Distributed data monitoring method for supply chain intelligent collaborative management platform

Abstract

The present invention discloses a distributed data monitoring method for a supply chain intelligent collaborative management platform, which relates to the field of distributed data analysis technology and specifically includes the following steps: constructing a supply chain intelligent collaborative management platform; installing data monitoring equipment, collecting feature data of different intermediate links, arranging them into time series vectors, and respectively calculating the dynamic correlation of different intermediate links in the time dimension; constructing a dynamic association graph, performing causal testing on the basis of the constructed dynamic association graph, and forming a directed dynamic association graph; collecting feature data of different nodes in real time, predicting the impact of changes in feature data of nodes on other nodes based on the constructed directed dynamic association graph, and sorting high bottleneck risk nodes in real time; the present invention calculates dynamic correlation based on time series vectors and introduces time lag parameters to construct a directed dynamic association graph, accurately capturing the temporal dependence and causal relationship between each link, and making up for the defect of traditional monitoring in insufficient mining of link connections.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed data analysis, and in particular to a distributed data monitoring method for a supply chain intelligent collaborative management platform. Background Art

[0002] In today's burgeoning global economy, supply chain systems continue to expand in scale and become increasingly complex, encompassing numerous intermediary links, from raw material supply and product manufacturing to transportation and distribution, warehousing management, and sales and delivery. Each link is closely interconnected and mutually influential. Fluctuations in one link can quickly propagate throughout the supply chain, triggering a chain reaction that can impact the efficiency, cost, and stability of the entire supply chain.

[0003] Traditional supply chain data monitoring methods often employ a centralized architecture, with data transmitted to a central server for processing and analysis. However, this approach presents numerous drawbacks when dealing with large-scale distributed data. For one thing, the centralized transmission of large amounts of data can easily cause network congestion, leading to data transmission delays and impacting the real-time nature of information. Furthermore, a failure of the central server can paralyze the entire monitoring system, resulting in poor system reliability and fault tolerance. Furthermore, traditional monitoring methods often only provide simple monitoring of a single link or localized data, making it difficult to effectively capture the complex temporal dependencies and dynamic relationships between links, making it impossible to collaboratively manage the supply chain and provide early warning of risks from a holistic perspective.

[0004] With the continuous development and application of emerging technologies such as the Internet of Things, big data, and artificial intelligence, supply chain management is moving towards intelligent and collaborative processes. To achieve data interconnection and intelligent analysis across various supply chain links, thereby improving the supply chain's overall efficiency and risk resilience, an efficient and reliable distributed data monitoring method is urgently needed. This method must be able to collect characteristic data from each link in real time, deeply explore the dynamic correlations and causal relationships between data, and accurately predict the impact of changes in each link on the entire supply chain. This allows for the timely identification of high-risk bottlenecks and provides strong support for intelligent and collaborative supply chain management. Summary of the Invention

[0005] The purpose of the present invention is to provide a distributed data monitoring method for a supply chain intelligent collaborative management platform to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a distributed data monitoring method for a supply chain intelligent collaborative management platform, the distributed data monitoring method specifically comprising the following steps:

[0007] Step S100: Build a supply chain intelligent collaborative management platform to achieve data interconnection and intelligent analysis between various intermediate links.

[0008] Step S200: Install data monitoring equipment at all intermediate links of the supply chain intelligent collaborative management platform, wherein the data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time;

[0009] Step S300: Based on the collected time series vector, the dynamic correlation of different intermediate links in the time dimension is calculated respectively to reflect the temporal dependency between the links.

[0010] Step S400: construct a dynamic association graph with the intermediate links as nodes and the dynamic correlation of the different intermediate links in the time dimension as edge weights, perform causal testing on the basis of the constructed dynamic association graph, and form a directed dynamic association graph; form a graph structure data model for modeling the relationship between the intermediate links.

[0011] Step S500: collecting feature data of different nodes in real time, predicting the impact of feature data changes of nodes on other nodes based on the constructed directed dynamic association graph, and sorting high bottleneck risk nodes in real time;

[0012] In step S100, a supply chain intelligent collaborative management platform is constructed, specifically:

[0013] The supply chain intelligent collaborative management platform includes a basic layer, an application layer and a collaborative layer;

[0014] The basic layer includes warehouse local servers, logistics center edge gateways, supplier hosting servers, and supplier API gateways;

[0015] The application layer includes order processing microservices, in-vehicle IoT devices, logistics status and supplier contract management system APIs;

[0016] The collaborative layer includes a quality inspection system, a transportation plan management system and a collaborative control system.

