Intelligent replenishment method and system based on automobile detection data

Abstract

The application discloses an intelligent replenishment method and system based on automobile detection data, relates to the technical field of automobile detection, combines multi-source data fusion with digital twinning, realizes the leap from data-driven prediction to mechanism enhancement, aligns multi-source heterogeneous data with driving mileage as a unified benchmark, provides standardized input conforming to physical degradation rules, and fuses real-time data and physical mechanism models in digital twinning, still has extrapolation ability under extreme working conditions that cannot be covered by pure data models, overcomes the defect that traditional black box models are not interpretable, connects multi-agent negotiation game with reinforcement learning, realizes the evolution of replenishment from independent calculation to global cooperation, undertakes the life probability distribution of each component, independently games to reach Nash equilibrium under global fund and storage constraints, and solves the resource competition conflict problem when multiple components trigger replenishment simultaneously.

Description

Technical Field

[0001] This invention relates to the field of automotive inspection technology, specifically to an intelligent replenishment method and system based on automotive inspection data. Background Technology

[0002] Automotive parts replenishment is a core part of automotive aftermarket supply chain management. Existing replenishment solutions generally rely solely on real-time operating parameters or fault codes collected by on-board diagnostic interfaces, neglecting existing data such as historical repair work orders that contain prior knowledge. Furthermore, the time bases of different data sources are inconsistent, making it difficult to comprehensively depict the status of parts. Existing solutions mostly use pure data-driven deep learning models to predict remaining lifetime. However, the reliability of these black-box models drops sharply under extreme conditions, and the results are not traceable, making it impossible to provide verifiable evidence for replenishment decisions. In terms of demand forecasting, existing technologies calculate replenishment demand for each component independently. When multiple components are triggered simultaneously, the total demand is likely to exceed the limits of funds or warehousing, requiring manual intervention for adjustment, which disrupts the automated closed loop. At the decision-making level, existing solutions execute replenishment strategies in one go without considering dynamic factors such as supplier capacity fluctuations and logistics delays. This leads to a disconnect between decision-making and actual execution, and a lack of adaptability. At the execution level, the heterogeneous systems and insufficient mutual trust among multiple participants in the supply chain, coupled with the risk of data tampering and difficulties in reconciliation among multiple parties in centralized orders, restrict the rapid and reliable execution of replenishment instructions. In summary, existing technologies urgently need an intelligent replenishment solution that can achieve multi-source data fusion, physical mechanism and data-driven collaborative prediction, global collaborative game, dynamic adaptive response and reliable automated execution. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent replenishment method and system based on vehicle inspection data to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent replenishment method based on vehicle inspection data, comprising the following steps: S1. Collect real-time operating parameters, fault self-diagnosis messages and historical maintenance work order data, and perform spatiotemporal alignment and fusion of the three types of data according to a unified time axis. The data format after alignment and fusion is a multidimensional detection data tensor. S2. Receive multidimensional detection data tensors, construct digital twins for each component, drive simulation with physical degradation mechanism model, quantify uncertainty with Monte Carlo, and output degradation trajectory and remaining life probability distribution of each component. S3. Encapsulate the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment needs. S4. Obtain the set of replenishment needs, and use the reinforcement learning model to output the optimal replenishment action based on real-time supply and logistics status. After the smart contract verifies the business rules, it automatically generates electronic orders and broadcasts them on the blockchain.

