Data-driven adaptive optimization method and system for vehicle manufacturing supply chain

Abstract

The present invention discloses a vehicle manufacturing supply chain adaptive optimization method and system based on data-driven, which involves supply chain management related fields. The method includes: retrieving historical manufacturing full cycle data sets to obtain real-time manufacturing data sets; performing integrated learning, performing correlation analysis on supply chain nodes according to learning results, and constructing dynamic interaction simulation networks between chains; defining manufacturing constraints, activating dynamic interaction simulation networks between chains to perform demand forecasting, and performing production analysis according to forecasting results; simulating and executing production scheduling strategies to perform multi-objective optimization, generating labels to be optimized, performing dynamic monitoring, and updating production scheduling strategies; synchronizing the global optimization strategy of production scheduling to the vehicle manufacturing supply chain for verification, and performing adaptive optimization according to the verification results. The method solves the problems of lack of real-time dynamics, inability to respond to changes in a timely manner, and inaccurate scheduling in supply chain management, and achieves the effects of improving real-time dynamics, responding to changes in a timely manner, and improving scheduling accuracy.

Description

Technical Field

[0001] The present application relates to supply chain management-related fields, and in particular to a data-driven vehicle manufacturing supply chain adaptive optimization method and system. Background Art

[0002] In the vehicle manufacturing industry, supply chain optimization is a key factor in improving production efficiency, reducing costs, and enhancing market competitiveness. In the current vehicle manufacturing supply chain management, the system can only provide static data display and simple data analysis functions. For multiple nodes in the supply chain, there is a lack of real-time, dynamic correlation analysis and prediction capabilities. In addition, the existing production scheduling strategy is based on fixed rules and models, which is difficult to flexibly adjust according to real-time changing market demand and supply chain status.

[0003] Among the current relevant technologies, vehicle manufacturing supply chain management lacks real-time and dynamic features and is unable to respond promptly to market demand and changes in supply chain status, leading to technical problems such as inaccurate production scheduling. Summary of the invention

[0004] The present application provides a data-driven adaptive optimization method and system for a vehicle manufacturing supply chain. The method records data by traversing the vehicle manufacturing supply chain, retrieves historical manufacturing full-cycle data sets, and combines real-time sensor data to build a dynamic interactive simulation network between chains, defines manufacturing constraints, conducts demand forecasting and production analysis, formulates production scheduling strategies, performs multi-objective optimization by simulating the execution of production scheduling strategies, generates labels to be optimized, and performs dynamic monitoring and optimization to obtain a global optimization strategy for production scheduling. The vehicle manufacturing supply chain is adaptively optimized according to the verification results, and other technical means are used to achieve the technical effect of improving real-time and dynamic performance, responding to changes in market demand and supply chain status in a timely manner, and thereby improving the accuracy of production scheduling.

[0005] The present application provides a data-driven adaptive optimization method for a vehicle manufacturing supply chain, including: traversing a vehicle manufacturing supply chain to record data, retrieving a historical manufacturing full-cycle data set of a vehicle manufacturing supply chain, performing real-time sensing according to multiple supply chain nodes of the vehicle manufacturing supply chain, and obtaining a real-time manufacturing data set; performing integrated learning based on the historical manufacturing full-cycle data set and the real-time manufacturing data set, performing correlation analysis on multiple supply chain nodes according to the learning results, and constructing an inter-chain dynamic interaction simulation network; defining manufacturing constraints of a vehicle manufacturing supply chain, activating the inter-chain dynamic interaction simulation network to perform demand forecasting, performing production analysis according to the demand forecasting results, and formulating a production scheduling strategy; performing multi-objective optimization based on a vehicle manufacturing supply chain simulation and executing the production scheduling strategy, generating multiple labels to be optimized, dynamically monitoring the production scheduling strategy according to the multiple labels to be optimized, updating the production scheduling strategy, and obtaining a global optimization strategy for production scheduling; synchronizing the global optimization strategy for production scheduling to the vehicle manufacturing supply chain for verification, and adaptively optimizing the vehicle manufacturing supply chain according to the verification results; Among them, the manufacturing constraints of the vehicle manufacturing supply chain are defined, the inter-chain dynamic interactive simulation network is activated to perform demand forecasting, production analysis is performed based on the demand forecasting results, and a production scheduling strategy is formulated, including: manufacturing stability calculation is performed based on the vehicle manufacturing supply chain to obtain a manufacturing stability coefficient group, fluctuation extreme value analysis is performed based on the manufacturing stability coefficient group, and a stable upper limit value and a stable lower limit value are extracted; a stable numerical range is defined according to the stable upper limit value and the stable lower limit value, and the manufacturing constraints are constructed according to the numerical range; the inter-chain dynamic interactive simulation network is activated, and a plurality of supply chain nodes are traversed according to the multi-level dependency relationship to perform production scenario simulation to obtain a plurality of production scenario information; a demand analysis is performed on the vehicle manufacturing supply chain according to the manufacturing constraints combined with the plurality of production scenario information to generate a plurality of demand change trends; simulation prediction is performed based on the mapping of the plurality of demand change trends to the plurality of production scenario information to determine the demand forecasting results, a plurality of capacity bottleneck segments are identified according to the demand forecasting results, production analysis is performed according to the plurality of capacity bottleneck segments, and the production scheduling strategy is formulated.

[0006] In a possible implementation, integrated learning is performed based on the historical manufacturing full-cycle data set and the real-time manufacturing data set, and the following processing is performed: a timing analysis is performed on the vehicle manufacturing supply chain according to multiple supply chain nodes to determine the node timestamps; the historical manufacturing full-cycle data set and the real-time manufacturing data set are aligned according to the node timestamps to generate a data alignment matrix; feature analysis is performed based on the data alignment matrix to extract multiple eigenvalues, and the multiple eigenvalues ​​include historical eigenvalues ​​and real-time eigenvalues; the data alignment matrix is ​​randomly sampled based on the historical eigenvalues ​​to obtain a first sample group, and the data alignment matrix is ​​randomly sampled based on the real-time eigenvalues ​​to obtain a second sample group; integrated training is performed based on the first sample group and the second sample group to construct a gradient boosting machine, and tuning learning is performed through the gradient boosting machine to obtain the learning result.

[0007] In a possible implementation, integrated training is performed based on the first sample group and the second sample group to construct a gradient boosting machine, tuning learning is performed through the gradient boosting machine to obtain the learning result, and the following processing is performed: integrated learning is performed based on the first sample group to obtain multiple first weak classifiers, and integrated learning is performed based on the second sample group to obtain multiple second weak classifiers; a loss function is introduced to iteratively train the multiple first weak classifiers to obtain a first loss gradient value; the multiple second weak classifiers are iteratively trained according to the loss function to obtain a second loss gradient value; a weighted analysis is performed according to the first loss gradient value and the second loss gradient value according to the target gradient direction to construct the gradient boosting machine; the gradient boosting machine traverses the multiple first weak classifiers and the multiple second weak classifiers to perform contribution calculation to obtain multiple contribution coefficients, and the multiple contribution coefficients are added to the learning result.

