Risk intelligent early warning method and system for chemical raw material supply chain

Abstract

The application discloses a kind of chemical raw material supply chain risk intelligent early warning method and system, belong to supply chain data processing and risk early warning technical field.The method obtains procurement, transportation, inventory, warehousing, quality inspection and production plan consumption and other multi-source data, and establishes the space-time correlation with raw material batch as core after cleaning standardization;Combined with material attribute table to generate dangerous attribute constraint features, predict batch availability, available quantity and dynamic supplyable duration;Build a heterogeneous time-series supply chain graph with dangerous attribute gating and adjust the risk propagation weight, identify the risk source, propagation path, affected batch, storage location or tank, production device and the earliest impact time, generate risk warning information, improve the prediction and disposal of supply interruption.

Description

Technical Field

[0001] This invention belongs to the field of supply chain data processing and risk early warning technology, specifically relating to a method and system for intelligent risk early warning of the chemical raw material supply chain. Background Technology

[0002] Continuous chemical production enterprises typically require a continuous and stable supply of various chemical raw materials, auxiliary materials, catalysts, solvents, absorbents, packaging materials, or other production support materials during the production process. For critical chemical raw materials, supply delays, transportation anomalies, abnormal storage conditions, quality inspection lags, insufficient inventory, or fluctuations in supplier performance may lead to reduced production load, line shutdowns, or adjustments to production plans.

[0003] Existing supply chain risk warning methods typically collect data on suppliers, purchase orders, inventory, and logistics, and output risk alerts such as order delays, insufficient inventory, or logistics anomalies according to preset thresholds or rules. While some supply chain control towers, knowledge graphs, or complex network risk propagation methods can perform correlation analysis on suppliers, orders, transportation nodes, and inventory nodes, they usually focus on orders, suppliers, or transportation nodes as the main management objects, lacking a unified spatiotemporal correlation for the procurement, transportation, quality inspection, warehousing, storage tanks, and production consumption of chemical raw materials batches. This makes it difficult to track the status changes of specific raw material batches at each stage of the supply chain.

[0004] Compared to ordinary commodity supply chains, chemical raw material supply chains have stricter requirements for batch management, safe storage and transportation, and production continuity. Enterprise systems such as ERP, WMS, TMS, LIMS, and DCS typically record purchase orders, transportation routes, inventory batches, quality inspection releases, storage tank environments, and production requisition data separately. Existing methods struggle to uniformly convert released inventory, inventory awaiting inspection, restricted-use inventory, frozen inventory, estimated arrival quantities in transit, estimated quality inspection release times, production plan consumption rates, and minimum replenishment cycles into real, available inventory and dynamic supply duration for production units.

[0005] Meanwhile, existing supply chain risk propagation methods typically do not fully integrate the hazard categories, storage and transportation requirements, storage conditions, transportation qualifications, deviations from storage environments, incompatible combinations, quality inspection release status, and tank mixing factors recorded in the SDS of chemical raw materials. This makes it difficult to convert the aforementioned chemical safety constraints into batch availability constraints and risk propagation weights, resulting in early warning results that are difficult to form an interpretable propagation path from affected batches to affected storage locations or tanks to affected production units, and also make it difficult to accurately give the earliest impact time. Since procurement, transportation, quality inspection, warehousing, and production consumption data are stored in ERP, TMS, WMS, LIMS, and DCS systems respectively, existing methods can usually only output order delay, insufficient inventory, or logistics anomaly alerts, and cannot determine whether a batch in transit or awaiting inspection will affect a specific tank or production unit at a specific time.

[0006] Therefore, there is an urgent need for a risk intelligent early warning method and system for the chemical raw material supply chain, which can correlate procurement, transportation, quality inspection, warehousing and production consumption data at the batch level, and combine hazardous attribute constraints, dynamic supply availability duration and risk propagation weight to identify supply disruption risks, thereby improving risk positioning, early warning and targeted handling. Summary of the Invention

[0007] The purpose of this invention is to provide a risk intelligent early warning method and system for the chemical raw material supply chain, which solves the problems of existing supply chain early warning methods that cannot link purchase orders, transportation tracks, quality inspection release, storage tank environment and production unit consumption at the batch level of raw materials, and cannot use the hazard categories, storage and transportation requirements, storage and transportation qualifications, storage environment deviations and incompatible relationships recorded in the SDS as risk propagation weights to participate in the prediction of supply disruption risks, resulting in the early warning being unable to locate the affected batches, storage tanks and production units and the early warning lead time being insufficient.

[0008] To address the aforementioned technical problems, this invention provides a method for intelligent risk early warning in the chemical raw material supply chain, comprising the following steps: Acquire multi-source data of the target chemical raw material supply chain, including purchase order data, transportation trajectory data, inventory batch data, warehousing environment data, quality inspection release data, and production plan consumption data; The multi-source data is cleaned and standardized, and a batch-level spatiotemporal correlation is established based on the multi-source data and timestamps to generate batch-level supply chain status data with raw material batches as the core. A material attribute table is constructed for the target chemical raw material, and the storage environment and transportation trajectory data are matched with the material attribute table to generate the hazard attribute constraint characteristics of each batch of raw material. Based on the batch-level supply chain status data and the hazardous attribute constraint features, a batch-level supply chain status vector is generated. Based on the inventory batch data, transportation trajectory data, hazardous attribute constraint features and production plan consumption data, the availability status, predicted availability quantity and dynamic supply duration of the target chemical raw material for each raw material batch within the prediction window are determined. Based on the batch-level supply chain status data, hazard attribute constraint characteristics, and dynamic availability duration, a heterogeneous time-series supply chain graph is constructed. In the heterogeneous time-series supply chain graph, a hazard attribute gating factor is determined according to the hazard attribute constraint characteristics, and the hazard attribute gating factor is used to adjust the risk propagation weight of different edge types. Based on the adjusted risk propagation weights, the heterogeneous time-series supply chain diagram is processed for risk calculation, identifying risk sources, risk propagation paths, affected raw material batches, affected storage locations or tanks, affected production units, and the earliest impact time, and generating risk warning information to trigger supply chain risk management.

[0009] Furthermore, the cleaning and standardization process includes unifying material codes, supplier codes, storage location or tank numbers, and production unit numbers, as well as unifying time formats and units of measurement, and verifying, correcting, or marking missing, abnormal, or duplicate data; the batch-level spatiotemporal correlation is established based on order numbers, waybill numbers, supplier batch numbers, internal enterprise batch numbers, quality inspection numbers, storage location or tank numbers, production unit numbers, and timestamps.