[0005] Preferably, step S2 specifically includes the following steps: S201. Obtain any vehicle from the target vehicle set. ,Sure Internal component collection ,in , This indicates the total number of parts on the vehicle. Indicates the first The component is selected. Components Construct its corresponding digital twin instance. ,in , This indicates a static property library. Represents a dynamic state library. Representing environmental exposure profiles, derived from multidimensional detection data tensors. In China, by component dimension Perform slicing to identify components Detection feature sequences across all mileage scales ,in , Indicates components In the At each mileage mark 3D feature vectors Indicates the total length of the mileage sequence. Represents the set of real numbers. This represents the feature dimension of each component. Indicates the serial number; S202, will Each element is injected according to its data category. Dynamic state library and environmental exposure archives Complete the state synchronization update of the digital twin; S203. Start the virtual degradation simulation engine for the digital twin instance. Perform independent evolution simulation, with pre-defined components in the simulation engine. Physical degradation mechanism model of this type This physical degradation mechanism model The model is retrieved from the pre-built model library by component type code, and the simulation step size is set. Total number of simulation steps ,in , This indicates the pre-set upper limit of the total simulated mileage. This represents the floor function; S204, with components Current cumulative mileage Starting with a dynamic state library The current state value in the file is the initial condition, obtained through the environment exposure archive. The historical environmental sequence and future operating condition prediction sequence are used as driving inputs, and the physical degradation mechanism model is followed. The cumulative damage is calculated step by step in the simulation.

[0006] Preferably, step S2 further includes the following steps: S205, Statistical Components In the Cumulative damage at the end of each simulation step ,in , Indicates the simulation step number. Indicates the first The cumulative damage at the end of each simulation step. Indicates the first Damage increment within a simulation step; S206. Set the set of uncertainty parameters ,in , Indicates components The One uncertain parameter, This represents the total number of uncertain parameters. This represents the sequence number. The distribution type and statistical parameters are obtained through maximum likelihood estimation using historical replacement data of the same model of parts for a specified vehicle. A preset probability distribution is then determined based on the distribution type and statistical parameters. Each uncertain parameter... It follows a pre-defined probability distribution; S207, Regarding the set of uncertainty parameters conduct Independent random sampling. This represents the preset number of Monte Carlo samplings, yielding... Group sampling results ,in Indicates the first The parameter combination obtained from the sampling; S208, Utilize each set of parameter combinations Repeatedly execute the virtual degradation simulation to obtain the first... Components under degradation trajectory Failure Mileage ,in , Indicates components Mileage at the current moment, Indicates the simulation step number. Indicates a fixed simulation step size. Indicates the cumulative damage amount. Indicates the failure threshold. This represents the total number of simulation steps. If the total simulation mileage is within the range of... None of them exceeded the failure threshold. Then mark the simulation result. No invalidity occurred; S209, Failure mileage results for all simulation trajectories Construct a sample set Determine the current time Lower parts Remaining life random variable For any given remaining lifetime threshold Generate the corresponding probability distribution, and perform kernel density estimation on the probability distribution to obtain the parts. The remaining lifetime probability density function, and output The probability of failure within a future preset mileage window.

[0007] Preferably, step S3 specifically includes the following steps: S301. Each generated digital twin of a component is encapsulated as an independent intelligent agent. Each intelligent agent has the ability to perceive its own degradation state, submit replenishment proposals, and respond to global constraints. S302. Each agent calculates its individual replenishment needs according to its own remaining lifetime probability distribution. When the failure probability of a component within a future preset mileage window exceeds the trigger threshold, the agent generates an initial replenishment proposal, which includes the suggested replenishment quantity and the expected delivery time window. S303. After all agents submit their initial replenishment proposals, the proposals are aggregated and checked to see if they meet the global constraints. The global constraints include the upper limit of the replenishment budget, the upper limit of the storage capacity, and the maximum supply of each supplier. If all are met, the results are output directly. If there are any violations, a multi-round negotiation game is initiated.

[0008] Preferably, step S3 further includes the following steps: S304. In each round of negotiation, calculate the concession coefficient of each agent. The coefficient is determined by two factors: the safety criticality level of the component and the current urgency of loss. The current urgency of loss is specifically the ratio of the average remaining lifespan to the supplier's delivery cycle. The smaller the ratio, the more urgent the need for replenishment. Agents with larger concession coefficients enjoy higher retention priority and bear less reduction in the game. S305. Each agent iteratively adjusts its replenishment quantity and expected time window based on its own concession coefficient and the current global over-provision ratio. Agents with lower concession coefficients prioritize reducing quantity and delaying delivery time. After each round of adjustment, the global constraints are re-checked until any agent unilaterally changes its own proposal and cannot improve its own goal without harming the interests of other agents. At this point, Nash equilibrium is determined to be reached. The set of replenishment requirements output at this time is the collaborative optimization result under global constraints.

