Supply chain risk assessment decision method and system

Abstract

The invention relates to a supply chain risk assessment decision method and system. The method comprises the following steps: carrying out single-factor evaluation modality and multi-factor weighted evaluation modality on a risk original data set of each link of a supply chain, and creating a multi-dimensional risk quantification model; according to the multi-dimensional risk quantification model, performing multi-stage time scale evaluation on the risk original data set to obtain a multi-stage risk evaluation result; based on the multi-stage risk assessment result, modeling a supply chain network into a directed graph, and constructing an integer programming model for solving to obtain a risk-oriented path allocation scheme; and according to the risk-oriented path distribution scheme, risk index calculation and early warning grade division are carried out on the multi-stage warehousing system to obtain a collaborative early warning strategy of the multi-stage warehousing system. According to the invention, reasonable distribution of supply chain resources is realized, the congestion risk of logistics nodes is effectively relieved, and the operation efficiency of the whole supply chain is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of supply chain risk assessment, and in particular to a supply chain risk assessment and decision-making method and system. Background Art

[0002] In the cross-border logistics environment, system congestion caused by a large accumulation of orders, uneven resource allocation across different logistics nodes, and downstream congestion caused by upstream supplier priorities have become key factors affecting the stable operation of the supply chain. Existing risk assessment frameworks often suffer from large fluctuations in assessment results, untimely warnings, and a single response strategy when faced with these complex situations.

[0003] Multi-tiered warehousing systems within cross-border supply chains constitute crucial nodes in the global logistics network. The coordinated operation of these nodes is crucial for overall supply chain risk management. However, current risk management systems lack comprehensive assessments of risks across different timescales, making it difficult to establish an effective linkage between long-term strategic planning, mid-term tactical adjustments, and short-term emergency response. As a result, risk prevention and control measures often lag behind the occurrence of risks, making it impossible to achieve early warning and precise risk prevention and control. Summary of the Invention

[0004] The main purpose of the present invention is to provide a supply chain risk assessment and decision-making method and system. The present invention realizes the rational allocation of supply chain resources, effectively alleviates the congestion risk of logistics nodes, and improves the operating efficiency of the overall supply chain.

[0005] To achieve the above objectives, the present invention provides a supply chain risk assessment and decision-making method, comprising the following steps: Conduct single-factor and multi-factor weighted assessments on the original risk data sets of each link in the supply chain to create a multi-dimensional risk quantification model; Performing a multi-stage time scale assessment on the original risk data set according to the multi-dimensional risk quantification model to obtain a multi-stage risk assessment result; Based on the multi-stage risk assessment results, the supply chain network is modeled as a directed graph and an integer programming model is constructed to solve it, thereby obtaining a risk-oriented path allocation plan; According to the risk-oriented path allocation scheme, the risk index of the multi-level warehousing system is calculated and the warning level is divided, and a collaborative warning strategy for the multi-level warehousing system is obtained.

[0006] The present invention also provides a supply chain risk assessment and decision-making system, comprising: Create a module to perform single-factor and multi-factor weighted assessment on the original risk data sets of each link in the supply chain, and create a multi-dimensional risk quantification model; An evaluation module, configured to perform a multi-stage time scale evaluation on the risk original data set according to the multi-dimensional risk quantification model to obtain a multi-stage risk evaluation result; A solution module, configured to model the supply chain network as a directed graph and construct an integer programming model to solve the problem based on the multi-stage risk assessment results, thereby obtaining a risk-oriented path allocation solution; The calculation module is used to calculate the risk index and divide the warning level of the multi-level warehousing system according to the risk-oriented path allocation plan, and obtain a collaborative warning strategy for the multi-level warehousing system.

[0007] In summary, the technical solution provided by the present invention, through the establishment of a dual-modal integrated assessment framework, combines single-factor and multi-factor weighted assessment modalities to achieve more comprehensive and accurate risk assessment results, avoiding the one-sidedness that can result from a single assessment method and enabling more accurate identification and quantification of risks across the supply chain. The multidimensional risk quantification model considers the interaction between logistics risk and cost risk and uses a time-weighted time decay function to time-weight risk data, allowing risk assessment to focus more on recent data and improving the timeliness and accuracy of risk assessment. Through a three-stage timescale assessment, prediction, correction, and response are performed for long-term, medium-term, and short-term risks, achieving full-time risk management coverage and simultaneously meeting the needs of strategic planning, tactical adjustments, and emergency response. Based on a risk-based path allocation scheme, the supply chain network is modeled as a directed graph and applied with mixed-integer linear programming and a variable neighborhood search algorithm. This approach can efficiently solve large-scale path allocation problems, reduce computational complexity, and improve solution efficiency. By calculating risk indices and categorizing warning levels for multi-level warehousing systems, a collaborative early warning mechanism is established among different warehousing facilities, enabling preventive measures to be taken before risks occur, reducing the probability and impact of risk events. The risk-oriented path allocation strategy takes into account the capacity constraints and risk distribution of each logistics node, realizes the rational allocation of supply chain resources, effectively alleviates the congestion risk of logistics nodes, and improves the operating efficiency of the overall supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a schematic diagram of the steps of a supply chain risk assessment and decision-making method according to one embodiment of the present invention; Figure 2 This is a structural block diagram of a supply chain risk assessment and decision-making system in one embodiment of the present invention.

