Intelligent estimation system for dynamic cost of automobile parts fused with supply chain fluctuation

Abstract

The application belongs to the technical field of supply chain risk and cost intelligent monitoring, and relates to a dynamic cost intelligent estimation system for automobile parts fused with supply chain fluctuations, which comprises: a data acquisition and alignment module for generating a multimodal supply chain original data set; a semantic correlation analysis module for quantitatively generating a standardized supply chain risk signal flow; a causal graph construction module for constructing a dynamic tensor causal graph by analyzing a bill of materials; a graph dynamic adjustment module for real-time parameter space adjustment; a cost transmission calculation module for driving the transmission calculation of a dynamic cost increment; an uncertainty quantification module for generating a cost confidence interval distribution function by using a simulation method; and a monitoring and instruction output module for generating a cost report and outputting a procurement strategy suggestion. The application solves the problems of low integration of multi-source heterogeneous information, unclear cost transmission mechanism and difficulty in quantifying cost fluctuation uncertainty in a complex supply chain environment.

Description

Technical Field

[0001] This invention belongs to the technical field of intelligent monitoring of supply chain risks and costs, and relates to an intelligent dynamic cost estimation system for automotive parts that integrates supply chain fluctuations. Background Technology

[0002] In the current globalized supply chain environment, the automotive parts supply chain faces various uncertainties, including fluctuations in commodity prices, geopolitical events, and logistical changes. These factors are intertwined and impact the manufacturing costs of automotive parts. Therefore, timely perception and quantification of the transmission effects of external events on supply chain costs are crucial for companies to address market challenges.

[0003] Currently, the industry typically uses statistical analysis based on historical data, expert judgment, and simple linear models for cost forecasting. Supply chain planning and cost accounting largely rely on periodic data updates and static bills of materials for simulation, and some solutions also use early warning systems based on established rules to process supply chain data.

[0004] The aforementioned methods utilizing historical statistics and static rules have limitations in integrating real-time, multi-source, heterogeneous data, particularly in semantic understanding and risk quantification of unstructured information. Traditional models struggle to effectively characterize the nonlinearity and time lag of cost transmission within the supply chain, as well as the elastic buffering effects between different stages, leading to potential discrepancies between predictions and actual outcomes. Furthermore, the lack of an uncertainty quantification assessment mechanism limits the understanding of the scope of potential risks. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an intelligent dynamic cost estimation system for automotive parts that integrates supply chain fluctuations.

[0006] An intelligent dynamic cost estimation system for automotive parts that integrates supply chain fluctuations includes:

[0007] The data acquisition and alignment module is used to collect raw data from multi-source heterogeneous supply chains, perform timestamp alignment processing, and generate a multimodal supply chain raw dataset.

[0008] The semantic association analysis module is used to perform semantic association analysis on the original dataset of the multimodal supply chain and quantify and generate a standardized supply chain risk signal stream that includes event confidence factors.

[0009] The causal graph construction module is used to parse the bill of materials of the target automotive parts, extract core nodes and configure transmission parameters to construct a dynamic tensor causal graph.

[0010] The graph dynamic adjustment module uses the event confidence factor in the standardized supply chain risk signal flow to perform real-time parameter space adjustment on the dynamic tensor causal graph.

[0011] The cost transfer calculation module is used to drive the modulated dynamic tensor causal graph to perform numerical transfer and calculate the dynamic cost increment of the target automotive parts.

[0012] The uncertainty quantification module uses Monte Carlo simulation to iteratively sample dynamic cost increments and fits the distribution function of cost confidence intervals.

[0013] The monitoring and instruction output module is used to summarize dynamic cost increments and cost confidence interval distribution functions to generate cost reports, trigger early warnings based on preset thresholds, and output procurement strategy suggestions.

[0014] A further aspect of the present invention includes a data acquisition and alignment module, used to perform the following steps:

[0015] Real-time data on raw material futures prices, energy market indices, and real-time freight rates from logistics platforms can be captured using application programming interfaces (APIs).

[0016] Unstructured text data is obtained through web crawling, and structured tuples containing time, location, and event type are extracted using natural language processing techniques.

[0017] Real-time freight rate data and structured tuples are uniformly mapped to a preset standard time axis, forming a multi-dimensional data structure indexed by the time axis.

[0018] A further aspect of this invention includes a semantic association analysis module, which performs the following steps:

[0019] The original dataset of the multimodal supply chain is input into a pre-defined semantic network model to identify the causal relationships between data of different dimensions and cluster them into primary supply chain event tuples.

[0020] A multimodal cross-validation engine is used to compare the information consistency of primary supply chain event tuples under different confidence sources, and an information consistency index is generated.

[0021] By combining the preset information source authority, information consistency index, and historical event correlation, an event confidence factor is calculated, which includes the confidence probability of the event occurring, the potential impact range, and the timeliness of the information.

[0022] A further aspect of the present invention includes a causal graph construction module, which performs the following operations:

[0023] The types of raw materials, processing steps, logistics routes, and energy consumption quotas in the raw material list are used as the core nodes of the dynamic tensor causal graph.

[0024] A nonlinear cost transmission function is established between core nodes. The nonlinear cost transmission function includes marginal transmission coefficients determined by a piecewise function.

[0025] Configure time delay parameters that characterize the physical transmission cycle among core nodes, and configure elastic buffer factors that quantify the ability of each node to absorb cost shocks.

[0026] A further aspect of the present invention includes a spectrum dynamic adjustment module, used to perform the following operations:

[0027] Based on the entity types involved in the primary supply chain event tuples, match the corresponding core nodes in the dynamic tensor causal graph;

[0028] If the confidence probability of an event is greater than a preset threshold, the direct causal path in the dynamic tensor causal graph is forcibly activated, and the random perturbation parameter of the nonlinear cost transmission function is narrowed proportionally.