[0010] Furthermore, when generating the aforementioned hazard attribute constraint features, the following are extracted from the material attribute table: hazard category, storage condition threshold, transportation qualification requirements, transportation vehicle requirements, prohibited material set, quality inspection release cycle, alternative materials, and process importance level of production equipment. The warehousing environment data, transportation trajectory data, quality inspection release data, and production plan consumption data corresponding to each batch of raw materials are matched with the extracted results to form the hazard attribute constraint features of the corresponding batch.

[0011] Furthermore, when determining the availability status and predicted availability quantity of each raw material batch, the batch status, estimated arrival time, estimated quality inspection release time, storage environment status, and hazardous attribute constraint characteristics are read according to the time step. When the corresponding time step is not earlier than the estimated arrival time and estimated quality inspection release time, and the storage conditions and hazardous attribute constraints are met, the batch quantity is included in the predicted availability quantity. For batches with restricted use, the smaller value between the inventory quantity and the upper limit of restricted use is included. For frozen, unqualified, or unmet constraints batches, they are recorded as zero.

[0012] Furthermore, when determining the dynamic available supply duration, the predicted available quantity of each batch of raw materials is summarized according to time step, and the cumulative planned consumption from the current time to the corresponding time step is deducted to obtain the predicted available inventory. When the predicted available inventory is lower than the minimum continuous operating demand of the production unit for the first time, the time interval between the time step and the current time is determined as the dynamic available supply duration. The dynamic available supply duration is compared with the shortest replenishment cycle to determine the risk of insufficient supply.

[0013] Furthermore, constructing the heterogeneous time-series supply chain graph includes establishing nodes for suppliers, purchase orders, transportation, raw material batches, quality inspection, storage locations or tanks, and production units; establishing timestamped associated edges according to supply, fulfillment, batch attribution, inspection, warehousing occupancy, and material supply relationships; writing the batch availability status into the raw material batch node, and writing the dynamic supply duration into the production unit node.

[0014] Furthermore, the hazard attribute gating factors are calculated according to the edge type by the corresponding hazard attribute constraint features. Among them, the associated edge from the transportation node to the raw material batch node adopts the abnormal features of transportation qualification, transportation vehicle and transportation route; the associated edge from the raw material batch to the quality inspection adopts the quality inspection release status and the remaining amount of the maximum allowable storage time; the associated edge from the raw material batch to the storage location or tank adopts the features of storage environment deviation and incompatible relationship; and the associated edge from the storage location or tank to the production unit adopts the process importance level of the production unit and the degree of supply tension.

[0015] Furthermore, the risk direction of the risk attribute constraint features of each associated edge is normalized, and a risk attribute gating factor with a value of 0 to 1 is generated based on the normalization result and the gating parameter corresponding to the edge type. When the risk attribute gating factor reaches the first threshold, the basic risk propagation weight is increased; when the risk attribute gating factor is not higher than the second threshold, the basic risk propagation weight is decreased; and in other cases, the basic risk propagation weight remains unchanged. The first threshold is greater than the second threshold.

[0016] Furthermore, the risk calculation process includes: calculating the cumulative risk value of candidate propagation paths along the time sequence of the heterogeneous time-series supply chain diagram; screening candidate propagation paths whose timestamps satisfy the sequential order of supplier, transportation, raw material batch, quality inspection, storage location or storage tank to production unit; when the cumulative risk value reaches the propagation threshold and the dynamic supply duration of the corresponding production unit is lower than the duration threshold, determining the candidate propagation path as the risk propagation path, determining the time step that first meets the condition as the earliest impact time, and writing the risk propagation path and the earliest impact time into the risk warning information.

[0017] This invention also provides a risk intelligent early warning system for the chemical raw material supply chain, including a data acquisition module, a data cleaning and standardization module, a batch-level correlation modeling module, a hazardous attribute feature generation module, a batch availability prediction module, a dynamic supply duration calculation module, a heterogeneous time-series supply chain diagram construction module, a risk propagation identification module, a risk early warning generation module, a disposal suggestion generation module, and a result display module. The data acquisition module is used to acquire data consumed in procurement, transportation, inventory, warehousing, quality inspection, and production planning. The data cleaning and standardization module is used to unify fields, units, and times, and to handle abnormal data; The batch-level association modeling module is used to establish batch-level spatiotemporal associations; The hazardous attribute feature generation module is used to generate hazardous attribute constraint features; The batch availability prediction module and the dynamic availability calculation module are used to determine the predicted available quantity and the dynamic availability duration. The heterogeneous time-series supply chain graph construction module and the risk propagation identification module are used to construct heterogeneous time-series supply chain graphs, adjust risk propagation weights, and identify risk propagation paths. The risk warning generation module is used to generate risk warning information; The disposal suggestion generation module is used to generate candidate disposal suggestions; The results display module is used to show the risk warning information and candidate treatment suggestions.

[0018] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. Based on order number, waybill number, supplier batch number, internal enterprise batch number, quality inspection number, storage location number or tank number, production device number and timestamp, this invention establishes a batch-level spatiotemporal correlation between procurement, transportation, quality inspection, warehousing and production consumption, enabling the supply chain status to be located to specific raw material batches, storage locations or tanks and production devices, thereby improving the accuracy and traceability of risk positioning in the chemical raw material supply chain; 2. This invention transforms the hazard categories, storage and transportation requirements, storage condition thresholds, transportation qualification requirements, and incompatible relationships recorded in the SDS into hazard attribute constraint features, and participates in batch availability judgment and related edge risk propagation weight adjustment. This allows the same transportation delay or quality inspection lag event to obtain different propagation weights under different hazard attributes and production continuity constraints, thereby improving the differentiated identification capability and interpretability of risk warning. 3. Based on batch status, estimated arrival time, estimated quality inspection release time, storage environment status, hazardous attribute constraints and production plan consumption data, this invention determines the available quantity and dynamic supply duration of each raw material batch within the prediction window. Compared with simply comparing the current inventory with the safety stock threshold, it can more accurately determine the ability of the target chemical raw material to maintain the continuous operation of the production unit in the future period, thereby identifying the risk of insufficient inventory and supply disruption in advance. 4. This invention constructs a heterogeneous time-series supply chain diagram based on batch-level supply chain status data, hazardous attribute constraint features, and dynamic supply availability duration. It also uses hazardous attribute gating factors to adjust risk propagation weights, which can identify risk sources, risk propagation paths, affected raw material batches, affected storage locations or tanks, affected production units, and the earliest time of impact, making the generated risk warning information more interpretable and targeted for action. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the intelligent risk early warning method for the chemical raw material supply chain of the present invention. Figure 2 This is a schematic diagram of the intelligent early warning system for risks in the chemical raw material supply chain of the present invention; Figure 3 This is a heterogeneous time-series supply chain diagram and risk propagation path diagram for the hazardous attribute gating of the present invention. Detailed Implementation

[0020] The technical solutions of 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.

