A spare part supply chain whole-process digital collaborative management system and method

Abstract

The present application relates to the technical field of supply chain management, in particular to a kind of spare parts supply chain whole process digitization collaborative management system and method, by building dynamic scoring model to evaluate supplier data, obtain supplier score;When supplier score is lower than predetermined threshold, generate multi-source purchase plan, according to supplier score dynamic adjustment purchase distribution;Build in and out warehouse spare parts intelligent verification model to analyze the out-of-stock spare parts and the in-warehouse spare parts, generate acceptance report and block chain record association, when difference is found, trigger alarm and freeze order;By supplier performance rate, logistics data and market demand fluctuation, build supply chain risk topology, real-time identification and early warning high-risk node;Based on high-risk node, emergency purchase is initiated to supplier score high supplier, and based on geographic proximity, optimize inventory scheduling;Based on production line state and in-transit logistics data, real-time optimization inventory level, dynamically adjust safety stock threshold to respond to supply chain interruption.

Description

Technical Field

[0001] This invention relates to the field of supply chain management technology, specifically to a digital collaborative management system and method for the entire process of spare parts supply chain. Background Technology

[0002] With the rapid development of digital technology, enterprises are increasingly relying on computer information systems to achieve efficient, accurate, and collaborative operation management in their supply chain management. This is particularly true in the medical device industry, where the unique characteristics of the products and stringent regulatory requirements make full lifecycle management of the component supply chain crucial. Currently, most companies' supply chain management systems are limited to data recording and inventory scheduling at a single stage, lacking dynamic monitoring and collaborative management of the entire process from procurement, warehousing, storage, and outbound delivery to final distribution and after-sales service, making it difficult to establish a complete product traceability system. Addressing this issue, how to achieve full lifecycle management, ensure the supply chain's resilience in the face of unforeseen circumstances, and leverage these capabilities to provide reliable and transparent supply chain management support services to downstream component procurement customers has become a pressing technical challenge for the industry.

[0003] While some companies have begun to automate warehouse management using technologies such as barcodes, RFID, and the Internet of Things (IoT) in existing digital management systems, the lack of standardized data and the prevalence of information silos among systems hinder real-time data sharing and collaborative decision-making across the supply chain. This not only affects inventory accuracy and order fulfillment efficiency but also increases the risks faced by the medical device supply chain, such as abnormal temperature and humidity, logistics delays, and quality issues. Furthermore, the multi-stage regulatory requirements for medical device products necessitate systems that support end-to-end traceability and data logging to meet industry standards such as GMP and GSP. Therefore, building a comprehensive data collaboration platform covering the entire process of procurement, warehousing, inventory, outbound delivery, distribution, and after-sales service—one that not only optimizes internal management but also provides procurement customers with the necessary data visibility and service guarantees—is crucial for improving supply chain resilience and ensuring product quality.

[0004] Furthermore, supply chain resilience, a crucial factor in ensuring the continuity of medical device supply, hinges on real-time monitoring, dynamic adjustment, and risk warning across all stages of the supply chain through digital means. This enables rapid response and timely adjustments in the face of emergencies such as logistical disruptions and equipment failures, achieving efficient resource allocation. Currently, there is a lack of a digital collaborative platform that integrates full lifecycle management and supply chain resilience assurance. This platform not only requires technical capabilities such as data interoperability, intelligent early warning, and decision support, and the ability to make some management functions or information transparent, providing value-added supply chain management services to component procurement customers, but also necessitates a management model that achieves collaborative operations and full-process transparency across all stages, meeting the high standards of the medical device industry for safe, timely, and reliable supply.

[0005] To address this, a digital collaborative management system and method for the entire spare parts supply chain are proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a digital collaborative management system and method for the entire process of spare parts supply chain. This system involves constructing a dynamic scoring model to evaluate supplier data and obtain supplier scores; when a supplier score falls below a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier score; an intelligent verification model for outbound and inbound spare parts is constructed to analyze outbound and inbound spare parts, generating acceptance reports and linking them to blockchain records; when discrepancies are detected, alarms are triggered and orders are frozen; a supply chain risk topology map is constructed using supplier fulfillment rates, logistics data, and market demand fluctuations to identify and warn of high-risk nodes in real time; emergency procurement is initiated from suppliers with high scores based on high-risk nodes, and inventory scheduling is optimized based on geographical proximity; and inventory levels are optimized in real time based on the production line status and in-transit logistics data of outbound and inbound spare parts, dynamically adjusting safety stock thresholds to cope with supply chain disruptions.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A digital collaborative management system and method for the entire process of a spare parts supply chain, comprising:

[0009] The multi-dimensional dynamic scoring module constructs a dynamic scoring model to evaluate supplier data and obtain supplier scores; when a supplier score is lower than a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier score.

