Supply chain service type resilience dynamic monitoring system based on multi-source data

Abstract

The application relates to the technical field of supply chain data processing and risk control, and discloses a supply chain service type resilience dynamic monitoring system based on multi-source data, which comprises a business access module, a parameter mapping module, a boundary and disturbance extraction module, a damping reset and evolution module, a numerical integral calculation module and a failure tracking and early warning module. When the state verification evolution module determines that the transient deviation of a preset service node exceeds the lower limit constraint vector or the upper limit constraint vector, the corresponding coefficient value in the initial damping matrix is processed to be zero, and a second-order system dynamics description model is constructed by using the zero-updated dynamic damping matrix. The model changes the limitation of the traditional scheme which only adopts linear diffusion logic, accurately simulates the nonlinear mutation process of the supply chain elastic service performance buffer being completely depleted and entering the denial of service state, and deeply reflects the dynamic evolution characteristics of the elastic performance mechanism being invalid and the service attenuation window being closed due to overload of a specific service node.

Description

Technical Field

[0001] This invention relates to the field of supply chain data processing and risk control technology, specifically a dynamic monitoring system for supply chain service resilience based on multi-source data. Background Technology

[0002] Supply chain resilience monitoring refers to the process of collecting and analyzing data from service nodes to assess the resilience and recovery capabilities of the supply chain network in the face of supply and demand fluctuations. It involves real-time monitoring of the status of upstream and downstream enterprise nodes and the transmission of risks, and is crucial for ensuring the continuous flow of goods and the security of the capital chain.

[0003] Existing monitoring technologies typically rely on the underlying system to capture basic business indicators such as inventory capacity, funding amount, goods in transit, or delivery date to build an early warning rule base. In practical applications, fixed thresholds are often set, such as setting a safety stock warning line. When a single node indicator exceeds the limit, an alarm is triggered. In addition, some solutions combine graph network characteristics to calculate the static risk assessment score of the supply chain through the frequency of order transactions to assist in scheduling.

[0004] However, existing technologies have the drawback of being unable to track the cascading failure risk caused by node buffer failures. Enterprise nodes have physical limits on their buffer resources, such as safety stock or delivery redundancy. When external disturbances accumulate and deplete the buffer resources of a specific node, the node's ability to mitigate business fluctuations fails, and the fluctuations are directly transmitted to adjacent nodes. Traditional solutions only use threshold alarms or linear diffusion logic, failing to reflect the dynamic evolution process after buffer depletion. These shortcomings prevent existing systems from accurately calculating the dissipative transfer range of fluctuations and extracting rigid transmission topology paths, resulting in delayed cascading failure warnings and inaccurate risk identification. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a dynamic monitoring system for supply chain service resilience based on multi-source data, which solves the problem that existing systems have difficulty extracting the non-dissipative rigid transmission path of fluctuations after node buffers are exhausted.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a supply chain service resilience dynamic monitoring system based on multi-source data, including a data acquisition module for accessing the business databases of each preset service node in the supply chain service flow network and acquiring business data, wherein the business data includes static capacity parameters, dynamic business flow parameters and node association parameters; The parameter mapping module receives the business data and maps the supply chain service flow network into a physical parameter matrix containing a node inertia matrix, an initial damping matrix, and a directed network stiffness matrix. It establishes the corresponding node inertia matrix, the initial damping matrix, and the directed network stiffness matrix based on the static capacity parameters, the dynamic business flow parameters, and the node association parameters, respectively. The boundary disturbance extraction module is used to extract the extreme boundary of the preset service node to generate the lower limit constraint vector and the upper limit constraint vector; at the same time, it captures external abnormal disturbances and converts them into time-dependent excitation vectors. The state verification evolution module receives the physical parameter matrix, the lower limit constraint vector, the upper limit constraint vector, and the time-dependent excitation vector, performs calculus and difference solutions, and when it is determined that the transient deviation of the preset service node exceeds the lower limit constraint vector or the upper limit constraint vector, the corresponding coefficient value in the initial damping matrix is ​​set to zero, and a second-order system dynamic description model is constructed using the updated dynamic damping matrix. The numerical integration module is used to receive and iteratively solve the second-order system dynamics description model; The link tracing module is used to extract the preset service node sequence corresponding to the occurrence coefficient value being zero from the dynamic damping matrix after it is updated to zero, determine it as a rigid transmission topology path, map the rigid transmission topology path back to the cascade service failure and broken link in the supply chain service flow network, and output a supply chain service failure alarm signal.

[0007] Preferably, the static capacity parameters include the absolute capacity limit parameter of the preset service node, the capital reserve limit parameter, the physical storage capacity limit, the maximum design capacity of the production line, the maximum design purchase batch per transaction, and the maximum advance payment amount of the letter of credit. The dynamic business flow parameters include the current inventory of the preset service node, the amount of materials in transit, the safety stock buffer, the delivery time redundancy, the internal material self-consumption rate, the internal production cycle, the lower limit zero inventory threshold, and the working capital depletion point. The node association parameters include the forward order binding rate and reverse order binding rate between the upstream and downstream preset service nodes, the forward average supply cycle and the reverse average supply cycle, the reverse fund settlement cycle, and the limit value of the order delay tolerance period agreed in the business contract.

[0008] Preferably, the parameter mapping module specifically includes: An inertial calculation unit is used to independently perform dimensionless normalization processing on the absolute capacity limit parameter and the capital reserve limit parameter, perform linear weighted calculation using preset weight coefficients to obtain the equivalent inertial coefficient of the preset service node, and construct the diagonalized node inertial matrix according to the preset service node arrangement sequence. The damping calculation unit is used to perform dimensionless normalization processing on the safety stock buffer amount and the delivery time redundancy amount, and to perform linear weighted calculation using the set damping weight coefficient to obtain the initial damping coefficient of the preset service node, and to construct the diagonalized initial damping matrix. The stiffness calculation unit is used to perform a multiplication operation on the positive order binding rate and the reciprocal of the positive average delivery cycle which is limited by a very small time precision threshold to obtain the stiffness coefficient, and to obtain the internal node stiffness parameters based on the internal material self-consumption rate and the reciprocal of the internal production cycle, and to assemble and construct the asymmetric directed network stiffness matrix.

[0009] Preferably, before establishing the physical parameter matrix, the parameter mapping module performs cleaning and time alignment processing on the collected business data, and uses a preset time resampling period to uniformly map the asynchronously acquired multi-source business data streams to the same timestamp sequence; for discrete business data with abrupt state changes, a forward filling algorithm is used to fill in the gaps, and for smoothly changing continuous numerical business data, a linear interpolation algorithm is used to fill in the gaps, and the set time resampling period is consistent with the time step of the calculus difference solution stage.

[0010] Preferably, the boundary disturbance extraction module extracts the lower limit zero inventory threshold, the working capital depletion point, and the order delay tolerance time limit as their original values, calculates the difference between these values ​​and the initial state benchmark value of the preset service node, and performs a dimensionless transformation to generate a lower limit hard limit parameter to construct the lower limit constraint vector. The same process is then performed on the upper limit of the warehouse physical capacity, the maximum design capacity of the production line, and the maximum advance payment amount of the letter of credit to generate upper limit hard limit parameters to construct the upper limit constraint vector. Simultaneously, for external abnormal disturbance quantities containing different physical dimensions, the corresponding extreme value reference standard is used as the calculation denominator to perform proportional conversion, obtaining a dimensionless excitation value of the positive scalar value to construct the time-dependent excitation vector.