[0009] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0011] Reference Figure 1 This embodiment provides a supply chain risk assessment and decision-making method, including the following steps: S1, conducts single-factor assessment and multi-factor weighted assessment on the original risk data sets of each link in the supply chain to create a multi-dimensional risk quantification model; Data on logistics delays, cost fluctuations, inventory levels, supplier performance, and market demand fluctuations are collected from various links in the supply chain. This data is collected from historical records, real-time monitoring systems, and supplier feedback. By integrating these multiple data sources, a raw risk dataset is generated, reflecting the risk profile of supply chain operations. Individual risk factors within the raw risk dataset are quantitatively scored to assess their impact on the supply chain. For example, logistics delay data is scored based on the duration and frequency of delays, as well as their impact on overall supply chain operations, while cost fluctuation data is scored based on the magnitude and frequency of cost fluctuations. The scoring criteria for each factor are determined based on actual conditions and industry experience. The resulting single-factor assessment results reflect the specific performance of each risk factor at different links. Each risk factor is weighted to produce a comprehensive multi-factor assessment. During the weighting calculation, each risk factor is assigned a different weight based on the actual situation. Once the weights are determined, a comprehensive score is generated through weighted summation. Decision thresholds T1 and T2 are set for the single-factor and multi-factor comprehensive assessment results, respectively. When the score of a single risk factor exceeds threshold T1, it means that the risk of that factor exceeds the acceptable range. Similarly, if the combined score of multiple factors exceeds threshold T2, the overall risk level of the supply chain is too high, impacting its stability and efficiency. The threshold setting depends on historical data analysis, industry standards, and the company's risk tolerance. A multi-dimensional assessment framework is constructed based on risk warning signals. This framework comprises serial, parallel, feedback, and feedforward assessment units. The serial assessment unit is suitable for assessing situations where risks accumulate gradually, such as the transmission of risk from one link to another. The parallel assessment unit is suitable for simultaneously assessing the impact of multiple independent risk sources on the supply chain, such as situations where multiple links are independent but have cumulative impacts. The feedback assessment unit assesses the impact of risks already experienced in historical data on the existing system, focusing on the ongoing impact of past supply chain risk events on the stability of the existing system. The feedforward assessment unit predicts supply chain risks based on current and future data, allowing for proactive preparation by anticipating potential risks. By combining these assessment units, different types of supply chain risks can be flexibly addressed and responded to promptly. Through the collaboration of these assessment units, a bimodal integrated assessment framework is formed. This framework automatically selects assessment units based on different risk types to achieve the optimal assessment results. Based on the bimodal integrated assessment framework, mathematical modeling of single-factor scores and multi-factor comprehensive scores is carried out to obtain a multi-dimensional risk quantification model.

[0012] Each individual factor assessment result is parameterized. Multiple variables, including the original risk measurement, scaling coefficient, and sensitivity parameter, are defined to capture the specific impact of each individual factor on supply chain risk. For example, individual risk factors such as logistics delays, inventory fluctuations, and supplier performance manifest themselves differently. Appropriate scaling coefficients are set to describe the strength of these factors' impact, while sensitivity parameters reflect their sensitivity to supply chain stability. By setting these parameters, a nonlinear risk function is constructed. This function not only accounts for the impact of individual factors on risk but also addresses the complex variations of these factors under varying conditions, resulting in a mathematical model for single-factor risk. Based on the evaluation results of the multi-factor comprehensive score, a multi-factor risk foundation model is established. Weighting coefficients are assigned to all relevant risk factors. These weighting coefficients are based on their importance, frequency, and potential impact in actual operations. By multiplying each risk factor's score by its corresponding weighting coefficient and then summing the products, a comprehensive risk score is generated, reflecting the multidimensional risks of each link in the supply chain. Quantitative analysis of the correlations between different risk factors in the multi-factor risk-based model reveals whether there are interdependencies or amplifying effects between them. For example, logistics delays and changes in inventory levels can influence each other, amplifying overall supply chain risk. By quantifying these correlations, the multi-factor risk-based model is modified to produce a modified risk model. This modified model comprehensively considers the interactive effects between risk factors, providing more realistic risk assessment results. The modified risk model calculates logistics risk and cost risk, reflecting their impact on the overall stability of the supply chain, resulting in a more comprehensive and representative overall risk index. The temporal attributes of the overall risk index are differentiated. A time-weighted risk assessment value is calculated to reflect the changing trends of supply chain risk over time. Based on this time-weighted risk assessment value, risk threshold parameters are adaptively adjusted. Risk thresholds are automatically optimized based on actual risk conditions and changes over time, ensuring that the risk management system remains optimal.

[0013] S2, based on the multi-dimensional risk quantification model, conducts a multi-stage time scale assessment on the original risk data set to obtain a multi-stage risk assessment result; Specifically, for the original risk data set within the first preset time window, an autoregressive integrated moving average analysis method is used to perform long-term risk forecasting. The autoregressive integrated moving average model is a time series forecasting method that effectively captures the time dependence and trend changes in the data by combining the autoregressive component of historical data with the moving average component. In long-term risk forecasting, the autoregressive integrated moving average model uses past data trends to infer future risk trends, helping companies identify potential risk events and risk levels. This step results in a set of long-term risk forecasts that reflect the potential evolution of risks across various links in the supply chain over an extended period of time. For the original risk data set within the second preset time window, a model predictive control method is used to perform medium-term risk correction. Model predictive control effectively controls system behavior by predicting future system states and making optimized decisions. In supply chain risk management, model predictive control methods leverage historical data and forecast results to dynamically adjust supply chain routes and correct medium-term risks in real time. By applying the model predictive control model, medium-term risk factors are corrected, and different logistics routes are classified into risk levels based on the correction results. This step assigns different risk levels to different logistics routes, identifying areas and time periods with higher risk. A medium-term risk distribution is generated, identifying high-risk areas and time periods. For real-time risk monitoring data within the third preset time window, the difference between the current risk value and the preset safety threshold is calculated. The monitoring system collects various risk indicators within the supply chain in real time, such as logistics delays, inventory fluctuations, and supplier status, and compares them with the preset safety threshold. When the difference in the monitored data is greater than zero, it indicates that the current risk level has exceeded the preset safety range, triggering a risk response. Short-term risk response conditions are generated based on this difference calculation. A larger difference indicates a more severe deviation from normal risk, increasing the urgency and intensity of the response. Based on the generated short-term risk response conditions, specific short-term risk response strategies are formulated, including route adjustments, resource reallocation, and contingency plans. When the risk response conditions are met, the company immediately adjusts logistics routes or production processes to ensure that resources can be quickly deployed to high-risk areas to reduce potential losses. The intensity of the response is proportional to the risk difference: the larger the risk difference, the more drastic the response measures and the higher the priority of the action. An adaptive response approach based on risk differentials ensures that the supply chain can respond quickly and effectively to unexpected risks, minimizing the likelihood of supply chain disruptions. This multi-stage risk assessment is generated by integrating long-term risk prediction results, mid-term risk distribution, and short-term risk response strategies.