[0010] The full lifecycle data chain module constructs an intelligent verification model for outbound and inbound spare parts, analyzes outbound and inbound spare parts, generates acceptance reports and links them with blockchain records, and triggers alarms and freezes orders when discrepancies are found.

[0011] The supply chain risk assessment module constructs a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, identifies and warns of high-risk nodes in real time, initiates emergency procurement to suppliers with high supplier ratings based on high-risk nodes, and optimizes inventory scheduling based on geographical proximity.

[0012] The inventory optimization module optimizes inventory levels in real time based on the production line status and in-transit logistics data of outbound and inbound spare parts, and dynamically adjusts the safety stock threshold to cope with supply chain disruptions.

[0013] Preferably, the supplier data includes the supplier's historical performance data, real-time geographical risk data, and capacity flexibility data; a unique digital identifier is assigned to the outbound and inbound spare parts; the unique digital identifier includes the product batch number, production date, quality inspection information, and logistics tracking code.

[0014] Preferably, the dynamic scoring model includes a data input processing layer, a data feature extraction layer, a weight update layer, and a supplier score generation layer;

[0015] The data input processing layer preprocesses supplier historical performance data, real-time geographic risk data, and capacity flexibility data.

[0016] The data feature extraction layer analyzes preprocessed supplier historical performance data, real-time geographic risk data, and capacity elasticity data to obtain on-time delivery rate trends, disaster probability, and redundancy indicators.

[0017] The weight update layer learns and generates weights for on-time delivery rate trends, disaster probability, and redundancy indicators based on historical and real-time supplier data inputs.

[0018] The supplier rating generation layer generates supplier ratings by weighted combination calculation of weights and features.

[0019] Preferably, the intelligent verification model for inbound and outbound spare parts includes an intelligent verification and analysis layer for outbound spare parts and an intelligent verification and analysis layer for inbound spare parts;

[0020] The intelligent verification and analysis layer for outbound spare parts uses RFID technology to identify the identity information of outbound spare parts; uses image recognition technology to verify the physical characteristics of outbound spare parts; compares the identified identity information and physical characteristics with the spare parts information in the sales order; and triggers an alarm and freezes the relevant order when a discrepancy is found.

[0021] The intelligent verification and analysis layer for incoming spare parts uses AI vision technology to inspect the appearance and quality of the incoming spare parts; compares the inspection results with the spare parts specifications in the purchase order; automatically generates an acceptance report based on the comparison results; and links the acceptance report with blockchain records.

[0022] Preferably, the specific steps for constructing the supply chain risk topology map are as follows:

[0023] By analyzing the supplier fulfillment rate, logistics data, and market demand fluctuations, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are obtained.

[0024] Construct a supply chain diagram, where nodes represent suppliers, manufacturers, and logistics hubs, and edges represent the flow of materials and / or information;

[0025] The supply chain diagram, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are input into a graph neural network. The graph neural network extracts node features, predicts the risk probability of each node, and identifies nodes whose risk probability exceeds a predetermined threshold, thus obtaining high-risk nodes.

[0026] Preferably, the specific steps for obtaining the dynamically adjusted safety stock threshold include:

[0027] Production line status data and in-transit logistics data are preprocessed;

[0028] By combining historical inventory data, production plans, market demand, and real-time data input, and using transfer learning, a demand forecasting model is constructed to generate forecast results. Based on the forecast results, the current inventory level is evaluated, the difference between inventory and future demand is calculated, and the safety stock threshold in inventory management is automatically adjusted based on the calculation results to identify inventory risks. A dynamic inventory optimization algorithm is introduced to obtain inventory optimization results. Based on the inventory optimization results, a supply chain adjustment plan is generated.

[0029] A digital collaborative management method for the entire spare parts supply chain includes:

[0030] A dynamic scoring model is constructed to evaluate supplier data and obtain supplier scores; when a supplier score is lower than a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier score.

[0031] A smart verification model for outbound and inbound spare parts is constructed to analyze outbound and inbound spare parts, generate acceptance reports and link them with blockchain records, and trigger alarms and freeze orders when discrepancies are found.

[0032] By constructing a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, high-risk nodes can be identified and warned in real time; emergency procurement can be initiated from suppliers with high supplier ratings based on high-risk nodes, and inventory scheduling can be optimized based on geographical proximity.

[0033] Based on the production line status and in-transit logistics data of outbound and inbound spare parts, inventory levels are optimized in real time, and safety stock thresholds are dynamically adjusted to cope with supply chain disruptions.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1. This invention employs a multi-dimensional dynamic scoring module. Through comprehensive preprocessing, feature extraction, and weight updates of supplier historical performance, real-time geographical risk, and capacity flexibility data, it generates accurate supplier scores in real time. This invention reduces the uncertainty risks caused by information lag and incomplete data by enabling dynamic feedback on supplier risk and supply capacity. When the score falls below a preset threshold, a multi-source procurement plan is automatically generated, and procurement allocation is intelligently adjusted based on the score to ensure stable supply of key components and optimal procurement costs. This technology improves supply chain transparency and responsiveness, driving the intelligent upgrade of supply chain management.