[0011] Preferably, the state verification evolution module has a built-in piecewise nonlinear decision function, which executes the following logic within each discrete time step: Calculate the transient deviation of each preset service node and construct the state vector at the current moment; when it is determined that the transient deviation is greater than the lower hard limit parameter and less than the upper hard limit parameter, assign the dynamic damping coefficient of the corresponding preset service node at the current moment to the corresponding coefficient value in the initial damping matrix; When it is determined that the transient deviation is less than or equal to the lower hard limit parameter, or greater than or equal to the upper hard limit parameter, the dynamic damping coefficient of the corresponding preset service node at the current moment is forcibly assigned to a value of zero. The updated dynamic damping coefficients are used to generate the zeroed-updated dynamic damping matrix, and the set of matrix parameters containing this matrix is ​​transmitted to the numerical integration calculation module.

[0012] Preferably, the second-order system dynamics description model includes a second-order rate of change vector, a first-order rate of change vector, the state vector, the directed network stiffness matrix, the nodal inertia matrix, the zeroed-updated dynamic damping matrix, and the time-dependent excitation vector. The diagonal elements with zero values ​​in the zeroed-updated dynamic damping matrix cause the calculation of the first derivative dissipative drag of the transient deviation to be stopped at the corresponding preset service node, allowing the undiminished value of the transient deviation to be directly transmitted to the downstream preset service node through the directed network stiffness matrix.

[0013] Preferably, the numerical integration calculation module is specifically used for: A matrix reduction substitution is performed to eliminate the second-order differential term corresponding to the second-order rate of change vector in the second-order system dynamics description model. The state vector is extracted and concatenated with the first-order rate of change vector to form an extended state vector. A matrix inversion operation is performed on the node inertia matrix to obtain the inverse inertia matrix. This inverse matrix is ​​then used to simultaneously establish the zeroed-updated dynamic damping matrix and the directed network stiffness matrix to generate the reduced-order first-order state-space equation. A fourth-order Runge-Kutta algorithm is used to perform numerical integration derivation to obtain the updated extended state vector corresponding to the next time step.

[0014] Preferably, the link tracing module is specifically used for: Filter out the non-dissipative anomaly preset service nodes corresponding to the target diagonal position index with a value of zero in the dynamically damped matrix after being zeroed and updated; The graph traversal connected component search algorithm is called to cluster adjacent non-dissipative anomaly preset service nodes with connected paths into directed connected subgraphs, and the directed connecting edges in the subgraphs are extracted and marked as anomaly rigid propagation topology paths. Starting from the endpoint of the abnormal rigid transmission topology path, a topology traversal search is performed. The state transmission ratio is calculated using the directed network stiffness matrix and the internal node stiffness parameters of the affected preset service node. The transient deviation increment transmitted to the affected preset service node is obtained and the predicted state value is calculated. The preset service nodes affected by the predicted state values ​​exceeding the lower or upper constraint vectors are used as new traversal starting points for iterative splicing until no new failure breakdown state nodes are added in the supply chain service flow network, thus generating the complete cascading service failure and breakage link.

[0015] Preferably, the boundary disturbance extraction module further performs boundary recalculation at a preset update cycle, wherein the preset update cycle is set to an integer multiple of the time step set for performing the calculus difference solution; the boundary disturbance extraction module dynamically modifies the element values ​​in the lower limit constraint vector and the upper limit constraint vector according to the contract adjustment data or capacity change data fed back by the data acquisition module as the business data, so as to achieve dynamic consistency correction between the hard limit boundary and the actual business capability.

[0016] This invention provides a dynamic monitoring system for supply chain service resilience based on multi-source data. It has the following beneficial effects: 1. This invention determines, through the state verification evolution module, when the transient deviation of a preset service node exceeds the lower or upper limit constraint vector, that the corresponding coefficient value in the initial damping matrix is ​​set to zero and a second-order system dynamic description model is constructed using the updated dynamic damping matrix. This changes the limitation of the traditional scheme that only uses linear diffusion logic, accurately simulates the nonlinear mutation process of the supply chain elastic service fulfillment buffer being completely exhausted and entering a denial-of-service state, and deeply reflects the dynamic evolution characteristics of a specific service node causing the elastic fulfillment mechanism to fail and the service decay window to close due to overload.

[0017] 2. This invention extracts the sequence of preset service nodes corresponding to the zeroing of occurrence coefficient values ​​from the updated dynamic damping matrix using a link tracing module, and determines it to be a rigid transmission topology path. Combined with the predicted state values ​​of the affected preset service nodes, it performs iterative splicing of new starting points, which makes up for the deficiency of existing systems in extracting rigid transmission topology paths. It maps and restores the rigid transmission topology path to the cascading service failure and breakage link in the supply chain service flow network, and solves the problems of delayed cascading failure early warning and inaccurate risk positioning.

[0018] 3. This invention receives business data through a parameter mapping module and maps the supply chain service flow network into a physical parameter matrix containing node inertia matrix, initial damping matrix, and directed network stiffness matrix. The boundary disturbance extraction module performs boundary recalculation and dynamically modifies the element values ​​in the lower and upper constraint vectors at integer multiples of the time step of calculus difference solution. This overcomes the shortcomings of traditional technology that uses a fixed threshold management method and achieves consistency correction between hard limit boundaries and actual business capabilities. Attached Figure Description

[0019] Figure 1 This is a block diagram of the module structure of the present invention; Figure 2 This is a structural diagram of the parameter mapping module of the present invention; Figure 3 This is a schematic diagram of the process of the present invention.

[0020] Among them, 100 is the data acquisition module; 200 is the parameter mapping module; 210 is the inertial calculation unit; 220 is the damping calculation unit; 230 is the stiffness calculation unit; 300 is the boundary disturbance extraction module; 400 is the state verification evolution module; 500 is the numerical integration calculation module; and 600 is the link tracing module. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example: Please see the appendix Figure 1 -Appendix Figure 3 This invention provides a dynamic monitoring system for supply chain service resilience based on multi-source data, including: a data acquisition module 100, a parameter mapping module 200, a boundary disturbance extraction module 300, a state verification evolution module 400, a numerical integration calculation module 500, and a link tracing module 600.

[0023] The data acquisition module 100 is used to access the underlying business database of each service node (including suppliers, manufacturers, warehousing centers and logistics nodes) in the supply chain service flow network to obtain static capacity parameters, dynamic business flow parameters and correlation parameters between nodes.

[0024] The parameter mapping module 200 receives the collected business data and maps the supply chain topology network into a physical parameter matrix. Specifically, it includes: the inertia calculation unit 210 establishing a node inertia matrix based on static capacity parameters; the damping calculation unit 220 establishing an initial damping matrix based on buffer inventory and delivery time margins; and the stiffness calculation unit 230 establishing a directed network stiffness matrix based on correlation parameters. It should be noted that although the system in this embodiment uses physical dynamic parameters (and physical indicators) for calculation, each of them has a unique corresponding meaning in the digital supply chain regarding service flow and performance quality.

[0025] Safety stock buffer and delivery time redundancy are mapped to the elastic service fulfillment window of nodes at the business level; absolute capacity limit and warehouse capacity limit are mapped to the maximum service response throughput capacity. When the system determines that the transient deviation of a node exceeds the hard limit boundary and forcibly sets the dynamic damping coefficient to zero, at the business level, it essentially indicates that the node has entered a denial-of-service or service paralysis state due to overload, resulting in service latency fluctuations generating dissipationless rigid penetration and being transferred to downstream. Therefore, the cascading failure link and failure alarms ultimately output by this system are essentially the risk links of non-performance of cascading services in the supply chain.

[0026] The boundary disturbance extraction module 300 extracts the extreme boundary (lower limit zero inventory threshold and upper limit of physical warehouse capacity) of the business network physical space to generate lower and upper limit constraint vectors; at the same time, it captures external abnormal disturbances and converts them into time-dependent excitation vectors through numerical normalization.