[0021] In the description of this invention, the term "target chemical raw material" refers to the main raw material, auxiliary material, catalyst, solvent, absorbent, packaging material or other production support material used in the ethylene glycol production process; the target chemical raw material can be a single material or a combination of multiple materials; The term "batch-level supply chain status data" refers to data formed by linking purchase orders, shipping documents, supplier batch numbers, internal enterprise batch numbers, quality inspection status, storage location or tank, tank environment, and production consumption data, with raw material batches as the core. The term "dynamic availability duration" refers to the predicted duration for which a target chemical feedstock can sustain continuous operation of a production facility after taking into account released inventory, inventory awaiting inspection, restricted use inventory, frozen inventory, expected arrivals in transit, expected quality inspection release time, and production plan consumption rate. The term "hazard attribute gating factor" refers to a parameter used to adjust the risk propagation weight in a heterogeneous time-series supply chain diagram, generated based on factors such as the SDS hazard category of chemical raw materials, the degree of deviation of the storage environment, the degree of matching of transportation qualifications, the risk of incompatible relationships, the quality inspection release status, the feasibility of substitution, and the criticality of production.

[0022] The term "intelligent risk warning" refers to a computer-based data processing process that automatically generates risk warning information according to preset processing rules based on batch-level spatiotemporal correlation, hazard attribute constraints, dynamic availability duration, and risk propagation weight. It does not require training of neural network models, deep learning models, or machine learning models as a necessary step.

[0023] Example 1: This embodiment provides a method for intelligent early warning of risks in the chemical raw material supply chain, such as... Figure 1 As shown, the method includes the following steps: S1, Multi-source data acquisition Collect multi-source data from the target chemical raw material supply chain; the target chemical raw materials are key raw materials, auxiliary materials, catalysts, solvents, absorbents, packaging materials or other production support materials used by ethylene glycol producers in the production process.

[0024] Multi-source data includes purchase order data, supplier performance data, supplier equipment operation data, transportation trajectory data, inventory batch data, warehousing environment data, quality inspection and release data, production plan consumption data, market price data, and regional abnormal event data.

[0025] Purchase order data includes purchase order number, material code, material name, specification grade, purchase quantity, planned delivery time, supplier code, contract number, purchase price, and order status.

[0026] Supplier performance data includes the supplier's historical on-time delivery rate, historical quality pass rate, historical return rate, supplier capacity, supplier equipment operating status, supplier equipment maintenance plan, supplier location, and supplier access status.

[0027] The transportation trajectory data includes waybill number, carrier, vehicle number, tank container number, anonymized driver identifier, driver's professional qualification status, dangerous goods transportation qualification, planned route, actual GPS trajectory, current location, real-time speed, estimated arrival time, stop duration, deviation information, transportation temperature, transportation pressure, and abnormal transportation status. Driver-related data is only used to verify dangerous goods transportation qualifications and associated waybills, and is not used for driver personal credit evaluation, personnel disciplinary action, or any purpose unrelated to supply chain risk warnings; vehicle GPS trajectory data is used to determine transportation timeliness, route deviation, and the impact of regional anomalies.

[0028] Inventory batch data includes supplier batch number, internal enterprise batch number, warehousing time, inventory quantity, available quantity, frozen quantity, restricted usage quantity, storage location number, storage tank number, batch status, and batch expiration date.

[0029] Storage environment data includes tank liquid level, tank temperature, tank humidity, tank pressure, gas concentration, storage duration, tank capacity, tank maintenance status, and material information of adjacent storage locations.

[0030] Quality inspection release data includes sampling time, inspection items, inspection status, inspection results, estimated release time, actual release time, re-inspection status, reasons for non-compliance, and concession acceptance status.

[0031] Production planning data includes production unit number, production plan, shift plan, target chemical raw material consumption quota, real-time consumption rate, minimum continuous operating requirement of production unit, unit load, and planned maintenance time.

[0032] Market price data includes the target chemical raw material market price, price fluctuation range, regional price difference, and supply tightness index. In this embodiment, market price data serves as an external disturbance characteristic of supply tightness risk and is not used as the sole output of business decision-making results. The market price data, regional price difference, and supply tightness index are only used to adjust for the impact of supply tightness risk, abnormal supply performance risk, or regional abnormal events, and are not used to determine purchase transaction prices, supplier commercial rankings, transaction terms, or operating profit decisions.

[0033] Regional anomaly data includes extreme weather, road closures, port congestion, safety incidents, environmental production restrictions, public emergencies, or other events that may affect supply chain fulfillment in the supplier's location or transportation route.

[0034] In this embodiment, data on transport personnel, vehicle trajectory, supplier performance, and market prices are collected and processed according to the principle of minimum necessity. Transport personnel data is de-identified and used only to verify the qualifications of dangerous goods transport personnel for the corresponding waybill; vehicle trajectory data is used only to determine transport timeliness, route deviation, estimated arrival time, and the impact of regional anomalies; supplier performance and market price data serve only as characteristics of supply chain risk disturbances and are not used to automatically determine purchase transaction prices, supplier commercial rankings, or transaction terms.

[0035] S2, Multi-source data cleaning, standardization, and batch-level spatiotemporal correlation The collected multi-source data is cleaned and standardized.

[0036] Specifically, material names, material codes, supplier codes, tank numbers, and production unit numbers from different systems are converted into unified codes; time fields from different sources are converted into a unified time format; different units of measurement are converted into a unified unit of measurement; missing, abnormal, and duplicate fields are validated, corrected, or marked; and data credibility weights are set for conflicting data sources.

[0037] In one alternative implementation, the data reliability weight is determined according to the real-time nature, accuracy, and business priority of the data source; for example, real-time location data in a transportation GPS system has a higher priority than manually entered transportation status data; quality inspection release results in a LIMS system have a higher priority than expected quality inspection status in a procurement system; and tank level data in a DCS or WMS system has a higher priority than manually counted data.

[0038] After cleaning and standardization are completed, a batch-level spatiotemporal relationship is established based on the order number, waybill number, supplier batch number, internal enterprise batch number, warehouse entry number, quality inspection number, storage location number, storage tank number, production unit number, and timestamp.

[0039] The batch-level spatiotemporal correlations include: the correlation between purchase orders and supplier batches; the correlation between supplier batches and transport orders; the correlation between transport orders and transport tracks; the correlation between transport orders and batches in transit; the correlation between batches in transit and estimated arrival time; the correlation between batches arriving at the factory and quality inspection reports; the correlation between quality inspection release batches and storage locations or tanks; the correlation between storage locations or tanks and production requisition records; and the consumption correlation between raw material batches and production units.