[0036] 2. The full lifecycle data chain module and intelligent verification model for spare parts constructed in this invention utilize multiple identification technologies such as RFID, image recognition, and AI vision to automatically verify the identity and quality of spare parts entering and leaving the warehouse, and accurately compare them with sales and purchase order data. This invention enables full traceability of spare parts information, timely detection of discrepancies between inbound / outbound data and order information, automatic triggering of alarms and freezing of relevant orders, effectively preventing erroneous transfers and quality risks. Simultaneously, linking acceptance reports with blockchain records achieves tamper-proof data and real-time sharing, significantly improving supply chain transparency and trust, and reducing errors and delays caused by manual verification.

[0037] 3. The inventory optimization module of this invention collects real-time production line status and in-transit logistics data, combines historical inventory data, production plans, and market demand information, and utilizes transfer learning to construct a demand forecasting model to accurately assess the difference between current inventory levels and future demand. Based on this assessment result, it automatically adjusts the safety stock threshold in inventory management, introduces a dynamic inventory optimization algorithm, and generates inventory optimization results and supply chain adjustment plans, thereby effectively mitigating the risk of supply chain disruptions. This invention possesses real-time and intelligent advantages, enabling it to rapidly respond to changes at each stage of the supply chain, dynamically balance inventory costs and supply security, and significantly improve operational efficiency and market competitiveness. Attached Figure Description

[0038] Figure 1 A schematic diagram of the structure of a full-process digital collaborative management system for spare parts supply chain provided by the present invention;

[0039] Figure 2 A flowchart illustrating a digital collaborative management method for the entire spare parts supply chain provided by this invention;

[0040] Figure 3 This is a schematic diagram of the dynamic scoring model structure provided in an embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of the entire supply chain structure provided for an embodiment of the present invention. Detailed Implementation

[0042] 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.

[0043] Example 1:

[0044] Please see Figures 1 to 4 This invention provides a digital collaborative management system for the entire process of a spare parts supply chain, applied to a digital collaborative management method for the entire process of a spare parts supply chain. The digital collaborative management system for the entire process of a spare parts supply chain includes a multi-dimensional dynamic scoring module, a full lifecycle data chain module, a supply chain risk assessment module, and an inventory optimization module. See details below. Figure 1 For the module correspondence method and steps in a digital collaborative management system for the entire spare parts supply chain, please refer to [link / reference]. Figure 2 The specific technical solution is as follows:

[0045] A multi-dimensional dynamic scoring module constructs a dynamic scoring model to evaluate supplier data and obtain supplier scores. When a supplier score is lower than a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier score. The supplier data includes historical performance data, real-time geographical risk data, and capacity flexibility data.

[0046] Furthermore, the dynamic scoring model includes a data input processing layer, a data feature extraction layer, a weight update layer, and a supplier score generation layer, see [link / reference]. Figure 3 ;

[0047] The data input processing layer preprocesses supplier historical performance data, real-time geographic risk data, and capacity flexibility data.

[0048] The data feature extraction layer analyzes preprocessed supplier historical performance data, real-time geographic risk data, and capacity elasticity data to obtain on-time delivery rate trends, disaster probability, and redundancy indicators.

[0049] The weight update layer learns and generates weights for on-time delivery rate trends, disaster probability, and redundancy indicators based on historical and real-time supplier data inputs.

[0050] The supplier rating generation layer generates supplier ratings by weighted combination calculation of weights and features.

[0051] In this embodiment, by constructing a multi-dimensional dynamic scoring module, the present invention can comprehensively evaluate a supplier's historical performance, real-time geographical risks, and capacity elasticity, achieving accurate calculation of supplier scores. The synergistic effect of the data input processing layer, data feature extraction layer, weight update layer, and supplier score generation layer ensures the dynamic nature and adaptability of the scoring. Specifically, the data feature extraction layer can extract key features from supplier data, such as on-time delivery rate trends, disaster probability, and redundancy indicators, making supplier evaluation more comprehensive. The weight update layer combines historical and real-time data for adaptive learning, enabling the scoring results to dynamically adjust with changes in the supply chain environment, improving the system's responsiveness to sudden risks. When a supplier's score falls below a predetermined threshold, the system can automatically generate a multi-source procurement plan and dynamically adjust the procurement allocation scheme based on changes in the supplier's score, thereby reducing the risk of supply chain disruptions and improving the flexibility and reliability of procurement decisions.