[0029] If the confidence probability of an event is not greater than a preset threshold, activate alternative paths or elastic nodes in the dynamic tensor causal graph and proportionally increase the random perturbation parameters.

[0030] A further aspect of the present invention includes a cost transmission calculation module, used to perform the following operations:

[0031] Starting from the source node of the dynamic tensor causal graph, local cost changes are propagated downstream layer by layer, and the local cost changes are attenuated according to the elastic buffer factor.

[0032] By using the real-time marginal transfer coefficient operator and combining the value proportion weight of each core node in the bill of materials, the impact of the source disturbance after being transmitted through the activation path is weighted and aggregated to generate the total cost change rate.

[0033] By accumulating time delay parameters along the activation path in the dynamic tensor causal graph, the specific time point at which the total cost change rate reaches the target automotive component finished product is determined.

[0034] A further aspect of the present invention includes an uncertainty quantification module, used to perform the following operations:

[0035] Establish a probabilistic model of random variables based on the modulated random perturbation parameters and the potential impact amplitude;

[0036] Based on a probabilistic model of random variables, multiple iterative simulations were performed, and the frequency distribution of the rate of change of the simulated total cost generated in each iteration was statistically analyzed.

[0037] The frequency distribution is smoothly fitted using the kernel density estimation method, and the upper and lower bounds of cost fluctuations are calculated based on the preset confidence level.

[0038] A further aspect of the present invention includes a monitoring and command output module, used to perform the following operations:

[0039] Compare the upper bound of the total cost change rate or cost fluctuation with the preset safety deviation threshold;

[0040] If the safety deviation threshold is exceeded, the preset strategy rule base is retrieved based on the event type in the standardized supply chain risk signal flow;

[0041] Match the response strategy to the event type and automatically generate procurement strategy suggestions, including suggestions for price lock-in agreements, safety stock adjustments, or alternative path switching.

[0042] A further aspect of this invention involves sending the generated cost report and procurement strategy recommendations to a mobile terminal via a secure transmission protocol, and initiating an internal approval process based on the decision instructions received from the mobile terminal.

[0043] A further aspect of this invention involves the synergistic effect of standardized supply chain risk signal flow and dynamic tensor causal graph, which dynamically adjusts the topological activation state and perturbation range of physical transmission parameters of the dynamic tensor causal graph through event confidence factors.

[0044] In summary, the present invention has the following beneficial technical effects:

[0045] 1. This invention identifies causal relationships between data from different dimensions by real-time collection and timestamp alignment of global financial, logistics, and unstructured text data, using semantic association analysis technology. This mechanism can transform fragmented raw information into a standardized risk signal stream containing event confidence factors, helping to improve the objectivity and timeliness of perception of potential supply chain risks and providing structured data support for subsequent cost estimation.

[0046] 2. This invention transforms a static bill of materials into a dynamic logical topology by constructing a dynamic tensor causal graph that includes a cost transmission function, time delay parameters, and elastic buffer factors. By combining event confidence factors to adjust the graph's parameters in real time, the transmission path can be dynamically activated or the disturbance parameters adjusted according to the veracity of external risk events, thereby more accurately characterizing the nonlinearity and time delay effects of cost transmission in complex supply chains.

[0047] 3. This invention, based on cost propagation calculation, introduces an uncertainty quantification mechanism based on Monte Carlo simulation. This mechanism can not only calculate the dynamic cost increment point values ​​of target components, but also fit and generate a cost confidence interval distribution function. This combination of point value estimation and interval estimation quantitatively assesses the possible range and probability distribution of cost fluctuations, providing enterprises with more meaningful cost forecasting information and helping to improve the robustness of decision-making.

[0048] 4. This invention achieves intelligent output of procurement strategy suggestions by comparing the estimated results with safety deviation thresholds in real time and automatically matching response solutions using a strategy rule base. Through real-time access via mobile terminals and integration with internal approval processes, it shortens the enterprise's response cycle to supply chain cost risks and improves the automation level and management efficiency of cost monitoring. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0050] Figure 1 This is a schematic diagram of the framework in the embodiments of this application.

[0051] Figure 2 This is a flowchart illustrating an embodiment of this application. Detailed Implementation

[0052] The following is in conjunction with the appendix Figure 1 - Figure 2 A preferred description of the present invention is provided below.

[0053] See attached document Figure 1 - Figure 2 This invention proposes an intelligent dynamic cost estimation system for automotive parts that integrates supply chain fluctuations, comprising the following modules:

[0054] The data acquisition and alignment module is used to collect raw data from multi-source heterogeneous supply chains, perform timestamp alignment processing, and generate a multimodal supply chain raw dataset.

[0055] The semantic association analysis module is used to perform semantic association analysis on the original dataset of the multimodal supply chain and quantify and generate a standardized supply chain risk signal stream that includes event confidence factors.

[0056] The causal graph construction module is used to parse the bill of materials of the target automotive parts, extract core nodes and configure transmission parameters to construct a dynamic tensor causal graph.

[0057] The graph dynamic adjustment module uses the event confidence factor in the standardized supply chain risk signal flow to perform real-time parameter space adjustment on the dynamic tensor causal graph.

[0058] The cost transfer calculation module is used to drive the modulated dynamic tensor causal graph to perform numerical transfer and calculate the dynamic cost increment of the target automotive parts.

[0059] The uncertainty quantification module uses Monte Carlo simulation to iteratively sample dynamic cost increments and fits the distribution function of cost confidence intervals.

[0060] The monitoring and instruction output module is used to summarize dynamic cost increments and cost confidence interval distribution functions to generate cost reports, trigger early warnings based on preset thresholds, and output procurement strategy suggestions.