[0040] Through the above association, the system generates batch-level supply chain status data with raw material batches as the core; this batch-level supply chain status data can reflect the entire process status of a target chemical raw material from purchase order, supplier preparation, transportation, arrival at the factory, quality inspection, warehousing, storage to production use.

[0041] S3. Construct the material property table and generate hazardous attribute constraint features. For each target chemical raw material, construct a material property table.

[0042] The material attribute table includes material code, material name, specification grade, SDS hazard category, storage temperature threshold, storage humidity threshold, storage pressure threshold, incompatible material set, compatible material set, transportation qualification requirements, carrier vehicle or tank container requirements, quality inspection item set, quality inspection release cycle, maximum allowable storage time, alternative material set, and process importance level to the ethylene glycol production unit.

[0043] Based on the material property table, the hazardous property constraints of each batch of target chemical raw materials are generated.

[0044] Hazardous attribute constraints include hazard level characteristics, storage condition deviation characteristics, transportation qualification matching characteristics, transportation vehicle matching characteristics, incompatible relationship risk characteristics, quality inspection and release cycle characteristics, alternative feasibility characteristics, and production criticality characteristics.

[0045] Among them, the storage condition deviation feature is determined based on the deviation between the actual storage environment parameters and the storage condition threshold in the material attribute table; for example, if the actual storage temperature of a certain batch in the storage tank exceeds the upper limit of the storage temperature in the material attribute table, the system generates a temperature deviation feature and determines the degree of storage abnormality of the batch based on the duration and magnitude of the deviation.

[0046] The transportation qualification matching feature is determined based on the degree of matching between the carrier's qualifications, vehicle qualifications, tank container requirements, and the transportation qualification requirements of the target chemical raw materials. If the target chemical raw materials require dangerous goods transportation qualifications, but the qualifications of the carrier's vehicles or tank containers do not meet the requirements, the transportation qualification matching feature is marked as mismatch or low match.

[0047] The risk characteristics of the incompatible material set of the target chemical raw material are determined based on the relationship between the incompatible material set of the target chemical raw material and the materials in the actual adjacent storage locations, materials transported in the same vehicle, or materials in the same tank area; if there is an incompatible relationship between the target chemical raw material and the materials in the adjacent storage locations, then the incompatible material risk characteristics are generated.

[0048] The feasibility characteristics of substitution are determined based on the set of substitute materials for the target chemical raw material, the specifications and grades of the substitute materials, their quality inspection status, available inventory, storage and transportation conditions, and compatibility with the production equipment.

[0049] S4. Generate batch-level supply chain state vectors For each batch of each target chemical raw material, a batch-level supply chain state vector is generated based on batch-level supply chain state data and hazardous attribute constraint characteristics.

[0050] The batch-level supply chain status vector includes batch identifier, corresponding purchase order, corresponding supplier, corresponding waybill, current stage, planned arrival time, estimated arrival time, actual arrival time, current inventory quantity, released quantity, quantity awaiting inspection, frozen quantity, restricted usage quantity, degree of deviation of warehousing environment, quality inspection release status, degree of deviation of transportation trajectory, supplier performance stability, supplier equipment operation status, production plan occupancy, estimated availability time and batch availability index.

[0051] The current stage includes: ordered, supplier prepared, shipped, in transit, awaiting inspection at the factory, under quality inspection, released, put into storage, restricted use, frozen, used in production, and closed due to abnormality.

[0052] Batch availability index is used to characterize the probability that a batch can be actually used by the production unit within the forecast window. The batch availability index is determined according to preset rules based on batch status, quality inspection status, storage environment status, transportation status, and hazardous attribute constraints.

[0053] In one optional implementation, the batch availability index is determined according to the following preset rules: batches that have been released and are stored in a normal environment correspond to a first availability coefficient; batches that have arrived at the factory for inspection or are undergoing quality inspection correspond to a second availability coefficient based on the expected quality inspection release time; batches in transit correspond to a third availability coefficient based on the expected arrival time, abnormal transportation status, and expected quality inspection release time; batches with restricted use correspond to a fourth availability coefficient based on the degree of deviation from the storage environment and the re-inspection status; and frozen batches, unqualified batches, or batches whose hazardous attribute constraints are not met correspond to a zero availability coefficient.

[0054] Among them, the first available coefficient is greater than the second, third, and fourth available coefficients.

[0055] S5, Batch Availability Prediction and Dynamic Availability Calculation The system predicts the availability status and quantity of each batch within the prediction window based on inventory batch data, in-transit transportation data, quality inspection and release data, warehousing environment data, and production plan consumption data.

[0056] Batch availability status includes released and available, arrived at the factory awaiting inspection, under quality inspection, in transit and expected to be available, restricted use, frozen and unavailable, unqualified and unavailable, delayed arrival, and warehouse abnormality pending confirmation.

[0057] For the target chemical raw material, the system calculates the predicted available inventory at each time point within the prediction window according to the time series.

[0058] In one alternative implementation, for the target chemical raw material m, the system predicts the reaction time step by step within the prediction window. Calculate and forecast available inventory : ;in, For batch b at time step The predicted available quantity, From the current moment to the time step The cumulative planned consumption of the target chemical raw material m by the production unit during the period.

[0059] Predicted available quantity of batch b Based on batch inventory quantity Limited usage limit The quality inspection status, storage environment status, incompatible delivery relationships, transportation qualification status, and time availability status are determined. Specifically, batches that have undergone quality inspection and release, whose storage environment meets requirements, and whose hazardous attribute constraints are satisfied... Take the current available inventory quantity; restricted usage batch, Get the current inventory quantity and the restricted usage limit. The smaller of the values; batches pending confirmation, only... Batches arriving no earlier than the estimated arrival time and estimated quality inspection release time, and for which the storage environment meets the requirements, are included in the predicted available quantity; batches that are frozen, unqualified, fail re-inspection, do not meet storage conditions, do not meet transportation qualifications, or have prohibited combinations are not included. Take 0.

[0060] when When the demand first falls below the minimum continuous operating requirement of the production unit, the interval between that time step and the current time is determined as the dynamic available supply duration; if within the forecast window... If the demand is not lower than the minimum continuous operating demand of the production unit, the dynamic supply duration will be marked as greater than the forecast window length.

[0061] Furthermore, the system determines the shortest replenishment cycle for the target chemical raw materials based on the qualified supplier's procurement cycle, supplier's preparation cycle, transportation cycle, arrival and weighing time, quality inspection and release cycle, and warehousing time.

[0062] When the dynamic available supply time is less than the shortest replenishment cycle, the system determines that the target chemical raw material is at risk of insufficient inventory or supply disruption.