[0052] The full lifecycle data chain module constructs an intelligent verification model for outbound and inbound spare parts to analyze outbound and inbound spare parts, generate acceptance reports and associate them with blockchain records, and trigger alarms and freeze orders when discrepancies are found; a unique digital identifier is assigned to the outbound and inbound spare parts; the unique digital identifier includes product batch number, production date, quality inspection information and logistics tracking code;

[0053] Furthermore, the intelligent verification model for inbound and outbound spare parts includes an intelligent verification and analysis layer for outbound spare parts and an intelligent verification and analysis layer for inbound spare parts;

[0054] The intelligent verification and analysis layer for outbound spare parts uses RFID technology to identify the identity information of outbound spare parts; uses image recognition technology to verify the physical characteristics of outbound spare parts; compares the identified identity information and physical characteristics with the spare parts information in the sales order; and triggers an alarm and freezes the relevant order when a discrepancy is found.

[0055] The intelligent verification and analysis layer for incoming spare parts uses AI vision technology to inspect the appearance and quality of the incoming spare parts; compares the inspection results with the spare parts specifications in the purchase order; automatically generates an acceptance report based on the comparison results; and links the acceptance report with blockchain records.

[0056] In this embodiment, the present invention achieves intelligent verification of inbound and outbound spare parts through the construction of a full lifecycle data chain module, improving the accuracy and security of spare parts management. By assigning unique digital identifiers to outbound and inbound spare parts and combining RFID technology, image recognition technology, and AI vision technology, the traceability of spare parts information throughout the entire process can be ensured. The intelligent verification and analysis layer for outbound spare parts can effectively identify the identity information of outbound spare parts and ensure consistency with sales orders through physical feature comparison, avoiding the risks of mis-shipment, omission, or counterfeiting; the intelligent verification and analysis layer for inbound spare parts uses AI vision technology to inspect the appearance and quality of spare parts and automatically generates acceptance reports, which are linked to blockchain records to ensure the credibility and immutability of acceptance data. When the system detects discrepancies between inbound and outbound spare parts, it can promptly trigger alarms and freeze relevant orders, thereby quickly detecting and preventing abnormal circulation and reducing losses and risks in supply chain management. This method can significantly improve the intelligence level of spare parts circulation and optimize inventory management processes. By combining RFID and AI vision technology with blockchain records, this invention reduces the error rate of incoming batches from 8.2% to 0.5%, and improves the efficiency of acceptance report generation and freeze response by 95.8%, effectively solving the problems of inefficiency and high error rate of traditional manual verification. See Table 1 for specific data.

[0057] Table 1. Effect of intelligent verification technology on error rate reduction in inbound and outbound inventory.

[0058] Technical indicators Traditional manual verification This invention is an intelligent verification system. Error rate reduction Batch matching error rate 8.2％ 2.3% (RFID + AI Vision) -72.0％ Quality missed detection rate 12％ 2.7% (Blockchain-related acceptance) -77.5％ Order freeze response time Manual processing (2 hours) Automatic trigger (<20 minutes) -83.3％

[0059] The supply chain risk assessment module constructs a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, and identifies and warns of high-risk nodes in real time; it initiates emergency procurement to suppliers with high supplier ratings based on high-risk nodes, and optimizes inventory scheduling based on geographical proximity; at the same time, it uses transfer learning and dynamic clustering combination to predict the demand for spare parts in intermittent demand patterns.

[0060] Furthermore, the specific steps for constructing the supply chain risk topology map are as follows:

[0061] By analyzing the supplier fulfillment rate, logistics data, and market demand fluctuations, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are obtained.

[0062] Construct a supply chain diagram, where nodes represent suppliers, manufacturers, and logistics hubs, and edges represent the flow of materials and / or information;

[0063] The supply chain diagram, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are input into a graph neural network. The graph neural network extracts node features, predicts the risk probability of each node, and identifies nodes whose risk probability exceeds a predetermined threshold, thus obtaining high-risk nodes.

[0064] In this embodiment, the present invention, through the construction of a supply chain risk assessment module, achieves intelligent identification and dynamic early warning of global supply chain risks, thereby improving the security and responsiveness of supply chain management. Based on supplier fulfillment rates, logistics data, and market demand fluctuations, the system can construct a supply chain risk topology map and use graph neural networks to extract features and predict risks of suppliers, manufacturers, and logistics hubs in the supply chain, thereby identifying high-risk nodes in real time. When the risk probability of a high-risk node exceeds a predetermined threshold, the system can automatically trigger an early warning mechanism and take targeted countermeasures, such as initiating emergency procurement from suppliers with higher supplier ratings, to reduce the risk of supply chain disruption. Simultaneously, inventory scheduling is optimized based on geographical proximity to ensure the efficiency of spare parts supply. Furthermore, a method combining transfer learning and dynamic clustering is used to predict spare parts demand in intermittent demand patterns, enabling the supply chain to adapt more accurately to market fluctuations and improving the flexibility and accuracy of inventory management. This method not only enhances the stability and resilience of the supply chain but also optimizes supply chain resource allocation and improves overall operational efficiency.