[0061] In one embodiment of the present invention, the data acquisition and alignment module is used to perform the following steps:

[0062] The system utilizes application programming interfaces (APIs) to capture real-time raw material futures prices, energy market indices, and real-time freight rate data from logistics platforms. It also acquires unstructured text data through web crawlers and extracts structured tuples containing time, location, and event type using natural language processing techniques. Finally, it maps real-time freight rate data and structured tuples to a preset standard timeline, forming a multi-dimensional data structure indexed by the timeline.

[0063] First, various types of data are captured in real time using application programming interfaces (APIs). In this embodiment, the software module acting as a futures index data acquisition agent periodically initiates RESTful API requests to the real-time market data interfaces of major global commodity exchanges, such as the London Metal Exchange and the New York Mercantile Exchange, to obtain raw material futures price data for commodities including copper, aluminum, and crude oil, as well as energy market index data such as the CRB Commodity Index. In this embodiment, an application programming interface (API) refers to a well-defined set of software interaction specifications that allows data exchange and function calls between different software systems, acquiring real-time data programmatically without manual intervention. RESTful API requests are based on the HTTP / HTTPS protocol, using a Uniform Resource Identifier (URI) to locate resources and employing HTTP methods such as GET and POST for operation.

[0064] Simultaneously, the software module of the logistics freight rate data collection agent sends requests to the public API of the international logistics platform to obtain real-time freight rate data for sea, land, and air transport. API requests are securely transmitted via HTTPS, and the data response format is typically JSON or XML. Each data record includes a timestamp, specific commodity, index, route identifier, and corresponding numerical value. JSON and XML are two common data exchange formats used to describe structured data. The futures index data collection agent and the logistics freight rate data collection agent periodically perform data fetching operations at a preset polling interval. The polling interval refers to the frequency at which the system checks for updates to the data source, typically set between 30 seconds and 5 minutes, depending on the update frequency of the data source and the system's real-time requirements.

[0065] Secondly, to handle unstructured information, text data is acquired through web crawlers, and key information is extracted using natural language processing (NLP) technology. A distributed web crawler cluster, comprised of a supply chain information crawler module, periodically accesses multiple preset data sources, including but not limited to international news agency websites, industry analysis report platforms, and official announcement pages of major suppliers. It should be noted that a web crawler is a program that automatically browses the World Wide Web and retrieves page content. This solution employs a distributed architecture to improve crawling efficiency and fault tolerance; the distributed web crawler cluster contains multiple crawler nodes running in parallel, working collaboratively through task queues and a result storage sharing mechanism. The web crawler cluster requests geopolitical news text, industry early warning announcements, and supplier capacity reports via the HTTP / S protocol, storing them as raw text files. The HTTP / S protocol is used to transmit hypertext or encrypted hypertext, ensuring the security and integrity of data transmission.

[0066] Subsequently, the software component of the unstructured information processing engine receives the text file and invokes pre-trained named entity recognition (NAME) and event extraction models. The NAME recognition model identifies time entities, location entities, and related product or company entities in the text. This model is a natural language processing technique designed to identify and classify entities with specific meanings from unstructured text; in this scheme, it focuses on identifying time, location, and event-related entities. The event extraction model identifies specific event types from the text. This model aims to identify and structure specific types of events from the text, extracting information such as participants, time, and location. It should be understood that both the NAME recognition and event extraction models are fine-tuned based on pre-trained language models with a Transformer architecture, such as BERT or its variants. The Transformer architecture is a self-attention deep learning model architecture that exhibits excellent performance due to its parallel processing capabilities and ability to capture long-range dependencies. The BERT model learns language representations through unsupervised pre-training on massive corpora and then fine-tunes them for downstream tasks. The model ultimately outputs standardized unstructured information tuples containing time, location, and event type.

[0067] Finally, all the acquired heterogeneous data are timestamped to generate the final multimodal supply chain raw dataset. The data alignment and fusion unit's software components receive raw material futures price data streams, energy market index data streams, real-time freight rate data streams, and unstructured information tuple streams. The data alignment and fusion unit first converts the timestamps of all data streams to Coordinated Universal Time (UTC). Using UTC as a unified time base avoids data confusion caused by time zone differences. For numerical time series data, this unit uses linear interpolation or nearest neighbor interpolation methods to resample and map it to a preset standard time axis. Timestamp alignment makes time series data from different frequencies or sources comparable on a unified time axis through resampling, interpolation, and other methods, solving the problem of inconsistent data granularity. Linear interpolation is a method of estimating the value of an intermediate point by drawing a straight line between two known points; nearest neighbor interpolation selects the value of the nearest known data point as the estimate.

[0068] A standard timeline is defined as a discrete sequence of time points, tracing back a certain time range from the current date, with each hour as the smallest granularity. This granularity can be set according to business needs. For unstructured information tuples, their time entities are parsed and mapped to the corresponding most recent time point on the standard timeline. If multiple unstructured events exist at a given time point, they are aggregated at that point. After all time series are aligned, the data alignment and fusion unit integrates raw material futures prices, energy market indices, real-time freight rate data, and time-mapped unstructured information to form a time-indexed multidimensional data structure—the multimodal supply chain raw dataset. This dataset integrates raw supply chain data of various types and sources. Its data structure is, for example, a relational database table or a table in a time series database, containing timestamps, various numerical indicators, and standardized event fields.

[0069] In one embodiment of the present invention, the semantic association analysis module is used to perform the following steps:

[0070] The original multimodal supply chain dataset is input into a pre-defined semantic network model to identify causal relationships between data from different dimensions and cluster them into primary supply chain event tuples. A multimodal cross-validation engine is used to compare the information consistency of the primary supply chain event tuples under different confidence sources to generate an information consistency index. Combining the pre-defined information source authority, information consistency index, and historical event correlation, an event confidence factor is calculated, which includes the confidence probability of event occurrence, potential impact magnitude, and information timeliness.

