Abstract:To explore the potential traffic characteristics of large-scale and complex networks,it is insufficient to capture the spatial characteristics of traffic flows only considering the total traffic volume.It is also necessary to consider the intensity distribution of traffic flows in different origin-destination(OD)directions.To this end,the paper proposes a network partitioning method with OD directions of traffic sub-regions,to better capture the intensity distribution of traffic flows in different OD directions.Firstly,with the objectives of minimizing the variance of vehicle speed and regional compactness in each region,the original network is partitioned into multiple initial regions.On this basis,with the objectives of minimizing the vehicle speed variance within regions,maximizing network compactness,and minimizing the difference of region scales,a mathematical optimization model is established to merge the initial regions.The Technique for Order Preference by Similarity to an Ideal Solution is used to assess these three objectives,and the optimal partition result is obtained using the genetic algorithm.Vehicle GPS data is processed to obtain vehicle trajectories.Based on these trajectory data,a method is proposed to calculate the speed variance in each region,which involves storing the speed of vehicle ODs and link IDs.Finally,the effectiveness of the proposed partitioning method is verified using bus GPS data in Shenzhen.Results show that,compared with traditional methods,the proposed method achieves more compact regions with even sizes,improving partitioning efficiency.Moreover,by constructing a macroscopic fundamental diagram(MFD)with OD directions of traffic sub-regions,the spatial evolution characteristics of traffic flows in the network are better captured.

Abstract:This paper develops a manufacturer-led Stackelberg game model to examine the strategic choices of manufacturers and retailers in selling products with different levels of carbon emission reduction under a dual-channel supply chain consisting of direct sales and distribution.Results show that:1)when consumers preference for low-carbon products and the carbon reduction level of products increase,the optimal prices for both manufacturers and the retailers increase.2)When the carbon reduction level of products is either very low or very high,manufacturers tend to sell low-carbon products through the direct channel and distribute traditional products via the retail channel;otherwise,manufacturers sell traditional products directly and distribute low-carbon products through retailers.3)The equilibrium sales strategy of manufacturers is always detrimental to the retailer profits.4)A higher carbon reduction level of products is more beneficial to both manufacturers and retailers.This paper aims to provide insights for manufacturer strategies for selling carbon reduction products in a dual-channel supply chain.

Abstract:For a two-tier supply chain with a manufacturer and a retailer,this paper analyzes the impact of manufacturer preference for consumer surplus on the strategy of retailer private brand invasion,considering whether the business objectives of manufacturers take consumer surplus into account.The decisions and profits of supply chain members are compared between the different manufacturer business objectives of profit maximization and a hybrid of public welfare and profit.The findings reveal that the retailer invasion strategies can be classified into three types:ineffective,threatening,and mixed,according to the competitiveness of the private brand.Compared to profit maximization,the hybrid objective of manufacturers expands the region of the ineffective invasion strategy of retailers to hinder the private brand invasion;even in the case of a successful invasion,it can reduce potential profit losses.Regardless of whether the invasion occurs,manufacturer preference for consumer surplus weakens the double marginalization effect in the supply chain,reducing its own profit while improving the retailer profit,overall supply chain performance,and total consumer surplus.

Abstract:In a two-level supply chain comprised of a capital-constrained retailer with risk-aversion and a supplier,this paper adopts the conditional value-at-risk(CVaR)to characterize the retailer risk-aversion and develop a game model.Under supplier credit guarantees,option pricing of the supplier and the optimal ordering decisions of the retailer are studied with two financing modes:bank loans and delayed payments.Furthermore,the choice of financing modes for the retailer is explored,while the impact of retailer risk-aversion degrees and financing interest rates on equilibrium decisions is analyzed.Results show that the retailer risk-aversion degree affects its financing choice.If the risk-aversion coefficient is below a certain threshold,the retailer prefers partial credit loans,whereas a higher coefficient leads to the choice of delayed payments.The optimal option prices and ordering quantities increase with the rise of the risk-aversion coefficient under both financing modes.The CVaR of the retailer decreases with the increase of the bank loan interest rate but is independent of delayed payment interest rates.In addition,the retailer risk-aversion must be within a certain range,otherwise,the supply chain operation cannot be sustainable.Finally,the theoretical results are verified by numerical analysis.