[0009] Preferably, step S4 specifically includes the following steps: S401. Map replenishment demand to reinforcement learning decision states. The decision states include the inventory level of each component, the estimated arrival time of orders in transit, the dynamic supply capacity of each supplier, and real-time logistics status information. The dynamic supply capacity of each supplier is obtained in real time from the supplier through an API interface, including the current available capacity and promised delivery rate. The real-time logistics status is obtained by accessing a third-party logistics platform to obtain the transportation time and abnormal warning information. S402. Input the current decision state into the pre-trained deep reinforcement learning decision model. The model aims to minimize the total supply chain cost. The total cost comprehensively considers inventory holding costs, stockout loss costs, and emergency restocking premium costs. The model outputs the optimal replenishment action, which is the actual replenishment quantity of each component and the suggested order placement time.

[0010] Preferably, step S4 further includes the following steps: S403. The replenishment action output by the decision model is submitted to a pre-deployed smart contract on the blockchain in a structured data format. The smart contract is used to verify whether the decision meets the preset business rules. If it does, the electronic purchase order template is automatically filled in, the digital signatures of each participant are called to complete the signing, and a legally valid electronic purchase order is generated. Otherwise, automatic correction is executed in sequence. S404. The generated order is automatically broadcast to the supplier, logistics provider and relevant internal approval nodes through the blockchain network by the smart contract. All participants update the order execution status in real time on the unified distributed ledger, including order confirmation, shipment, in transit, receipt and settlement, realizing the full-link automated closed loop and reliable traceability of replenishment instructions from generation to fulfillment.

[0011] The intelligent replenishment system based on vehicle inspection data includes a data acquisition unit, a simulation driving unit, a negotiation replenishment unit, and an order output unit. The data acquisition unit extracts real-time operating parameters, fault self-diagnosis messages and historical maintenance work order data, and performs spatiotemporal alignment and fusion of the three types of data according to a unified time axis. The data format after alignment and fusion is a multidimensional detection data tensor. The simulation driving unit receives multidimensional detection data tensors, constructs digital twins for each component, drives the simulation with a physical degradation mechanism model, quantifies uncertainty using Monte Carlo, and outputs the degradation trajectory and remaining life probability distribution of each component. The negotiated replenishment unit encapsulates the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment demands. The order output unit obtains the replenishment demand set, and the reinforcement learning model combines real-time supply and logistics status to output the optimal replenishment action. After the smart contract verifies the business rules, it automatically generates an electronic order and broadcasts it on the blockchain.

[0012] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention combines multi-source data fusion with digital twins, realizing a leap from data-driven to mechanism-enhanced prediction. It aligns multi-source heterogeneous data with driving mileage as a unified benchmark, providing standardized input that conforms to the laws of physical degradation. The digital twin integrates real-time data and physical mechanism models, and still has extrapolation capabilities under extreme conditions that cannot be covered by pure data models, overcoming the inexplicable defects of traditional black box models. 2. This invention connects multi-agent negotiation game with reinforcement learning, realizing the evolution of replenishment from independent calculation to global collaboration. It takes into account the probability distribution of the lifetime of each component and achieves Nash equilibrium through autonomous game under global capital and storage constraints. It solves the resource competition conflict problem when multiple components trigger replenishment at the same time. After receiving the collaborative request, it combines the real-time dynamic optimization of the supply chain to optimize the specific replenishment action, overcoming the shortcomings of traditional static planning that cannot respond to environmental changes. 3. This invention organically couples the various stages, realizing a comprehensive upgrade of the automotive parts supply chain from passive response to proactive prediction, from independent decision-making to collaborative game theory, from static planning to dynamic adaptation, and from manual execution to reliable automation. This significantly reduces the risks of inventory capital occupation and stockouts, and improves the overall response efficiency and intelligence level of the supply chain. Attached Figure Description

[0013] Figure 1 An overall method flowchart is provided for embodiments of the present invention. Detailed Implementation

[0014] 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.