[0071] First, the original multimodal supply chain dataset is input into a pre-defined semantic network model to identify causal relationships between data from different dimensions, and clusters highly correlated data into primary supply chain event tuples. In this embodiment, the semantic network model is a knowledge representation system built on graph data structures and semantic parsing techniques. It describes domain knowledge by defining entities, relationships, and attributes, and supports reasoning based on this knowledge. The model pre-defines an ontology for the automotive supply chain, including entity types such as raw materials, parts, suppliers, transporters, regions, policies, and natural disasters, as well as the causal relationships between them. Its nodes represent entities in the supply chain, such as raw material A, port B, supplier C, and geopolitical event D, while edges represent predefined relationships between entities, such as supply relationships, transportation routes, and influencing factors. Numerical time-series data in the original multimodal supply chain dataset is parsed into continuous attribute updates, while unstructured information tuples are mapped to discrete event nodes.

[0072] The inference engine employs a combination of rule-based and statistical learning methods to continuously scan data patterns in the original multimodal supply chain dataset. For example, when it detects an abnormal upward trend in the futures price of a certain raw material category within a short period, and simultaneously, unstructured information reveals strikes or policy changes related to major copper-producing regions, the semantic network model infers a strong causal relationship between the two based on pre-defined causal rules, such as strikes in producing regions leading to supply shortages, thereby pushing up futures prices. These causal rules can be defined by domain experts or obtained through association rule mining and machine learning training on historical big data; for example, rising prices of key raw materials can increase component costs. For data points or events with strong correlations, the semantic network model clusters them into independent primary supply chain event tuples. These tuples contain the event's subject, object, impact type, initial timestamp, and references to the original data involved.

[0073] Secondly, a multimodal cross-validation engine is used to compare the consistency of information from different confidence sources in the primary supply chain event tuple to assess the authenticity of the events. The multimodal cross-validation engine receives the primary supply chain event tuple as input. For each event in the tuple, the engine initiates multiple parallel data validation threads. Each validation thread is responsible for retrieving evidence from the confidence source type it is interested in. For example, if the primary supply chain event tuple involves a strike, the engine will retrieve, for example, official reports from international news agencies, content posted by high-authority users on social media platforms, and official announcements from relevant industry associations. The multimodal cross-validation engine employs algorithms based on semantic similarity matching and information aggregation within a time window to compare information content obtained from different sources. Semantic similarity matching is a method for measuring the similarity of text content. In this scheme, a pre-trained word vector model is used to convert words into high-dimensional vectors, and the cosine similarity between vectors is calculated to quantify text similarity. For example, if multiple news reports describe key information such as the time, location, number of participants, and demands of the strike with high consistency and match the initial information in the primary supply chain event tuple, the information is considered to have high consistency. The engine assigns an initial information consistency index based on the consistency of information from these confidence sources. It should be noted that the information consistency index is a floating-point number between 0 and 1, reflecting the degree of similarity in the descriptions of the same event from different information sources. For example, 0.9 indicates high consistency, while 0.3 indicates significant disagreement.

[0074] Finally, based on the authority of the information source, the information consistency index, and the correlation of historical events, a multi-dimensional event confidence factor is calculated for each primary supply chain event tuple, and a standardized supply chain risk signal stream is output. The software component of the confidence calculation module receives the primary supply chain event tuple, the information consistency index, and the correlation of historical events as input. The correlation of historical events is a floating-point number between 0 and 1, using queries to the historical supply chain event database to measure the similarity and potential impact trend of the current event with known high-impact events in the past in terms of type, region, and time. The authority of the information source is preset; for example, official government announcements have higher authority than ordinary industry forums. These values ​​are based on industry consensus and are subject to expert evaluation and periodic updates based on the accuracy of historical information.

[0075] The confidence calculation module uses weighted averages or machine learning models, such as decision trees or neural networks, to synthesize these factors and calculate multidimensional event confidence factors. . It is a vector whose components contain the confidence probability of an event occurring. Potential impact range and information timeliness The confidence calculation module will These are appended to the primary supply chain event tuples to form a standardized supply chain risk signal stream. The standardized supply chain risk signal stream is output in a structured data format, such as JSON, where each entry includes event type, subject, region, main impact, time window, and multi-dimensional event confidence factors. .

[0076] The relevant calculation formulas are as follows:

[0077]

[0078]

[0079]

[0080]

[0081] in, The confidence factor for an event is a vector containing multiple dimensions. The confidence probability of an event occurring represents the likelihood that the event will actually happen, and its value ranges from [value missing]. Its calculation method is a function ,function For example, the Sigmoid activation function. ,in This represents a comprehensive score reflecting the consistency and relevance of information obtained through weighted calculation. and To calibrate the curve steepness and offset parameters based on historical sample distribution, through The center point of the weighted score is mapped to the linear sensitivity region of the Sigmoid function, and then... Adjust the output dynamic range to ensure that the output probability values ​​can cover... Interval.

[0082] Representing the potential impact, it indicates the extent of the impact the event might have on the supply chain if it occurs; its value ranges from [value missing]. Its calculation method is a function. ,function This can be a linear combination or a more complex nonlinear model, such as a decision tree, that comprehensively assesses the potential impact based on information consistency, historical correlation, and event type. ,in The range of values ​​is ; This represents the upper limit truncation function, ensuring that the final calculated potential impact does not exceed 1. If the calculated result is greater than 1, then it is set to 1.

[0083] Represents the timeliness of information, indicating the effective lifespan of the information, and its value range is: ,in This indicates that the information is completely fresh and valid. This indicates that the information has expired; its calculation method is a function. ,function It can be defined as This represents a linear model of information decay over time, but in practical applications, it has limitations. ;

[0084] The authority of an information source refers to the reliability assessment of the information source, with a value range of [value range missing]. For example, government reports are well-known media for The forum post is These are the system's preset static parameters, which are updated periodically.

[0085] The information consistency index is calculated through semantic similarity matching and multi-source information aggregation, and its value range is [value range missing]. This index is calculated in real time by a multimodal cross-validation engine.