[0027] The state verification evolution module 400 receives the parameter matrix, constraint vector, and excitation vector, and performs calculus and differential solutions. When the transient deviation of the judgment node exceeds the lower or upper limit constraint vector, the coefficient value of the node corresponding to the initial damping matrix is ​​set to zero. This module uses the updated dynamic damping matrix to construct a second-order system dynamic description model and transmits it to the numerical integration calculation module 500 for iterative solution.

[0028] The link tracing module 600 extracts the sequence of nodes where the damping coefficient value is continuously set to zero, determines it to be a rigid transmission topology path, maps it back to the cascade service failure and broken link in the supply chain service flow network, and outputs a non-attenuated transmission supply chain service failure alarm signal.

[0029] The dynamic monitoring method also provided by this invention relies on the aforementioned system and its workflow includes: Step S1: The data acquisition module 100 synchronously extracts the data streams of each preset service node and establishes an initial system state vector. Step S2: The parameter mapping module 200 performs time alignment processing on the input data, extracts the absolute capacity limit parameter and the capital reserve limit parameter to construct the diagonal node inertia matrix, extracts the safety stock buffer amount and the delivery time redundancy amount to construct the initial damping matrix, and extracts the order binding rate and the reciprocal of the positive average supply cycle to construct the directed network stiffness matrix. Step S3: The boundary disturbance extraction module 300 obtains the zero inventory threshold and the upper limit of the physical storage capacity, and integrates them into a bilateral hard limit constraint vector; simultaneously, it converts the external business breakpoint signal into an excitation force vector applied to a specific node. Step S4: The state verification evolution module 400 performs state verification at discrete time points, compares the relative difference between the current state vector of the node and the bilateral hard limit constraint vector. When the difference exceeds the limit range, the diagonal elements of the damping matrix are modified according to the piecewise function to forcibly clear the buffer damping coefficient of the specific node. In step S5, the state verification evolution module 400 substitutes the real-time updated system dynamic matrix and excitation force vector into the second-order ordinary differential equation in the state space, calls the numerical solver to perform iterative calculations, and obtains the cascade diffusion response data. In step S6, the link tracing module 600 performs a graph network search based on the calculated high-order response data, marks the damped failure connected subgraph, extracts the transmission direction of nodes with non-dissipative attenuation of fluctuations, generates cascaded broken topology links, and sends a risk location alarm to the business monitoring terminal.

[0030] The data acquisition module 100 establishes a communication connection with the underlying business database of each preset service node in the supply chain entity network. Each preset service node includes a supplier node, a manufacturer node, a warehousing center node, and a logistics node. The data acquisition module 100 extracts multi-source heterogeneous business data according to a set polling cycle.

[0031] The business data extracted by the data acquisition module 100 includes static capacity parameters, dynamic business flow parameters, and node association parameters. Static capacity parameters include the absolute capacity limit of service nodes, the capital reserve limit (i.e., the total working capital reserve), the physical storage capacity limit, the maximum designed production capacity, the maximum single designed purchase batch, and the maximum advance payment amount of letters of credit. Dynamic business flow parameters include the current inventory of nodes, the amount of materials in transit, the safety stock buffer, the delivery time redundancy, the internal material self-consumption rate, the internal production cycle, the lower limit zero inventory threshold, and the working capital depletion point. Node association parameters include the forward and reverse order binding rate between upstream and downstream service nodes, the average supply cycle, the reverse fund settlement cycle, and the limit value of the order delay tolerance period stipulated in the business contract.

[0032] The data acquisition module 100 transmits the extracted business data to the parameter mapping module 200. The parameter mapping module 200 receives the business data and performs data cleaning operations, identifies and removes null fields and out-of-bounds fields in the data packets, and marks the time points corresponding to the removed fields as data missing points.

[0033] The parameter mapping module 200 performs time alignment processing on the cleaned business data. Since the data reporting frequency of each preset service node has a different sampling frequency, the parameter mapping module 200 uses a preset time resampling period to uniformly map the asynchronously acquired multi-source business data streams to the same timestamp sequence.

[0034] When performing time resampling operations, the parameter mapping module 200 uses a forward filling algorithm to fill in discrete business data with abrupt state changes, and a linear interpolation algorithm to fill in continuously numerical business data with smooth changes, eliminating data gaps caused by frequency differences and data cleaning. The set time resampling period is consistent with the time step of the subsequent state-space calculus equation solving stage.

[0035] The parameter mapping module 200 assigns a fixed-dimensional state identifier to each service node in the entity network based on the synchronization data aligned with the completion time. The parameter mapping module 200 combines the state identifiers to establish an initial system state vector with a fixed number of nodes. The initial system state vector represents the initial state baseline value of the supply chain topology network at the beginning of the evolution time, which is used as the underlying data source for the subsequent construction of the dynamic matrix.

[0036] The inertial calculation unit 210 inside the parameter mapping module 200 receives static capacity parameters from the data acquisition module 100.

[0037] The inertial computing unit 210 separates the absolute capacity limit parameter and the capital reserve limit parameter of each preset service node in the supply chain entity network from the static capacity parameter. The absolute capacity limit parameter represents the maximum physical limit of material throughput of the node per unit time, and the capital reserve limit parameter represents the maximum order advance amount of the node under the condition of no external capital injection.

[0038] For any node in the network topology graph The inertial computing unit 210 extracts the absolute capacity ceiling parameter. With upper limit parameter of funds reserves Inertial computing unit 210 for absolute capacity upper limit parameter With upper limit parameter of funds reserves Perform dimensionless normalization processing independently to obtain normalized capacity parameters. With normalized funding parameters .

[0039] The inertial calculation unit 210 uses preset weighting coefficients to normalize the production capacity parameters. With normalized funding parameters Perform linear weighted calculation to obtain the nodes. Equivalent inertia coefficient .

[0040] The formula for calculating the equivalent inertia coefficient is as follows: in, and These represent the weights allocated to the capacity dimension and the capital dimension, respectively, and satisfy the following conditions: The weight values ​​are set according to the actual business type of each node in the supply chain.

[0041] The inertial calculation unit 210 traverses all nodes in the supply chain entity network and calculates the corresponding set of equivalent inertial coefficients.

[0042] After obtaining the equivalent inertia coefficient set, the inertial calculation unit 210 constructs a dimension-based system according to the fixed arrangement sequence of all service nodes in the initial system state vector. Diagonalization of the matrix generates the nodal inertia matrix. .

[0043] Node inertia matrix The mathematical expression is as follows: in, This represents the total number of preset service nodes included in the network. This represents the diagonal matrix construction function and the nodal inertia matrix. All off-diagonal elements in the array are set to zero.

[0044] The nodal inertia matrix generated by the inertial calculation unit 210 In the state-space dynamic equations, the numerical elements on the diagonal are used to constrain the second-order rate of change of the corresponding nodes. The larger the equivalent inertia coefficient value of a node, the smaller the calculated nodal state acceleration response value when receiving an external excitation vector input of equal intensity.

[0045] The inertial calculation unit 210 will generate the nodal inertial matrix. Stored in the system memory for use by the state verification evolution module 400 when performing calculus difference solutions with a set time step.

[0046] The damping calculation unit 220 inside the parameter mapping module 200 receives the dynamic service flow parameters extracted by the data acquisition module 100.

[0047] The damping calculation unit 220 separates the safety stock buffer and delivery time redundancy of each preset service node in the supply chain entity network from the dynamic business flow parameters. The safety stock buffer represents the physical surplus of the node in response to material shortages, and the delivery time redundancy represents the time tolerance of the node in response to logistics delays.