[0065] The inventory optimization module optimizes inventory levels in real time based on the production line status and in-transit logistics data of outbound and inbound spare parts, and dynamically adjusts the safety stock threshold to cope with supply chain disruptions.

[0066] The specific steps for dynamically adjusting the safety stock threshold include:

[0067] Production line status data and in-transit logistics data are preprocessed;

[0068] By combining historical inventory data, production plans, market demand, and real-time data input, and using transfer learning, a demand forecasting model is constructed to generate forecast results. Based on the forecast results, the current inventory level is evaluated, the difference between inventory and future demand is calculated, and the safety stock threshold in inventory management is automatically adjusted based on the calculation results to identify inventory risks. A dynamic inventory optimization algorithm is introduced to obtain inventory optimization results. Based on the inventory optimization results, a supply chain adjustment plan is generated.

[0069] In this embodiment, the invention achieves intelligent and dynamic optimization of inventory management through the construction of an inventory optimization module, improving the stability and responsiveness of the supply chain. Based on real-time monitoring of production line status and in-transit logistics data, the system can dynamically adjust the safety stock threshold to effectively address supply chain disruption risks. By comprehensively analyzing production line status data, in-transit logistics data, and historical inventory data, and combining transfer learning to build a demand forecasting model, the system can accurately predict future inventory demand, ensuring that inventory levels match production needs. The system can assess inventory levels based on forecast results, calculate the difference between inventory and future demand, and automatically adjust the safety stock threshold accordingly, identifying potential inventory risks in advance. Simultaneously, a dynamic inventory optimization algorithm is introduced to optimize inventory management strategies, ensuring the rationality of inventory allocation. Finally, the system generates supply chain adjustment plans based on the optimization results, achieving precise allocation of inventory resources, reducing inventory backlog and shortage risks, and improving the overall operational efficiency of the supply chain.

[0070] This invention achieves intelligent and refined supply chain management by constructing multi-dimensional dynamic scoring, a full lifecycle data chain, supply chain risk assessment, and inventory optimization modules, thereby improving the stability, flexibility, and resilience of the supply chain. The multi-dimensional dynamic scoring module comprehensively evaluates suppliers' historical performance, geographical risk, and capacity elasticity, ensuring the accuracy and adaptability of supplier scoring and dynamically adjusting procurement allocation plans based on the scores, thus reducing the risk of supply chain disruptions. The full lifecycle data chain module utilizes unique digital identifiers, RFID, AI vision, and other technologies to intelligently verify inbound and outbound spare parts, ensuring the authenticity and immutability of data and improving the security and transparency of spare parts management. The supply chain risk assessment module constructs a supply chain risk topology map based on graph neural networks, identifies high-risk nodes in real time, and enhances supply chain resilience through emergency procurement and optimized inventory scheduling. The inventory optimization module utilizes production line status and in-transit logistics data, combined with transfer learning to predict demand, and dynamically adjusts safety stock thresholds to optimize inventory management and reduce the risk of supply chain disruptions. (See details...) Figure 4 .

[0071] Example 2:

[0072] A multi-dimensional dynamic scoring module constructs a dynamic scoring model to evaluate supplier data and obtain supplier scores. When a supplier's score falls below a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier's score. The supplier data includes historical performance data, real-time geographic risk data, and capacity flexibility data. Historical performance data includes the supplier's past on-time delivery rate and product quality pass rate, obtained through enterprise resource planning systems or historical order records. Real-time geographic risk data includes the natural disaster risk and political instability index of the supplier's location, collected through third-party geographic information systems (GIS) or real-time news data interfaces. Capacity flexibility data includes the supplier's capacity utilization rate and reserve capacity, obtained through production reports provided by the supplier or real-time monitoring by IoT devices.

[0073] Furthermore, the dynamic scoring model includes a data input processing layer, a data feature extraction layer, a weight update layer, and a supplier score generation layer;

[0074] The data input processing layer preprocesses supplier historical performance data, real-time geographic risk data, and capacity elasticity data. The preprocessing includes data cleaning, outlier removal, and normalization.

[0075] The data feature extraction layer analyzes preprocessed supplier historical performance data, real-time geographical risk data, and capacity elasticity data to obtain on-time delivery rate trends, disaster probability, and redundancy indicators. The on-time delivery rate trend is calculated using time series analysis to determine the on-time delivery rate change trend over the past 6 months. The disaster probability is estimated based on historical disaster data using a Poisson distribution to determine the disaster probability for the next 30 days. The redundancy indicator calculates redundancy based on capacity utilization and standby capacity: Redundancy = Standby Capacity / (Total Capacity × Utilization Rate).

[0076] The weight update layer is based on historical supplier data and real-time supplier data input, and uses machine learning algorithms to learn and generate weights for on-time delivery rate trends, disaster probability, and redundancy indicators.