[0017] Specifically:

[0018] The intermediate links include several supply ends, manufacturing ends, transportation ends, several storage ends, several sales ends and several delivery ends.

[0019] Specifically:

[0020] Among them, as preferred,

[0021] Warehouse local server: deployed at the storage end (such as a regional warehouse), collecting data such as temperature, humidity, and inventory levels at the storage end in real time;

[0022] Logistics center edge gateway: Installed at the transportation end, it collects data such as transportation routes, cargo status, and fuel consumption through on-board IoT devices (GPS, temperature and humidity sensors);

[0023] Supplier hosting server: connects to the supply side (raw material suppliers) and synchronizes data such as raw material batches, price fluctuations, and delivery cycles through the supplier API gateway;

[0024] Provide distributed data monitoring equipment;

[0025] The data monitoring device operates independently;

[0026] The data monitoring devices are collaboratively interconnected.

[0027] In step S200, data monitoring equipment is installed in all intermediate links of the supply chain intelligent collaborative management platform. The data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time, specifically:

[0028] Screening of characteristic data based on statistical analysis;

[0029] Preferably, the screening of characteristic data based on statistical analysis is specifically as follows:

[0030] Principal Component Analysis (PCA): Dimensionality reduction to extract the main components that explain supply chain fluctuations (supplier default, manufacturing machine anomalies, shipping timeouts, storage data update anomalies, etc.);

[0031] Mutual information: quantifies the nonlinear correlation between the main component data and the target (delivery delay), filters out low-correlation features, and screens out feature data;

[0032] The specific steps of arranging the characteristic data of the different intermediate links into time series vectors according to the collection time are as follows:

[0033] Step S211: Based on the filtered feature data, feature data of different intermediate links are collected at a certain frequency;

[0034] Step S212: perform data preprocessing; the data preprocessing includes data cleaning and data standardization;

[0035] Preferably, the data is normalized using the Z-score method;

[0036] The mean and standard deviation in the Z-score method can be the mean and standard deviation of a period of historical data;

[0037] Step S213: aligning data of different frequencies to the same time unit;

[0038] Step S214: Generate time series vectors of different intermediate links.

[0039] Specifically:

[0040] Set a timestamp; the timestamp is the time when the characteristic data is collected, and is used to align data of different frequencies to the same time unit.

[0041] In step S300, the dynamic correlation of different intermediate links in the time dimension is calculated based on the collected time series vector, specifically:

[0042] Step S301: real-time collection of time series vectors of different intermediate links;

[0043] Step S302: determining the window length according to the dimension of the time series vector, and calculating the mean of the feature data of different windows;

[0044] Step S303: Calculate the dynamic correlation of different intermediate links in the time dimension according to the determined window length and the windowed Pearson correlation coefficient method.

[0045] Preferably, the calculation formula for the dynamic correlation of different intermediate links in the time dimension is represented as follows:

[0046] ;

[0047] in, represents the dynamic correlation between intermediate link X and intermediate link Y; k represents the window index; W represents the window length; S represents the time step, that is, the same time unit; x t Represents the t-th feature data in the time series vector of the intermediate link X; Represents the t-th feature data in the time series vector of the intermediate link Y; Represents the mean of the characteristic data of the kth window in the time series vector of the intermediate link X; It represents the mean value of the feature data of the kth window in the time series vector of the intermediate link Y; t represents the quantity label of the feature data.

[0048] In step S400, a dynamic association graph is constructed with the intermediate links as nodes and the dynamic correlations of the different intermediate links in the time dimension as edge weights. A causal test is performed on the basis of the constructed dynamic association graph to form a directed dynamic association graph, specifically:

[0049] Step S401: construct a dynamic association graph by taking the intermediate links as nodes, the characteristic data transmission relationships between the intermediate links as edges, and the dynamic correlations of the different intermediate links in the time dimension as edge weights;

[0050] Step S402: Perform causal tests between different nodes in the dynamic association graph to determine the dominant direction between the nodes and optimize edge weights;

[0051] Step S403: Optimizing the constructed dynamic association graph into a directed dynamic association graph;

[0052] As a preferred method, a time lag parameter τ is introduced on the basis of the original dynamic correlation calculation to perform causal tests between different nodes, and the correlations in different directions are calculated respectively, specifically:

[0053] Positive correlation (XY) means predicting the time series vector of node Y through the historical time series vector before the time lag of node X:

[0054] ;

[0055] Reverse correlation (YX): This means predicting the time series vector of node X through the historical time series vector before the time lag of node Y:

[0056] ;

[0057] Where, τ represents the time lag parameter; Represents the directed dynamic dependency from intermediate link X to intermediate link Y; Indicates the directed dynamic dependency from intermediate link Y to intermediate link X; represents the mean of the characteristic data of the kth window in the historical time series before the intermediate link X time lag; represents the mean of the characteristic data of the kth window in the historical time series before the time lag of the intermediate link Y;

[0058] Among them, when τ=1, it means using the historical time series vector of X to predict the current time series vector of Y;

[0059] Determine the maximum time lag τmax based on the business scenario;

[0060] Calculate different and ;

[0061] The direction with the largest absolute value of the directed dynamic correlation is taken as the dominant direction, and the edge weight is updated to the directed dynamic correlation with the maximum absolute value;

[0062] The directed dynamic association graph is updated periodically.

[0063] In step S500, feature data of different nodes are collected in real time. Based on the constructed directed dynamic association graph, the impact of feature data changes of nodes on other nodes is predicted, and high bottleneck risk nodes are sorted in real time. The specific steps are as follows:

[0064] Step S501: Collect feature data of different nodes in real time to generate a time series vector;

[0065] Step S502: construct an influence propagation formula based on the change in the time series vector of the node in the constructed directed dynamic association graph and the edge weight, and predict the impact of the change in the node's feature data on the other nodes;

[0066] Preferably, the influence propagation formula is specifically:

[0067] ;

[0068] in, Represents the change in the time series vector of node X; Represents the impact propagation value of the time series vector on node Y; Represents the normalized value of the in-degree edge weight of node Y; the normalization method is Max-Min normalization method;

[0069] Step S503: Sort high bottleneck risk nodes in real time based on the current feature data of the prediction nodes and the predicted impact propagation value;

[0070] Preferably, the bottleneck risk prediction formula is:

[0071] ;

[0072] in, represents the bottleneck risk of node i; N represents the number of nodes; d represents the damping coefficient, which represents the probability of node influence transmission; j represents all nodes pointing to node i; represents the edge weight from node j to node i; represents the delay time of influence from node j to i; Represents the normalized value of the sum of all outgoing edge weights of node j; It represents the bottleneck risk of the initial node j, and the initial value is the average distribution basic risk 1 / N;

[0073] Calculate the bottleneck risk of each node separately, sort them in descending order, and check the nodes according to the sorting based on the actual situation.

[0074] Compared with the existing technology, the beneficial effects of the present invention are: the distributed data monitoring method of the supply chain intelligent collaborative management platform of the present invention, by installing independent and collaborative monitoring equipment in each intermediate link of the supply chain, uses statistical analysis methods to screen and preprocess feature data, effectively solves the drawbacks of traditional centralized architecture data transmission and processing, and improves data processing efficiency; based on time series vector calculation of dynamic correlation and introduction of time lag parameters to construct a directed dynamic association graph, accurately captures the temporal dependency and causal relationship between each link, and makes up for the defect of traditional monitoring in insufficient mining of link connections; based on the constructed graph model and related formulas, the impact of node changes is predicted in real time and high bottleneck risk nodes are sorted to achieve dynamic risk control, overcoming the problem that traditional monitoring is difficult to supervise globally; finally, through the graph structure data model, promotes collaboration among various links, optimizes resource allocation, enhances system fault tolerance and dynamic adaptability, effectively improves the overall operation efficiency, stability and response speed of the supply chain, and enhances the market competitiveness of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 This is a structural diagram of a distributed data monitoring method for a supply chain intelligent collaborative management platform of the present invention. DETAILED DESCRIPTION

[0076] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0077] Example: Figure 1 As shown, the present invention provides a technical solution, a distributed data monitoring method for a supply chain intelligent collaborative management platform, which specifically includes the following steps:

[0078] Step S100: Build a supply chain intelligent collaborative management platform to achieve data interconnection and intelligent analysis between various intermediate links.