[0008] In a possible implementation, association analysis is performed on multiple supply chain nodes based on the learning results, an inter-chain dynamic interaction simulation network is constructed, and the following processing is performed: multiple supply chain nodes of the vehicle manufacturing supply chain are traversed and matched according to the multiple contribution coefficients to generate a node matching array; multiple supply chain nodes are identified based on the node matching array to determine multiple node contribution level labels; causal relationship inference is performed on multiple supply chain nodes according to the multiple node contribution level labels, and an information flow diagram of the multiple supply chain nodes is drawn according to the inference results; interaction flow analysis is performed based on the information flow diagram to identify multi-level dependency relationships of multiple supply chain nodes, and multiple supply chain nodes are connected according to the multi-level dependency relationships to construct the inter-chain dynamic interaction simulation network of the vehicle manufacturing supply chain.

[0009] In a possible implementation, simulation prediction is performed based on the mapping of the multiple demand change trends to the multiple production scenario information to determine the demand forecast result, multiple capacity bottleneck segments are identified according to the demand forecast result, production analysis is performed according to the multiple capacity bottleneck segments, the production scheduling strategy is formulated, and the following processing is performed: based on the mapping of the multiple demand change trends to the multiple production scenario information, multiple supply chain nodes are traversed and collected to build a standard demand database; the standard demand database is indexed according to the time series of multiple supply chain nodes to obtain demand forecast results, and multiple demand change feature sets are extracted according to the demand forecast results, and the multiple demand change features include a short-term demand change feature set and a long-term demand change feature set; scheduling balance calculation is performed according to the short-term demand change feature set to obtain a short-term balance coefficient, and scheduling balance calculation is performed according to the long-term demand change feature set to obtain a long-term balance coefficient; the short-term balance coefficient and the long-term balance coefficient are jointly integrated, and the multiple capacity bottleneck segments are identified according to the integration result.

[0010] In a possible implementation, the production scheduling strategy is executed based on a vehicle manufacturing supply chain simulation to perform multi-objective optimization, generate multiple labels to be optimized, and perform the following processing: simulate the production scheduling execution of the vehicle manufacturing supply chain according to the production scheduling strategy, perform scheduling calculations based on the scheduling simulation execution results, and determine multiple scheduling indicators; perform multi-objective search on the production scheduling strategy according to the multiple scheduling indicators to determine a scheduling target set; perform scheduling state analysis based on the scheduling target set to construct a scheduling state space; perform scheduling action analysis based on the scheduling target set to construct a scheduling action space, introduce a reward function, and perform reinforcement learning on the scheduling target set in combination with the scheduling state space and the scheduling action space to generate a scheduling learning result; perform production risk assessment on the scheduling learning result according to the multiple production capacity bottleneck segments to generate multi-level production risk data, match and identify the production scheduling strategy based on the multi-level production risk data, and generate the multiple labels to be optimized.

[0011] The present application also provides a data-driven vehicle manufacturing supply chain adaptive optimization system, including: a data recording and sensing module, which is used to traverse the vehicle manufacturing supply chain for data recording, retrieve the historical manufacturing full-cycle data set of the vehicle manufacturing supply chain, perform real-time sensing according to multiple supply chain nodes of the vehicle manufacturing supply chain, and obtain a real-time manufacturing data set; an integrated association module, which is used to perform integrated learning based on the historical manufacturing full-cycle data set and the real-time manufacturing data set, perform association analysis on multiple supply chain nodes according to the learning results, and construct a dynamic interaction simulation network between chains; a production scheduling strategy formulation module, which is used to define manufacturing constraints of the vehicle manufacturing supply chain, activate the dynamic interaction simulation network between chains to predict demand, perform production analysis according to the demand prediction results, and formulate a production scheduling strategy; a production scheduling strategy update module, which is used to perform multi-objective optimization based on the vehicle manufacturing supply chain simulation execution of the production scheduling strategy, generate multiple labels to be optimized, dynamically monitor the production scheduling strategy according to the multiple labels to be optimized, update the production scheduling strategy, and obtain a global optimization strategy for production scheduling; a supply chain adaptive optimization module, which is used to synchronize the global optimization strategy for production scheduling to the vehicle manufacturing supply chain for verification, and adaptively optimize the vehicle manufacturing supply chain according to the verification results.

[0012] The data-driven vehicle manufacturing supply chain adaptive optimization method and system proposed in this application first traverses the vehicle manufacturing supply chain to record data, retrieves the historical manufacturing full cycle data set of the vehicle manufacturing supply chain, performs real-time sensing according to multiple supply chain nodes of the vehicle manufacturing supply chain, obtains the real-time manufacturing data set, then performs integrated learning based on the historical manufacturing full cycle data set and the real-time manufacturing data set, performs correlation analysis on multiple supply chain nodes according to the learning results, builds a dynamic interaction simulation network between chains, and then defines the manufacturing constraints of the vehicle manufacturing supply chain, activates the dynamic interaction simulation network between chains to predict demand, performs production analysis according to the demand prediction results, formulates a production scheduling strategy, and then simulates and executes the production scheduling strategy based on the vehicle manufacturing supply chain for multi-objective optimization, generates multiple tags to be optimized, dynamically monitors the production scheduling strategy according to multiple tags to be optimized, updates the production scheduling strategy, obtains the global optimization strategy of production scheduling, and finally synchronizes the global optimization strategy of production scheduling to the vehicle manufacturing supply chain for verification, and adaptively optimizes the vehicle manufacturing supply chain according to the verification results. The technical effect of improving real-time and dynamic performance, responding to changes in market demand and supply chain status in a timely manner, and thus improving the accuracy of production scheduling is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0014] Figure 1 A schematic flow chart of a data-driven vehicle manufacturing supply chain adaptive optimization method provided in an embodiment of the present application.

[0015] Figure 2 A schematic diagram of the structure of a data-driven vehicle manufacturing supply chain adaptive optimization system provided in an embodiment of the present application.

[0016] Explanation of the reference numerals: data recording and sensing module 10 , integration and association module 20 , production scheduling strategy formulation module 30 , production scheduling strategy updating module 40 , supply chain adaptive optimization module 50 . DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] The present application embodiment provides a data-driven vehicle manufacturing supply chain adaptive optimization method, such as Figure 1 As shown, the method includes:

[0021] Step S100, traverse the vehicle manufacturing supply chain to record data, retrieve the historical manufacturing full cycle data set of the vehicle manufacturing supply chain, perform real-time sensing based on multiple supply chain nodes of the vehicle manufacturing supply chain, and obtain a real-time manufacturing data set.