[0077] The supplier rating generation layer generates supplier ratings by weighted combination of weights and features;

[0078] The specific formula for calculating the supplier rating is as follows:

[0079] S = ω1T + ω2(1-P) + ω3R;

[0080] Where S is the supplier score, ω1 is the on-time delivery rate trend weight, T is the on-time delivery rate trend, ω2 is the geographical safety factor weight, P is the disaster probability, ω3 is the redundancy index weight, and R is the redundancy index.

[0081] The full lifecycle data chain module constructs an intelligent verification model for outbound and inbound spare parts to analyze outbound and inbound spare parts, generate acceptance reports and associate them with blockchain records, and trigger alarms and freeze orders when discrepancies are found; a unique digital identifier is assigned to the outbound and inbound spare parts; the unique digital identifier includes product batch number, production date, quality inspection information and logistics tracking code;

[0082] Furthermore, the intelligent verification model for inbound and outbound spare parts includes an intelligent verification and analysis layer for outbound spare parts and an intelligent verification and analysis layer for inbound spare parts;

[0083] The intelligent verification and analysis layer for outbound spare parts uses RFID technology to identify the identity information of outbound spare parts; it uses image recognition technology to verify the physical characteristics of outbound spare parts, including size, color, and quantity; it compares the identified identity information and physical characteristics with the spare parts information in the sales order; when a discrepancy is found, an alarm is triggered and the relevant order is frozen.

[0084] The outbound management process includes sales order processing, pre-entry of outbound shipments, delisting management, warehouse packing, generation of waybills, and receipt confirmation. This ensures that spare parts are delivered from the warehouse to the customer accurately according to the sales order requirements. Through digital and intelligent technologies, such as RFID for identity verification and image recognition for physical feature verification, the system can accurately match sales orders with outbound spare parts, avoiding shipping errors. It also records outbound status in real time, such as packing information and waybill numbers, for easy tracking. Furthermore, it promptly triggers alarms when discrepancies are detected (such as differences in spare parts size or batch numbers) to prevent problematic spare parts from leaving the warehouse.

[0085] The intelligent verification and analysis layer for incoming spare parts uses AI vision technology to inspect the appearance and quality of the incoming spare parts; compares the inspection results with the spare parts specifications in the purchase order; automatically generates an acceptance report based on the comparison results; and links the acceptance report with blockchain records.

[0086] The warehousing management process includes purchase order, inventory counting, pre-shelving of inventory, inventory verification, pre-arrival notification, pre-entry of warehousing data, acceptance, warehouse shelving, and coordination to ensure that purchased spare parts meet quality standards and are shelved in a timely manner, preparing for subsequent production or sales. Through AI vision technology and data comparison, the system can inspect the appearance and quality of incoming spare parts, such as surface scratches and specification deviations, to ensure compliance; verify quantities and specifications against purchase orders to reduce errors; and automatically generate acceptance reports and record them via blockchain to ensure data transparency and immutability.

[0087] The supply chain risk assessment module constructs a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, and identifies and warns of high-risk nodes in real time; it initiates emergency procurement to suppliers with high supplier ratings based on high-risk nodes, and optimizes inventory scheduling based on geographical proximity; at the same time, it uses transfer learning and dynamic clustering combination to predict the demand for spare parts in intermittent demand patterns.

[0088] Furthermore, the specific steps for constructing the supply chain risk topology map are as follows:

[0089] By analyzing the supplier fulfillment rate, logistics data, and market demand fluctuations, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are obtained.

[0090] Construct a supply chain diagram where nodes represent suppliers, manufacturers, and logistics hubs, and edges represent material and / or information flows. Nodes represent suppliers (e.g., S1, S2), manufacturers (e.g., M1), and logistics hubs (e.g., L1), and edges represent material flows (e.g., S1→M1) or information flows (e.g., M1→L1).

[0091] The supply chain diagram, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are input into a graph neural network. The graph neural network extracts node features and predicts the risk probability of each node. Nodes with risk probabilities exceeding a predetermined threshold are identified as high-risk nodes. The supply chain diagram and data are input into the graph neural network (GNN) to extract node features (such as low fulfillment rate and high delay) and predict risk probabilities. If the risk probability of node S1 is 0.85 (the threshold is set to 0.7 in this embodiment), it is identified as a high-risk node.

[0092] The formula for calculating the node risk probability is:

[0093] P risk (i)=σ(β1(1-F i )+β2L i +β3M i +b);

[0094] Among them, P risk (i) represents the risk probability of node i, β1 is the supplier fulfillment rate weight, and F i β1 represents the supplier fulfillment rate corresponding to node i (the lower the fulfillment rate, the higher the risk), β2 represents the weight of the logistics delay risk factor, and L represents the supplier fulfillment rate corresponding to node i. i β3 represents the logistics delay risk factor corresponding to node i (the higher the delay, the greater the risk), β3 represents the weight of the market demand fluctuation risk factor, and M represents the risk factor. i Let be the market demand volatility risk factor corresponding to node i (the greater the volatility, the higher the risk), b be the bias term, and σ() be the Sigmoid function used to map the weighted sum to the interval 0 to 1.