[0015] Example 1:

[0016] Please see Figure 1 This invention provides a technical solution: an intelligent replenishment method based on vehicle inspection data, comprising the following steps: S1. Collect real-time operating parameters, fault self-diagnosis messages and historical maintenance work order data, and perform spatiotemporal alignment and fusion of the three types of data according to a unified time axis. The data format after alignment and fusion is a multidimensional detection data tensor. S2. Receive multidimensional detection data tensors, construct digital twins for each component, drive simulation with physical degradation mechanism model, quantify uncertainty with Monte Carlo, and output degradation trajectory and remaining life probability distribution of each component. S3. Encapsulate the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment needs. S4. Obtain the set of replenishment needs, and use the reinforcement learning model to output the optimal replenishment action based on real-time supply and logistics status. After the smart contract verifies the business rules, it automatically generates electronic orders and broadcasts them on the blockchain.

[0017] S2 specifically includes the following steps: S201. Obtain any vehicle from the target vehicle set. ,Sure Internal component collection ,in , This indicates the total number of parts on the vehicle. Indicates the first The component is selected. Components Construct its corresponding digital twin instance. ,in , This indicates a static property library. Represents a dynamic state library. Representing environmental exposure profiles, derived from multidimensional detection data tensors. In China, by component dimension Perform slicing to identify components Detection feature sequences across all mileage scales ,in , Indicates components In the At each mileage mark 3D feature vectors Indicates the total length of the mileage sequence. Represents the set of real numbers. This represents the feature dimension of each component. Indicates the serial number; S202, will Each element is injected according to its data category. Dynamic state library and environmental exposure archives Complete the state synchronization update of the digital twin; S203. Start the virtual degradation simulation engine for the digital twin instance. Perform independent evolution simulation, with pre-defined components in the simulation engine. Physical degradation mechanism model of this type This physical degradation mechanism model The model is retrieved from a pre-built model library by component type. This library includes, but is not limited to, fatigue cumulative damage models, Arcard wear models, Arrhenius aging models, and Lundberg-Palmgren life models. The simulation step size is then set. Total number of simulation steps ,in , This indicates the pre-set upper limit of the total simulated mileage. This represents the floor function; S204, with components Current cumulative mileage Starting with a dynamic state library The current state value in the file is the initial condition, obtained through the environment exposure archive. The historical environmental sequence and future operating condition prediction sequence are used as driving inputs, and the physical degradation mechanism model is followed. Calculate the cumulative damage amount step by step in the simulation. S2 further includes the following steps: S205, Statistical Components In the Cumulative damage at the end of each simulation step ,in , Indicates the simulation step number. Indicates the first The cumulative damage at the end of each simulation step. Indicates the first Damage increment within a simulation step; S206. Set the set of uncertainty parameters ,in , Indicates components The One uncertain parameter, This represents the total number of uncertain parameters. This represents the sequence number. The distribution type and statistical parameters are obtained through maximum likelihood estimation using historical replacement data of the same model of parts for a specified vehicle. A preset probability distribution is then determined based on the distribution type and statistical parameters. Each uncertain parameter... It follows a pre-defined probability distribution; S207, Regarding the set of uncertainty parameters conduct Independent random sampling. This represents the preset number of Monte Carlo samplings, yielding... Group sampling results ,in Indicates the first The parameter combination obtained from the sampling; S208, Utilize each set of parameter combinations Repeatedly execute the virtual degradation simulation to obtain the first... Components under degradation trajectory Failure Mileage ,in , Indicates components Mileage at the current moment, Indicates the simulation step number. Indicates a fixed simulation step size. Indicates the cumulative damage amount. Indicates the failure threshold. This represents the total number of simulation steps. If the total simulation mileage is within the range of... None of them exceeded the failure threshold. Then mark the simulation result. No invalidity occurred; S209, Failure mileage results for all simulation trajectories Construct a sample set Determine the current time Lower parts Remaining life random variable For any given remaining lifetime threshold Generate the corresponding probability distribution, and perform kernel density estimation on the probability distribution to obtain the parts. The remaining lifetime probability density function, and output The probability distribution of failure within a future preset mileage window is as follows:

[0018] in, Represents the characteristic function, Indicates the serial number; S3 specifically includes the following steps: S301. Each generated digital twin of a component is encapsulated as an independent intelligent agent. Each intelligent agent has the ability to perceive its own degradation state, submit replenishment proposals, and respond to global constraints. S302. Each agent calculates its individual replenishment needs according to its own remaining lifetime probability distribution. When the failure probability of a component within a future preset mileage window exceeds the trigger threshold, the agent generates an initial replenishment proposal, which includes the suggested replenishment quantity and the expected delivery time window. The principle for determining the expected delivery time window is that the delivery time is no later than the time corresponding to a specific quantile in the remaining lifetime distribution, so as to ensure that the replacement is completed before the component fails. S303. After all agents submit their initial replenishment proposals, the global constraints are aggregated and checked to see if they are met. The global constraints include the upper limit of the replenishment budget, the upper limit of the storage capacity, and the maximum supply of each supplier. If all are met, the result is output directly. If there is a violation, a multi-round negotiation game is initiated. S3 also includes the following steps: S304. In each round of negotiation, calculate the concession coefficient of each agent. The coefficient is determined by two factors: the safety criticality level of the component and the current urgency of loss. The current urgency of loss is specifically the ratio of the average remaining lifespan to the supplier's delivery cycle. The smaller the ratio, the more urgent the need for replenishment. Agents with larger concession coefficients enjoy higher retention priority and bear less reduction in the game. S305. Each agent iteratively adjusts its replenishment quantity and expected time window based on its own concession coefficient and the current global over-provision ratio. Agents with lower concession coefficients prioritize reducing quantity and delaying delivery time. After each round of adjustment, the global constraints are re-checked until any agent unilaterally changes its own proposal and cannot improve its own goal without harming the interests of other agents. At this point, Nash equilibrium is determined to be reached. The set of replenishment requirements output at this time is the collaborative optimization result under global constraints. S4 specifically includes the following steps: S401. Replenishment demand is mapped to a reinforcement learning decision state. The decision state includes the inventory level of each component, the estimated arrival time of orders in transit, the dynamic supply capacity of each supplier, and real-time logistics status information. The dynamic supply capacity of each supplier is obtained in real time from the supplier through the API interface, including the current available capacity and promised delivery rate. The real-time logistics status is obtained by connecting to a third-party logistics platform to obtain the transportation time and abnormal warning information. S402. Input the current decision state into a pre-trained deep reinforcement learning decision model. This model aims to minimize the total supply chain cost, which comprehensively considers inventory holding costs, stockout loss costs, and emergency restocking premium costs. The model outputs the optimal replenishment action, namely the actual replenishment quantity and suggested order placement time for each component. The deep reinforcement learning decision model is specifically as follows: S4021. Construct a reinforcement learning environment that includes a state space, an action space, and a reward function; the state space includes at least the current inventory status, the status of orders in transit, the dynamic supply capacity of each supplier, and the real-time logistics status; the action space is the combination decision of the replenishment quantity and replenishment time of each component; the reward function is set as the weighted negative sum of inventory holding cost, stockout loss cost, and emergency restocking cost, with the goal of minimizing the total supply chain cost. S4022. The reinforcement learning model is trained using a near-end policy optimization algorithm, and the training data comes from historical replenishment logs and corresponding actual cost records. S4023. When making a decision, input the current state into the trained deep reinforcement learning model and output the replenishment action that maximizes the expected cumulative reward. S4 also includes the following steps: S403. The replenishment action output by the decision model is submitted to a pre-deployed smart contract on the blockchain in a structured data format. The smart contract is used to verify whether the decision meets the preset business rules. If it does, the electronic purchase order template is automatically filled in, the digital signatures of each participant are called to complete the signing, and a legally valid electronic purchase order is generated. Otherwise, automatic correction is executed in sequence. S404. The generated order is automatically broadcast to the supplier, logistics provider and relevant internal approval nodes through the blockchain network by the smart contract. All participants update the order execution status in real time on the unified distributed ledger, including order confirmation, shipment, in transit, receipt and settlement, realizing the full-link automated closed loop and reliable traceability of replenishment instructions from generation to fulfillment.