[0014] S3, based on the multi-stage risk assessment results, models the supply chain network as a directed graph and constructs an integer programming model to solve it, obtaining a risk-oriented path allocation plan; It should be noted that the supply chain network is modeled using a directed graph. The core of a supply chain network lies in the connectivity between its nodes. In this model, the node set V represents key components of the supply chain, such as suppliers, warehouses, distribution centers, and customer locations, while the edge set E represents the logistics paths connecting these nodes. This modeling approach abstracts the entire supply chain structure into a graphical model, where nodes represent entities and edges represent logistics flow channels. Based on the network graph, a risk value is assigned to each path based on the results of a multi-stage risk assessment. The path risk value not only reflects the potential risk of the path itself but also considers the overall risk profile of the supply chain link within which the path resides. Decision variables are defined to represent the path selection process; these variables take values ​​of 0 or 1, indicating whether a path is selected. This process yields a set of path risk allocation results. A nonlinear integer programming model is constructed. The model's objective is to minimize the overall supply chain risk—that is, to reduce overall risk exposure by selecting appropriate path allocation schemes. The objective function is the sum of the risk value of each path multiplied by the path flow, with the goal of minimizing the total risk of all selected paths. When constructing the objective function, multiple constraints are incorporated, including flow conservation, capacity, and demand satisfaction. Flow conservation ensures that the incoming and outgoing flows of each node are balanced, capacity ensures that the transport capacity of each route is not exceeded, and demand satisfaction ensures that end-customer needs are met. These constraints ensure the rationality and feasibility of route selection and resource allocation. A mixed-integer linear programming approach is applied to the nonlinear integer programming model to obtain a linearized solution model. The linearized solution model is then subjected to an iterative optimization using a variable neighborhood search algorithm (VNSA) with multiple neighborhood structures, resulting in a risk-based route allocation solution. The VNSA is a metaheuristic algorithm for combinatorial optimization problems with strong global search capabilities. By setting multiple neighborhood structures, the VNSA explores the solution space. At each iteration, the algorithm selects a neighborhood for search based on the local structure of the current solution and then updates the current solution based on the search results. Through continuous iterative optimization, the algorithm gradually converges to a more optimal route allocation solution.

[0015] An initial solution is set for the supply chain routing problem, and K neighborhood structures of varying sizes are constructed to form a neighborhood search space. The number of neighborhood structures is set to K (where K is a positive integer). The size of the neighborhood determines the algorithm's search scope and the diversity of solutions. By constructing multiple neighborhood structures, the variable neighborhood search algorithm can flexibly search within solution spaces of varying sizes, thereby avoiding the constraints of local optimal solutions on the algorithm's performance. After constructing the neighborhood search space, the algorithm perturbs the current solution by randomly swapping the flow distribution between two paths to generate a new solution within the neighborhood. Two different paths within the current solution are randomly selected and some or all of the logistics flows on these two paths are swapped. This perturbation mechanism generates a variety of new solutions within the neighborhood search space, preventing the algorithm from prematurely falling into local optimal solutions and enhancing its ability to explore the global optimal solution. The objective function is then calculated for these new solutions. The objective function is primarily based on risk assessment values ​​and is specifically formulated by summing the product of the flow on each path and the path risk value to obtain the overall supply chain risk index. The objective function value of the new solution is compared with that of the current solution. If the new solution's objective function value is significantly better than the current solution, meaning that the overall risk index is significantly lower, the new solution is accepted and used as the new current solution for the next iterative search. Otherwise, it is rejected or accepted based on a certain probability to prevent the algorithm from getting stuck in a local optimum. Through a continuous perturbation and comparison mechanism, the algorithm gradually iterates, gradually moving towards a lower risk solution. Furthermore, to control the overall efficiency and optimization effect of the variable neighborhood search algorithm, several termination conditions are set in advance. These termination conditions include a maximum number of iterations, a maximum computational time, or no significant improvement in the solution after M consecutive iterations. When the number of iterations reaches the set upper limit, the algorithm has fully explored and is no longer significantly improving the current solution. When the computational time reaches the maximum allowable limit, the algorithm has exhausted its resources and is no longer searching. If no new solution is found after M consecutive iterations that is better than the current solution, the search process has reached a local optimum and should be terminated. When any of these conditions are met, the optimization process is considered complete, the algorithm terminates, and the optimal solution at that point is output as the final solution of the variable neighborhood search. The final solution of the variable neighborhood search algorithm is converted into a risk-based path allocation plan. The specific numerical values ​​output by the mathematical model and the optimized path flow allocation plan are converted into practical and actionable decision recommendations, including specific logistics path selection plans, resource reallocation plans for warehouses and distribution centers, and corresponding contingency plans.