Abstract:The four-way shuttle-based high-density storage system represents a new generation of stereoscopic warehouse solutions in high-density and complex environments.This paper aims to solve the problem that four-way shuttles tend to experience frequent turns and visit many invalid nodes during operation with defective areas,resulting in long task execution time and low algorithm search efficiency.By establishing a topological map of the storage environment,the search performance of different heuristic estimation functions in the four-way shuttle topological map is compared and analyzed.In scenarios with defective areas that may lead to significant deviations between heuristic estimates and actual path cost,a triangle-based heuristic method is introduced to improve the A*algorithm.A simulation platform for the four-way shuttle-based high-density storage system is constructed to verify the effectiveness of the proposed method.Simulation results show that the improved A* algorithm can effectively reduce the number of explored nodes and turning points,thereby improving the search efficiency.

Abstract:The development of modern industry demands factories to operate under a flexible production mode involving multi-variety and mixed small-to-large batch sizes.It requires precise real-time quality monitoring methods to address the risk of quality fluctuations due to frequent process adjustments,ensuring consistent product quality.This study aims to integrate statistical process control(SPC)for monitoring quality anomalies in processes with unknown parameter distributions.A non-parametric GWMA SR control chart is developed for real-time monitoring of quality parameters.To achieve anomaly recognition,a PCA-RF-GA model is proposed for feature selection and fusion.Feature dimensionality is reduced by principal component analysis(PCA),and a random forest(RF)algorithm optimized by genetic algorithm(GA)is proposed for pattern classification.Simulation tests compare the recognition speed and accuracy across feature combinations and classification methods,demonstrating that the proposed method achieves superior performance.The proposed method is further applied to UO2 pellet diameter monitoring as an actual example.Results show that the combination of GWMA SR control chart and PCA-RF-GA model effectively identifies abnormal quality fluctuations in actual production with unknown parameter distributions.

Abstract:To cope with the forecast errors caused by non-normal distributions of data in highway vehicle speed forecasting,this paper proposes a hybrid probabilistic forecasting method that integrates a deep temporal convolutional network(DeepTCN)and Copula theory.A DeepTCN framework is established first to give a deterministic forecast considering multiple features.Then,an appropriate Copula function is fitted based on the conditional probability distribution of prediction errors according to the obtained forecasts by DeepTCN.Finally,probabilistic results are provided by error compensation.Real-world traffic data collected from a highway road in Guangzhou,China are utilized to verify the effectiveness of the proposed hybrid method.Data analysis and experimental results show that real-world data reflects significant randomness in traffic flows,with this randomness being relatively mild at both low and high vehicle densities,but more evident at medium densities;DeepTCN demonstrates superior performance in handling long-term time series information compared with various existing methods,making it well-suited for highway vehicle speed forecasting;by incorporating Copula functions,it can compensate for forecast errors caused by data randomness to some extent,further improving forecast accuracy.

Abstract:To reduce the total risk and cost related to the mixed storage and transportation of hazardous waste,a multi-objective optimization model under uncertain demand is proposed,which aims to jointly optimize decisions on facility locations,inventory control and route planning.Considering the risk derived from mixed storage of multiple types of hazardous waste,an environmental impact disutility function of treatment stations is formulated by introducing a toxicity coefficient and an odor factor.The resident disutility effect of transfer stations is considered to develop a corresponding risk assessment model.To cope with the uncertain demand of hazardous waste in storage and transportation,a computation method of total generation amount is modeled according to probability distribution functions,and the maximum storage capacity of facilities is estimated.A location and transportation model for hazardous waste of mixed storage and transportation is developed with the objective of minimizing total risk and cost.Given the complexity and multi-objective nature of the proposed model,an improved NSGA-II algorithm is designed to solve the problem.Finally,several tests are provided to demonstrate the effectiveness of the proposed model and algorithm.Computational results show that,multiple effective location-routing plans can be provided by the proposed method.Comparing to traditional risk models,the integrated assessment model can generate solutions with better trade-offs between cost and risk,achieving both personnel safety and environmental protection.Comparing to general multi-objective optimization methods,the improved algorithm can reduce the computation time by 38.79%,and solve problems of various scales within 1700 seconds while maintaining high computational stability.