[0079] In step S100, a supply chain intelligent collaborative management platform is constructed, specifically:

[0080] The supply chain intelligent collaborative management platform includes the basic layer, application layer and collaborative layer;

[0081] The basic layer includes warehouse local servers, logistics center edge gateways, supplier hosting servers, and supplier API gateways;

[0082] The application layer includes order processing microservices, in-vehicle IoT devices, logistics status, and supplier contract management system APIs;

[0083] The collaborative layer includes the quality inspection system, the transportation planning management system and the collaborative control system.

[0084] Specifically:

[0085] The intermediate links include several supply ends, manufacturing ends, transportation ends, several storage ends, several sales ends and several delivery ends.

[0086] Specifically:

[0087] Warehouse local server: deployed at the storage end (such as a regional warehouse), collecting data such as temperature, humidity, and inventory levels at the storage end in real time;

[0088] Logistics center edge gateway: Installed at the transportation end, it collects data such as transportation routes, cargo status, and fuel consumption through on-board IoT devices (GPS, temperature and humidity sensors);

[0089] Supplier hosting server: connects to the supply side (raw material suppliers) and synchronizes data such as raw material batches, price fluctuations, and delivery cycles through the supplier API gateway;

[0090] Provide distributed data monitoring equipment;

[0091] Data monitoring equipment is designed based on an embedded system architecture, integrating high-performance processors, large-capacity storage modules, multiple sensor interfaces, and communication modules, providing independent data collection, processing, and storage capabilities. For example, the edge gateway deployed at the logistics center at the transportation end is equipped with an ARM processor, enabling real-time processing of data such as transportation routes, cargo status, and fuel consumption collected by onboard IoT devices (GPS, temperature and humidity sensors, etc.). Built-in large-capacity flash memory can cache collected data for a certain period of time, allowing continuous data recording even during network outages, ensuring continuous data collection and enabling independent operation. Furthermore, each device is equipped with an independent power management system, supporting both battery-powered and external power supply modes, ensuring stable operation even in complex environments.

[0092] The data monitoring device operates independently;

[0093] The data monitoring devices are collaboratively interconnected.

[0094] Step S200: Install data monitoring equipment at all intermediate links of the supply chain intelligent collaborative management platform. The data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time.

[0095] In step S200, data monitoring equipment is installed in all intermediate links of the supply chain intelligent collaborative management platform. The data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time, specifically:

[0096] Screening of characteristic data based on statistical analysis;

[0097] Preferably, the screening of characteristic data based on statistical analysis is specifically as follows:

[0098] Principal Component Analysis (PCA): Dimensionality reduction to extract the main components that explain supply chain fluctuations (supplier default, manufacturing machine anomalies, shipping timeouts, storage data update anomalies, etc.);

[0099] Mutual information: quantifies the nonlinear correlation between the main component data and the target (delivery delay), filters out low-correlation features, and screens out feature data;

[0100] Principal component analysis (PCA) is used to reduce the dimensionality of data from each link, extracting the main component data that explain supply chain fluctuations, such as supplier delivery delays, supplier price fluctuations, manufacturing machine downtime, transportation timeouts, storage-end inventory data update delays, and abnormal fluctuations in sales-end order volume. Mutual information is then used to quantify the nonlinear correlation between these main component data and the delivery delay target, filtering out low-correlation features to obtain the filtered feature data.

[0101] Raw material related data: raw material batch number, used to trace the source and quality of raw materials; raw material component content, reflecting the quality of raw materials; raw material price fluctuation data, reflecting the trend of market price changes; raw material inventory quantity, to understand the supply side stocking situation.

[0102] Supplier delivery cycle, that is, the time from order placement to raw material delivery; the number of supplier defaults, such as delayed delivery, substandard delivery quality, etc.; the failure rate of supplier production equipment, which affects the stability of raw material supply.

[0103] Production process data: production plan completion rate, which measures the execution of production tasks; manufacturing machine operating status data, including startup time, shutdown times, and fault alarm information; product qualification rate, which reflects the production quality level; production line capacity utilization rate, which reflects the degree of production resource utilization.

[0104] Raw material input, statistics on the amount of raw materials used in the production process; parts loss rate, analysis of the consumption of parts in the production process; energy consumption data, such as the use of electricity, water, natural gas, etc.

[0105] Transportation status data: transportation route information, real-time recording of cargo transportation routes; transportation vehicle location (GPS coordinates), precise location of cargo; real-time temperature and humidity of cargo (for cargo with temperature and humidity requirements), ensuring a suitable transportation environment; transportation speed, reflecting transportation efficiency.