[0016] S4, based on the risk-oriented path allocation scheme, the risk index of the multi-level warehousing system is calculated and the warning level is divided to obtain the collaborative warning strategy of the multi-level warehousing system.

[0017] Specifically, a multi-tiered warehousing system is divided into three main types of warehousing facilities: regional distribution centers, overseas warehouses, and terminal distribution sites. Regional distribution centers, located at the hub of the supply chain, are responsible for the storage and regional distribution of large-scale goods. Overseas warehouses are used for inventory turnover in international markets, reducing the uncertainty of cross-border logistics and improving local distribution efficiency. Terminal distribution sites are responsible for delivering to end customers, and their operational efficiency and stability directly impact the customer experience. By clarifying the hierarchical structure of the warehousing system, more targeted risk management strategies can be formulated to ensure that warehousing facilities at different levels can operate collaboratively in a risky environment and mitigate the impact of systemic risks. Based on a risk-based path allocation scheme, the risk level of each type of warehousing facility is quantified to generate a risk index for each facility. The risk index calculation takes into account multiple factors, including the logistics risk of the region where the warehousing facility is located, inventory turnover rate, upstream and downstream supply chain dependencies, historical fluctuations, and the operational stability of the warehousing facility itself. A comprehensive assessment of these factors yields a risk index for each warehousing facility. A collaborative risk threshold matrix is ​​then established. This matrix defines the risk linkage relationships and early warning triggering criteria between different warehousing facilities. This ensures that when the risk index of a particular warehousing facility exceeds a certain threshold, the risk response mechanisms of other associated warehousing facilities are automatically triggered. The core of the collaborative risk threshold matrix is ​​to analyze the degree of mutual influence between different warehousing facilities. For example, when the risk level of a regional distribution center increases, it can have a ripple effect on the stability of terminal distribution sites. Therefore, appropriate thresholds are set in the matrix to trigger corresponding early warning measures. This process forms a dynamic early warning system for the warehousing system, enabling warehousing facilities at all levels to coordinate when facing risks, thereby improving the overall risk resilience of the supply chain. Early warning levels are calculated based on the facility risk index values, and corresponding graded early warning strategies are developed for each level. For example, early warning levels are divided into three tiers: low, medium, and high, each corresponding to different levels of response measures. For low-risk levels, conventional inventory adjustments and route optimization strategies are implemented, while for medium-risk levels, additional inventory redeployment and the activation of redundant supply chain resources are required to ensure supply chain stability. For high-risk levels, more stringent measures are implemented, such as emergency inventory replenishment, temporary logistics route changes, and supplier substitutions, to prevent risk contagion and supply chain disruptions. Through refined early warning grading and strategy matching, we ensure that the supply chain can implement the most appropriate response measures at different risk levels. We calculate the expected risk reduction rate for each tiered early warning strategy to ensure that the final collaborative early warning strategy maximizes the overall risk reduction effect of the supply chain. The expected risk reduction rate is calculated based on historical data and simulation analysis. By evaluating the actual performance of different strategies in similar risk environments, we calculate their effectiveness in reducing the overall risk level of the warehousing system.Select a multi-level warehousing system collaborative early warning strategy that can maximize the expected risk reduction rate to ensure the stability and resilience of the supply chain in a complex environment.

[0018] In one example, the original risk data sets of each link in the supply chain were evaluated using single-factor and multi-factor weighted evaluation models to create a multi-dimensional risk quantification model, including: Collect data on logistics delays, cost fluctuations, inventory levels, supplier performance, and market demand fluctuations at all stages of the supply chain to obtain the original risk data set. Quantitatively score the single risk factor in the original risk data set to obtain the single factor assessment result, and assign different weights to each risk factor in the original risk data set to perform weighted calculation to obtain the multi-factor comprehensive assessment result; Decision thresholds T1 and T2 are set for the single-factor evaluation results and the multi-factor comprehensive evaluation results respectively. When the single-factor score exceeds the threshold T1 or the multi-factor comprehensive score exceeds the threshold T2, a risk warning signal is generated; According to the risk warning signal, a series evaluation unit, a parallel evaluation unit, a feedback evaluation unit, and a feedforward evaluation unit are constructed, and the evaluation units are automatically selected for different risk types to obtain a dual-modal integrated evaluation framework; Based on the bimodal integrated assessment framework, mathematical modeling of single-factor scores and multi-factor comprehensive scores was performed to obtain a multi-dimensional risk quantification model.

[0019] In this example, we collect data on logistics delays, cost fluctuations, inventory levels, supplier performance, and market demand fluctuations from the supply chain to build a risk original dataset. nodes, each node represents a link in the supply chain, such as a supplier, warehouse, or distribution center, while the path Represents the logistics transportation between two nodes. For each path, the following five types of risk data are defined: Logistics delay risk :path Average transportation time deviation, that is, the difference between actual transportation time and planned transportation time; cost fluctuation risk :path The fluctuation of unit transportation cost; inventory level risk :node The degree of deviation of inventory levels relative to safety stock; supplier performance risk :supplier The historical order fulfillment rate of the company. A low fulfillment rate indicates that the supplier's supply reliability is low and the risk is high; the risk of market demand fluctuations :time The degree of fluctuation of market demand relative to the mean reflects market uncertainty. After data normalization, the single-factor risk score of each node is calculated. After the risk warning is triggered, a bimodal integrated evaluation framework is constructed, which includes a serial evaluation unit, a parallel evaluation unit, a feedback evaluation unit, and a feedforward evaluation unit. The serial evaluation unit is used to analyze the cumulative propagation effect of risks in the supply chain, such as how downstream warehousing and logistics are affected when suppliers delay delivery. The parallel evaluation unit is used to analyze the risk accumulation on multiple logistics paths, such as the risk contribution to the overall supply chain when multiple logistics nodes are affected by market fluctuations at the same time. The feedback evaluation unit is used to analyze the impact of historical risks. For example, if a path has experienced serious delays in the past, will it affect the current logistics planning? The feedforward evaluation unit is used to predict future risks, such as whether inventory shortages will occur when market demand increases.