[0086] Represents the relevance of historical events, obtained by assessing the similarity with similar events and their impacts in historical databases, with a value range of [value missing]. This index is generated by the semantic network model based on its internal knowledge base.

[0087] , , It is used for calculation The weighted coefficients correspond to the authority of the information source, the consistency index of information, and the relevance of historical events, respectively, and their sum is... In this embodiment, typical values ​​are as follows: These coefficients can be calibrated based on the experience of experts in the field and the results of cross-validation of historical prediction accuracy.

[0088] The event type correction factor is based on the nature of the event itself. Preset parameters for correction. For example, regional major natural disasters. Can be set to General labor disputes Can be set to It reflects the inherent potential impact of the event itself;

[0089] Represents the current system time;

[0090] It represents the time when the event occurred or was reported.

[0091] The effective duration threshold represents the threshold value for this type of event information. It depends on the type of event. For example, for daily fluctuations in international crude oil prices, It could be 24 hours; regarding the shutdown of a major port, its These thresholds, which may be 14 days, reflect the decay period of information value for different types of events.

[0092] In one embodiment of the present invention, the causal graph construction module is configured to perform the following steps:

[0093] The types of raw materials, processing steps, logistics routes, and energy consumption quotas in the raw material list are used as core nodes of the dynamic tensor causal graph. A nonlinear cost transmission function is established between the core nodes, which includes a marginal transmission coefficient determined by a piecewise function. A time delay parameter characterizing the physical transmission cycle is configured between the core nodes, and an elastic buffer factor is configured to quantify the ability of each node to absorb cost shocks.

[0094] First, the core nodes for constructing the BOM (Bill of Materials) graph are extracted by parsing the BOM file of the target automotive parts. The software component of the BOM parser module receives the BOM file of the target automotive parts, which is typically stored in XML, JSON, or a specific database format. The BOM is a technical document describing the product structure, detailing all materials, parts, components, and sub-components required for the product, as well as their assembly relationships and quantities. The BOM parser module uses industry-standard parsing libraries to process structured BOM data, such as XML-based CALS or STEP file parsers, or uses custom parsing scripts to process unstructured or semi-structured BOM information, such as Excel or TXT files.

[0095] The BOM parser module traverses the BOM structure, identifying and extracting the direct constituent materials of the components, such as metal sheets, electronic chips, and plastic particles, defining them as raw material type nodes. Simultaneously, the module identifies the various processing steps required to produce the components, such as stamping, welding, and injection molding. For each processing step, it also extracts its associated energy consumption quota node. Furthermore, from the BOM information or associated supply chain data, nodes detailing the logistics path of materials from suppliers to the production plant are also extracted, such as sea routes and land transportation segments. These extracted entities, including raw material types, processing steps, logistics paths, and energy consumption quotas, are instantiated as core nodes in a dynamic tensor causal graph, each node possessing a unique identifier and type attribute.

[0096] Secondly, a nonlinear cost transmission function is established between each core node, defining the marginal transmission coefficients of raw material price fluctuations and process cost changes on the final component cost. The software component of the cost transmission function generator traverses the core nodes and establishes causal edges between nodes according to preset association rules. For example, a procurement cost transmission edge is established from the raw material type node to the processing step node using that raw material, representing the transmission relationship of raw material procurement price fluctuations to downstream production links; an energy cost transmission edge is established from the energy consumption quota node to the energy-consuming processing step node, representing the impact of energy price or consumption changes on production costs; and a value-added cost transmission edge is established from the processing step node to its subsequent processing step nodes or final component nodes, representing how cost changes in a certain processing step affect the cost of subsequent steps or the final product.

[0097] For each established causal edge, the cost transmission function generator will configure a nonlinear cost transmission function for it. This function describes the upstream cost variables. How to non-linearly influence the cost of downstream nodes In this model, Defined as the percentage change in cost at upstream nodes, dimensionless; for example, 0.05 represents a 5% change. The output is also the cost change rate of the downstream node. Nonlinear cost transmission function. Based on historical cost data, supply chain contract terms, and expert knowledge in the field, parameters are determined through regression analysis or curve fitting methods. For example, when raw material prices fluctuate slightly, their impact on downstream costs may be linear; however, when fluctuations exceed a certain threshold, they may trigger supply chain disruptions, the activation of alternative solutions, etc., leading to an exponential or segmented, jump-like increase in impact.

[0098] Each function defines the marginal transitivity, or transitivity, on the causal edge it belongs to. The function form is:

[0099] ·

[0100] in, This represents a nonlinear cost transmission function, describing the rate of change in upstream costs. Impact on downstream cost change rates. This represents the rate of change of upstream cost variables, such as the percentage increase in raw material prices; it is a dimensionless quantity. and This is a preset threshold in the cost transmission function, representing the critical point of the upstream cost change rate. For example, Can be set to , Can be set to These thresholds reflect the differences in strategies adopted by companies in different price fluctuation ranges. , , The cost transmission coefficient is a dimensionless quantity. , It is an exponential or power parameter and is a dimensionless quantity. and It is a piecewise function at the threshold and The connection constant is determined by the left-hand limit value to eliminate numerical jumps at the segmentation points and ensure the continuity of the function. It should be noted that the above parameters can be determined by performing polynomial regression and piecewise regression analysis on at least 5000 historical purchase orders and production cost records from the past three years to capture the cost transmission characteristics with large, medium, and small fluctuations.