[0048] For any node in the network topology graph Damping calculation unit 220 extracts safety stock buffer amount Redundancy with delivery time Damping calculation unit 220 for safety stock buffer Redundancy with delivery time Independently perform dimensionless normalization processing. Specifically, the damping calculation unit 220 uses the set maximum physical inventory capacity to buffer the safety stock. Execution ratio conversion, utilizing the set maximum tolerance for default time to account for delivery time redundancy. Perform proportional conversion to obtain the normalized inventory buffer parameter. With normalized time redundancy parameter .

[0049] Damping calculation unit 220 uses the set damping weight coefficient to normalize the inventory buffer parameter. With normalized time redundancy parameter Perform linear weighted calculation to obtain the nodes. initial damping coefficient .

[0050] The formula for calculating the initial damping coefficient is as follows: in, and These represent the weights assigned to the inventory dimension and the time dimension, respectively, and satisfy the following conditions: The rules for setting the weight parameters are as follows: when the preset service node is a warehouse center node, the inventory dimension is assigned weights. The value is greater than the weight assigned to the time dimension. The numerical value; when the preset service node is a logistics node, the weight is assigned to the time dimension. The value is greater than the weight assigned to the inventory dimension. The value.

[0051] The damping calculation unit 220 traverses all preset service nodes in the supply chain entity network and calculates the corresponding set of initial damping coefficients.

[0052] After obtaining the initial set of damping coefficients, the damping calculation unit 220 constructs a dimension-based system according to the fixed arrangement sequence of all preset service nodes in the initial system state vector. Diagonalization of the matrix generates the initial damping matrix. ; Initial damping matrix The mathematical expression is as follows: in, This represents the total number of preset service nodes included in the network. This represents the diagonal matrix construction function, and the initial damping matrix. All off-diagonal elements in the array are set to zero.

[0053] The initial damping matrix generated by the damping calculation unit 220 In the equation, the numerical elements on the diagonal are used to provide dissipation coefficients in the state difference calculation process in the state-space dynamic equation, and constrain the first-order rate of change of the state deviation of the corresponding node.

[0054] The damping calculation unit 220 will generate the initial damping matrix. It is stored in the system memory and used as a reference parameter by the state verification evolution module 400 when performing calculus difference solution with a set time step.

[0055] The stiffness calculation unit 230 inside the parameter mapping module 200 receives the node association parameters and dynamic business flow parameters extracted by the data acquisition module 100.

[0056] The stiffness calculation unit 230 separates the order binding rate and average delivery cycle between any two connected preset service nodes in the entity network from the node association parameters. The order binding rate represents the ratio of the material flow between connected nodes to the total throughput of the nodes.

[0057] For the starting node in the physical network where there is a business interaction relationship With the target node Stiffness calculation unit 230 extracts the starting node For the target node Positive order binding rate Stiffness calculation unit 230 synchronously extracts the starting node. To the target node Positive average delivery time of delivered materials .

[0058] Stiffness calculation unit 230 will forward average delivery cycle Compare with the set minimum time precision threshold; if the positive average delivery cycle... If the time precision is less than the minimum time precision threshold, the stiffness calculation unit 230 assigns the minimum time precision threshold to the positive average delivery cycle. .

[0059] Stiffness calculation unit 230 calculates the forward average delivery cycle. The reciprocal of the stiffness calculation unit 230 will be the positive order binding rate. Multiply by the reciprocal of the positive average delivery cycle to obtain the stiffness coefficient, which characterizes the strength of the unidirectional wave transmission effect. .

[0060] The formula for calculating the stiffness coefficient is as follows: in, Indicates the starting node State fluctuations cause target node Numerical parameter representing the intensity of the effect of state change.

[0061] Stiffness calculation unit 230 uses the same logic as described above to calculate the target node. Point to the starting node reverse stiffness coefficient reverse stiffness coefficient The computation depends on the target node. For the starting node Reverse order binding rate and reverse fund settlement cycle And simultaneously perform threshold constraint verification with extremely low time precision.

[0062] Due to the unidirectional transmission difference in the upstream and downstream supply relationships between pre-set service nodes, the positive order binding rate... Reverse order binding rate The values ​​differ, and the forward supply time of materials differs from the reverse settlement time of funds; therefore, the stiffness coefficients calculated by stiffness calculation unit 230 are different. With reverse stiffness coefficient They are not equal.

[0063] For the main diagonal elements of the matrix, the stiffness calculation unit 230 separates the target node from the dynamic business flow parameters. Internal material self-consumption rate With internal production cycle Stiffness calculation unit 230 calculates the internal production cycle. The reciprocal of the internal material self-consumption rate. With internal production cycle Perform multiplication on the reciprocal to obtain the target node. Internal nodal stiffness parameters Internal production cycle Synchronization is limited by the aforementioned extremely low time precision threshold.

[0064] The stiffness calculation unit 230 traverses all preset service node interaction paths and independent node internal parameters in the entity network topology diagram to calculate the set of stiffness coefficients for the entire network.

[0065] After obtaining the set of stiffness coefficients across the entire network, the stiffness calculation unit 230 assembles the stiffness coefficient set into a dimension-1 array according to the fixed arrangement order of all preset service nodes in the initial system state vector. Directed network stiffness matrix .

[0066] Directed network stiffness matrix The mathematical expression is as follows: in, This represents the total number of pre-defined service nodes included in the physical network. Directed network stiffness matrix For an asymmetric matrix, the corresponding matrix elements on the off-diagonal lines are... and For two pre-defined service nodes that are not numerically symmetric and have no direct business interaction, the corresponding cross stiffness coefficients are both assigned to zero. The matrix elements on the diagonal... These are the calculated internal node stiffness parameters.

[0067] The stiffness calculation unit 230 will generate the directed network stiffness matrix. Stored in the system storage medium, the directed network stiffness matrix When the state verification evolution module 400 performs differential calculation of the second-order ordinary differential equation with a set time step, it calls the core state matrix as the characterization of the risk network transmission path. As a result, the parameter mapping module 200 establishes a quantitative mapping relationship between the abstract service response time, service elasticity buffer and service contract coupling degree in the supply chain network and the physical matrix by introducing physical dynamic parameters.

[0068] The boundary disturbance extraction module 300 receives the external monitoring data stream transmitted by the data acquisition module 100. The boundary disturbance extraction module 300 performs a benchmark threshold comparison on the external monitoring data stream and extracts abnormal disturbance data that deviates from the normal business benchmark range.

[0069] The abnormal disturbance data comes from different business scenarios and includes numerical parameters with different physical dimensions. The abnormal disturbance data includes the number of days of logistics delay representing the time dimension, the quantity of raw material shortage representing the physical dimension, and the amount of overdue payment representing the financial dimension.

[0070] The boundary disturbance extraction module 300 performs dimensionless processing on the extracted abnormal disturbance data. For abnormal disturbance data of different business dimensions, the boundary disturbance extraction module 300 calls the corresponding extreme value reference standard preset in the system as the calculation denominator.

[0071] Specifically, when the abnormal disturbance data is the number of days of logistics delay, the boundary disturbance extraction module 300 extracts the limit value of the order delay tolerance time limit defined in the supply chain agreement as the denominator of the division operation. When the abnormal disturbance data is the quantity of raw material shortage, the boundary disturbance extraction module 300 extracts the maximum single design purchase batch of the corresponding node as the denominator of the division operation. When the abnormal disturbance data is the amount of overdue payment, the boundary disturbance extraction module 300 extracts the upper limit parameter of the node's capital reserve as the denominator of the division operation. The boundary disturbance extraction module 300 obtains the dimensionless excitation value through proportional conversion. The generated dimensionless excitation value is uniformly set as a positive scalar value to characterize the degree of deviation from the initial equilibrium state of the node.