[0106] Inventory management data: inventory quantity, showing the current number of inventory goods; inventory turnover rate, measuring the turnover speed of inventory goods; inventory goods in and out records, including time, quantity, operator and other information; inventory goods shelf life warning data to prevent goods from expiration.

[0107] Sales business data: order quantity, reflecting market demand; order delivery time requirements, guiding supply chain production and distribution; customer return rate, analyzing product market acceptance.

[0108] Delivery execution data: delivery personnel location information for real-time customer inquiries; estimated delivery arrival time to enhance customer experience; goods receipt status to confirm delivery completion; and records of delivery anomalies.

[0109] The specific steps of arranging the characteristic data of different intermediate links into time series vectors according to the collection time are as follows:

[0110] Step S211: Based on the filtered feature data, feature data of different intermediate links are collected at a certain frequency;

[0111] Step S212: perform data preprocessing; data preprocessing includes data cleaning and data standardization;

[0112] Among them, the Z-score method was used for data standardization;

[0113] The mean and standard deviation in the Z-score method can be selected from the mean and standard deviation of a period of historical data;

[0114] Step S213: aligning data of different frequencies to the same time unit;

[0115] Step S214: Generate time series vectors of different intermediate links.

[0116] Specifically:

[0117] Set the timestamp; the timestamp records the time when the feature data was collected and is used to align data of different frequencies to the same time unit.

[0118] Step S300: Based on the collected time series vectors, the dynamic correlations of different intermediate links in the time dimension are calculated respectively, so as to reflect the temporal dependency between the links.

[0119] In step S300, the dynamic correlation of different intermediate links in the time dimension is calculated based on the collected time series vectors, specifically:

[0120] Step S301: real-time collection of time series vectors of different intermediate links;

[0121] Step S302: Determine the window length according to the dimension of the time series vector, and calculate the mean of the feature data of different windows;

[0122] Step S303: Calculate the dynamic correlation of different intermediate links in the time dimension according to the determined window length and the windowed Pearson correlation coefficient method.

[0123] Preferably, the calculation formula for the dynamic correlation of different intermediate links in the time dimension is represented as follows:

[0124] ;

[0125] in, represents the dynamic correlation between intermediate link X and intermediate link Y; k represents the window index; W represents the window length; S represents the time step, that is, the same time unit; x t Represents the t-th feature data in the time series vector of the intermediate link X; Represents the t-th feature data in the time series vector of the intermediate link Y; Represents the mean of the characteristic data of the kth window in the time series vector of the intermediate link X; It represents the mean value of the feature data of the kth window in the time series vector of the intermediate link Y; t represents the quantity label of the feature data.

[0126] Step S400: construct a dynamic association graph with the intermediate links as nodes and the dynamic correlation of different intermediate links in the time dimension as edge weights, perform causal testing on the basis of the constructed dynamic association graph, and form a directed dynamic association graph; form a graph structure data model for modeling the relationship between intermediate links.

[0127] In step S400, a dynamic association graph is constructed with intermediate links as nodes and the dynamic correlations of different intermediate links in the time dimension as edge weights. A causal test is performed on the basis of the constructed dynamic association graph to form a directed dynamic association graph, specifically:

[0128] Step S401: construct a dynamic association graph by taking intermediate links as nodes, the characteristic data transmission relationships between the intermediate links as edges, and the dynamic correlations of different intermediate links in the time dimension as edge weights;

[0129] Step S402: Perform causal tests between different nodes in the dynamic association graph to determine the dominant direction between the nodes and optimize edge weights;

[0130] Step S403: Optimizing the constructed dynamic association graph into a directed dynamic association graph;

[0131] On the basis of the original dynamic correlation calculation, the time lag parameter τ is introduced to perform causal tests between different nodes, and the correlations in different directions are calculated respectively, specifically:

[0132] Positive correlation (XY) means predicting the time series vector of node Y through the historical time series vector before the time lag of node X:

[0133] ;

[0134] Reverse correlation (YX): This means predicting the time series vector of node X through the historical time series vector before the time lag of node Y:

[0135] ;

[0136] Where, τ represents the time lag parameter; Represents the directed dynamic dependency from intermediate link X to intermediate link Y; Indicates the directed dynamic dependency from intermediate link Y to intermediate link X; represents the mean of the characteristic data of the kth window in the historical time series before the intermediate link X time lag; represents the mean of the characteristic data of the kth window in the historical time series before the time lag of the intermediate link Y;

[0137] Among them, when τ=1, it means using the historical time series vector of X to predict the current time series vector of Y;

[0138] Determine the maximum time lag τmax based on the business scenario;

[0139] Calculate different and ;

[0140] The direction with the largest absolute value of the directed dynamic correlation is taken as the dominant direction, and the edge weight is updated to the directed dynamic correlation with the maximum absolute value;

[0141] Regularly update the directed dynamic association graph.