Abstract:To address the limitations of fault diagnosis performance based on a single sensor and a single fault feature,a two-stage multi-source sensor information fusion method for bearing fault diagnosis is proposed.First,considering sample imbalance characteristics and multi-feature correlation,an improved Stacking algorithm is proposed to construct a one-dimensional residual feature fusion network.This network improves the training of multi-sensor features and achieves fault feature fusion.Then,to cope with the uncertainty of multi-source sensor information,an improved evidence fusion rule considering the amount of evidence and its reliability is proposed to realize fault decision fusion based on Dempster-Shafer evidence theory.A case study on rolling bearings shows that the proposed method achieve all above 0.98 on average accuracy,precision,recall and f1-score under different working conditions.Compared with other seven methods,the proposed method shows superior performance and good generalization.

Abstract:To examine the impact of different integration behaviors adopted by firms with exploratory and exploitative M&A targets,this study categorizes spin-off behaviors into"additive"and"subtractive"ones from a network perspective.Integrating complex adaptive system theory,we develop an analytical framework that links M&A targets,integration behaviors,and outcomes.Additionally,an NK model is employed to compare the performance and evolution of interfirm networks for different spin-offs with the two types of M&A targets.Results show that:additive spin-offs increase betweenness centrality,reduce the constraint degree,global clustering coefficient,and overall network efficiency,rendering the ego network more open and facilitating the explorative target.Whereas the subtractive spin-offs improve the constraint degree and betweenness centrality while reducing the global clustering coefficient and overall network efficiency,making the ego network more closed,which is conducive to the exploitative target.The higher decision complexity makes a positive effect of spin-offs on M&A performance more pronounced than that of no spin-off behavior.

Abstract:As a critical piece of equipment in engineering manufacturing,the stability and reliability of welding guns are crucial for the continuity of production lines and the quality of products.To address the challenge of data imbalance in welding gun fault prediction,a welding gun fault prediction method based on an improved Jaya-RUSBoost model is proposed.By combining undersampling,ensemble learning,and parameter setting optimization,this method achieves data balance and improves the accuracy of fault prediction.First,a RUSBoost fault prediction model is constructed,and experiments are designed to evaluate the impact of hyperparameters on model performance,thereby determining the optimal range of model parameters.Subsequently,the Jaya metaheuristic algorithm is employed to iteratively optimize the parameters of the RUSBoost model to obtain the optimal parameter configuration of fault prediction.The results of the case study show that compared with the traditional RUSBoost algorithm,the proposed algorithm improves the average fault prediction accuracy and F1-score by 9.43%and 8.41%,respectively,across five welding guns.Moreover,compared with various machine learning models,the accuracy and other indicators are also significantly improved.The proposed method in this paper offers high practical value and broad prospects for promotion,providing effective support for the intelligent maintenance of welding equipment.

Abstract:As a core component of new energy vehicles,the thermal performance of lithium batteries directly affects the overall operational efficiency and safety of the vehicle.Enhancing battery thermal performance is therefore of critical importance.An Online Robust Parameter Design(ORPD)method based on observable noise information is proposed,which integrates offline design with online adjustment.The proposed method optimizes the parameters of thermal performance,improving both optimality and robustness.Considering the dynamic behavior of noise factors in actual production processes,a time series-based noise factor model is developed.Then,control variables are classified into online controllable and offline controllable categories,and a response surface model of the quality characteristic is constructed,with analytical expressions for its mean and variance being derived.Finally,a two-stage online adjustment strategy is introduced,aiming to minimize quality loss.After determining offline control variables,online control variables are dynamically adjusted based on real-time observations of noise factors,enabling compensation for both performance optimality and stability of quality characteristics.The effectiveness and robustness of the proposed ORPD approach are demonstrated through a case study involving the liquid cooling performance of lithium batteries in new energy vehicles.The results provide methodological support for robust optimization of battery thermal performance.