[0019] Example 2:

[0020] The present invention also provides an intelligent replenishment system based on vehicle inspection data, including a data acquisition unit, a simulation driving unit, a negotiation replenishment unit, and an order output unit; The data acquisition unit extracts real-time operating parameters, fault self-diagnosis messages, and historical maintenance work order data. These three types of data are then spatiotemporally aligned and fused according to a unified timeline. The fused data format is a multidimensional detection data tensor. The real-time operating parameters include analog and digital signals for engine speed, coolant temperature, fuel pressure, brake pad wear voltage, battery terminal voltage, and transmission output shaft speed. The fault self-diagnosis messages are specifically compiled by each ECU according to SAE J1979 and ISO standards. The diagnostic fault codes output by the 15765 protocol each include a DTC code and freeze frame data at the time of the fault occurrence. The historical repair work order data includes parts replacement records, replacement mileage, repair operation codes, and repair notes. The spatiotemporal alignment and fusion specifically uses the vehicle's mileage as a unified time axis benchmark. Real-time operating parameters are mapped to mileage coordinates according to the sampling time. The fault occurrence mileage field carried in the fault self-diagnosis message is directly mapped to mileage coordinates. Repair records in the historical repair work order data are mapped to mileage coordinates according to the repair mileage. For data with missing mileage markers, linear interpolation is used to complete the mileage coordinates. The fused multidimensional detection data tensor format is as follows: ,in Indicates the length of the mileage sequence. Indicates the number of parts on the target vehicle. This represents the feature dimension of each component; The simulation-driven unit receives multidimensional detection data tensors, constructs digital twins for each component, drives the simulation with a physical degradation mechanism model, quantifies uncertainties using Monte Carlo methods, and outputs the degradation trajectory and remaining life probability distribution of each component. The negotiation replenishment unit encapsulates the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment needs. The order output unit obtains the replenishment demand set, and the reinforcement learning model combines real-time supply and logistics status to output the optimal replenishment action. After the smart contract verifies the business rules, it automatically generates electronic orders and broadcasts them on the blockchain.

[0021] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0022] 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. An intelligent replenishment method based on vehicle inspection data, characterized in that, The method includes the following steps: S1. Collect real-time operating parameters, fault self-diagnosis messages and historical maintenance work order data, and perform spatiotemporal alignment and fusion of the three types of data according to a unified time axis. The data format after alignment and fusion is a multidimensional detection data tensor. S2. Receive multidimensional detection data tensors, construct digital twins for each component, drive simulation with physical degradation mechanism model, quantify uncertainty with Monte Carlo, and output degradation trajectory and remaining life probability distribution of each component. S3. Encapsulate the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment needs. S4. Obtain the set of replenishment needs, and use the reinforcement learning model to output the optimal replenishment action based on real-time supply and logistics status. After the smart contract verifies the business rules, it automatically generates electronic orders and broadcasts them on the blockchain.