[0022] Specifically, the entire vehicle manufacturing supply chain is combed in detail, including raw material suppliers, parts manufacturers, assembly plants, logistics and transportation, sales networks and other links. The historical manufacturing full cycle data sets of the vehicle manufacturing supply chain are retrieved, which include historical production volume, inventory volume, order volume, transportation time, production cost, etc. These data are stored in a database or data warehouse and can be obtained through SQL queries or other data extraction tools. Sensors or data acquisition devices are installed at multiple key nodes of the supply chain (such as production lines, warehouses, transport vehicles, etc.) to collect data in the manufacturing process in real time, such as production progress, inventory status, transportation location, etc. These data are transmitted to the data center for processing through the Internet of Things (IoT) technology. Among them, the historical manufacturing full cycle data set refers to all relevant data generated by the vehicle manufacturing supply chain in the past period of time, including historical records of production, inventory, sales, etc. The real-time manufacturing data set is data in the supply chain collected in real time using sensors or data acquisition devices for real-time monitoring and analysis.

[0023] Step S200, performing integrated learning based on the historical manufacturing full-cycle data set and the real-time manufacturing data set, performing association analysis on multiple supply chain nodes according to the learning results, and constructing an inter-chain dynamic interaction simulation network.

[0024] Specifically, the historical manufacturing full cycle data set is merged with the real-time manufacturing data set, and machine learning algorithms (such as gradient boosting trees) are used for ensemble learning to extract patterns and regularities in the data. Ensemble learning is a machine learning paradigm that improves prediction performance by training multiple models and combining their results. Based on the learning results, multiple nodes in the supply chain are analyzed for association through network analysis, cluster analysis and other methods to find out the mutual influence relationship between them. Using the results of the association analysis, a simulation network that can simulate the dynamic interaction between each node in the supply chain is constructed, namely the inter-chain dynamic interaction simulation network. This network is used to predict and analyze the behavior of the supply chain, and can be used to predict bottleneck nodes or shortage risks that may occur in the supply chain.

[0025] In a possible implementation, based on the integrated learning of the historical manufacturing full cycle data set and the real-time manufacturing data set, step S200 further includes step S210, performing a time series analysis on the vehicle manufacturing supply chain according to multiple supply chain nodes to determine the node timestamp. Specifically, historical operation data of supply chain nodes are collected, including the operation start time, end time, operation cycle, etc. These data are analyzed to determine the typical operation time pattern of each node. A unique timestamp is assigned to each node to identify the operation time point or cycle of the node.

[0026] Step S220, align the historical manufacturing full cycle data set with the real-time manufacturing data set according to the node timestamp to generate a data alignment matrix. Specifically, according to the node timestamp determined in step S210, the historical manufacturing full cycle data set and the real-time manufacturing data set are time synchronized. For missing data points, interpolation, mean filling or other appropriate methods are used for processing. The processed data is organized into a two-dimensional matrix according to the nodes and timestamps, namely the data alignment matrix, in which the rows represent the supply chain nodes, the columns represent the timestamps, and the matrix elements are the data values ​​of the corresponding nodes at the corresponding time points.

[0027] Step S230, feature analysis is performed based on the data alignment matrix to extract multiple eigenvalues, wherein the multiple eigenvalues ​​include historical eigenvalues ​​and real-time eigenvalues. Specifically, a statistical description is performed on the data alignment matrix, and statistics such as the mean, variance, maximum value, and minimum value of each node are calculated as historical eigenvalues. For real-time manufacturing data sets, data within the current time window is extracted as real-time eigenvalues, such as current inventory, production progress, etc.

[0028] Step S240, randomly sampling the data alignment matrix based on the historical eigenvalues ​​to obtain a first sample group, and randomly sampling the data alignment matrix based on the real-time eigenvalues ​​to obtain a second sample group. Specifically, randomly sampling the historical eigenvalues ​​ensures that each sample contains historical data of multiple nodes to form the first sample group. Randomly sampling the real-time eigenvalues ​​ensures that each sample contains real-time data within the current time window to form the second sample group.

[0029] Step S250, perform integrated training according to the first sample group and the second sample group, construct a gradient boosting machine, perform tuning learning through the gradient boosting machine, and obtain the learning result. Specifically, use the first sample group to perform initial training on the gradient boosting machine to obtain a preliminary model. Use the second sample group to verify and adjust the preliminary model, and perform tuning learning through methods such as cross-validation and grid search. During the tuning process, the parameters and structure of the gradient boosting machine are continuously iterated and updated until the predetermined performance indicators or convergence conditions are reached. The final learning results are used for subsequent association analysis and the construction of a dynamic interaction simulation network between chains. This implementation method ensures the consistency and validity of the data by aligning and feature analyzing the historical manufacturing full-cycle data set and the real-time manufacturing data set, obtains diversified training samples through random sampling, and combines the gradient boosting machine integrated learning method to improve the generalization ability and prediction accuracy of the model.

[0030] In a possible implementation, integrated training is performed according to the first sample group and the second sample group, a gradient boosting machine is constructed, and tuning learning is performed through the gradient boosting machine to obtain the learning result. Step S250 further includes step S251, based on the first sample group, integrated learning is performed to obtain multiple first weak classifiers, and based on the second sample group, integrated learning is performed to obtain multiple second weak classifiers. Specifically, using methods such as AdaBoost, integrated learning is performed using the first sample group to train multiple first weak classifiers. Similarly, integrated learning is performed using the second sample group to train multiple second weak classifiers, and the types, quantities and training methods of these weak classifiers are the same as those of the first sample group, but the input data is different. Among them, the weak classifier is a classifier whose performance is slightly better than random guessing, and is a base learner in integrated learning.

[0031] Step S252, introducing a loss function to iteratively train the multiple first weak classifiers to obtain a first loss gradient value; Step S253, iteratively training the multiple second weak classifiers according to the loss function to obtain a second loss gradient value.

[0032] Specifically, for multiple first weak classifiers, a loss function (such as mean square error) is introduced for iterative training. In each iteration, the difference between the prediction result of the current weak classifier and the actual value is calculated, and the parameters of the weak classifier are updated according to the gradient information of the loss function. Through multiple iterations, the first loss gradient value is obtained, which reflects the contribution of the historical feature value to the prediction result. Similarly, for multiple second weak classifiers, iterative training is also performed to obtain the second loss gradient value, which reflects the contribution of the real-time feature value to the prediction result.

[0033] Step S254, weighted analysis is performed according to the first loss gradient value and the second loss gradient value in the target gradient direction to construct the gradient boosting machine. Specifically, the weight of each weak classifier is calculated according to the first loss gradient value and the second loss gradient value. The size of the weight reflects the contribution of the weak classifier to the overall prediction result. Multiple weighted weak classifiers are combined to construct a gradient boosting machine. The gradient boosting machine is an iterative decision tree algorithm that corrects the residual of the previous model by continuously adding new decision trees. During prediction, the gradient boosting machine will make a comprehensive judgment based on the weight of each weak classifier and the prediction result to give the final prediction value.