[0101] Finally, latency parameters and elastic buffer factors are configured between core nodes to characterize the physical transmission process of cost fluctuations in the supply chain, forming a dynamic tensor causal graph. The established causal graph is further parameterized by a software component of the physical transmission parameter configurator. For each causal edge, a latency parameter is configured. This represents the time required for upstream cost changes to propagate to downstream nodes; the delay parameter. This data originates from KPI data in the supply chain management system, such as purchase order to warehousing time, production work order completion time, and finished product outbound time. The average value is then calculated after statistical averaging and variance analysis. For example, if a raw material takes an average of 15 days from procurement to final integration into a product, then its corresponding... It is 15 days. The calculation formula is:

[0102]

[0103] in, This represents the time delay parameter, quantified in units of time.

[0104] Represents the production and processing cycle; Represents the logistics and transportation cycle; This represents safety stock or buffer time; all are expressed in physical time units, such as hours.

[0105] This is the time granularity conversion factor, used to map physical time to system time steps. For example, if the physical time unit is hours, and the system is set to a time step of 24 hours, i.e., simulated on a daily basis, then... , This represents the function for rounding up.

[0106] At the same time, an elastic buffer factor is configured for each raw material type node or processing step node. This factor quantifies the node's ability to absorb or mitigate upstream cost shocks, such as inventory levels, the number of alternative suppliers, and process flexibility. Representing the elastic buffer factor, it is a dimensionless quantity, and its value range is... The calculation formula is:

[0107]

[0108] in, It represents the inventory level, expressed as inventory days or inventory value percentage, and can be expressed as the ratio of target safety stock to actual inventory. The substitutability or switching cost of an alternative supplier can be quantified as the reciprocal of the time required to switch to an alternative supplier. This represents the percentage of redundancy in the production or logistics process. This is a nonlinear function used to calculate the elastic buffer factor by considering various factors. In this embodiment, the function can be a multilayer perceptron model, whose training data comes from the actual buffering effect of different inventory levels, backup plans, and redundancy capabilities on cost fluctuations during historical supply chain disruption events. The output value of this model is normalized to the interval [0,1], where 0 represents no buffering capability and 1 represents full buffering.

[0109] By using core nodes, nonlinear cost transmission functions, and time delay parameters and elastic buffer factor When integrated into a unified data structure, a dynamic tensor causal graph is formed, which is a multi-dimensional graph data structure in which nodes and edges carry rich dynamic attribute information.

[0110] In one embodiment of the present invention, the spectrum dynamic adjustment module is used to perform the following operations:

[0111] Based on the entity type involved in the primary supply chain event tuple, the corresponding core node in the dynamic tensor causal graph is matched; if the confidence probability of the event is greater than a preset threshold, the direct causal path in the dynamic tensor causal graph is forcibly activated, and the random perturbation parameter of the nonlinear cost transmission function is narrowed proportionally; if the confidence probability of the event is not greater than the preset threshold, the alternative path or elastic node in the dynamic tensor causal graph is activated, and the random perturbation parameter is expanded proportionally.

[0112] First, the standardized supply chain risk signal stream is input into a dynamic tensor causal graph. The standardized supply chain risk signal stream is received by the graph modulator's software component. For each primary supply chain event tuple in the signal stream, the graph modulator parses the event confidence factors it contains. The modulator determines the type of event, such as raw material price fluctuations, logistical disruptions, or factory shutdowns, the specific entities involved, and the confidence probability of the event occurring. and potential impact The graph modulator matches the event with core nodes in the dynamic tensor causal graph. Upon successful matching, the graph modulator determines the activation priority of the corresponding topological path based on the event confidence factor. The topological path activation priority is a composite score, calculated... The higher the score, the more immediate and emphasized the path driven by the corresponding event in the graph, meaning that the associated causal path will be given priority in cost transmission calculation.

[0113] Secondly, for events with a confidence factor greater than a preset threshold... For high-confidence events, the graph modulator forcibly activates direct causal paths in the dynamic tensor causal graph and narrows the random perturbation parameters of the causal transfer function. Preset threshold. Used to distinguish between high-confidence and low-confidence events, its typical value ranges from 0.7 to 0.85. This threshold is selected based on historical data backtesting to maximize the accuracy of the prediction model. A direct causal path refers to the shortest and most direct cost transmission chain from the source of an event to the cost impact on the target component. The confidence probability of an event occurring... Exceeding the preset threshold At that time, the graph modulator will identify and activate the causal path directly related to the event.

[0114] In addition, the spectrum modulator will according to The values ​​are dynamically adjusted to control the nonlinear cost transfer functions on these activated paths. The random perturbation parameters. Specifically, each Functions typically contain an implicit random error term, which follows a mean of . Standard deviation is It follows a normal distribution. Baseline standard deviation The numerical range is arrive This value is obtained by fitting the data using the maximum likelihood estimation method through residual analysis of historical cost transmission data, and typically reflects the inherent uncertainty level of this stage. For high-confidence events, the spectrum modulator will use this standard deviation. Adjusted to ,in, This is the narrowing coefficient; The function is used to ensure that the standard deviation is always positive. For a very small positive number, such as This means that when a high-confidence event occurs, the system considers its cost transmission to be more certain, thus narrowing the range of random fluctuations in the nonlinear cost transmission function.

[0115] Finally, for events with a confidence factor less than a preset threshold... For low-confidence events, the graph modulator activates alternative paths or resilient nodes in the dynamic tensor causal graph and simultaneously expands the sampling range of random perturbation parameters. Alternative paths refer to alternative solution links that can be activated when the main supply chain path is blocked, such as switching to a higher-cost alternative supplier; resilient nodes refer to nodes with high resilience buffer factors, such as inventory nodes, whose activation implies that their buffering effect is taken into account. The confidence probability of the event occurrence is... Not greater than the preset threshold At that time, the graph modulator will activate the alternative paths or resilient nodes associated with the event. Alternative paths may include alternative supplier switching cost paths, alternative transportation route cost paths, such as the cost of air freight replacing sea freight; resilient nodes may refer to nodes with high resilience buffer factors. The activation of inventory nodes or redundant capacity nodes means that their buffering capacity will participate in the cost transmission calculation. Simultaneously, the graph modulator will process all nonlinear cost transmission functions on these activated paths or nodes. The random perturbation parameters are adjusted to ,in This is an expansion factor. This indicates that the system believes there is greater uncertainty regarding the potential impact of low-confidence events, thus expanding the range of random fluctuations in the nonlinear cost transmission function to reflect a wider range of possible outcomes.