2. The intelligent replenishment method based on vehicle inspection data according to claim 1, characterized in that: S2 specifically includes the following steps: S201. Obtain any vehicle from the target vehicle set. ,Sure Internal component collection ,in , This indicates the total number of parts on the vehicle. Indicates the first The component is selected. Components Construct its corresponding digital twin instance. ,in , This indicates a static property library. Represents a dynamic state library. Representing environmental exposure profiles, derived from multidimensional detection data tensors. In China, by component dimension Perform slicing to identify components Detection feature sequences across all mileage scales ,in , Indicates components In the At each mileage mark 3D feature vectors Indicates the total length of the mileage sequence. Represents the set of real numbers. This represents the feature dimension of each component. Indicates the serial number; S202, will Each element is injected according to its data category. Dynamic state library and environmental exposure archives Complete the state synchronization update of the digital twin; S203. Start the virtual degradation simulation engine for the digital twin instance. Perform independent evolution simulation, with pre-defined components in the simulation engine. Physical degradation mechanism model of this type This physical degradation mechanism model The model is retrieved from the pre-built model library by component type code, and the simulation step size is set. Total number of simulation steps ,in , This indicates the pre-set upper limit of the total simulated mileage. This represents the floor function; S204, with components Current cumulative mileage Starting with a dynamic state library The current state value in the file is the initial condition, obtained through the environment exposure archive. The historical environmental sequence and future operating condition prediction sequence are used as driving inputs, and the physical degradation mechanism model is followed. The cumulative damage is calculated step by step in the simulation.

3. The intelligent replenishment method based on vehicle inspection data according to claim 2, characterized in that: S2 further includes the following steps: S205, Statistical Components In the Cumulative damage at the end of each simulation step ,in , Indicates the simulation step number. Indicates the first The cumulative damage at the end of each simulation step. Indicates the first Damage increment within a simulation step; S206. Set the set of uncertainty parameters ,in , Indicates components The One uncertain parameter, This represents the total number of uncertain parameters. This represents the sequence number. The distribution type and statistical parameters are obtained through maximum likelihood estimation using historical replacement data of the same model of parts for a specified vehicle. A preset probability distribution is then determined based on the distribution type and statistical parameters. Each uncertain parameter... It follows a pre-defined probability distribution; S207, Regarding the set of uncertainty parameters conduct Independent random sampling. This represents the preset number of Monte Carlo samplings, yielding... Group sampling results ,in Indicates the first The parameter combination obtained from the sampling; S208, Utilizing each set of parameter combinations Repeatedly execute the virtual degradation simulation to obtain the first... Components under degradation trajectory Failure Mileage ,in , Indicates components Mileage at the current moment, Indicates the simulation step number. Indicates a fixed simulation step size. Indicates the cumulative damage amount. Indicates the failure threshold. This represents the total number of simulation steps. If the total simulation mileage is within the range of... None of them exceeded the failure threshold. Then mark the simulation result. No invalidation occurred; S209, Failure mileage results for all simulation trajectories Construct a sample set Determine the current time Lower parts Remaining life random variable For any given remaining lifetime threshold Generate the corresponding probability distribution, and perform kernel density estimation on the probability distribution to obtain the parts. The remaining lifetime probability density function, and output The probability of failure within a future preset mileage window.

4. The intelligent replenishment method based on vehicle inspection data according to claim 1, characterized in that: S3 specifically includes the following steps: S301. Each generated digital twin of a component is encapsulated as an independent intelligent agent. Each intelligent agent has the ability to perceive its own degradation state, submit replenishment proposals, and respond to global constraints. S302. Each agent calculates its individual replenishment needs according to its own remaining lifetime probability distribution. When the failure probability of a component within a future preset mileage window exceeds the trigger threshold, the agent generates an initial replenishment proposal, which includes the suggested replenishment quantity and the expected delivery time window. S303. After all agents submit their initial replenishment proposals, the proposals are aggregated and checked to see if they meet the global constraints. The global constraints include the upper limit of the replenishment budget, the upper limit of the storage capacity, and the maximum supply of each supplier. If all are met, the results are output directly. If there are any violations, a multi-round negotiation game is initiated.