[0095] As described in the supply chain risk assessment module, topology analysis based on graph neural networks (GNN) can identify global supply chain node risks in real time. Compared with expert systems, the risk prediction accuracy is improved by 35.3%, and the response speed is improved by leaps and bounds from 48 hours to real time. See Table 2 for specific data.

[0096] Table 2 Comparison of Response Time between Supply Chain Risk Topology Map and Traditional Risk Assessment Methods

[0097] Evaluation Dimensions Expert Risk Assessment System Graph Neural Network Model Efficiency Improvement Risk identification delay 48 hours Real-time dynamic early warning +100％ Node association analysis depth Local node analysis Global topology mining +80％ High-risk prediction accuracy 68％ 92% (GNN feature extraction) +35.3％

[0098] The inventory optimization module optimizes inventory levels in real time based on the production line status and in-transit logistics data of outbound and inbound spare parts, and dynamically adjusts the safety stock threshold to cope with supply chain disruptions.

[0099] The specific steps for dynamically adjusting the safety stock threshold include:

[0100] Production line status data and in-transit logistics data are collected and preprocessed. Production line status data is collected by assessing the operating status of production equipment, production efficiency, and downtime. Preprocessing steps include data cleaning, outlier detection, and standardization. In-transit logistics data is collected by obtaining the location of goods in transit, estimated arrival time, and transportation delays.

[0101] By combining historical inventory data, production plans, market demand, and real-time data input, and using transfer learning, a demand forecasting model is constructed to generate forecast results. Based on the forecast results, the current inventory level is evaluated, the difference between inventory and future demand is calculated, and the safety stock threshold in inventory management is automatically adjusted based on the calculation results to identify inventory risks. A dynamic inventory optimization algorithm is introduced to obtain inventory optimization results. Based on the inventory optimization results, a supply chain adjustment plan is generated.

[0102] Model building approach: Utilize historical inventory data, production plans, market demand data, and real-time updated production line and logistics data as inputs; employ transfer learning techniques to quickly adapt to the data characteristics of the current environment, improving prediction accuracy, based on existing prediction models in the source domain (e.g., similar production environments or market contexts).

[0103] Prediction result generation: By using a trained demand forecasting model, the demand volume and demand fluctuation trend in the future are predicted, forming quantitative demand forecast results.

[0104] Inventory assessment: Compare real-time inventory levels with forecasted demand to calculate the difference between inventory and future demand (i.e., inventory gap or inventory surplus); conduct a preliminary assessment of inventory risk to identify potential inventory shortages or overstocking risks.

[0105] Dynamically adjust safety stock threshold: Automatically update the safety stock threshold based on the calculated inventory gap.

[0106] Optimization objective: The objective is to balance inventory costs and inventory risks, ensuring that neither production interruptions due to insufficient inventory nor increased warehousing costs due to inventory backlog.

[0107] Algorithm Implementation: A dynamic inventory optimization algorithm is introduced, which comprehensively considers the current inventory status, safety stock threshold, in-transit logistics situation and market demand to generate an inventory optimization plan;

[0108] The algorithm outputs key indicators such as optimal inventory level, reorder point, and replenishment quantity suggestions, which guide actual inventory management and adjustments to procurement plans.

[0109] Solution Development: Based on inventory optimization results, develop specific supply chain adjustment plans, which may cover aspects such as procurement strategies, production scheduling, and logistics optimization. The goal of the plan is to quickly respond to demand forecasts and inventory assessments, adjust procurement and production plans in a timely manner, and ensure the stability and efficiency of the overall supply chain operation.

[0110] Risk monitoring and feedback: After the plan is formulated, a real-time monitoring mechanism needs to be established to track inventory changes and market dynamics, and to provide timely feedback and optimize the dynamic adjustment model of the inventory safety threshold.

[0111] As described in the inventory optimization module, transfer learning combined with dynamic clustering significantly improves the accuracy of intermittent demand forecasting to 85% and shortens the training cycle by 78.6%, overcoming the limitations of the traditional ARIMA model under complex demand patterns. See Table 3 for details.