[0072] The boundary disturbance extraction module 300 extracts the node numbers of the preset service nodes that generate abnormal disturbance data in the system topology network. Based on the fixed arrangement order of all preset service nodes in the initial system state vector, the boundary disturbance extraction module 300 fills the calculated dimensionless excitation values ​​into the corresponding coordinate dimensions of the set column vector. 3D coordinates and system topology network numbered The preset service nodes have a unique and definite mapping relationship.

[0073] For a preset service node that has not experienced any business anomalies at the current timestamp, the boundary disturbance extraction module 300 assigns a value of zero to the coordinate position corresponding to the column vector.

[0074] After the boundary disturbance extraction module 300 completes the traversal and coordinate assignment of all nodes, it generates a time-dependent excitation vector matching the dimension of the state space equation. .

[0075] Time-dependent excitation vector The mathematical expression is: in, This represents the total number of pre-defined service nodes included in the physical network. For the first Each node at time... Receive and convert the dimensionless excitation values.

[0076] The boundary disturbance extraction module 300 will generate a time-dependent excitation vector. Stored in the data buffer register, time-dependent excitation vector The state verification evolution module 400 is called as an external dynamic excitation term of the second-order nonlinear ordinary differential equation system, and serves as an input parameter for subsequent network state difference calculation.

[0077] The boundary disturbance extraction module 300 acquires the static capacity parameters and dynamic business flow parameters transmitted by the data acquisition module 100. Based on the functional type of the preset service node, the boundary disturbance extraction module 300 determines the business evolution limit of the node state offset. For the preset service node whose functional type is a warehouse center, the boundary disturbance extraction module 300 prioritizes extracting the maximum physical warehouse capacity as the upper limit original value; for the preset service node whose functional type is a manufacturer, the boundary disturbance extraction module 300 prioritizes extracting the maximum design capacity of the production line as the upper limit original value.

[0078] For any node in the physical network Boundary disturbance extraction module extracts 300 nodes Lower limit hard limit parameter Lower limit hard limit parameter To determine the threshold for identifying node service interruptions, the boundary disturbance extraction module 300 sets the lower limit zero inventory threshold, the working capital depletion point, and the limit values ​​of the order delay tolerance period as the hard limit parameters for calculating the lower limit. The original value.

[0079] Boundary disturbance extraction module 300 synchronous extraction nodes Upper limit hard limit parameter Upper limit hard limit parameter To determine the threshold for node resource saturation, the boundary disturbance extraction module 300 sets the upper limit of physical storage capacity, the maximum design capacity of the production line, and the maximum advance payment limit of the letter of credit in the static capacity parameters as hard limit parameters for calculation. The original value.

[0080] The boundary disturbance extraction module 300 calculates the difference between the original value and the initial state baseline value, performs a dimensionless transformation on the difference, obtains the limit value of the offset dimension, and makes the lower limit hard limit parameter. With upper limit hard limit parameter The mapping scale and the system state vector To maintain consistency, the boundary disturbance extraction module 300 defines an increase in service resources as a positive offset and a decrease in service resources as a negative offset. The boundary disturbance extraction module 300 arranges the lower limit hard limit parameters of all preset service nodes in order of node number to construct... dimensional lower bound constraint vector .

[0081] The boundary disturbance extraction module 300 arranges the upper limit hard limit parameters of all preset service nodes in order of node number, and constructs... Upper limit constraint vector Lower bound constraint vector With upper limit constraint vector Together they constitute the state-space constraint interval of the evolution of the supply chain topology network state space.

[0082] Because the operational capabilities of supply chain nodes are affected by cyclical fluctuations, the boundary disturbance extraction module 300 performs boundary recalculation at a preset update cycle. The preset update cycle is set to an integer multiple of the time step of the calculus difference solution. The boundary disturbance extraction module 300 dynamically modifies the lower limit constraint vector based on the contract adjustment data or capacity change data fed back by the data acquisition module 100. With upper limit constraint vector The element values ​​in the data are used to achieve consistency correction between hard limit boundaries and actual business capabilities.

[0083] The boundary perturbation extraction module 300 will generate the lower bound constraint vector. With upper limit constraint vector The aforementioned lower bound constraint vector is transmitted to the state verification evolution module 400. With the aforementioned upper limit constraint vector It serves as a criterion for determining the switching of nonlinear damping coefficients, and is used to constrain the state update of the damping matrix in each time step of the subsequent state equation.

[0084] The state verification evolution module 400 obtains the initial system state vector established by the parameter mapping module 200. The state verification evolution module 400 sets the initial system state vector as the zero-time reference value for iterative calculation, and sets the transient deviation of all preset service nodes to zero value at time zero.

[0085] At each discrete time step Internally, the state verification evolution module 400 reads the state vector from the system register at the previous moment. First-order rate of change vector and the second-order rate of change vector .

[0086] The state verification evolution module 400 uses the second-order Taylor discrete iterative operator to calculate the transient deviation at the current time step. The aforementioned discrete calculation formula is as follows: Among them, transient deviation Representing the Each preset service node at time The dimensionless numerical difference relative to the corresponding baseline value in the initial system state vector.

[0087] The state verification evolution module 400 linearly combines the transient deviations of all preset service nodes to construct the state vector at the current moment. State vector It reflects the transient offset state of all nodes in the supply chain topology network under the pressure of business fluctuations.

[0088] The state verification evolution module 400 will convert the state vector The input is a built-in boundary alignment unit, which will process the state vector. Each element in The lower bound constraint vector output by the boundary perturbation extraction module 300 and upper limit constraint vector Perform a numerical comparison on the corresponding elements in the table.

[0089] State verification evolution module 400 maintains state vector The update frequency is synchronized with the frequency of calculus difference solution. Within each discrete time step, the state verification evolution module 400 completes the solution cycle of state extraction, offset calculation and boundary comparison.

[0090] The comparison result output by the state verification evolution module 400 serves as the control signal for triggering subsequent nonlinear damped saturation triggering logic. The state verification evolution module 400 ensures the state vector... The numerical accuracy meets the switching requirements of the dynamic damping matrix in the subsequent dynamic equations.

[0091] The state verification evolution module 400 acquires the numerical comparison results output by the boundary comparison unit, and extracts the current discrete time step. Internal state vector Each transient deviation in .

[0092] The state verification evolution module 400 synchronously reads the initial damping matrix constructed by the parameter mapping module 200. Initial damping matrix The elements on the diagonal represent the initial damping coefficients of each pre-defined service node in the supply chain entity network. .

[0093] The State Verification Evolution Module 400 incorporates a piecewise nonlinear decision function, which is applied to the state vector. arbitrary transient deviation in Perform independent state range attribution determination.

[0094] State verification evolution module 400 determines transient deviation. Hard limit parameter greater than the lower limit And less than the upper limit hard limit parameter At that time, determine the first If the preset service nodes are in the normal dissipation range, the state verification evolution module 400 will... The dynamic damping coefficient of each preset service node at the current moment The corresponding coefficient values ​​in the initial damping matrix .

[0095] State verification evolution module 400 determines transient deviation. Less than or equal to the lower limit hard limit parameter At that time, or to determine the transient deviation. Greater than or equal to the upper limit hard limit parameter At that time, determine the first The preset service node is in the damped saturation breakdown range, and the state verification evolution module 400 will... The dynamic damping coefficient of each preset service node at the current moment By assigning a value of zero, and by forcibly setting the dynamic damping coefficient to zero when the transient deviation exceeds the hard limit boundary, this system can accurately simulate the nonlinear abrupt process of the complete failure of the physical buffer capacity of the supply chain at the computational level, significantly improving the real-world fit of the early warning model to cascading failure scenarios.

[0096] The mathematical expression for the piecewise nonlinear control logic executed by the state verification evolution module 400 is as follows: in, This represents the dynamic damping coefficient, which is subject to nonlinear constraints on state variables.