5. The intelligent replenishment method based on vehicle inspection data according to claim 4, characterized in that: S3 further includes the following steps: S304. In each round of negotiation, calculate the concession coefficient of each agent. The coefficient is determined by two factors: the safety criticality level of the component and the current urgency of loss. The current urgency of loss is specifically the ratio of the average remaining lifespan to the supplier's delivery cycle. The smaller the ratio, the more urgent the need for replenishment. Agents with larger concession coefficients enjoy higher retention priority and bear less reduction in the game. S305. Each agent iteratively adjusts its replenishment quantity and expected time window based on its own concession coefficient and the current global over-provision ratio. Agents with lower concession coefficients prioritize reducing quantity and delaying delivery time. After each round of adjustment, the global constraints are re-checked until any agent unilaterally changes its own proposal and cannot improve its own goal without harming the interests of other agents. At this point, Nash equilibrium is determined to be reached. The set of replenishment requirements output at this time is the collaborative optimization result under global constraints.

6. The intelligent replenishment method based on vehicle inspection data according to claim 1, characterized in that: S4 specifically includes the following steps: S401. Map replenishment demand to reinforcement learning decision states. The decision states include the inventory level of each component, the estimated arrival time of orders in transit, the dynamic supply capacity of each supplier, and real-time logistics status information. The dynamic supply capacity of each supplier is obtained in real time from the supplier through an API interface, including the current available capacity and promised delivery rate. The real-time logistics status is obtained by accessing a third-party logistics platform to obtain the transportation time and abnormal warning information. S402. Input the current decision state into the pre-trained deep reinforcement learning decision model. The model aims to minimize the total supply chain cost. The total cost comprehensively considers inventory holding costs, stockout loss costs, and emergency restocking premium costs. The model outputs the optimal replenishment action, which is the actual replenishment quantity of each component and the suggested order placement time.

7. The intelligent replenishment method based on vehicle inspection data according to claim 6, characterized in that: S4 further includes the following steps: S403. The replenishment action output by the decision model is submitted to a pre-deployed smart contract on the blockchain in a structured data format. The smart contract is used to verify whether the decision meets the preset business rules. If it does, the electronic purchase order template is automatically filled in, the digital signatures of each participant are called to complete the signing, and a legally valid electronic purchase order is generated. Otherwise, automatic correction is executed in sequence. S404. The generated order is automatically broadcast to the supplier, logistics provider and relevant internal approval nodes through the blockchain network by the smart contract. All participants update the order execution status in real time on the unified distributed ledger, including order confirmation, shipment, in transit, receipt and settlement, realizing the full-link automated closed loop and reliable traceability of replenishment instructions from generation to fulfillment.

8. An intelligent replenishment system based on vehicle inspection data, characterized in that, The intelligent replenishment system is applicable to the intelligent replenishment method based on vehicle detection data as described in any one of claims 1-7, and includes a data acquisition unit, a simulation driving unit, a negotiation replenishment unit, and an order output unit; The data acquisition unit extracts real-time operating parameters, fault self-diagnosis messages and historical maintenance work order data, and performs spatiotemporal alignment and fusion of the three types of data according to a unified time axis. The data format after alignment and fusion is a multidimensional detection data tensor. The simulation driving unit receives multidimensional detection data tensors, constructs digital twins for each component, drives the simulation with a physical degradation mechanism model, quantifies uncertainty using Monte Carlo, and outputs the degradation trajectory and remaining life probability distribution of each component. The negotiated replenishment unit encapsulates the digital twin as an independent intelligent agent. After each intelligent agent submits a replenishment proposal, it negotiates in multiple rounds under global funding and warehousing constraints until a Nash equilibrium is reached, and outputs a set of collaborative replenishment demands. The order output unit obtains the replenishment demand set, and the reinforcement learning model combines real-time supply and logistics status to output the optimal replenishment action. After the smart contract verifies the business rules, it automatically generates an electronic order and broadcasts it on the blockchain.

Images