[0034] Step S255, according to the gradient boosting machine, traverse the multiple first weak classifiers and the multiple second weak classifiers to perform contribution calculation, obtain multiple contribution coefficients, and add the multiple contribution coefficients to the learning result. Specifically, traverse multiple first weak classifiers and second weak classifiers, and calculate the contribution coefficient of each weak classifier according to their weights and prediction results in the gradient boosting machine. The contribution coefficient is an indicator reflecting the importance of each weak classifier in the overall prediction result. The calculated contribution coefficient is added to the learning result as a basis for subsequent analysis, optimization and decision-making. This implementation method can obtain a more powerful gradient boosting machine model by constructing multiple weak classifiers and performing iterative training. This model can more accurately predict bottleneck nodes or shortage risks in the supply chain.

[0035] In a possible implementation, multiple supply chain nodes are subjected to association analysis based on the learning results, and a dynamic interaction simulation network between chains is constructed. Step S200 further includes step S260, in which multiple supply chain nodes of the vehicle manufacturing supply chain are traversed and matched according to the multiple contribution coefficients to generate a node matching array. Specifically, all nodes in the vehicle manufacturing supply chain are traversed, and each node corresponds to a contribution coefficient. For each node, its contribution coefficient is matched with the contribution coefficients of other nodes. The matching process can be based on similarity calculation, threshold judgment or other statistical methods. Through matching, a node matching array is obtained, which contains the association strength and direction information between nodes. This array is a two-dimensional matrix, in which rows and columns represent nodes in the supply chain, respectively, and the values ​​in the matrix represent the association strength between nodes.

[0036] Step S270, identify multiple supply chain nodes based on the node matching array, and determine multiple node contribution level labels. Specifically, a series of level standards are set based on the threshold, ranking or other statistical indicators of the contribution coefficient. According to the association strength and direction information in the node matching array, and the set level standards, a contribution level label is determined for each node. This label can be numerical (such as 1-5 levels) or descriptive (such as "core node", "key node", "ordinary node", etc.). Finally, the determined level label is added to the node information for subsequent analysis and optimization.

[0037] Step S280, causal relationship inference is performed on multiple supply chain nodes according to the multiple node contribution level labels, and an information flow diagram of the multiple supply chain nodes is drawn according to the inference results. Specifically, based on the node contribution level labels and the node matching array, a causal inference method (such as Granger causality test, Bayesian network, etc.) is used to determine the causal relationship between the nodes. Based on the result of causal relationship inference, an information flow diagram is drawn, in which the nodes represent entities in the supply chain and the arrows represent the direction and path of information flow.

[0038] Step S290, performing interaction flow analysis according to the information flow diagram, identifying the multi-level dependencies of multiple supply chain nodes, connecting the multiple supply chain nodes according to the multi-level dependencies, and constructing the inter-chain dynamic interaction simulation network of the vehicle manufacturing supply chain. Specifically, based on the information flow diagram and the results of causal relationship inference, the multi-level dependencies between supply chain nodes are identified. These relationships can be direct (such as supplier-manufacturer relationship) or indirect (such as raw material supplier-distributor relationship). According to the identified multi-level dependencies, a dynamic interaction simulation network between chains is constructed using simulation software or tools. This network contains all nodes in the supply chain and the connection relationships between them, as well as the interaction rules and parameters between nodes. This implementation method can more accurately identify key nodes and potential risk points in the supply chain by calculating the contribution coefficient and performing node matching and grade division based on it, and combining the node matching array and the contribution grade label to perform causal relationship inference, thereby improving the accuracy and reliability of the inference. The simulation network is constructed based on the identified multi-level dependencies and causal relationship inference results, which can more accurately simulate the dynamic interaction process in the supply chain and improve the accuracy of the inter-chain dynamic interaction simulation network.

[0039] Step S300, defining the manufacturing constraints of the vehicle manufacturing supply chain, activating the inter-chain dynamic interaction simulation network to perform demand forecasting, performing production analysis based on the demand forecasting results, and formulating a production scheduling strategy, specifically including step S310, performing manufacturing stability calculation based on the vehicle manufacturing supply chain, obtaining a manufacturing stability coefficient group, performing fluctuation extreme value analysis based on the manufacturing stability coefficient group, and extracting a stable upper limit value and a stable lower limit value. Specifically, historical manufacturing data and real-time manufacturing data of the vehicle manufacturing supply chain are collected, including production volume, production efficiency, equipment failure rate, etc. By comparing the data differences between different time periods or different production batches, the manufacturing stability of the supply chain (the ability of the supply chain to maintain stable operation during the production process) is calculated using statistical methods (such as standard deviation, coefficient of variation, etc.). The calculated stability indicators are integrated into a stability coefficient group, which is a collection of multiple stability indicators and is used to measure the manufacturing stability of the supply chain.

[0040] Step S320, define a stable numerical interval according to the stable upper limit value and the stable lower limit value, and construct the manufacturing constraint condition according to the numerical interval. Specifically, use statistical methods (such as extreme value theory, quantile analysis, etc.) to perform fluctuation extreme value analysis on the stability coefficient group to determine the fluctuation upper limit and fluctuation lower limit of the supply chain during the production process. According to the fluctuation upper limit and fluctuation lower limit, define a stable numerical interval to represent the fluctuation range of the supply chain under normal production conditions. Use the stable numerical interval as a manufacturing constraint condition to limit fluctuations in the production process and ensure the stable operation of the supply chain.

[0041] Step S330, activate the inter-chain dynamic interaction simulation network, traverse multiple supply chain nodes according to the multi-level dependency relationship to simulate the production scenario, and obtain multiple production scenario information. Specifically, activate the inter-chain dynamic interaction simulation network to ensure that the network is in operation. Set different production conditions according to actual needs, such as production volume, production efficiency, equipment failure rate, etc. Use the simulation network to simulate the production scenario and record the supply chain response under different production conditions.

[0042] Step S340, performing demand analysis on the vehicle manufacturing supply chain according to the manufacturing constraints and the multiple production scenario information, and generating multiple demand change trends. Specifically, according to the production scenario simulation results and the manufacturing constraints, analyzing the demand changes of the supply chain under different production conditions. Integrate the analysis results into a demand change trend chart or table to intuitively display the demand changes of the supply chain.

[0043] Step S350, based on the mapping of the multiple demand change trends to the multiple production scenario information, simulation prediction is performed to determine the demand prediction results, multiple capacity bottleneck segments are identified according to the demand prediction results, production analysis is performed according to the multiple capacity bottleneck segments, and the production scheduling strategy is formulated. Specifically, according to the demand change trend, simulation prediction is performed in the inter-chain dynamic interaction simulation network, and the supply chain response under different demand conditions is recorded. By analyzing the simulation prediction results, the capacity bottleneck segments in the supply chain are identified, that is, those links that are prone to insufficient or excessive capacity in the production process. According to the capacity bottleneck segments and simulation prediction results, targeted production scheduling strategies are formulated to optimize the production efficiency and stability of the supply chain. This implementation method uses the inter-chain dynamic interaction simulation network to simulate and predict production scenarios, and comprehensively analyzes the response of the supply chain under different production conditions, so that a more reasonable production scheduling strategy can be formulated.