[0063] When the dynamic available supply period is greater than or equal to the shortest replenishment period but less than the preset safety buffer period, the system determines that there is a potential risk of insufficient supply of the target chemical raw material.

[0064] In this embodiment, instead of simply comparing the current inventory level with the safety stock threshold, the system combines batch status, in-transit status, quality inspection status, warehousing status, and production consumption status to obtain a supply duration that is closer to the actual production situation.

[0065] S6. Construct a heterogeneous time-series supply chain diagram with hazardous attribute gating. like Figure 3 As shown, a heterogeneous time-series supply chain diagram is constructed based on batch-level supply chain status data, hazardous attribute constraints, and dynamic availability duration.

[0066] The nodes in the heterogeneous time-series supply chain diagram include supplier nodes, supplier production equipment nodes, purchase order nodes, transportation nodes, raw material batch nodes, quality inspection nodes, storage location or tank nodes, warehousing environment nodes, production equipment nodes, market price nodes, and regional abnormal event nodes.

[0067] The edges in a heterogeneous temporal supply chain graph include: supply relationship edges between supplier nodes and purchase order nodes; performance impact edges between supplier production unit nodes and purchase order nodes; performance relationship edges between purchase order nodes and transportation nodes; path relationship edges between transportation nodes and transportation trajectories; batch attribution relationship edges between transportation nodes and raw material batch nodes; inspection relationship edges between raw material batch nodes and quality inspection nodes; warehousing occupation relationship edges between raw material batch nodes and storage location or tank nodes; material supply relationship edges between storage location or tank nodes and production unit nodes; consumption relationship edges between production plans and raw material batch nodes; impact relationship edges between regional abnormal event nodes and supplier nodes, transportation nodes, or warehousing nodes; and supply tension relationship edges between market price nodes and supplier nodes or purchase order nodes.

[0068] In a heterogeneous time-series supply chain diagram, node characteristics include node type, node status, timestamp, corresponding material code, corresponding batch status, hazardous attribute constraint characteristics, dynamic available supply duration, and historical risk labels.

[0069] Edge characteristics include edge type, association strength, time interval, performance deviation, path deviation, warehousing environment deviation, quality inspection release status, production consumption intensity, and hazard attribute gating factor.

[0070] The system uses the raw material batch availability status obtained from S5 as the status feature of the raw material batch node, and the dynamic supply duration as the status feature of the production unit node. The raw material batch availability status and dynamic supply duration are used to reflect the actual impact of hazardous attribute constraints, quality inspection release status, storage environment status, and production plan consumption on the continuous operation of the production unit.

[0071] In this embodiment, the system sets hazard attribute gating factors for the edges in the heterogeneous time-series supply chain graph. The hazard attribute gating factors are determined based on the SDS hazard category, the degree of deviation from the storage environment, the degree of matching of transportation qualifications, the degree of matching of carrier vehicles or tank containers, the risk of incompatible relationships, the quality inspection release status, the maximum allowable storage time, the feasibility of substitution, and the production criticality.

[0072] Hazard attribute gating factors are used to adjust the weight of risk propagation between nodes in the supply chain. For example, when the target chemical raw material has a high hazard level, low matching degree of transportation qualifications, and regional abnormal events occur along the transportation route, the risk weight propagated from the transportation node to the raw material batch node and the production unit node is increased. As another example, when the storage environment of a certain inventory batch continuously deviates from SDS storage conditions, and this batch is a batch recently planned for use by the production unit, the risk weight propagated from this inventory batch node to the production unit node is increased.

[0073] In one alternative implementation, a basic risk propagation weight is assigned to each associated edge: transport to batch, batch to quality inspection, batch to storage location or tank, and storage location or tank to production unit. The system is based on edge type. Select the corresponding hazard attribute constraint features, and normalize each feature into a numerical feature according to the risk direction, where the larger the value, the higher the corresponding risk.

[0074] For the links from transportation to batch, at least the degree of mismatch in transportation qualifications, the degree of mismatch in transportation tools, and the degree of abnormality in transportation routes should be used; for the links from batch to quality inspection, at least the quality inspection release status and the remaining maximum allowable storage time should be used; for the links from batch to storage location or tank, at least the degree of deviation from the storage environment and the risk of incompatible relationships should be used; for the links from storage location or tank to production unit, at least the degree of supply tightness corresponding to the process importance level of the production unit and the dynamic available supply time should be used.

[0075] The system calculates the hazard attribute gating factor based on the numerical characteristics corresponding to the edge type and the preset weights. Specifically, for the time step... The associated edges below ,in and These are the nodes at both ends of the associated edge. As the edge type, the system will classify the edge type. The corresponding hazardous attribute constraint features are normalized into gating feature vectors. , among which The value ranges from 0 to 1, with a larger value indicating a higher risk.

[0076] Among them, discrete state characteristics such as mismatched transportation qualifications, mismatched transportation tools, prohibited delivery relationships, and failure to release after quality inspection are normalized according to whether the corresponding constraints are met; continuous characteristics such as deviation of warehousing environment, abnormal transportation route, remaining maximum allowable storage time, and supply tightness corresponding to dynamic supply duration are normalized according to deviation magnitude, remaining time, or supply tightness.

[0077] Hazardous attribute gating factor Calculate using the following formula: ;in, edge type The corresponding basic gating parameters, edge type The gating weight corresponding to the i-th hazardous attribute constraint feature, and All parameters are between 0 and 1, and No greater than 1.

[0078] For associated edges Basic risk transmission weight The system uses a hazard attribute gating factor. Determine the adjusted risk propagation weights : when hour, ; when hour, ; when hour, ; in, edge type The corresponding first threshold, edge type The corresponding second threshold, Greater than ; This is the weighting factor. For weighting coefficients, and All parameters are between 0 and 1.

[0079] Therefore, the same event of transportation delay, quality inspection lag, or insufficient inventory has a higher risk propagation weight when there is a mismatch in transportation qualifications, a deviation in the warehousing environment, the existence of a non-distribution relationship, or a high demand for continuous operation of production facilities; and a lower risk propagation weight when alternatives are feasible, the warehousing environment meets the requirements, and the impact on production is low.

[0080] S7, Risk Calculation and Processing The system processes batch-level supply chain status data, hazardous attribute constraints, dynamic supply duration, and node and edge features of heterogeneous time-series supply chain graphs according to preset risk calculation rules, and outputs the risk identification results of the target chemical raw materials.

[0081] The risk propagation identification module sequentially performs batch availability judgment, transportation timeliness anomaly identification, warehousing environment anomaly identification, supplier performance stability judgment, dynamic supply duration judgment, risk propagation path identification, and handling suggestion constraint matching according to preset risk calculation rules, and outputs risk type, risk level, risk source, affected batch, affected storage location or tank, affected production unit, earliest impact time, and risk propagation path.