Abstract:To address the integrated optimization of layout and scheduling in fiber optic gyroscope assembly workshops,this paper incorporates the principle of lean logistics into the solution process.A mixed-integer programming model is developed with the objectives of minimizing the makespan,lean logistics distances and workstation layout rationality.Based on the specific problem features,an adaptive non-dominated sorting genetic algorithm III(NSGA-III)is designed to solve this problem.A crossover operator based on independent evolution is designed to better preserve high-quality encoding segments representing lean solutions without detours,reflows,or production waiting,thereby enhancing the exploratory capability of the algorithm.A"substitution-elimination"strategy is proposed to repair infeasible individuals generated during crossover.To reduce the waste of detours in logistics routes,a swap mutation operator is designed based on the lean logistics principle.Finally,an instance analysis is conducted on the assembly task of fiber optic gyroscope products in an enterprise to verify the effectiveness of the proposed approach.

Abstract:The balancing problem of type-I assembly lines focuses on finding an assembly line layout with the optimal cost under a given cycle time constraint.Traditional global search-based solution methods are computationally complex and time-consuming.This study addresses a practical scenario where customers are only concerned with the optimal cost rather than the specific layout plan.To this end,the search problem is transformed into a prediction problem,and artificial intelligence algorithms are adopted to predict the optimal cost.The bottleneck of the key factors that constrain the optimal cost cannot be quickly found using feature importance ranking analysis.Firstly,in the case of varying worker cost,a new simulation dataset is constructed by solving an integer linear programming model;secondly,based on this dataset,random forest regression,decision tree,and XGBoost algorithm models are trained to predict the optimal cost using seven parameters including type-I and type-II worker cost;finally,three different methods are applied to rank the importance of features in order to identify the key factors that constrain the optimal cost.The comprehensive performance of three regression algorithms is evaluated using four indicators:R2,RMSLE,EV,and MPE.Results show that the XGBoost algorithm has the highest MPE of 5.12%,while the random forest regression algorithm achieves the lowest MPE of 4.09%,confirming the feasibility of using intelligent algorithms to predict the optimal cost.This study provides a new method for optimal cost prediction using intelligent algorithms.The results of feature importance ranking indicate that the working time of type-I worker has a significantly greater impact on the optimal cost than other factors.

Abstract:To optimize the ambiguity resolution in real-time kinematic(RTK)positioning,a GNSS ambiguity resolution method based on both data and model-driven approaches is proposed.This algorithm focuses on selecting the optimal subset of ambiguities to increase the success rate of ambiguity fixing.First,considering the influence of ambiguity subsets on the accuracy of baseline solutions,a subset method based on the ambiguity level is adopted to ensure a high success rate of ambiguity fixing and high accuracy of baseline solutions.Model-and data-driven approaches are combined during partial ambiguity resolution,incorporating long short-term memory(LSTM)neural networks to ensure the stability and reliability of ambiguity fixing solutions.Results show that,while ensuring high reliability of ambiguity fixing,compared to existing algorithms,the LSTM algorithm based on resolution rules improves the fix rate and fix success rate of the selected dataset from 73.59%,72.83%,93.07%,and 87.57%to 99.23%,95.51%,98.67%,and 92.74%,respectively.Both the accuracy and robustness of RTK positioning are thus enhanced.

Abstract:To address the reliability assessment requirements of multi-stage mission systems,this study proposes an optimized design framework for accelerated degradation testing based on the Wiener degradation model.A mission success probability model is established to analyze the joint effects of load heterogeneity and temporal asymmetry across mission stages.Utilizing the large-sample approximation method,the asymptotic variance of parameter estimation is derived as the optimization objective to determine the optimal test plan.Numerical case studies reveal that high-load stages dominate resource allocation due to their nonlinear degradation characteristics,necessitating increased component allocation to capture rapid degradation patterns.Reduced stage duration amplifies the variance of parameter estimation,requiring a higher proportion of high-load group resources to compensate for data insufficiency.Two compromise rules are further proposed,and quantitative analyses using relative efficiency metrics demonstrate a positive correlation between mission complexity and efficiency loss.The resource-biased strategy achieves superior balance between estimation accuracy and practical constraints,providing actionable suboptimal solutions for engineering applications.