[0020] In one example, based on a bimodal integrated assessment framework, mathematical modeling of single-factor scores and multi-factor comprehensive scores was performed to obtain a multi-dimensional risk quantification model, including: By setting multiple variables such as original risk measurement value, proportional coefficient and sensitivity parameter, the single factor score in the single factor assessment result is parameterized and a nonlinear risk function is constructed to obtain a single factor risk mathematical model; A weight coefficient is set for each risk factor in the multi-factor comprehensive score in the multi-factor comprehensive assessment result, and the multi-factor risk basic model is obtained by calculating the sum of the product of the weight coefficient and the corresponding risk factor score; Quantitatively analyze the correlation between different risk factors in the multi-factor risk basic model to obtain a modified risk model. Then, calculate the logistics risk and cost risk through the modified risk model to obtain the overall risk index. The time attribute of the overall risk index is differentiated to obtain a time-weighted risk assessment value. Based on the time-weighted risk assessment value, the risk threshold parameters are adaptively adjusted to obtain a multi-dimensional risk quantification model.

[0021] In this example, the single factor score is parameterized and a nonlinear risk function is constructed to obtain a single factor risk mathematical model. In the supply chain system, the single factor score is affected by the original risk measurement value, the proportionality coefficient and the sensitivity parameter. A single factor risk function is set. As a node The risk value at the point where the risk occurs depends on different risk factors, such as logistics delays, cost fluctuations, inventory levels, supplier performance, and market demand fluctuations. By constructing a nonlinear risk function, the impact of different risk factors in the supply chain system on the overall risk can be more realistically reflected. A weight coefficient is set for each risk factor in the multi-factor comprehensive evaluation results, and the multi-factor risk basic model is calculated using the sum of the product of the weight coefficient and the corresponding risk factor score. Based on the modified risk model, the logistics risk and cost risk are calculated to obtain the overall risk index. Since supply chain risk has a dynamic characteristic in time, the time attribute of the overall risk index is differentiated to obtain a time-weighted risk assessment value. Based on the time-weighted risk assessment value, the risk threshold parameters are adaptively adjusted to obtain a multi-dimensional risk quantification model.

[0022] In one example, based on the multi-dimensional risk quantification model, a multi-stage time scale assessment is performed on the original risk dataset to obtain a multi-stage risk assessment result, including: Perform autoregressive integrated moving average analysis on the original risk data set within the first preset time window to obtain long-term risk prediction results; Perform model prediction and control on the original risk data set within the second preset time window to obtain a mid-term risk correction result. Based on the mid-term risk correction result, risk levels are assigned to different logistics routes to generate a mid-term risk distribution that identifies high-risk areas and time periods. For the real-time risk monitoring data within the third preset time window, the difference between the current risk value and the preset safety threshold is calculated, and a risk response is triggered when the difference is greater than zero, thereby obtaining a short-term risk response condition; Based on the short-term risk response conditions, formulate path adjustment, resource reallocation and emergency response plans. The response intensity is proportional to the risk difference, and the short-term risk response strategy is obtained; Integrate long-term risk forecast results, medium-term risk distribution and short-term risk response strategies to form multi-stage risk assessment results.

[0023] In this example, within the first preset time window, the autoregressive integrated moving average analysis method is used to combine the trend changes and seasonal characteristics of historical data to predict future supply chain risks. The risk data at each time point were regressed and combined with The error term is adjusted at each time point to produce a long-term risk forecast. Within the second preset time window, the goal of the medium-term risk assessment is to use model predictive control methods to modify risk in order to optimize the risk distribution along the supply chain. By solving the optimization problem, the medium-term risk correction results are obtained. These corrections are then used to assign risk levels to different logistics routes, generating a medium-term risk distribution for high-risk areas and time periods. Within the third preset time window, the core of the short-term risk assessment is to calculate the difference between the current risk value and the preset safety threshold in real time. When the difference exceeds zero, a risk response is triggered. To address short-term risks, route adjustments, resource reallocation, and emergency response plans are formulated. The long-term risk prediction results, medium-term risk distribution, and short-term risk response strategies are integrated to form a multi-stage risk assessment. Long-term risk predictions are used to formulate supply chain strategic adjustments, such as supplier selection and long-term inventory adjustments. Medium-term risk corrections are used to dynamically optimize the supply chain network, such as adjusting transportation modes and optimizing inventory distribution. Short-term risk responses enable real-time interventions, such as replanning logistics routes and activating emergency stockpiling.

[0024] In one example, based on the results of a multi-stage risk assessment, the supply chain network was modeled as a directed graph and an integer programming model was constructed to solve it. This resulted in a risk-based path allocation solution, including: The supply chain network is modeled as a directed graph, where the node set V represents suppliers, warehouses, distribution centers, and customer points, and the edge set E represents the logistics paths connecting the nodes, thus obtaining a supply chain network graph model; Based on the multi-stage risk assessment results, a risk value is assigned to each path and decision variables are defined to obtain the path risk allocation result; A nonlinear integer programming model is constructed. The objective function of the nonlinear integer programming model is set to minimize the overall risk and set flow conservation constraints, capacity constraints, and demand satisfaction constraints. Perform mixed integer linear programming on the nonlinear integer programming model to obtain a linearized solution model; The variable neighborhood search algorithm is applied to the linearized solution model, and multiple neighborhood structures are set for iterative optimization to obtain a risk-oriented path allocation solution.