[0097] The State Verification Evolution Module 400 completes the verification of all entities in the network. The interval determination and coefficient assignment of each preset service node are performed. The state verification and evolution module 400 then combines the updated dynamic damping coefficients into matrix elements according to the fixed arrangement order of the preset service nodes in the initial system state vector to generate the dynamic damping matrix at the current time step. The state verification evolution module 400 will use the dynamic damping matrix. All off-diagonal elements in the array are assigned the value zero.

[0098] Dynamic damping matrix For dimension The diagonalized matrix, due to the introduction of a piecewise nonlinear decision function, the dynamic damping matrix The diagonal elements in the equation change discretely in value as the system state evolves.

[0099] The state verification evolution module 400 will generate a dynamic damping matrix. The dynamic damping matrix is ​​input as a time-varying dissipation parameter into the system-level second-order state-space ordinary differential equation. It replaces the constant damping term in the original differential equation, providing real-time matrix input for the subsequent numerical integration module 500 to perform equation order reduction.

[0100] The state verification evolution module 400 receives a set of dynamic damping coefficients within a discrete time step. The state verification evolution module 400 selects zero-value damping parameters with values ​​equal to zero from the set of dynamic damping coefficients and extracts the preset service node number corresponding to the aforementioned zero-value damping parameters.

[0101] The state verification evolution module 400 locates the corresponding diagonal position index in the dynamic damping matrix based on the preset service node number, and then performs a zeroing operation on the numerical elements of the corresponding diagonal position index.

[0102] For non-zero coefficients in the dynamic damping coefficient set whose values ​​are equal to the corresponding coefficient values ​​in the initial damping matrix, the state verification evolution module 400 maintains the diagonal values ​​of the non-zero coefficients in the dynamic damping matrix unchanged, and performs a discrete value replacement operation to generate the transiently zeroed updated dynamic damping matrix. .

[0103] State verification evolution module 400 limits the dynamic damping matrix after transient zeroing update. All off-diagonal elements in the matrix are zero, and the dynamic damping matrix is ​​updated after being transiently reset to zero. It includes the initial damping coefficient value and discrete numerical zeros.

[0104] The state verification evolution module 400 updates the dynamic damping matrix by setting the transient state to zero. In the second-order state-space ordinary differential equation of the input system, the state verification evolution module 400 establishes a dynamic damping matrix including the transient zeroing update. The system-level second-order state-space ordinary differential equation is expressed as follows: in, It is a second-order rate of change vector. It is a first-order rate of change vector. The stiffness matrix of the directed network. For state vectors, Let be the time-dependent excitation vector and the nodal inertia matrix. The parameter mapping module 200 pre-generates a diagonal matrix with constant coefficients based on the static capacity parameters of each preset service node.

[0105] Dynamic damping matrix after transient zeroing update Replace the initial damping matrix in the second-order state-space ordinary differential equation .

[0106] In the solution logic of numerical integration, the dynamic damping matrix is ​​updated after transient zeroing. The diagonal elements with a value of zero will clear the result of the first-order rate of change product of the corresponding preset service node, thereby stopping the calculation of the first-order derivative dissipation resistance of the transient deviation.

[0107] The undecayed values ​​of node state offsets are directly passed through the directed network stiffness matrix. It is transmitted to the downstream preset service node with an associated path in the physical topology network; at the business level, it represents the complete paralysis of the elastic fulfillment service of the node, causing the upstream service fluctuation to generate dissipationless rigid penetration, directly transferring and breaking through the service carrying capacity limit of the downstream node.

[0108] The state verification evolution module 400 will include the dynamic damping matrix updated after transient zeroing. The set of matrix parameters is transmitted to the numerical integration calculation module 500, which performs the order reduction processing and numerical iterative solution of the dynamic equation. The numerical integration calculation module 500 then performs state deviation data inference calculation for the next time step based on the set of matrix parameters.

[0109] The numerical integration module 500 receives the set of matrix parameters generated by the state verification evolution module 400 and establishes a system-level second-order state-space ordinary differential equation, which includes the nodal inertia matrix. The dynamic damping matrix after transient zeroing update Directed network stiffness matrix and time-dependent excitation vector .

[0110] The numerical integration module 500 performs matrix order reduction substitution to eliminate the second-order differential terms in the aforementioned system-level second-order state-space ordinary differential equations. The numerical integration module 500 also extracts the state vectors from the system-level second-order state-space ordinary differential equations. With the first-order rate of change vector .

[0111] Numerical integration module 500 constructs extended state vectors Extended state vector The dimension is From the state vector With the first-order rate of change vector Concatenate and expand the state vector The mathematical expression is .

[0112] Numerical integration module 500 pairs of extended state vectors Perform the differentiation operation to obtain the first-order extended derivative vector. First-order extended derivative vector From the first-order rate of change vector With the second-order rate of change vector Composed of spliced ​​elements, a first-order extended derivative vector The mathematical expression is .

[0113] Numerical integration module 500 extracts nodal inertia matrix Node inertia matrix The static fixed asset values ​​on the diagonal are all set to real numbers that are strictly greater than zero to ensure the node inertia matrix. It is a non-singular matrix that satisfies the necessary and sufficient mathematical condition for matrix inversion. The numerical integration module calculates the inertia matrix of 500 pairs of nodes. Perform matrix inversion to obtain the inertial inverse matrix. .

[0114] The numerical integration module 500 utilizes the inverse inertial matrix. The dynamic damping matrix after transient zeroing update and the directed network stiffness matrix Construct the state transition matrix State transition matrix The mathematical expression is: in, For dimension is The identity matrix, For dimension is The zero matrix.

[0115] The numerical integration module 500 utilizes the inverse inertial matrix. Construct the input matrix Input matrix The mathematical expression is: Among them, matrix elements For dimension is The zero matrix.

[0116] The numerical integration module 500 utilizes extended state vectors. First-order extended derivative vector State transition matrix Input matrix and time-dependent excitation vector By simultaneously solving the equations, we can generate the reduced first-order state-space equation. The mathematical expression of the reduced first-order state-space equation is as follows: The numerical integration calculation module 500 stores the reconstructed first-order state-space equation into the solution buffer. The first-order state-space equation serves as the basic solution model for the numerical integration calculation module 500 to perform discrete-time step integration derivation, and is used to calculate the transient deviation and first-order rate of change of each preset service node in the next time step.

[0117] The numerical integration module 500 retrieves the first-order state-space equations within the solution buffer and sets the discrete time step. and get the current time Extended state vector .

[0118] The numerical integration module 500 constructs the state derivative function corresponding to the first-order state-space equation. State derivative function The mathematical expression is ,in, The state derivative function, For the current moment, To expand the state vector, Here is the state transition matrix. For the input matrix, Let be the time-dependent excitation vector, in a single discrete time step. During the internal integration process, the numerical integration module 500 sets the state transition matrix. With input matrix The value remains constant and is not updated with the internal substep size.

[0119] The numerical integration calculation module 500 uses the fourth-order Runge-Kutta algorithm to perform numerical integration derivation calculations, so as to ensure that after the complex high-order system dynamic equations are reduced in order and reconstructed, the solution efficiency of long step-size calculations and the numerical stability of discrete difference integrals can be effectively balanced. The numerical integration calculation module 500 calculates the four derivative slope parameters within the current discrete time step based on the various system matrices at the current time.

[0120] For time-dependent excitation vectors generated by discrete service monitoring The numerical integration calculation module 500 utilizes the current time... With the next time step The excitation vector value is obtained by using a linear interpolation algorithm to calculate the half-step time. numerical excitation vector This is to meet the data retrieval requirements of multi-step slope evaluation.