[0044] In a possible implementation, based on the mapping of the multiple demand change trends to the multiple production scenario information, simulation prediction is performed to determine the demand prediction results, multiple capacity bottleneck segments are identified according to the demand prediction results, production analysis is performed according to the multiple capacity bottleneck segments, and the production scheduling strategy is formulated. Step S350 further includes step S351, based on the mapping of the multiple demand change trends to the multiple production scenario information, multiple supply chain nodes are traversed and collected to build a standard demand database. Specifically, multiple demand change trends are mapped to multiple production scenario information to ensure that each supply chain node has corresponding demand data. Each supply chain node is traversed and collected to collect its historical demand data and real-time demand data. Historical demand data includes demand volume, demand change trend, etc. in the past period of time; real-time demand data includes demand volume, order status, etc. at the current moment or in the near future. The collected data is integrated into a standard demand database for subsequent analysis and prediction.

[0045] Step S352, indexing the standard demand database according to the time series of multiple supply chain nodes to obtain demand forecast results, and extracting multiple demand change feature sets based on the demand forecast results, wherein the multiple demand change features include a short-term demand change feature set and a long-term demand change feature set. Specifically, using time series analysis technology, the standard demand database is indexed to obtain demand forecast results for each supply chain node. Based on the demand forecast results, short-term demand change feature sets and long-term demand change feature sets are extracted. The short-term demand change feature set includes recent demand fluctuations, order changes, etc.; the long-term demand change feature set includes demand trends and seasonal changes over a longer period of time.

[0046] Step S353, perform scheduling balance calculation according to the short-term demand change feature set to obtain a short-term balance coefficient, and perform scheduling balance calculation according to the long-term demand change feature set to obtain a long-term balance coefficient. Specifically, the scheduling balance coefficient of the vehicle manufacturing supply chain in the short term is calculated using the short-term demand change feature set, and this coefficient reflects the ability of the supply chain to cope with short-term demand fluctuations. Similarly, the scheduling balance coefficient of the supply chain in a longer period of time is calculated using the long-term demand change feature set, and this coefficient reflects the ability of the supply chain to cope with long-term demand trend changes.

[0047] Step S354, the short-term balance coefficient and the long-term balance coefficient are jointly integrated, and the multiple capacity bottleneck segments are identified according to the integration results. Specifically, the short-term balance coefficient and the long-term balance coefficient are jointly integrated. If the two coefficients converge, it is considered that both the short-term and long-term are balanced; if the coefficients differ or fluctuate greatly, it is considered that neither the short-term nor the long-term is balanced. According to the integration results, those supply chain nodes or links that are not balanced in the short-term or long-term are identified, and these nodes or links are the capacity bottleneck segments. This implementation method accurately predicts the demand changes in the vehicle manufacturing supply chain by constructing a standard demand database and using time series analysis technology to forecast demand. By extracting the short-term and long-term demand change feature sets and performing scheduling balance calculations, the capacity bottleneck segments in the supply chain are accurately identified, thereby improving the pertinence of production scheduling strategy formulation.

[0048] Step S400, based on the vehicle manufacturing supply chain simulation, the production scheduling strategy is executed to perform multi-objective optimization, a plurality of labels to be optimized are generated, the production scheduling strategy is dynamically monitored according to the plurality of labels to be optimized, the production scheduling strategy is updated, and a global optimization strategy for production scheduling is obtained.

[0049] Specifically, the formulated production scheduling strategy is simulated and executed in the inter-chain dynamic interaction simulation network to observe its impact on the supply chain. Based on the simulation results, multiple tags to be optimized are generated. The tags to be optimized are indicators or tags used to measure and optimize the performance of the production scheduling strategy, such as production cost, production efficiency, inventory turnover rate, etc. The production scheduling strategy is dynamically monitored according to these tags to be optimized, and the strategy is adjusted in real time to optimize the target. Based on the monitoring results and feedback, the production scheduling strategy is continuously updated and optimized until the global optimum is reached.

[0050] In a possible implementation, the production scheduling strategy is executed based on the simulation of the vehicle manufacturing supply chain to perform multi-objective optimization, and multiple labels to be optimized are generated. Step S400 further includes step S410, simulating the production scheduling execution of the vehicle manufacturing supply chain according to the production scheduling strategy, performing scheduling calculations according to the scheduling simulation execution results, and determining multiple scheduling indicators. Specifically, computer simulation technology is used to simulate the production scheduling execution of the vehicle manufacturing supply chain according to the production scheduling strategy. The simulation process includes each node in the supply chain, from the procurement of raw materials to the delivery of finished products. Through simulation, the changes in key indicators such as the operating status of the supply chain, inventory levels, and production speed under different production scheduling strategies are analyzed. During the execution process, detailed scheduling calculations are performed based on the simulation results to determine multiple key scheduling indicators, such as production efficiency, inventory turnover rate, production cycle, etc.

[0051] Step S420, a multi-objective search is performed on the production scheduling strategy according to the multiple scheduling indicators to determine a scheduling target set. Specifically, after obtaining multiple scheduling indicators, a multi-objective search algorithm is used to find an optimal solution or a near-optimal solution in the solution space formed by the scheduling indicators. This process is used to find one or more production scheduling strategies so that indicators such as production efficiency and inventory turnover rate reach an optimal or near-optimal state. Through multi-objective search, a scheduling target set is determined, which includes multiple possible optimal or near-optimal production scheduling strategies.

[0052] Step S430, based on the scheduling target set, a scheduling state analysis is performed to construct a scheduling state space, based on the scheduling target set, a scheduling action analysis is performed to construct a scheduling action space, a reward function is introduced, and the scheduling target set is reinforced learned in combination with the scheduling state space and the scheduling action space to generate a scheduling learning result. Specifically, after the scheduling target set is determined, the scheduling target set is trained using a reinforcement learning algorithm. Reinforcement learning is a machine learning algorithm that learns strategies by interacting with the environment. Here, the environment is the vehicle manufacturing supply chain, and the strategy is the production scheduling strategy. A scheduling state space and a scheduling action space are constructed. The scheduling state space contains the state of the supply chain at different time points, such as inventory level, production speed, etc.; the scheduling action space contains possible scheduling actions, such as increasing or decreasing production volume, adjusting inventory strategy, etc. By introducing a reward function, the advantages and disadvantages of different scheduling actions under a given state are evaluated. The design of the reward function is based on economic indicators of supply chain operations, such as cost, revenue, etc. In the reinforcement learning process, the algorithm will continuously try different scheduling actions and adjust the strategy according to the reward function to maximize the long-term reward. Finally, the reinforcement learning algorithm generates a scheduling learning result, that is, an optimized production scheduling strategy.