[0025] In this example, the entire supply chain system is modeled as a directed graph to clarify the connection between each link in the supply chain. Suppose the supply chain network consists of a set of nodes and edge sets Composition, of which It includes supply chain links such as suppliers, warehouses, distribution centers and customer points. Represents the logistics path between nodes, that is, the transportation channel from supplier to warehouse, from warehouse to distribution center, and from distribution center to customer. In this model, each node Represents an entity in the supply chain, and each path Represents the slave node in the logistics network To Node Possible transport routes. For each path , set the transport capacity , represents the maximum amount of cargo that the route can carry, and defines the flow variable , indicating that in the path The actual number of goods transported on the supply chain is calculated. Based on the multi-stage risk assessment results, a risk value is assigned to each path to quantify the relative risk level of each path in the supply chain network. Mixed integer linear programming takes a long time to solve on large-scale supply chain networks, so a variable neighborhood search algorithm is used for optimization. This algorithm dynamically adjusts the search space and iteratively optimizes within multiple neighborhood structures to find the optimal path allocation solution.

[0026] Optionally, before modeling the supply chain network as a directed graph based on the multi-stage risk assessment results, the following steps are also included: constructing a data matrix for the multi-stage risk assessment results, treating each risk assessment indicator as an observation variable, constructing an observation matrix containing data from multiple time points, and obtaining a supply chain risk observation data set; performing singular value decomposition on the supply chain risk observation data set, calculating the system's singular value sequence and drawing a singular value distribution graph, and obtaining a preliminary estimation result of the system dimension; constructing multiple state space models of different dimensions based on the preliminary estimation result of the system dimension, each model corresponding to a different state space dimension, and obtaining a candidate model set; calculating the cross-Gramian matrix for each model in the candidate model set, which captures the interaction relationship and input / output dynamic characteristics between different risk states in the supply chain system, and obtaining the state interaction relationship Quantify the results; based on the quantification results of the state interaction relationship, calculate the risk estimation energy error of each candidate model, and the error quantification model fits the actual observation data to obtain the model quality evaluation index; set the model selection threshold according to the model quality evaluation index, and determine the optimal model dimension when the energy error is lower than the threshold and the error improvement is not significant after increasing the model dimension, and obtain the autonomous model dimension selection result; apply the autonomous model dimension selection result to the dynamic characteristics analysis of the supply chain network, determine the necessary number of state variables in the network modeling process, and obtain the optimized supply chain network modeling parameters; based on the optimized supply chain network modeling parameters, construct a supply chain network directed graph model with the optimal state space dimension, improve the model's expression accuracy of the actual supply chain dynamic characteristics, and provide a more accurate network topology structure for the subsequent integer programming model solution.

[0027] In one example, a variable neighborhood search algorithm is applied to a linearized solution model, and multiple neighborhood structures are set for iterative optimization to obtain a risk-oriented path allocation solution, including: Set an initial solution for the supply chain path allocation problem and construct K neighborhood structures of different sizes, where K is a positive integer, to obtain a neighborhood search space; Perturb the initial solution and randomly swap the traffic distribution of the two paths based on the current solution to obtain a new solution in the neighborhood. Calculate the objective function value of the new solution and compare it with the objective function value of the current optimal solution to obtain a local search result. Based on the local search result, accept the non-improved solution according to the set probability to jump out of the local optimum and obtain a global search strategy. Set termination conditions for the search process, including the maximum number of iterations, the maximum computation time, or no improvement after M consecutive iterations. When any of these conditions is met, the search is stopped and the algorithm termination judgment is obtained. The final solution of the variable neighborhood search algorithm is converted into a risk-oriented path allocation scheme, which includes a logistics flow allocation table and execution time schedule between nodes.

[0028] In this example, an initial solution is set , which represents the logistics flow allocation scheme in the supply chain network. Representative Path The cargo flow on the supply chain should meet the basic constraints of the supply chain, such as flow conservation, path capacity and customer demand. After determining the initial solution, construct Neighborhood structures of different sizes, each neighborhood structure corresponds to a different solution space, where is a positive integer representing the range of different solution sets that the algorithm can search. In order to optimize the initial solution, a perturbation is performed on the basis of the current solution, that is, the traffic distribution of the two paths is randomly exchanged to explore new feasible solutions.

[0029] In one example, based on a risk-oriented path allocation scheme, a risk index is calculated and an early warning level is divided for a multi-level warehousing system, resulting in a collaborative early warning strategy for the multi-level warehousing system, including: The multi-level warehousing system is divided into multiple types of warehousing facilities, including regional distribution centers, overseas warehouses, and terminal distribution stations, to obtain the hierarchical structure of the warehousing system; Based on the risk-oriented path allocation scheme, the facility risk index is calculated for each type of storage facility to obtain the risk index value of each facility; Set up a collaborative risk threshold matrix, which is used to trigger collaborative early warnings between different storage facilities and obtain early warning triggering standards; Calculate the warning level based on the facility risk index value, and generate a graded warning strategy for each warning level; Calculate the expected risk reduction rate of the hierarchical early warning strategy and select the multi-level warehousing system collaborative early warning strategy that maximizes the expected risk reduction rate.