[0121] Numerical integration module 500 calculates the first slope parameter. The first slope parameter The calculation formula is as follows: The numerical integration module 500 utilizes the first slope parameter Calculate the second slope parameter : The numerical integration module 500 utilizes the second slope parameter. Calculate the third slope parameter : The numerical integration module 500 utilizes the third slope parameter. Calculate the fourth slope parameter : The numerical integration module 500 performs a weighted summation on the four slope parameters obtained from the calculation to calculate the next time step. The corresponding updated extended state vector Update the extended state vector The calculation formula is as follows: in, To update the extended state vector, For discrete time steps, This is the first slope parameter. This is the second slope parameter. This is the third slope parameter. This is the fourth slope parameter.

[0122] Numerical integration module 500 updates extended state vector The execution dimension is decomposed, and the numerical integration calculation module 500 extracts and updates the extended state vector. The former The dimensional elements constitute the state vector for the next time step. The numerical integration module 500 extracts and updates the extended state vector. After The elements form the first-order rate of change vector for the next time step. ,in, The total number of service nodes and the state vector are preset for the entire network. It includes the transient deviation of all preset service nodes in the network after the solution is completed at the current time step.

[0123] The numerical integration calculation module 500 will calculate the state vector. With the first-order rate of change vector Synchronously transmitted to the state verification evolution module 400, state vector As the input criterion for the next discrete time step boundary comparison and damping parameter switching, it enables the continuous time step recursive calculation of the differential equation.

[0124] The link tracing module 600 acquires the state vector output by the numerical integration calculation module 500. and the dynamic damping matrix updated after transient zeroing output by the state verification evolution module 400. The link tracing module 600 synchronously acquires the state vector output by the numerical integration calculation module 500. .

[0125] The link tracing module 600 extracts the dynamic damping matrix after the transient zeroing update. The link tracing module 600 filters out the target diagonal position index with a value of zero from all diagonal numerical elements in the data, and then extracts the corresponding preset service node number based on the target diagonal position index. .

[0126] The link tracing module 600 marks the preset service node corresponding to the aforementioned target diagonal position index as a non-dissipation abnormal preset service node. The state deviation dissipation parameter of the non-dissipation abnormal preset service node is zero. The link tracing module 600 obtains the entity network topology diagram constructed by the parameter mapping module 200.

[0127] Based on the connection edge data in the entity network topology graph, the link tracing module 600 determines the network adjacency relationship between each non-dissipative anomaly preset service node. The link tracing module 600 calls the graph traversal connected component search algorithm to cluster the non-dissipative anomaly preset service nodes with connected paths into directed connected subgraphs for adjacent non-dissipative anomaly preset service nodes with direct directed connection edges, and divides each directed connected subgraph into an independent non-dissipative rigid conduction region.

[0128] The link tracing module 600 traverses all the non-dissipation anomaly preset service nodes in the system and completes the connectivity determination operation. The link tracing module 600 generates a set of rigid conduction regions within the current discrete time step. The set of rigid conduction regions contains at least one topological subgraph composed of non-dissipation anomaly preset service nodes.

[0129] Within a dissipationless rigid transmission region, the transmission of transient deviations does not involve damping attenuation calculations. The state deviation data is only constrained by the stiffness coefficients in the directed network stiffness matrix. The link tracing module 600 extracts all directed connection edges that constitute the topological subgraph within the set of rigid transmission regions.

[0130] The link tracing module 600 marks the aforementioned directed connection edges as abnormal rigid propagation topology paths. These abnormal rigid propagation topology paths characterize the topology transmission relationship where state deviation data undergoes a non-fading transfer to downstream nodes in the physical topology network. The link tracing module 600 stores the generated abnormal rigid propagation topology path data into the system's topology state register. Simultaneously, the link tracing module 600 also stores the state vector... The transient deviation value of the corresponding non-dissipative anomaly preset service node is associated and mapped with the aforementioned anomaly rigid propagation topology path and stored as the analysis benchmark for subsequent risk diffusion evolution calculation.

[0131] The link tracing module 600 reads the abnormal rigid conduction topology path data stored in the topology status register, and simultaneously obtains the transient deviation value of the non-dissipative abnormal preset service node associated with the aforementioned abnormal rigid conduction topology path.

[0132] Based on the entity network topology diagram, the link tracing module 600 takes the endpoint node of the aforementioned abnormal rigid transmission topology path as the starting point and performs a topology traversal search along the direction of the directed connection edge. The link tracing module 600 extracts the downstream preset service nodes that have a direct directed connection relationship with the aforementioned endpoint node and marks the aforementioned downstream preset service nodes as affected nodes.

[0133] The link tracing module 600 extracts the directed network stiffness matrix constructed by the parameter mapping module 200. The link tracing module 600 extracts the stiffness coefficients between the aforementioned endpoint node and the affected node from the aforementioned directed network stiffness matrix, and simultaneously extracts the internal node stiffness parameters of the affected node.

[0134] Because the aforementioned abnormal rigid transmission topology path has no damped dissipation, the link tracing module 600 will determine the transient deviation value of the non-dissipative abnormal preset service node. As the transmission reference value of the aforementioned endpoint node, the link tracing module 600 divides the aforementioned stiffness coefficient by the aforementioned internal node stiffness parameter to obtain the dimensionless state transfer ratio. The link tracing module 600 performs a multiplication operation between the aforementioned transmission reference value and the aforementioned state transfer ratio to obtain the transient deviation increment transmitted to the affected node. The transient deviation increment represents the state change amount of the service fluctuation value transmitted to downstream nodes along the physical network topology.

[0135] The link tracing module 600 performs a summation calculation on the state offset increment and the state vector value of the affected node at the current discrete time step to obtain the predicted state value of the affected node. The link tracing module 600 splices the abnormal rigid propagation topology path with the directed connection edge where the affected node is located to generate a cascading failure directed path. The aforementioned cascading failure directed path represents the continuous propagation topology sequence of business risks in the supply chain network.

[0136] The link tracing module 600 retrieves the lower bound constraint vector output by the boundary perturbation extraction module 300. With upper limit constraint vector The link tracing module 600 extracts the corresponding affected nodes. Lower bound constraint vector With upper limit constraint vector .

[0137] The link tracing module 600 compares the predicted state value of the affected node with the aforementioned lower hard limit parameter and the aforementioned upper hard limit parameter. When the link tracing module 600 determines that the predicted state value of the affected node is less than or equal to the aforementioned lower hard limit parameter, or determines that the predicted state value of the affected node is greater than or equal to the aforementioned upper hard limit parameter, the link tracing module 600 determines that the affected node is in a failure breakdown state.

[0138] For affected nodes in a failure and breakdown state, the link tracing module 600 updates the affected nodes in the failure and breakdown state to the new traversal starting point, and returns to perform the operation of extracting downstream preset service nodes along the direction of the directed connection edge, until there are no newly added failure and breakdown state nodes in the entity network topology graph, and ends the iterative splicing of the cascaded failure directed path.

[0139] The link tracing module 600 generates a cascading failure warning signal for all nodes on the cascading failure directed path that has completed iterative splicing. The link tracing module 600 outputs the cascading failure directed path and the associated cascading failure warning signal to the system display and control terminal. The cascading failure warning signal is used to trigger resource scheduling or business flow blocking control commands for the corresponding service nodes.