[0053] Step S440, the scheduling learning result is subjected to production risk assessment according to the multiple production capacity bottleneck segments, multi-level production risk data is generated, the production scheduling strategy is matched and identified based on the multi-level production risk data, and the multiple labels to be optimized are generated. Specifically, after obtaining the scheduling learning result, it is applied to the simulation environment of the vehicle manufacturing supply chain to perform production risk assessment. The evaluation process mainly focuses on the risk level of the supply chain under different production scheduling strategies, such as production interruption, inventory backlog, etc. Based on the multi-level production risk data, the production scheduling strategy is matched and identified, and multiple labels to be optimized are generated. These labels identify the links in the supply chain that may have risks or need to be optimized, and provide a basis for subsequent strategy adjustments. This implementation method obtains multiple key scheduling indicators by simulating production scheduling execution and scheduling calculation, which provides a basis for subsequent multi-objective search and reinforcement learning, helps to find the optimal or approximately optimal production scheduling strategy, improves the operating efficiency of the supply chain, and can timely discover the links in the supply chain that may have risks or need to be optimized through production risk assessment and generation of labels to be optimized, and take corresponding measures to improve them, which helps to reduce production risks and costs and improve the overall competitiveness of the supply chain.

[0054] Step S500: synchronizing the production scheduling global optimization strategy to the vehicle manufacturing supply chain for verification, and adaptively optimizing the vehicle manufacturing supply chain according to the verification result.

[0055] Specifically, the global optimization strategy for production scheduling after multi-objective optimization is synchronized to the actual operation of the vehicle manufacturing supply chain. The effectiveness of the strategy is verified in actual operation, and its improvement on supply chain performance is observed. According to the verification results, the vehicle manufacturing supply chain is adaptively optimized, including adjusting the supply chain structure, improving the production process, optimizing inventory management and other aspects. The embodiment of the present application adopts traversing the vehicle manufacturing supply chain for data recording, retrieving the historical manufacturing full cycle data set, and combining real-time sensor data to build a dynamic interactive simulation network between chains, define manufacturing constraints, perform demand forecasting and production analysis, formulate production scheduling strategies, perform multi-objective optimization through simulating the execution of production scheduling strategies, generate labels to be optimized, and perform dynamic monitoring and optimization to obtain a global optimization strategy for production scheduling. According to the verification results, the vehicle manufacturing supply chain is adaptively optimized and other technical means are used to achieve the technical effect of improving real-time and dynamic performance, responding to changes in market demand and supply chain status in a timely manner, and thus improving the accuracy of production scheduling.

[0056] In the above, refer to Figure 1 The data-driven vehicle manufacturing supply chain adaptive optimization method according to an embodiment of the present invention is described in detail. Figure 2 A data-driven vehicle manufacturing supply chain adaptive optimization system according to an embodiment of the present invention is described.

[0057] The data-driven vehicle manufacturing supply chain adaptive optimization system according to the embodiment of the present invention is used to solve the technical problems of the existing technology that lack real-time and dynamics, cannot respond to changes in market demand and supply chain status in a timely manner, and lead to inaccurate production scheduling, so as to achieve the technical effect of improving real-time and dynamics, responding to changes in market demand and supply chain status in a timely manner, and thus improving the accuracy of production scheduling. The data-driven vehicle manufacturing supply chain adaptive optimization system includes: a data recording and sensing module 10, an integrated association module 20, a production scheduling strategy formulation module 30, a production scheduling strategy update module 40, and a supply chain adaptive optimization module 50.

[0058] The data recording and sensing module 10 is used to traverse the vehicle manufacturing supply chain to record data, retrieve the historical manufacturing full-cycle data set of the vehicle manufacturing supply chain, perform real-time sensing according to multiple supply chain nodes of the vehicle manufacturing supply chain, and obtain a real-time manufacturing data set; the integrated association module 20 is used to perform integrated learning based on the historical manufacturing full-cycle data set and the real-time manufacturing data set, perform association analysis on multiple supply chain nodes according to the learning results, and construct a dynamic interaction simulation network between chains; the production scheduling strategy formulation module 30 is used to define the manufacturing constraints of the vehicle manufacturing supply chain, activate the dynamic interaction simulation network between chains to predict demand, perform production analysis according to the demand prediction results, and formulate a production scheduling strategy; the production scheduling strategy update module 40 is used to perform multi-objective optimization based on the vehicle manufacturing supply chain simulation execution of the production scheduling strategy, generate multiple tags to be optimized, dynamically monitor the production scheduling strategy according to the multiple tags to be optimized, update the production scheduling strategy, and obtain a global optimization strategy for production scheduling; the supply chain adaptive optimization module 50 is used to synchronize the global optimization strategy for production scheduling to the vehicle manufacturing supply chain for verification, and adaptively optimize the vehicle manufacturing supply chain according to the verification results.

[0059] The specific configuration of the integrated association module 20 will be described in detail below. As described above, based on the integrated learning of the historical manufacturing full cycle data set and the real-time manufacturing data set, the integrated association module 20 may further include: a node timestamp determination unit for performing a timing analysis on the vehicle manufacturing supply chain according to multiple supply chain nodes to determine the node timestamp; an alignment processing unit for aligning the historical manufacturing full cycle data set with the real-time manufacturing data set according to the node timestamp to generate a data alignment matrix; a feature analysis unit for performing feature analysis based on the data alignment matrix to extract multiple eigenvalues, wherein the multiple eigenvalues ​​include historical eigenvalues ​​and real-time eigenvalues; a random sampling unit for randomly sampling the data alignment matrix based on the historical eigenvalues ​​to obtain a first sample group, and for randomly sampling the data alignment matrix based on the real-time eigenvalues ​​to obtain a second sample group; an integrated training unit for performing integrated training based on the first sample group and the second sample group, constructing a gradient boosting machine, and performing tuning learning through the gradient boosting machine to obtain the learning result.

[0060] Among them, integrated training is performed based on the first sample group and the second sample group to construct a gradient boosting machine, and tuning learning is performed through the gradient boosting machine to obtain the learning result. The integrated training unit may further include: an integrated learning subunit is used to perform integrated learning based on the first sample group to obtain multiple first weak classifiers, and to perform integrated learning based on the second sample group to obtain multiple second weak classifiers; an iterative training subunit is used to introduce a loss function to iteratively train the multiple first weak classifiers to obtain a first loss gradient value, and to iteratively train the multiple second weak classifiers according to the loss function to obtain a second loss gradient value; a weighted analysis subunit is used to perform weighted analysis according to the first loss gradient value and the second loss gradient value according to the target gradient direction to construct the gradient boosting machine; a contribution calculation subunit is used to perform contribution calculation according to the gradient boosting machine by traversing the multiple first weak classifiers and the multiple second weak classifiers to obtain multiple contribution coefficients, and add the multiple contribution coefficients to the learning result.