Abstract:With the increasing diversity of consumer market demand,the manufacturing industry is shifting from a single-variety and large-batch production mode to a multi-variety and small-batch one.Complex production systems like assembly systems are widely adopted to meet the market demand for diversified products.This paper conducts a transient performance analysis for assembly systems considering unreliable machines and limited buffers under a multi-type,small-batch production mode.First,based on the establishment of a standard mathematical model,we derive the exact expression of transient performance indicators using Markov analysis method.Then,to solve the state space explosion problem,a series of auxiliary production lines is introduced,and an efficient approximation method based on dynamic aggregation is proposed.Finally,the accuracy and effectiveness of the proposed approximation algorithm are validated through simulation-based comparison experiments.

Abstract:Driven by smart manufacturing in Industry 4.0 and the human-centric philosophy of Industry 5.0,there is an increasingly urgent demand in the discrete manufacturing industry to utilize precise spatial-temporal information of people,machines,and materials in the production process to enhance workshop logistics management.However,key challenges remain in accurately locating production resources in real-time and complex workshop environment,and utilizing spatial-temporal data to provide intelligent services in logistics operations for cost reduction and efficiency improvement.This paper analyzes the practical demand of logistics management in discrete manufacturing and explores the potential value of spatial-temporal data.Based on it,a framework of intelligent logistics management system driven by the indoor positioning technology is proposed.This framework combines the industrial internet of things(IIoT),cloud computing,digital twins(DT)and artificial intelligence(AI)to provide location-based intelligent services for various personnel in logistics operations.To accommodate complex workshop environment,a cell recognition indoor localization algorithm(CRILA)that integrates bluetooth low energy(BLE)and ultra-wideband(UWB)technologies is proposed,which can sense environmental changes and update positioning models online adaptively.It enables real-time positioning with high accuracy and stability,while also offering advantages of low cost,easy deployment,and strong scalability.Finally,the proposed system framework and algorithm are developed and validated in the mainframe production workshop of a computer manufacturer.

Abstract:In order to ensure the on-time delivery of machine tools in mixed-flow assembly shops,a scheduling optimization method based on improved deep multi-agent reinforcement learning is proposed,aiming to to address the low solution quality and slow training speed in minimizing production delays.A scheduling optimization model for mixed-flow assembly lines is constructed with the objective of minimizing delay time,where double deep Q network(DDQN)agents with decentralized execution are applied to learn the relationship between production information and scheduling objectives.The framework adopts the strategies of centralized training and decentralized execution,utilizing parameter sharing to deal with the non-stationary problem in multi-agent reinforcement learning.On this basis,a recurrent neural network is used to manage variable-length state and action representations,enabling agents to handle problems of arbitrary scale.A global/local reward function is also introduced to solve the reward sparsity problem in the training process.The optimal parameter combinations are identified through ablation experiments.Numerical experimental results show that,compared with the standard benchmarks,the proposed algorithm improves the average total delay of workpieces by 24.1%to 32.3%compared to before the improvement,and the training speed increased by 8.3%in terms of the achievement of the objective.

Abstract:Cold chain multimodal transport combines the particularity of cold chain transportation and the complexity of multimodal transportation.Compared with general transportation problems,its path optimization models are more complex and impose higher demand on solution algorithms.This paper presents a comprehensive review on the path optimization of cold chain multimodal transportation,focusing on models and algorithms.First,a knowledge graph visual analysis is conducted based on 369 Chinese and English literatures using a systematic literature review method.Then,40 key literatures among them that address both model construction and algorithm solutions are further examined.The selection of optimization objectives,cost calculation,and model-solving algorithms are summarized.It is found that most single-objective optimization studies focus on minimizing the total transportation cost,while multi-objective studies mainly minimize transportation cost,transportation time,and carbon emissions,maximize customer satisfaction.Although multi-objective optimization better reflects real-world scenarios,it often involves complex models and high computational complexity.Therefore,most studies still adopt single-objective optimization.For cost modeling,including refrigeration,cargo damage,and carbon emissions,most studies commonly use fixed cost coefficients and seldom consider the impact of different transportation conditions.Solution algorithms mainly fall into two categories:exact algorithms and heuristic algorithms.Exact algorithms are suitable for small-scale optimization problems,while machine learning-based heuristics are often combined with other algorithms and demonstrate greater advantages in solving large-scale practical problems.Finally,future development directions are discussed in terms of model construction,objective selection and computation,as well as algorithm implementation.