[0082] Specifically, batch availability is determined based on batch status, quality inspection status, storage environment status, transportation status, and hazardous attribute constraints; transportation timeliness anomaly identification is determined based on planned route, actual GPS trajectory, parking duration, deviation information, regional anomalies, and estimated arrival time; storage environment anomaly identification is determined based on tank temperature, humidity, pressure, liquid level, gas concentration, storage duration, and SDS storage conditions; and risk propagation path identification is based on the risk propagation weight adjusted according to the hazardous attribute gating factors corresponding to each time step and edge type. Proceed. For time steps nodes The system is based on the pointed-to node Calculate the propagation risk value of nodes in the preceding associated edges. : ;in, For nodes The predecessor node, For the reason point to Associated edges, This refers to the time interval or processing delay corresponding to the associated edge. Preorder node The propagation risk value at the corresponding time step. For time step Lower related edge Adjusted risk transmission weights.

[0083] For candidate paths formed from the risk source node through transportation nodes, raw material batch nodes, quality inspection nodes, storage location or tank nodes to the production unit node. The system calculates the cumulative risk value of a path according to the timestamps of each associated edge on the path. : .

[0084] When candidate path The timestamps on the data satisfy the sequential order of supplier, transportation, quality inspection, storage location or tank to production unit, and the cumulative risk value of the path. When the preset propagation threshold is reached and the dynamic supply time of the corresponding production unit is lower than the preset duration threshold, the system determines the candidate path as the risk propagation path and determines the time step that first meets the above conditions as the earliest impact time.

[0085] Hazardous attribute gating factors , , , , , The risk propagation threshold and preset duration threshold can be calibrated or verified based on historical arrival times, historical quality inspection results, historical inventory changes, historical production and requisition records, historical warehousing anomaly records, and risk handling feedback. This calibration or verification is used to determine the degree of influence of hazardous attribute constraint features on risk propagation weights under different edge types. It does not change the preset processing rules for batch-level spatiotemporal correlation, batch availability judgment, dynamic supply duration calculation, hazardous attribute gating, and risk propagation path identification, nor is it used as a training step for machine learning models.

[0086] S8. Risk Warning Information Generation When the risk identification results meet the preset warning conditions, the system generates risk warning information.

[0087] The preset early warning conditions include at least one of the following: the dynamic availability time is less than the shortest replenishment cycle; the dynamic availability time is less than the continuous operation safety threshold of the production unit; the estimated arrival time of the batch in transit is later than the planned arrival time by more than the preset time; the estimated release time of quality inspection is later than the production plan requisition time; the storage environment parameters exceed the SDS storage condition threshold; the carrier, vehicle or tank container qualifications do not meet the transportation qualification requirements; the supplier's equipment operation status is abnormal and affects the fulfillment of the corresponding purchase order; regional abnormal events affect the supplier's location, transportation route or storage node; the batch availability index is lower than the preset index threshold; the risk propagation intensity is higher than the preset propagation threshold.

[0088] Risk warning information includes risk level, risk type, risk source, impact on purchase orders, impact on waybills, impact on raw material batches, impact on storage locations or tanks, impact on production facilities, earliest impact time, dynamic supply availability duration, risk propagation path, and handling recommendations.

[0089] The risk level can be set as Level 1, Level 2, Level 3, and Level 4, or as low risk, medium risk, high risk, and major risk.

[0090] The risk propagation path can be represented as: supplier equipment malfunction, delay in fulfilling purchase orders, postponement of the expected arrival time of batches in transit, delay in the release time of batches awaiting inspection, the dynamic availability of target chemical raw materials being less than the minimum replenishment cycle, and the risk of reduced load on ethylene glycol production units.

[0091] Risk transmission paths can also be represented as: regional road control, detours for hazardous materials transport vehicles, extended transportation time, delayed arrival of key raw material batches at the plant, expected depletion of released inventory, and production facilities being affected at the expected time.

[0092] Risk propagation paths can also be represented as: abnormal tank environment, reduced batch availability, reduced released usable inventory, shortened dynamic supply time, and continuous operation risk of production facilities.

[0093] S9. Generation of Restrictive Handling Recommendations The system generates disposal recommendations based on risk type, risk level, affected batch, affected storage tank, affected production unit, and earliest time of impact.

[0094] Disposal recommendations are generated under the following constraints: material specification and grade constraints; hazard category and storage and transportation requirements listed in the SDS; compatibility constraints of alternative materials; supplier access status constraints; supplier capacity constraints; hazardous materials transportation qualification constraints; availability constraints of transport vehicles or tank containers; storage capacity constraints; incompatible matching constraints; quality inspection release cycle constraints; production plan requirements constraints; and minimum replenishment cycle constraints. If a candidate disposal recommendation does not meet any of the hard constraints listed in the SDS regarding hazard category and storage and transportation requirements, hazardous materials transportation qualification, transport vehicle or tank container requirements, storage capacity, incompatible matching, quality inspection release status, or continuous operation requirements of the production unit, the system will not output that candidate disposal recommendation as an executable recommendation.

[0095] Recommended actions include switching to qualified suppliers, allocating batches of the same specifications that have already been released, adjusting the inspection schedule without changing the inspection standards and release conditions, adjusting transportation routes, replacing carriers or transport vehicles with the appropriate qualifications, adjusting the priority of entry into the factory, adjusting storage locations or tanks, using alternative materials, adjusting the load on production equipment, triggering replenishment orders in advance, reminding the quality inspection department to expedite inspections, and reminding the warehousing department to handle abnormal storage environments.

[0096] Each disposal suggestion can generate an executability score. The executability score is determined based on disposal time, disposal cost, material matching degree, transportation qualification matching degree, storage capacity matching degree, quality inspection cycle, and the degree of impact on production plan.

[0097] In this embodiment, the disposal recommendations are generated under the constraints of material specifications, hazard categories and storage and transportation requirements recorded in the SDS, transportation qualifications, quality inspection and release status, tank capacity, incompatible relationships, and continuous operation requirements of the production unit. These candidate disposal recommendations do not include individual credit ratings, personnel disciplinary actions, procurement transaction price determination, supplier commercial rankings, or automatic transaction execution results. For matters involving supplier switching, carrier replacement, transportation route adjustments, storage location or tank adjustments, or production unit load adjustments, implementation is subject to review and confirmation by personnel with appropriate authority.