[0142] Step S500: collecting feature data of different nodes in real time, predicting the impact of feature data changes of nodes on other nodes based on the constructed directed dynamic association graph, and sorting high bottleneck risk nodes in real time;

[0143] In step S500, feature data of different nodes are collected in real time. Based on the constructed directed dynamic association graph, the impact of feature data changes of nodes on other nodes is predicted, and high bottleneck risk nodes are sorted in real time. The specific steps are as follows:

[0144] Step S501: Collect feature data of different nodes in real time to generate a time series vector;

[0145] Step S502: construct an influence propagation formula based on the change in the time series vector of the node in the constructed directed dynamic association graph and the edge weight, and predict the impact of the change in the node's feature data on the other nodes;

[0146] As a preferred embodiment, the influence propagation formula is specifically:

[0147] ;

[0148] in, Represents the change in the time series vector of node X; Represents the impact propagation value of the time series vector on node Y; Represents the normalized value of the in-degree edge weight of node Y; the normalization method is Max-Min normalization method;

[0149] Step S503: Sort high bottleneck risk nodes in real time based on the current feature data of the prediction nodes and the predicted impact propagation value;

[0150] Preferably, the bottleneck risk prediction formula is:

[0151] ;

[0152] in, represents the bottleneck risk of node i; N represents the number of nodes; d represents the damping coefficient, which represents the probability of node influence transmission; j represents all nodes pointing to node i; represents the edge weight from node j to node i; represents the delay time of influence from node j to i; Represents the normalized value of the sum of all outgoing edge weights of node j; It represents the bottleneck risk of the initial node j, and the initial value is the average distribution basic risk 1 / N;

[0153] Calculate the bottleneck risk of each node separately, sort them in descending order, and check the nodes according to the sorting based on the actual situation.