[0030] In this example, the entire warehousing system is divided into different hierarchies to facilitate precise control within supply chain management. The warehousing system is assumed to consist of three main tiers: regional distribution centers, overseas warehouses, and terminal distribution sites. Within this structure, regional distribution centers are responsible for large-scale inventory management and regional distribution, overseas warehouses are used to reduce cross-border logistics delays, and terminal distribution sites directly serve customers, impacting final delivery reliability. After determining the hierarchical structure of the warehousing system, a risk index is calculated for each type of warehousing facility based on a risk-based routing solution. By calculating the risk index for each warehousing facility, the risk of the facility is identified, allowing for the development of appropriate response strategies. Following the calculation of the facility risk index, a collaborative risk threshold matrix is ​​established to trigger collaborative early warnings across different warehousing facilities. Alert levels are calculated based on the facility risk index values, and tiered early warning strategies are developed for each level. For each level, a tiered early warning strategy is developed, and its expected risk reduction rate is calculated to select the optimal collaborative early warning strategy for the multi-tiered warehousing system. To select the optimal collaborative early warning strategy, the algorithm must identify the one that maximizes the overall risk reduction rate among all possible early warning scenarios.

[0031] In this embodiment, according to the collaborative early warning strategy of the multi-level warehousing system, the following steps are also included: multi-dimensional characterization of the risk data generated by the collaborative early warning strategy of the multi-level warehousing system, converting the risk index, time series and spatial distribution information into a numerical matrix to obtain a risk data characterization matrix; constructing a supply chain risk visualization network model, which adopts a multi-layer convolutional neural network architecture to extract and enhance the features of the risk data characterization matrix to obtain a risk feature mapping result; applying a super-resolution processing algorithm to the risk feature mapping result, improving the resolution and clarity of the risk hot spot area through upsampling and detail reconstruction technology, and obtaining a high-resolution risk distribution map; designing and applying a denoising filter for the noise and artifacts in the high-resolution risk distribution map to eliminate the visual interference caused by data fluctuations and measurement errors, and obtaining a purified risk visualization result; based on The purified risk visualization results are used to construct a real-time risk monitoring interface, which includes a supply chain network topology diagram, a risk heat map, and a time series change curve, to obtain a multi-level risk visualization system; the multi-level risk visualization system is optimized for memory and accelerated for computation, and through model compression and parallel computing technology, the system is ensured to run at a speed of no less than 25 frames per second on standard hardware, to obtain real-time risk monitoring capabilities; based on the real-time risk monitoring capabilities, an interactive decision-making assistance tool is designed, which allows decision makers to directly adjust risk thresholds, modify path allocation parameters, and simulate the effects of different decision-making plans through the interface, to obtain a visual decision support system; the visual decision support system is integrated with the early warning strategy execution module to form a closed-loop feedback mechanism, which displays the early warning strategy execution effect in real time and dynamically adjusts the strategy parameters according to the execution results, to obtain an adaptive risk management platform.

[0032] Reference Figure 2 This embodiment provides a supply chain risk assessment and decision-making system, including: Create a module to perform single-factor and multi-factor weighted assessment on the original risk data sets of each link in the supply chain, and create a multi-dimensional risk quantification model; The evaluation module is used to perform a multi-stage time scale evaluation on the original risk data set based on the multi-dimensional risk quantification model to obtain a multi-stage risk assessment result; The solution module is used to model the supply chain network as a directed graph and construct an integer programming model to solve the problem based on the multi-stage risk assessment results, thereby obtaining a risk-oriented path allocation solution; The calculation module is used to calculate the risk index and divide the warning level of the multi-level warehousing system according to the risk-oriented path allocation plan, and obtain the collaborative warning strategy of the multi-level warehousing system.

[0033] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0034] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, system, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, system, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, system, article, or method comprising the element.

[0035] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A supply chain risk assessment and decision-making method, characterized by: The following steps are involved: Conduct single-factor and multi-factor weighted assessments on the original risk data sets of each link in the supply chain to create a multi-dimensional risk quantification model; Performing a multi-stage time scale assessment on the original risk data set according to the multi-dimensional risk quantification model to obtain a multi-stage risk assessment result; Based on the multi-stage risk assessment results, the supply chain network is modeled as a directed graph and an integer programming model is constructed to solve it, thereby obtaining a risk-oriented path allocation plan; According to the risk-oriented path allocation scheme, the risk index of the multi-level warehousing system is calculated and the warning level is divided, and a collaborative warning strategy for the multi-level warehousing system is obtained.

2. The supply chain risk assessment and decision-making method according to claim 1 is characterized in that: The risk original data sets of each link in the supply chain are evaluated using single-factor assessment mode and multi-factor weighted assessment mode to create a multi-dimensional risk quantification model, including: Collect data on logistics delays, cost fluctuations, inventory levels, supplier performance, and market demand fluctuations at all stages of the supply chain to obtain the original risk data set. Quantitatively scoring a single risk factor in the original risk data set to obtain a single factor assessment result, and assigning different weights to each risk factor in the original risk data set to perform weighted calculation to obtain a multi-factor comprehensive assessment result; Setting decision thresholds T1 and T2 for the single-factor evaluation result and the multi-factor comprehensive evaluation result, respectively, and generating a risk warning signal when the single-factor score exceeds the threshold T1 or the multi-factor comprehensive score exceeds the threshold T2; According to the risk warning signal, a series evaluation unit, a parallel evaluation unit, a feedback evaluation unit and a feedforward evaluation unit are constructed, and the evaluation units are automatically selected for different risk types to obtain a dual-modal integrated evaluation framework; Based on the dual-modal integrated assessment framework, mathematical modeling is performed on single-factor scores and multi-factor comprehensive scores to obtain a multi-dimensional risk quantification model.