[0098] Example 2: This embodiment provides a risk intelligent early warning system for the chemical raw material supply chain, such as Figure 2As shown, the system includes a data acquisition module, a data cleaning and standardization module, a batch-level correlation modeling module, a hazardous attribute feature generation module, a batch availability prediction module, a dynamic supply duration calculation module, a heterogeneous time-series supply chain diagram construction module, a risk propagation identification module, a risk warning generation module, a disposal suggestion generation module, and a results display module.

[0099] The data acquisition module is used to collect purchase order data, supplier performance data, supplier equipment operation data, transportation trajectory data, inventory batch data, warehousing environment data, quality inspection release data, production plan consumption data, market price data, and regional abnormal event data.

[0100] The data cleaning and standardization module is used to perform field standardization, unit conversion, time synchronization, outlier handling, duplicate data merging, and data credibility labeling on data from different sources.

[0101] The batch-level association modeling module is used to establish spatiotemporal relationships between purchase orders, transport orders, raw material batches, quality inspection status, storage locations or tanks, and production consumption based on purchase order numbers, waybill numbers, supplier batch numbers, internal enterprise batch numbers, quality inspection numbers, storage location or tank numbers, production unit numbers, and timestamps.

[0102] The hazardous attribute feature generation module is used to generate hazardous attribute constraint features based on the hazard categories, storage and transportation requirements, storage condition thresholds, transportation qualification requirements, incompatible relationships, quality inspection cycles, and alternative material sets recorded in the SDS of the target chemical raw material.

[0103] The batch availability prediction module is used to predict the availability status and quantity of each batch within the prediction window based on the raw material batch status, estimated arrival time in transit, quality inspection release status, storage environment status, and production plan requirements.

[0104] The dynamic availability calculation module is used to calculate the dynamic availability of the target chemical raw material based on the released inventory, the inventory awaiting inspection, the estimated arrival quantity in transit, the estimated quality inspection release time, the production plan consumption rate, and the shortest replenishment cycle.

[0105] The heterogeneous time-series supply chain graph construction module is used to construct a heterogeneous time-series supply chain graph that includes supplier nodes, supplier production equipment nodes, purchase order nodes, transportation nodes, raw material batch nodes, quality inspection nodes, warehousing nodes, production equipment nodes, market price nodes, and regional abnormal event nodes.

[0106] The risk propagation identification module is used to identify the propagation path, propagation intensity, and affected scope of risks between supply chain nodes based on heterogeneous time-series supply chain diagrams and hazard attribute gating factors.

[0107] The risk warning generation module is used to generate risk level, risk type, risk source, impact on purchase orders, impact on waybills, impact on batches, impact on storage locations or tanks, impact on production units, earliest impact time, and risk propagation path based on the risk identification results.

[0108] The disposal suggestion generation module is used to generate candidate disposal suggestions and feasibility scores under the constraints of material specifications, hazard categories and storage and transportation requirements recorded in SDS, supplier access status, transportation qualifications, storage capacity, incompatible relationships, quality inspection cycle and production plan.

[0109] The results display module is used to present early warning results in the form of risk dashboards, supply chain maps, batch status lists, production unit impact views, risk propagation path diagrams, and disposal suggestion lists.

[0110] In one optional implementation, the system further includes a parameter verification module, used to verify and update the estimated arrival time, estimated quality inspection release time, batch availability index, risk identification threshold, and risk propagation weight based on the actual arrival time, actual quality inspection results, actual usage results, actual inventory changes, and risk handling feedback. The verification and updating only adjust preset parameters or thresholds, without changing the processing rules for batch-level spatiotemporal correlation, hazardous attribute constraints, batch availability judgment, dynamic supply duration calculation, and risk propagation path identification, and without generating new model structures, training sample labels, or loss functions.

[0111] Example 3: This embodiment provides a risk warning process for production impact caused by transportation delays.

[0112] An ethylene glycol production company purchased target chemical raw material A. The purchase order number is PO-001, the supplier is S1, the supplier batch number is B-S1-001, the transport document number is T-001, and the transport vehicle is V-001.

[0113] The system establishes the association between the order, supplier batch, and shipping document based on the purchase order number PO-001, supplier batch number B-S1-001, and shipping document number T-001.

[0114] During transportation, the system detected that the GPS trajectory of vehicle V-001 deviated from the planned route, and regional anomaly data showed that there was road closure near the planned path. The system adjusted the estimated arrival time based on the degree of deviation between the actual trajectory and the planned trajectory, the duration of the stop, the scope of road closures, and the vehicle's current speed.

[0115] Meanwhile, inventory batch data shows that the current released inventory of target chemical raw material A is M1, the inventory awaiting inspection is M2, and the estimated usable quantity of batches in transit is M3. Production plan consumption data shows that production unit D1 will continue to consume target chemical raw material A within the future forecast window.

[0116] The system combines the released inventory M1, the inventory awaiting inspection M2, the estimated available quantity of batches in transit M3, the revised estimated arrival time, the estimated quality inspection release time, and the consumption rate of production unit D1 to calculate the dynamic supply duration of the target chemical raw material A.

[0117] If the dynamic supply duration is less than the shortest replenishment cycle of the target chemical raw material A, the system determines that there is a risk of insufficient inventory or supply disruption.

[0118] The system identified the following risk propagation paths in the heterogeneous time-series supply chain diagram with hazardous attribute gating: abnormal transportation path, delayed arrival of batches in transit, delayed quality inspection release time, shortened dynamic supply duration, and risk of reduced load on production unit D1.

[0119] The system generates early warning information, including risk types of transportation delay risk and inventory shortage risk, risk level of high risk, risk sources of abnormal transportation routes and regional road control, affecting purchase order PO-001, affecting waybill T-001, affecting batch B-S1-001, affecting production unit D1, and the earliest time of impact is the expected time of supply shortage.

[0120] The system further generates disposal suggestions, including allocating batches of the same specifications that have already been released, adjusting transportation routes, reminding the quality inspection department to pre-arrange inspection resources, triggering backup supplier orders in advance, and evaluating the short-term load reduction plan for production unit D1.