[0154] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A distributed data monitoring method for a supply chain intelligent collaborative management platform, characterized by: The distributed data monitoring method specifically includes the following steps: Step S100: Building a supply chain intelligent collaborative management platform; Step S200: Install data monitoring equipment at all intermediate links of the supply chain intelligent collaborative management platform, wherein the data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time; Step S300: Based on the collected time series vector, the dynamic correlation of different intermediate links in the time dimension is calculated respectively, specifically: Step S301: real-time collection of time series vectors of different intermediate links; Step S302: determining the window length according to the dimension of the time series vector, and calculating the mean of the feature data of different windows; Step S303: Calculate the dynamic correlation of different intermediate links in the time dimension according to the determined window length and the windowed Pearson correlation coefficient method; The calculation formula of the dynamic correlation of different intermediate links in the time dimension is represented as follows: ； in, represents the dynamic correlation between intermediate link X and intermediate link Y; k represents the window index; W represents the window length; S represents the time step; Represents the t-th feature data in the time series vector of the intermediate link Y; Represents the mean of the characteristic data of the kth window in the time series vector of the intermediate link X; represents the mean value of the feature data of the kth window in the time series vector of the intermediate link Y; t represents the quantity label of the feature data; Step S400: construct a dynamic association graph with the intermediate links as nodes and the dynamic correlations of the different intermediate links in the time dimension as edge weights. Perform causal testing on the constructed dynamic association graph to form a directed dynamic association graph, specifically: Step S401: construct a dynamic association graph by taking the intermediate links as nodes, the characteristic data transmission relationships between the intermediate links as edges, and the dynamic correlations of the different intermediate links in the time dimension as edge weights; Step S402: Perform causal tests between different nodes in the dynamic association graph to determine the dominant direction between the nodes and optimize edge weights; Step S403: Optimizing the constructed dynamic association graph into a directed dynamic association graph; Regularly updating the directed dynamic association graph; Among them, based on the original dynamic correlation calculation, the time lag parameter τ is introduced to perform causal tests between different nodes, and the correlations in different directions are calculated respectively, specifically: Positive correlation X->Y means predicting the time series vector of node Y through the historical time series vector before the time lag of node X: ； The reverse correlation Y->X means predicting the time series vector of node X through the historical time series vector before the time lag of node Y: ； Where, τ represents the time lag parameter; Represents the directed dynamic dependency from intermediate link X to intermediate link Y; Indicates the directed dynamic dependency from intermediate link Y to intermediate link X; represents the mean of the characteristic data of the kth window in the historical time series before the intermediate link X time lag; represents the mean of the characteristic data of the kth window in the historical time series before the time lag of the intermediate link Y; Determine the maximum time lag τmax based on the business scenario; Calculate different and ; The direction with the largest absolute value of the directed dynamic correlation is taken as the dominant direction, and the edge weight is updated to the directed dynamic correlation with the maximum absolute value; Regularly updating the directed dynamic association graph; Step S500: collecting feature data of different nodes in real time, predicting the impact of feature data changes of nodes on other nodes based on the constructed directed dynamic association graph, and sorting high bottleneck risk nodes in real time; Step S501: Collect feature data of different nodes in real time to generate a time series vector; Step S502: construct an influence propagation formula based on the change in the time series vector of the node in the constructed directed dynamic association graph and the edge weight, and predict the impact of the change in the node's feature data on the other nodes; Step S503: Sort high bottleneck risk nodes in real time based on the current feature data of the prediction nodes and the predicted impact propagation value; The influence propagation formula is specifically: ； in, Represents the change in the time series vector of node X; Represents the impact propagation value of the time series vector on node Y; Represents the normalized value of the in-degree edge weight of node Y; the normalization method is Max-Min normalization method; The prediction formula of the bottleneck risk is specifically: ； in, represents the bottleneck risk of node i; N represents the number of nodes; d represents the damping coefficient, which represents the probability of node influence transmission; j represents all nodes pointing to node i; represents the edge weight from node j to node i; represents the delay time of influence from node j to i; Represents the normalized value of the sum of all outgoing edge weights of node j; It represents the bottleneck risk of the initial node j, and the initial value is the average distribution basic risk 1 / N; Calculate the bottleneck risk of each node separately, sort them in descending order, and check the nodes according to the sorting based on the actual situation.

2. The distributed data monitoring method for a supply chain intelligent collaborative management platform according to claim 1, characterized in that: In step S100, a supply chain intelligent collaborative management platform is constructed, specifically: The supply chain intelligent collaborative management platform includes a basic layer, an application layer and a collaborative layer; The basic layer includes warehouse local servers, logistics center edge gateways, supplier hosting servers, and supplier API gateways; The application layer includes order processing microservices, in-vehicle IoT devices, logistics status and supplier contract management system APIs; The collaborative layer includes a quality inspection system, a transportation plan management system and a collaborative control system.

3. The distributed data monitoring method for a supply chain intelligent collaborative management platform according to claim 1, characterized in that: Specifically: The intermediate links include several supply ends, manufacturing ends, transportation ends, several storage ends, several sales ends and several delivery ends.

4. The distributed data monitoring method for a supply chain intelligent collaborative management platform according to claim 3, characterized in that: Specifically: Provide distributed data monitoring equipment; The data monitoring device operates independently; The data monitoring devices are collaboratively interconnected.

5. The distributed data monitoring method for a supply chain intelligent collaborative management platform according to claim 4, characterized in that: In step S200, data monitoring equipment is installed in all intermediate links of the supply chain intelligent collaborative management platform. The data monitoring equipment is used to collect characteristic data of different intermediate links and arrange the characteristic data of different intermediate links into time series vectors according to the collection time, specifically: Screening of characteristic data based on statistical analysis; The specific steps of arranging the characteristic data of the different intermediate links into time series vectors according to the collection time are as follows: Step S211: Based on the filtered feature data, feature data of different intermediate links are collected at a certain frequency; Step S212: perform data preprocessing; the data preprocessing includes data cleaning and data standardization; Step S213: aligning data of different frequencies to the same time unit; Step S214: Generate time series vectors of different intermediate links.

6. The distributed data monitoring method for a supply chain intelligent collaborative management platform according to claim 5, characterized in that: Specifically: Set a timestamp; the timestamp is the time when the characteristic data is collected, and is used to align data of different frequencies to the same time unit.

Images