[0116] in: The standard deviation of random disturbances representing the inherent or baseline of the nonlinear cost transmission function, for example... Can be to Between, indicating cost changes arrive The inherent uncertainty. This standard deviation is fitted using historical data. The statistics were obtained in real time. The confidence probability of an event occurring is represented, and its value range is [value range missing]. . This represents a preset event confidence threshold used to distinguish between high-confidence and low-confidence events, with a value range of [value missing]. . Represents the narrowing coefficient, which is a positive real number, for example... This is used to control the narrowing effect of high-confidence events on random disturbance parameters. Simulation experiments verify that when a high-confidence event occurs, it will... Amplifying the extent of deviations exceeding the threshold by a factor of 2 reduces randomness and effectively improves model determinism. The multiplication factor is a positive real number, for example... This is used to control the amplification of random disturbance parameters by low-confidence events. This formula ensures... Higher than , The smaller; The lower , The larger the value, the more it is used to ensure that the randomness of low-confidence events is reasonably amplified in order to cover a larger space of uncertainty.

[0117] In one embodiment of the present invention, the cost propagation calculation module is used to perform the following steps:

[0118] Starting from the source node of the dynamic tensor causal graph, local cost changes are propagated downstream layer by layer, and the local cost changes are attenuated according to the elastic buffer factor. Using the real-time marginal transmission coefficient operator, combined with the value proportion weight of each core node in the bill of materials, the impact of the source disturbance after being transmitted through the activation path is weighted and aggregated to generate the total cost change rate. The time delay parameter is accumulated along the activation path in the dynamic tensor causal graph to determine the specific time point when the total cost change rate reaches the target automotive parts finished product.

[0119] First, the activated nodes in the dynamic tensor causal graph are driven to perform numerical transmission, calculating the local cost changes of each node based on the current raw material price fluctuations and a nonlinear cost transmission function. In this embodiment, the cost transmission calculation engine carrying out the computational tasks is a distributed computing framework, such as using Apache Spark or a similar parallel processing engine, to handle large-scale graph traversal and numerical computation. The software component of the cost transmission calculation engine receives the dynamic tensor causal graph and the latest event data from the standardized supply chain risk signal stream. The engine traverses all activated paths in the graph according to the set topology path activation priority and modulated random perturbation parameters.

[0120] For each activated causal edge The engine starts from its upstream node Get current cost fluctuation input For example, the percentage change in raw material prices directly stems from the magnitude of event impact in the standardized supply chain risk signaling stream, or is transmitted from local cost changes at more upstream nodes. The engine will... Substituting the nonlinear cost transmission function associated with the causal edge To account for uncertainties, the engine will... The mean is the standard deviation of the adjusted random disturbance parameter. Random variables are drawn from a normal distribution with standard deviation. Add it to The output shows that the local cost variation caused by this causal edge is reflected in the data. Calculated as Numerical propagation refers to the dissemination of cost change data across graph nodes and edges, similar to packet forwarding in a network. The engine uses a breadth-first search algorithm to propagate these local cost changes downstream, starting from the raw material nodes or macro-event nodes in the graph, ensuring that all paths with direct or indirect impact are considered. During propagation, if nodes have a resilient buffer factor... The corresponding cost changes will be based on Attenuation, i.e., the change in local cost after attenuation. .

[0121] Secondly, the cost transmission calculation engine aggregates the local cost changes of each node through a real-time marginal transmission coefficient operator to calculate the total cost change of the target automotive component relative to the baseline cost. The real-time marginal transmission coefficient operator performs weighted aggregation calculations based on the calculus chain rule. This operator analyzes the aforementioned nonlinear cost transmission function. The numerical derivative at the current running point is used to extract the marginal contribution of each local cost change to the final component cost. Specifically, this operator, at the graph aggregation layer, combines the current value of each node connected by a causal edge with the nonlinear cost transmission function at the current input value. derivative at point Calculate the sensitivity of each local cost change to the final cost, and dynamically calculate the marginal pass-through coefficient of each local cost change to the final component cost. Marginal transitivity coefficient By applying the nonlinear cost transmission function In the current input The numerical derivative at a given point is obtained by normalization, for example... The operator performs a weighted summation of local cost changes along all relevant paths to obtain the total cost change of the target automotive component. Change in total cost It is expressed as a percentage and is used to measure how the cost of a target automotive component changes relative to its preset baseline cost, such as historical average cost or standard cost.

[0122] The calculation formula is as follows:

[0123]

[0124]

[0125]

[0126]

[0127] in, The representative value is from the mean. Standard deviation is Random variables drawn from a normal distribution. Representative node The current cost fluctuation input can be the magnitude of the event impact from the standardized supply chain risk signal flow, or it can be a local cost change transmitted from upstream nodes. Represents the nodes passed through elastic buffer factor Local cost changes after attenuation. Nodes representing configuration The elastic buffer factor. This represents the change in the total cost of the target automotive component, expressed as a percentage. As representative node The value-based weighting factor in the bill of materials, which represents the proportion of the cost of a particular material or process to the total cost, is used to normalize local fluctuations into an overall impact. It represents the set of all activated paths from the source of the event to the target automotive component node. Representative set A specific path within it. Representing a path One of the edges in the middle. Representing the side The real-time marginal transfer coefficient, i.e. It is determined by the derivative of the nonlinear cost transmission function at the current operating point. This represents the weight of the component whose value is ultimately affected by this path in the bill of materials. Representative path The starting point, such as the initial disturbance magnitude at the source of raw materials or events. This represents the specific point in time at which the change in total cost reaches the target component / finished product, expressed in units of time. This represents the time of occurrence of events in a standardized supply chain risk signaling stream, where This sets a preset conversion factor for the system's time granularity units, such as a factor to convert hours to date offsets. The formula converts the accumulated latency parameters along the path into specific durations and then adds them to the event occurrence time. The final effective date can be obtained from the above. Edges representing configuration The delay parameter. This indicates that the path with the highest cumulative latency is selected from all active paths.