[0140] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic monitoring system for supply chain service resilience based on multi-source data, characterized in that, The data acquisition module is used to access the business databases of each preset service node in the supply chain service flow network and obtain business data, which includes static service carrying parameters, dynamic service flow parameters and service contract association parameters. The parameter mapping module receives the business data and maps the supply chain service flow network into a physical parameter matrix containing a node inertia matrix, an initial damping matrix, and a directed network stiffness matrix. It establishes the corresponding node inertia matrix, the initial damping matrix, and the directed network stiffness matrix based on the static capacity parameters, the dynamic business flow parameters, and the node association parameters, respectively. The boundary disturbance extraction module is used to extract the extreme boundary of the preset service node to generate the lower limit constraint vector and the upper limit constraint vector; at the same time, it captures external abnormal disturbances and converts them into time-dependent excitation vectors. The state verification evolution module receives the physical parameter matrix, the lower limit constraint vector, the upper limit constraint vector, and the time-dependent excitation vector, performs calculus and difference solutions, and when it is determined that the transient deviation of the preset service node exceeds the lower limit constraint vector or the upper limit constraint vector, the corresponding coefficient value in the initial damping matrix is ​​set to zero, and a second-order system dynamic description model is constructed using the updated dynamic damping matrix. The numerical integration module is used to receive and iteratively solve the second-order system dynamics description model; The link tracing module is used to extract the preset service node sequence corresponding to the occurrence coefficient value being zero from the dynamic damping matrix after it is updated to zero, determine it as a rigid transmission topology path, map the rigid transmission topology path back to the cascade service failure and broken link in the supply chain service flow network, and output a supply chain service failure alarm signal.

2. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 1, characterized in that, The static capacity parameters include the absolute capacity limit of the preset service node, the capital reserve limit, the physical storage capacity limit, the maximum designed production capacity, the maximum single design purchase batch, and the maximum advance payment amount of the letter of credit. The dynamic business flow parameters include the current inventory of the preset service node, the amount of materials in transit, the safety stock buffer, the delivery time redundancy, the internal material self-consumption rate, the internal production cycle, the lower limit zero inventory threshold, and the working capital depletion point. The node association parameters include the forward order binding rate and reverse order binding rate between the upstream and downstream preset service nodes, the forward average supply cycle and the reverse average supply cycle, the reverse fund settlement cycle, and the limit value of the order delay tolerance period agreed in the business contract.

3. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 2, characterized in that, The parameter mapping module specifically includes: An inertial calculation unit is used to independently perform dimensionless normalization processing on the absolute capacity limit parameter and the capital reserve limit parameter, perform linear weighted calculation using preset weight coefficients to obtain the equivalent inertial coefficient of the preset service node, and construct the diagonalized node inertial matrix according to the preset service node arrangement sequence. The damping calculation unit is used to perform dimensionless normalization processing on the safety stock buffer amount and the delivery time redundancy amount, and to perform linear weighted calculation using the set damping weight coefficient to obtain the initial damping coefficient of the preset service node, and to construct the diagonalized initial damping matrix. The stiffness calculation unit is used to perform a multiplication operation on the positive order binding rate and the reciprocal of the positive average delivery cycle which is limited by a very small time precision threshold to obtain the stiffness coefficient, and to obtain the internal node stiffness parameters based on the internal material self-consumption rate and the reciprocal of the internal production cycle, and to assemble and construct the asymmetric directed network stiffness matrix.

4. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 1, characterized in that, Before establishing the physical parameter matrix, the parameter mapping module performs cleaning and time alignment processing on the collected business data. It uses a preset time resampling period to uniformly map the asynchronously acquired multi-source business data streams to the same timestamp sequence. For discrete business data with abrupt state changes, a forward filling algorithm is used to fill in the gaps. For smoothly changing continuous numerical business data, a linear interpolation algorithm is used to fill in the gaps. The set time resampling period is consistent with the time step of the calculus difference solution stage.

5. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 2, characterized in that, The boundary disturbance extraction module extracts the lower limit zero inventory threshold, the working capital depletion point, and the order delay tolerance time limit as their original values, calculates the difference between these values ​​and the initial state benchmark value of the preset service node, and performs a dimensionless transformation to generate a lower limit hard limit parameter to construct the lower limit constraint vector. The same process is then applied to extract the upper limit of the warehouse physical capacity, the maximum design capacity of the production line, and the maximum advance payment amount of the letter of credit to generate upper limit hard limit parameters to construct the upper limit constraint vector. Simultaneously, for external abnormal disturbance quantities containing different physical dimensions, the corresponding extreme value reference standard is used as the denominator to perform proportional conversion, obtaining a dimensionless excitation value of the positive scalar value to construct the time-dependent excitation vector.

6. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 5, characterized in that, The state verification evolution module has a built-in piecewise nonlinear decision function, which executes the following logic within each discrete time step: Calculate the transient deviation of each preset service node and construct the state vector at the current moment; when it is determined that the transient deviation is greater than the lower hard limit parameter and less than the upper hard limit parameter, assign the dynamic damping coefficient of the corresponding preset service node at the current moment to the corresponding coefficient value in the initial damping matrix; When it is determined that the transient deviation is less than or equal to the lower hard limit parameter, or greater than or equal to the upper hard limit parameter, the dynamic damping coefficient of the corresponding preset service node at the current moment is forcibly assigned to a value of zero. The updated dynamic damping coefficients are used to generate the zeroed-updated dynamic damping matrix, and the set of matrix parameters containing this matrix is ​​transmitted to the numerical integration calculation module.

7. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 6, characterized in that, The second-order system dynamics description model includes a second-order rate of change vector, a first-order rate of change vector, the state vector, the directed network stiffness matrix, the nodal inertia matrix, the zeroed-updated dynamic damping matrix, and the time-dependent excitation vector. The diagonal elements with zero values ​​in the zeroed-updated dynamic damping matrix cause the calculation of the first derivative dissipative drag of the transient deviation to be stopped at the corresponding preset service node, allowing the undiminished value of the transient deviation to be directly transmitted to the downstream preset service node through the directed network stiffness matrix.

8. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 7, characterized in that, The numerical integration calculation module is specifically used for: A matrix reduction substitution is performed to eliminate the second-order differential term corresponding to the second-order rate of change vector in the second-order system dynamics description model. The state vector is extracted and concatenated with the first-order rate of change vector to form an extended state vector. A matrix inversion operation is performed on the node inertia matrix to obtain the inverse inertia matrix. This inverse matrix is ​​then used to simultaneously establish the zeroed-updated dynamic damping matrix and the directed network stiffness matrix to generate the reduced-order first-order state-space equation. A fourth-order Runge-Kutta algorithm is used to perform numerical integration derivation to obtain the updated extended state vector corresponding to the next time step.

9. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 3, characterized in that, The link tracing module is specifically used for: Filter out the non-dissipative anomaly preset service nodes corresponding to the target diagonal position index with a value of zero in the dynamically damped matrix after being zeroed and updated; The graph traversal connected component search algorithm is called to cluster adjacent non-dissipative anomaly preset service nodes with connected paths into directed connected subgraphs, and the directed connecting edges in the subgraphs are extracted and marked as anomaly rigid propagation topology paths. Starting from the endpoint of the abnormal rigid transmission topology path, a topology traversal search is performed. The state transmission ratio is calculated using the directed network stiffness matrix and the internal node stiffness parameters of the affected preset service node. The transient deviation increment transmitted to the affected preset service node is obtained and the predicted state value is calculated. The preset service nodes affected by the predicted state values ​​exceeding the lower or upper constraint vectors are used as new traversal starting points for iterative splicing until no new failure breakdown state nodes are added in the supply chain service flow network, thus generating the complete cascading service failure and breakage link.

10. The supply chain service resilience dynamic monitoring system based on multi-source data according to claim 5, characterized in that, The boundary disturbance extraction module also performs boundary recalculation at a preset update cycle, which is set to an integer multiple of the time step set for performing the calculus difference solution. The boundary disturbance extraction module dynamically modifies the element values ​​in the lower limit constraint vector and the upper limit constraint vector based on the contract adjustment data or capacity change data fed back by the data acquisition module as the business data, so as to achieve dynamic consistency correction between the hard limit boundary and the actual business capability.

Images