[0061] Among them, association analysis is performed on multiple supply chain nodes according to the learning results, and a dynamic interaction simulation network between chains is constructed. The integrated association module 20 may further include: a traversal matching unit is used to traverse and match multiple supply chain nodes of the vehicle manufacturing supply chain according to the multiple contribution coefficients to generate a node matching array; a node contribution level label determination unit is used to identify multiple supply chain nodes based on the node matching array, and determine multiple node contribution level labels; a causal relationship inference unit is used to perform causal relationship inference on multiple supply chain nodes according to the multiple node contribution level labels, and draw an information flow diagram of multiple supply chain nodes according to the inference results; an interactive flow analysis unit is used to perform interactive flow analysis according to the information flow diagram, identify the multi-level dependency relationships of multiple supply chain nodes, connect the multiple supply chain nodes according to the multi-level dependency relationships, and construct the dynamic interaction simulation network between chains of the vehicle manufacturing supply chain.

[0062] Next, the specific configuration of the production scheduling strategy formulation module 30 will be described in detail. As described above, the manufacturing constraints of the vehicle manufacturing supply chain are defined, the inter-chain dynamic interaction simulation network is activated to perform demand forecasting, production analysis is performed based on the demand forecasting results, and a production scheduling strategy is formulated. The production scheduling strategy formulation module 30 may further include: a manufacturing stability calculation unit is used to perform manufacturing stability calculation based on the vehicle manufacturing supply chain, obtain a manufacturing stability coefficient group, perform fluctuation extreme value analysis based on the manufacturing stability coefficient group, and extract a stable upper limit value and a stable lower limit value; a manufacturing constraint condition construction unit is used to define a stable numerical range according to the stable upper limit value and the stable lower limit value, and construct the manufacturing constraint condition according to the numerical range; a production scenario simulation unit is used to activate the inter-chain dynamic interaction simulation network, traverse multiple supply chain nodes according to the multi-level dependency relationship to perform production scenario simulation, and obtain multiple production scenario information; a demand analysis unit is used to perform demand analysis on the vehicle manufacturing supply chain according to the manufacturing constraints combined with the multiple production scenario information, and generate multiple demand change trends; a production scheduling strategy formulation unit is used to perform simulation prediction based on the multiple demand change trends mapped to the multiple production scenario information, determine the demand forecasting results, identify multiple capacity bottleneck segments according to the demand forecasting results, perform production analysis according to the multiple capacity bottleneck segments, and formulate the production scheduling strategy.

[0063] Among them, based on the mapping of the multiple demand change trends to the multiple production scenario information, simulation prediction is performed to determine the demand forecast result, multiple capacity bottleneck segments are identified according to the demand forecast result, production analysis is performed according to the multiple capacity bottleneck segments, and the production scheduling strategy is formulated. The production scheduling strategy formulation unit may further include: a standard demand database construction subunit is used to traverse and collect multiple supply chain nodes based on the mapping of the multiple demand change trends to the multiple production scenario information to construct a standard demand database; a demand change feature set extraction subunit is used to index the standard demand database according to the time series of multiple supply chain nodes to obtain demand forecast results, and extract multiple demand change feature sets according to the demand forecast results, and the multiple demand change features include short-term demand change feature sets and long-term demand change feature sets; a scheduling balance calculation subunit is used to perform scheduling balance calculation according to the short-term demand change feature set to obtain a short-term balance coefficient, and perform scheduling balance calculation according to the long-term demand change feature set to obtain a long-term balance coefficient; a joint integration subunit is used to jointly integrate the short-term balance coefficient with the long-term balance coefficient, and identify the multiple capacity bottleneck segments according to the integration result.

[0064] The specific configuration of the production scheduling strategy update module 40 will be described in detail below. As described above, based on the simulation execution of the vehicle manufacturing supply chain, the production scheduling strategy is optimized for multi-objective optimization, and multiple labels to be optimized are generated. The production scheduling strategy update module 40 may further include: a simulated production scheduling execution unit is used to simulate the production scheduling execution of the vehicle manufacturing supply chain according to the production scheduling strategy, perform scheduling calculation according to the scheduling simulation execution result, and determine multiple scheduling indicators; a multi-objective search unit is used to perform multi-objective search on the production scheduling strategy according to the multiple scheduling indicators to determine the scheduling target set; a reinforcement learning unit is used to perform scheduling state analysis based on the scheduling target set, construct a scheduling state space, perform scheduling action analysis based on the scheduling target set, construct a scheduling action space, introduce a reward function, and perform reinforcement learning on the scheduling target set in combination with the scheduling state space and the scheduling action space to generate a scheduling learning result; a label generation unit to be optimized is used to perform production risk assessment on the scheduling learning result according to the multiple production capacity bottleneck segments, generate multi-level production risk data, match and identify the production scheduling strategy based on the multi-level production risk data, and generate the multiple labels to be optimized.

[0065] The data-driven vehicle manufacturing supply chain adaptive optimization system provided in the embodiments of the present invention can execute the data-driven vehicle manufacturing supply chain adaptive optimization method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0066] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0067] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A data-driven adaptive optimization method for vehicle manufacturing supply chain, characterized in that: The method comprises: Traverse the vehicle manufacturing supply chain to record data, retrieve the historical manufacturing full cycle data set of the vehicle manufacturing supply chain, perform real-time sensing based on multiple supply chain nodes of the vehicle manufacturing supply chain, and obtain real-time manufacturing data sets; Based on the historical manufacturing full cycle data set and the real-time manufacturing data set, integrated learning is performed, and a correlation analysis is performed on multiple supply chain nodes according to the learning results to construct a dynamic interaction simulation network between chains; Define manufacturing constraints of the vehicle manufacturing supply chain, activate the dynamic interactive simulation network between the chains to perform demand forecasting, perform production analysis based on the demand forecasting results, and formulate a production scheduling strategy; Execute the production scheduling strategy to perform multi-objective optimization based on vehicle manufacturing supply chain simulation, generate multiple tags to be optimized, dynamically monitor the production scheduling strategy according to the multiple tags to be optimized, update the production scheduling strategy, and obtain a global optimization strategy for production scheduling; Synchronizing the production scheduling global optimization strategy to the vehicle manufacturing supply chain for verification, and adaptively optimizing the vehicle manufacturing supply chain according to the verification results; The manufacturing constraints of the vehicle manufacturing supply chain are defined, the dynamic interaction simulation network between the chains is activated to perform demand forecasting, production analysis is performed based on the demand forecasting results, and a production scheduling strategy is formulated, including: Calculate manufacturing stability based on the vehicle manufacturing supply chain to obtain a manufacturing stability coefficient group, perform fluctuation extreme value analysis based on the manufacturing stability coefficient group, and extract a stable upper limit value and a stable lower limit value; Delimiting a stable numerical range according to the stable upper limit value and the stable lower limit value, and constructing the manufacturing constraint condition according to the numerical range; Activate the inter-chain dynamic interaction simulation network, traverse multiple supply chain nodes according to multi-level dependency relationships to perform production scenario simulation, and obtain multiple production scenario information; Performing demand analysis on the vehicle manufacturing supply chain according to the manufacturing constraint conditions combined with the multiple production scenario information to generate multiple demand change trends; Based on the mapping of the multiple demand change trends to the multiple production scenario information, simulation prediction is performed to determine the demand prediction results, multiple production capacity bottleneck segments are identified according to the demand prediction results, production analysis is performed according to the multiple production capacity bottleneck segments, and the production scheduling strategy is formulated.