3. The supply chain risk assessment and decision-making method according to claim 2, characterized in that: Based on the dual-modal integrated assessment framework, mathematical modeling is performed on single-factor scores and multi-factor comprehensive scores to obtain a multi-dimensional risk quantification model, including: By setting multiple variables including original risk measurement value, proportional coefficient and sensitivity parameter, parameterizing the single factor score in the single factor assessment result and constructing a nonlinear risk function, a single factor risk mathematical model is obtained; Setting weight coefficients for each risk factor in the multi-factor comprehensive score in the multi-factor comprehensive assessment result, and calculating the sum of the products of the weight coefficients and the corresponding risk factor scores to obtain a multi-factor risk basic model; Quantitatively analyzing the correlation between different risk factors in the multi-factor risk basic model to obtain a modified risk model, and calculating the logistics risk and cost risk using the modified risk model to obtain an overall risk index; The time attribute of the overall risk index is differentiated to obtain a time-weighted risk assessment value, and based on the time-weighted risk assessment value, the risk threshold parameter is adaptively adjusted to obtain a multi-dimensional risk quantification model.

4. The supply chain risk assessment and decision-making method according to claim 1 is characterized in that: The multi-stage time scale evaluation of the original risk data set is performed according to the multi-dimensional risk quantification model to obtain a multi-stage risk evaluation result, including: Perform autoregressive integrated moving average analysis on the original risk data set within the first preset time window to obtain long-term risk prediction results; Performing model prediction control on the original risk data set within the second preset time window to obtain a mid-term risk correction result, and assigning risk levels to different logistics routes based on the mid-term risk correction result to generate a mid-term risk distribution that identifies high-risk areas and time periods; For the real-time risk monitoring data within the third preset time window, the difference between the current risk value and the preset safety threshold is calculated, and a risk response is triggered when the difference is greater than zero, thereby obtaining a short-term risk response condition; Based on the short-term risk response conditions, formulate path adjustment, resource reallocation and emergency response plans, with the response intensity proportional to the risk difference, to obtain a short-term risk response strategy; The long-term risk prediction results, the medium-term risk distribution and the short-term risk response strategy are integrated to form a multi-stage risk assessment result.

5. The supply chain risk assessment and decision-making method according to claim 1 is characterized in that: Based on the multi-stage risk assessment results, the supply chain network is modeled as a directed graph and an integer programming model is constructed to solve the problem, thereby obtaining a risk-oriented path allocation solution, including: The supply chain network is modeled as a directed graph, where the node set V represents suppliers, warehouses, distribution centers, and customer points, and the edge set E represents the logistics paths connecting the nodes, thus obtaining a supply chain network graph model; Based on the multi-stage risk assessment results, a risk value is assigned to each path and a decision variable is defined to obtain a path risk allocation result; Constructing a nonlinear integer programming model, wherein the objective function of the nonlinear integer programming model is set to minimize the overall risk and set flow conservation constraints, capacity constraints, and demand satisfaction constraints; Performing mixed integer linear programming on the nonlinear integer programming model to obtain a linearized solution model; A variable neighborhood search algorithm is applied to the linearized solution model, multiple neighborhood structures are set for iterative optimization, and a risk-oriented path allocation solution is obtained.

6. The supply chain risk assessment and decision-making method according to claim 5 is characterized in that: The variable neighborhood search algorithm is applied to the linearized solution model, multiple neighborhood structures are set for iterative optimization, and a risk-oriented path allocation scheme is obtained, including: Set an initial solution for the supply chain path allocation problem and construct K neighborhood structures of different sizes, where K is a positive integer, to obtain a neighborhood search space; Perturbing the initial solution, randomly exchanging the flow distribution of the two paths based on the current solution, and obtaining a new solution in the neighborhood; Calculating the objective function value of the new solution and comparing it with the objective function value of the current optimal solution to obtain a local search result, and based on the local search result, accepting non-improved solutions according to a set probability to jump out of the local optimum and obtain a global search strategy; Setting termination conditions for the search process, including the maximum number of iterations, the maximum computation time, or no improvement after M consecutive iterations. When any of the conditions is met, the search is stopped and the algorithm termination judgment is obtained; The final solution of the variable neighborhood search algorithm is converted into a risk-oriented path allocation scheme, which includes a logistics flow allocation table and an execution time schedule between nodes.

7. The supply chain risk assessment and decision-making method according to claim 1 is characterized in that: The risk index calculation and warning level classification of the multi-level warehousing system are performed according to the risk-oriented path allocation scheme to obtain a collaborative warning strategy for the multi-level warehousing system, including: The multi-level warehousing system is divided into multiple types of warehousing facilities, including regional distribution centers, overseas warehouses, and terminal distribution stations, to obtain the hierarchical structure of the warehousing system; Based on the risk-oriented path allocation scheme, a facility risk index is calculated for each type of storage facility to obtain a risk index value for each facility; Setting a collaborative risk threshold matrix, wherein the collaborative risk threshold matrix is ​​used to trigger collaborative early warnings among different storage facilities, and obtaining early warning triggering criteria; Calculate the warning level based on the facility risk index value, and generate a graded warning strategy for each warning level; Calculate the expected risk reduction rate of the hierarchical early warning strategy, and select a multi-level warehousing system collaborative early warning strategy that maximizes the expected risk reduction rate.

8. A supply chain risk assessment and decision-making system, characterized by: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: Create a module to perform single-factor and multi-factor weighted assessment on the original risk data sets of each link in the supply chain, and create a multi-dimensional risk quantification model; An evaluation module, configured to perform a multi-stage time scale evaluation on the risk original data set according to the multi-dimensional risk quantification model to obtain a multi-stage risk evaluation result; A solution module, configured to model the supply chain network as a directed graph and construct an integer programming model to solve the problem based on the multi-stage risk assessment results, thereby obtaining a risk-oriented path allocation solution; The calculation module is used to calculate the risk index and divide the warning level of the multi-level warehousing system according to the risk-oriented path allocation plan, and obtain a collaborative warning strategy for the multi-level warehousing system.

Images