveal the problem of vortex buildup at the top of the lobe well, and provide a novel guidance idea for improving the performance of high-speed fuel pumps.

A model for calculating the global overpressure time history of a single cloud detonation from overpressure time history of discrete positions in the range of single cloud detonation is to be proposed and verified. The overpressure distribution produced by multiple cloud detonation and the influence of cloud spacing and fuel mass of every cloud on the overpressure distribution are to be studied.

A calculation method is used to obtain the global overpressure field distribution after single cloud detonation from the overpressure time history of discrete distance to detonation center after single cloud detonation. On this basis, the overpressure distribution produced by multi-cloud under different cloud spacing and different fuel mass conditions is obtained.

The results show that for 150 kg fuel, when the spacing of three clouds is 40 m, 50 m, respectively, the overpressure range of larger than 0.1 MPa is 5496.48 mˆ2 and 6235.2 mˆ2, which is 2.89 times and 3.28 times of that of single cloud detonation. The superposition effect can be ignored when the spacing between the three clouds is greater than 60 m. In the case of fixed cloud spacing, once the overpressure forms continuous effective superposition, the marginal utility of fuel decreases.

A model for calculating the global overpressure time history of a single cloud detonation from overpressure time history of discrete positions in the range of single cloud detonation is proposed and verified. Based on this method, the global overpressure field of single cloud detonation is reconstructed, and the superimposed overpressure distribution characteristics of three cloud detonation are calculated and analyzed.