[0121] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent early warning of risks in a chemical raw material supply chain, characterized in that, Includes the following steps: Acquire multi-source data of the target chemical raw material supply chain, including purchase order data, transportation trajectory data, inventory batch data, warehousing environment data, quality inspection release data, and production plan consumption data; The multi-source data is cleaned and standardized, and a batch-level spatiotemporal correlation is established based on the multi-source data and timestamps to generate batch-level supply chain status data with raw material batches as the core. A material attribute table is constructed for the target chemical raw material, and the storage environment and transportation trajectory data are matched with the material attribute table to generate the hazard attribute constraint characteristics of each batch of raw material. Based on the batch-level supply chain status data and the hazardous attribute constraint features, a batch-level supply chain status vector is generated. Based on the inventory batch data, transportation trajectory data, hazardous attribute constraint features and production plan consumption data, the availability status, predicted availability quantity and dynamic supply duration of the target chemical raw material for each raw material batch within the prediction window are determined. Based on the batch-level supply chain status data, hazard attribute constraint characteristics, and dynamic availability duration, a heterogeneous time-series supply chain graph is constructed. In the heterogeneous time-series supply chain graph, a hazard attribute gating factor is determined according to the hazard attribute constraint characteristics, and the hazard attribute gating factor is used to adjust the risk propagation weight of different edge types. Based on the adjusted risk propagation weights, the heterogeneous time-series supply chain diagram is processed for risk calculation, identifying risk sources, risk propagation paths, affected raw material batches, affected storage locations or tanks, affected production units, and the earliest impact time, and generating risk warning information to trigger supply chain risk management.

2. The intelligent risk early warning method for the chemical raw material supply chain according to claim 1, characterized in that, The cleaning and standardization process includes unifying material codes, supplier codes, storage location or tank numbers, and production unit numbers, as well as unifying time formats and units of measurement, and verifying, correcting, or marking missing, abnormal, or duplicate data; the batch-level spatiotemporal correlation is established based on order number, waybill number, supplier batch number, internal enterprise batch number, quality inspection number, storage location or tank number, production unit number, and timestamp.

3. The intelligent risk early warning method for the chemical raw material supply chain according to claim 1, characterized in that, When generating the aforementioned hazard attribute constraint features, the following are extracted from the material attribute table: hazard category, storage condition threshold, transportation qualification requirements, transportation vehicle requirements, prohibited material set, quality inspection release cycle, alternative materials, and process importance level of production equipment. The warehousing environment data, transportation trajectory data, quality inspection release data, and production plan consumption data corresponding to each batch of raw materials are matched with the extracted results to form the hazard attribute constraint features of the corresponding batch.

4. The intelligent risk early warning method for the chemical raw material supply chain according to claim 1, characterized in that, When determining the availability status and predicted availability quantity of each raw material batch, the batch status, estimated arrival time, estimated quality inspection release time, storage environment status, and hazardous attribute constraint characteristics are read according to the time step. When the corresponding time step is not earlier than the estimated arrival time and estimated quality inspection release time, and the storage conditions and hazardous attribute constraints are met, the batch quantity is included in the predicted availability quantity. For batches with restricted use, the smaller value between the inventory quantity and the upper limit of restricted use is included. For frozen, unqualified, or unmet constraints batches, they are recorded as zero.

5. The intelligent risk early warning method for the chemical raw material supply chain according to claim 4, characterized in that, When determining the dynamic available supply duration, the predicted available quantity of each batch of raw materials is summarized according to time step, and the cumulative planned consumption from the current time to the corresponding time step is deducted to obtain the predicted available inventory. When the predicted available inventory first falls below the minimum continuous operating demand of the production unit, the time interval between the time step and the current time is determined as the dynamic available supply duration. The dynamic available supply duration is compared with the shortest replenishment cycle to determine the risk of insufficient supply.

6. The intelligent risk early warning method for the chemical raw material supply chain according to claim 1, characterized in that, Constructing the heterogeneous time-series supply chain graph includes establishing nodes for suppliers, purchase orders, transportation, raw material batches, quality inspection, storage locations or tanks, and production units; establishing timestamped associated edges according to supply, fulfillment, batch attribution, inspection, warehousing occupancy, and material supply relationships; writing the batch availability status into the raw material batch node, and writing the dynamic supply duration into the production unit node.

7. The intelligent risk early warning method for the chemical raw material supply chain according to claim 6, characterized in that, The hazard attribute gating factors are calculated according to the edge type by the corresponding hazard attribute constraint features. Among them, the associated edge from the transportation node to the raw material batch node adopts the abnormal features of transportation qualification, transportation vehicle and transportation route; the associated edge from the raw material batch to the quality inspection adopts the quality inspection release status and the remaining amount of the maximum allowable storage time; the associated edge from the raw material batch to the storage location or tank adopts the features of storage environment deviation and incompatible relationship; and the associated edge from the storage location or tank to the production unit adopts the process importance level of the production unit and the degree of supply tension.

8. The intelligent risk early warning method for the chemical raw material supply chain according to claim 7, characterized in that, The risk direction of the hazard attribute constraint features of each associated edge is normalized, and a hazard attribute gating factor with a value of 0 to 1 is generated based on the normalization result and the gating parameter corresponding to the edge type. When the hazard attribute gating factor reaches the first threshold, the basic risk propagation weight is increased; when the hazard attribute gating factor is not higher than the second threshold, the basic risk propagation weight is decreased; and in other cases, the basic risk propagation weight remains unchanged. The first threshold is greater than the second threshold.

9. The intelligent risk early warning method for the chemical raw material supply chain according to claim 8, characterized in that, The risk calculation process includes: calculating the cumulative risk value of candidate propagation paths along the time sequence of the heterogeneous time-series supply chain diagram; screening candidate propagation paths whose timestamps satisfy the sequential order of supplier, transportation, raw material batch, quality inspection, storage location or storage tank to production unit; when the cumulative risk value reaches the propagation threshold and the dynamic supply duration of the corresponding production unit is lower than the duration threshold, the candidate propagation path is determined as the risk propagation path, the time step that first meets the condition is determined as the earliest impact time, and the risk propagation path and the earliest impact time are written into the risk warning information.

10. A risk intelligent early warning system for a chemical raw material supply chain, characterized in that, It includes a data acquisition module, a data cleaning and standardization module, a batch-level correlation modeling module, a hazardous attribute feature generation module, a batch availability prediction module, a dynamic supply duration calculation module, a heterogeneous time-series supply chain diagram construction module, a risk propagation identification module, a risk warning generation module, a disposal suggestion generation module, and a results display module; The data acquisition module is used to acquire data consumed in procurement, transportation, inventory, warehousing, quality inspection, and production planning. The data cleaning and standardization module is used to unify fields, units, and times, and to handle abnormal data; The batch-level association modeling module is used to establish batch-level spatiotemporal associations; The hazardous attribute feature generation module is used to generate hazardous attribute constraint features; The batch availability prediction module and the dynamic availability calculation module are used to determine the predicted available quantity and the dynamic availability duration. The heterogeneous time-series supply chain graph construction module and the risk propagation identification module are used to construct heterogeneous time-series supply chain graphs, adjust risk propagation weights, and identify risk propagation paths. The risk warning generation module is used to generate risk warning information; The disposal suggestion generation module is used to generate candidate disposal suggestions; The results display module is used to show the risk warning information and candidate treatment suggestions.

Images