[0112] Table 3 Comparison of the accuracy improvement of transfer learning in demand forecasting

[0113] Predicting scenarios Traditional ARIMA model Transfer learning + dynamic clustering Improved accuracy Stable demand forecast 89％ 93％ +4.5％ Intermittent demand forecasting 52％ 85% (Dynamic Clustering Fit) +63.5％

[0114] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital collaborative management system for the entire spare parts supply chain, characterized in that, include: The multi-dimensional dynamic scoring module constructs a dynamic scoring model to evaluate supplier data and obtain supplier scores. When a supplier's score falls below a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier's score. Supplier data includes historical performance data, real-time geographic risk data, and capacity flexibility data. The dynamic scoring model includes a data input processing layer, a data feature extraction layer, a weight update layer, and a supplier score generation layer; The data input processing layer preprocesses supplier historical performance data, real-time geographic risk data, and capacity flexibility data; The data feature extraction layer analyzes preprocessed supplier historical performance data, real-time geographic risk data, and capacity elasticity data to obtain on-time delivery rate trends, disaster probability, and redundancy indicators. The weight update layer learns and generates weights for on-time delivery rate trends, disaster probability, and redundancy indicators based on historical and real-time supplier data inputs. The supplier rating generation layer generates supplier ratings by weighted combination of weights and features; The full lifecycle data chain module constructs an intelligent verification model for outbound and inbound spare parts, analyzes outbound and inbound spare parts, generates acceptance reports and links them with blockchain records, and triggers alarms and freezes orders when discrepancies are found. The supply chain risk assessment module constructs a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, and identifies and warns of high-risk nodes in real time. Emergency procurement is initiated from suppliers with high supplier ratings based on high-risk nodes, and inventory scheduling is optimized based on geographical proximity. The specific steps for constructing the supply chain risk topology map are as follows: By analyzing the supplier fulfillment rate, logistics data, and market demand fluctuations, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are obtained. Construct a supply chain diagram, where nodes represent suppliers, manufacturers, and logistics hubs, and edges represent the flow of materials and / or information; The supply chain map, supplier fulfillment rate, logistics delay data, and market demand fluctuation data are input into the graph neural network. The graph neural network extracts node features, predicts the risk probability of each node, and obtains nodes whose risk probability exceeds a predetermined threshold, thus identifying high-risk nodes. The inventory optimization module optimizes inventory levels in real time based on the production line status and in-transit logistics data of outbound and inbound spare parts, and dynamically adjusts the safety stock threshold to cope with supply chain disruptions.

2. The full-process digital collaborative management system for spare parts supply chain according to claim 1, characterized in that: Each outbound and inbound spare parts is assigned a unique digital identifier; the unique digital identifier includes the product batch number, production date, quality inspection information, and logistics tracking code.

3. The full-process digital collaborative management system for spare parts supply chain according to claim 1, characterized in that: The intelligent verification model for inbound and outbound spare parts includes an intelligent verification and analysis layer for outbound spare parts and an intelligent verification and analysis layer for inbound spare parts. The intelligent verification and analysis layer for outbound spare parts uses RFID technology to identify the identity information of outbound spare parts and uses image recognition technology to verify the physical characteristics of outbound spare parts. The system compares the identified identity information and physical characteristics with the spare parts information in the sales order; when a discrepancy is found, an alarm is triggered and the relevant order is frozen. The intelligent verification and analysis layer for incoming spare parts uses AI vision technology to inspect the appearance and quality of the incoming spare parts; compares the inspection results with the spare parts specifications in the purchase order; automatically generates an acceptance report based on the comparison results; and links the acceptance report with blockchain records.

4. The full-process digital collaborative management system for spare parts supply chain according to claim 1, characterized in that: The specific steps for dynamically adjusting the safety stock threshold include: Production line status data and in-transit logistics data are preprocessed; By combining historical inventory data, production plans, market demand, and real-time data input, and using transfer learning, a demand forecasting model is constructed to generate forecast results. Based on the forecast results, the current inventory level is evaluated, the difference between inventory and future demand is calculated, and the safety stock threshold in inventory management is automatically adjusted based on the calculation results to identify inventory risks. A dynamic inventory optimization algorithm is introduced to obtain inventory optimization results. Based on the inventory optimization results, a supply chain adjustment plan is generated.

5. A method for end-to-end digital collaborative management of the spare parts supply chain, applied to the end-to-end digital collaborative management system for the spare parts supply chain as described in any one of claims 1-4, characterized in that, include: Build a dynamic scoring model to evaluate supplier data and obtain supplier scores; When a supplier's score falls below a predetermined threshold, a multi-source procurement plan is generated, and procurement allocation is dynamically adjusted based on the supplier's score. A smart verification model for outbound and inbound spare parts is constructed to analyze outbound and inbound spare parts, generate acceptance reports and link them with blockchain records, and trigger alarms and freeze orders when discrepancies are found. By constructing a supply chain risk topology map based on supplier fulfillment rates, logistics data, and market demand fluctuations, high-risk nodes can be identified and warned in real time. Emergency procurement is initiated from suppliers with high supplier ratings based on high-risk nodes, and inventory scheduling is optimized based on geographical proximity. Based on the production line status and in-transit logistics data of outbound and inbound spare parts, inventory levels are optimized in real time, and safety stock thresholds are dynamically adjusted to cope with supply chain disruptions.

Images