2. The data-driven vehicle manufacturing supply chain adaptive optimization method according to claim 1, characterized in that: Based on the historical manufacturing full cycle data set and the real-time manufacturing data set, integrated learning is performed, and the method includes: Perform time series analysis on the vehicle manufacturing supply chain according to multiple supply chain nodes to determine the node timestamps; Aligning the historical manufacturing full cycle data set with the real-time manufacturing data set according to the node timestamps to generate a data alignment matrix; Performing feature analysis based on the data alignment matrix to extract multiple feature values, wherein the multiple feature values ​​include historical feature values ​​and real-time feature values; Randomly sampling the data alignment matrix based on the historical eigenvalues ​​to obtain a first sample group, and randomly sampling the data alignment matrix based on the real-time eigenvalues ​​to obtain a second sample group; An integrated training is performed according to the first sample group and the second sample group to construct a gradient boosting machine, and tuning learning is performed through the gradient boosting machine to obtain the learning result.

3. The data-driven vehicle manufacturing supply chain adaptive optimization method according to claim 2, characterized in that: Performing integrated training according to the first sample group and the second sample group, constructing a gradient boosting machine, performing tuning learning through the gradient boosting machine, and obtaining the learning result, the method comprising: Performing ensemble learning based on the first sample group to obtain a plurality of first weak classifiers, and performing ensemble learning based on the second sample group to obtain a plurality of second weak classifiers; Introducing a loss function to iteratively train the multiple first weak classifiers to obtain a first loss gradient value; Iteratively training the plurality of second weak classifiers according to the loss function to obtain a second loss gradient value; Performing weighted analysis according to the first loss gradient value and the second loss gradient value in accordance with the target gradient direction to construct the gradient boosting machine; The gradient boosting machine traverses the multiple first weak classifiers and the multiple second weak classifiers to perform contribution calculations, obtains multiple contribution coefficients, and adds the multiple contribution coefficients to the learning result.

4. The data-driven vehicle manufacturing supply chain adaptive optimization method according to claim 3, characterized in that: According to the learning results, multiple supply chain nodes are analyzed for correlation and a dynamic interaction simulation network between chains is constructed. The methods include: Traversing and matching multiple supply chain nodes of the vehicle manufacturing supply chain according to the multiple contribution coefficients to generate a node matching array; Identifying multiple supply chain nodes based on the node matching array and determining multiple node contribution level labels; Performing causal relationship inference on multiple supply chain nodes according to the multiple node contribution level labels, and drawing an information flow diagram of the multiple supply chain nodes according to the inference results; An interactive flow analysis is performed based on the information flow diagram to identify the multi-level dependency relationships of multiple supply chain nodes, and multiple supply chain nodes are connected according to the multi-level dependency relationships to construct the inter-chain dynamic interactive simulation network of the vehicle manufacturing supply chain.

5. The data-driven vehicle manufacturing supply chain adaptive optimization method according to claim 1, characterized in that: Based on the mapping of the multiple demand change trends to the multiple production scenario information, simulation prediction is performed to determine the demand prediction result, multiple production capacity bottleneck segments are identified according to the demand prediction result, production analysis is performed according to the multiple production capacity bottleneck segments, and the production scheduling strategy is formulated, the method includes: Based on the mapping of the multiple demand change trends to the multiple production scenario information, multiple supply chain nodes are traversed and collected to build a standard demand database; Indexing the standard demand database according to the time series of multiple supply chain nodes to obtain demand forecast results, and extracting multiple demand change feature sets according to the demand forecast results, wherein the multiple demand change features include a short-term demand change feature set and a long-term demand change feature set; Performing a dispatch balance calculation according to the short-term demand change feature set to obtain a short-term balance coefficient, and performing a dispatch balance calculation according to the long-term demand change feature set to obtain a long-term balance coefficient; The short-term balance coefficient and the long-term balance coefficient are jointly integrated, and the multiple production capacity bottleneck sections are identified according to the integration result.

6. The data-driven vehicle manufacturing supply chain adaptive optimization method according to claim 5, characterized in that: Based on the simulation of the vehicle manufacturing supply chain, the production scheduling strategy is executed to perform multi-objective optimization, and multiple labels to be optimized are generated. The method includes: Performing a simulated production scheduling execution on the vehicle manufacturing supply chain according to the production scheduling strategy, performing scheduling calculations based on the scheduling simulation execution results, and determining multiple scheduling indicators; Performing a multi-objective search on the production scheduling strategy according to the multiple scheduling indicators to determine a scheduling target set; Perform scheduling state analysis based on the scheduling target set to construct a scheduling state space, perform scheduling action analysis based on the scheduling target set to construct a scheduling action space, introduce a reward function, perform reinforcement learning on the scheduling target set in combination with the scheduling state space and the scheduling action space, and generate a scheduling learning result; The scheduling learning results are used to perform production risk assessment according to the multiple production capacity bottleneck segments to generate multi-level production risk data, and the production scheduling strategy is matched and identified based on the multi-level production risk data to generate the multiple labels to be optimized.

7. Data-driven vehicle manufacturing supply chain adaptive optimization system, characterized by: The system is used to implement the data-driven vehicle manufacturing supply chain adaptive optimization method according to any one of claims 1 to 6, and the system comprises: The data recording and sensing module is used to traverse the vehicle manufacturing supply chain to record data, retrieve the historical manufacturing full cycle data set of the vehicle manufacturing supply chain, perform real-time sensing based on multiple supply chain nodes of the vehicle manufacturing supply chain, and obtain real-time manufacturing data sets; An integrated association module, used for performing integrated learning based on the historical manufacturing full cycle data set and the real-time manufacturing data set, performing association analysis on multiple supply chain nodes according to the learning results, and constructing a dynamic interaction simulation network between chains; A production scheduling strategy formulation module is used to define manufacturing constraints of the vehicle manufacturing supply chain, activate the inter-chain dynamic interactive simulation network to perform demand forecasting, perform production analysis based on the demand forecasting results, and formulate a production scheduling strategy; A production scheduling strategy updating module is used to perform multi-objective optimization on the production scheduling strategy based on a vehicle manufacturing supply chain simulation, generate multiple tags to be optimized, dynamically monitor the production scheduling strategy according to the multiple tags to be optimized, update the production scheduling strategy, and obtain a global optimization strategy for production scheduling; The supply chain adaptive optimization module is used to synchronize the production scheduling global optimization strategy to the vehicle manufacturing supply chain for verification, and adaptively optimize the vehicle manufacturing supply chain according to the verification results.

Images