[0128] Finally, the cost propagation calculation engine, combining the time delay parameters in the dynamic tensor causal graph, predicts the specific time point at which the total cost change reaches the target component, generating a dynamic cost increment. For each cost propagation path that significantly impacts the target automotive component, the engine traces along that path and accumulates the time delay parameters associated with all causal edges along the path. Cumulative latency is the sum of all latency parameters along the cost propagation path, reflecting the total time required for cost fluctuations to propagate from the source to the target. Since cost events may affect target components through multiple paths, the engine identifies the critical path with the largest cumulative latency, or determines the final latency using a weighted average latency. Specific time points... Calculated as the time of occurrence of supply chain risk events Including accumulated time delay, the dynamic cost increment is determined by the change in total cost. and specific time points Together they form a structured data format for output.

[0129] In one embodiment of the present invention, the uncertainty quantification module is configured to perform the following steps:

[0130] A probabilistic model of random variables is established based on the modulated random disturbance parameters and the potential impact amplitude; multiple iterative simulations are performed based on the probabilistic model of random variables, and the frequency distribution of the simulated total cost change rate generated in each iteration is statistically analyzed; the kernel density estimation method is used to smoothly fit the frequency distribution, and the upper and lower bounds of cost fluctuations are calculated based on the preset confidence level.

[0131] First, the random perturbation parameters of the dynamic tensor causal graph after modulation are extracted, and combined with the obtained event confidence factors, a probabilistic model of random variables describing cost fluctuations is established. The software component of the uncertainty model builder receives the standard deviation of the adjusted random perturbation parameters on each activation path. Additionally, the component also receives the potential impact magnitude from the event confidence factors associated with these events. The uncertainty model builder will apply this to each activation path. Considered as an inherent source of randomness in the cost transmission process of this path, it is assumed to follow a mean of The distribution follows a normal pattern. Furthermore, the magnitude of the potential impact is considered. The uncertainty model builder models these independent sources of random disturbances through a composite stochastic process, constructing a model to describe the change in total cost. A comprehensive random variable probability model. Specific construction methods include, but are not limited to: [combining each...] The corresponding normal distribution is considered as the source of cost fluctuations along its corresponding path. Using the central limit theorem or convolution theorem, these independent or weakly correlated random variables are combined to form a composite probability distribution describing the total cost variation. If there is a strong correlation between paths, a multivariate normal distribution or a Copula function may be used to model their joint probability. This model is not simply a matter of stacking these... Instead, it considers their transmission paths, interdependencies, and... The resulting potential biases form a multidimensional joint probability distribution, whose parameters can dynamically reflect the certainty and intensity of the event's impact. In this embodiment, considering the nonlinear characteristics of the transfer function and its potential non-Gaussian distribution, the uncertainty model builder does not forcibly perform analytical merging of probability density functions, but instead utilizes Monte Carlo simulation. In each iteration, random sample values ​​on the path are generated based on the random perturbation parameters, and the accumulated error is propagated numerically. Finally, by statistically analyzing a large number of iteration results, such as using kernel density estimation (KDE), a true total cost distribution that reflects the nonlinear distortion is fitted.

[0132] Secondly, the calculated dynamic cost increment is iteratively sampled multiple times using Monte Carlo simulation to statistically determine the frequency distribution of the estimation results. Monte Carlo simulation is a computational method that estimates probability distributions or numerical problems that are difficult to solve analytically through repeated random sampling. The total cost change output is received by the software component of the Monte Carlo simulation engine. The current point value is estimated, and the probability model of random variables is built by the uncertainty model builder. The Monte Carlo simulation engine performs... Each independent simulation iteration. The number of iterations is typically set to 10,000 to 100,000 to ensure the statistical significance of the simulation results.

[0133] In each iteration, the engine re-extracts random perturbation terms for the nonlinear cost transmission function on all activated paths in the dynamic tensor causal graph, based on the probabilistic model of random variables. Its standard deviation is the modulated Then, the engine re-executes the cost propagation calculation, i.e., numerical propagation and marginal impact gradient kernel summation, to obtain the total cost change in this iteration of the simulation. The Monte Carlo simulation engine will obtain the results of each iteration. Store it. When After the next iteration, the engine will process all... Perform statistical analysis and classify them into pre-defined categories. There are several intervals, and the calculation of the value within each interval is performed. The frequency of occurrence is used to generate a frequency distribution of the estimation results. A frequency distribution is a statistical chart that divides continuous simulation results into discrete intervals, for example, using the square root rule or Sturges' formula. The quantity.

[0134] Finally, the cost confidence interval distribution function is fitted based on the frequency distribution, and the upper and lower bounds of cost fluctuations at the preset confidence level are calculated. The software component of the distribution fitting and interval calculation module receives the frequency distribution generated by the Monte Carlo simulation engine. The module employs non-parametric density estimation methods, such as kernel density estimation, or attempts to fit various parametric distributions, such as the normal distribution, gamma distribution, and Student's t-distribution, to determine the optimal cost confidence interval distribution function by minimizing the fitting error, such as through a chi-square test or a Kolmogorov-Smirnov test. Kernel density estimation is a nonparametric method used to estimate the probability density function of a random variable. It obtains a continuous density estimate by smoothing the histogram, typically using a Gaussian kernel function. Once the distribution function is fitted, the module obtains a pre-set confidence level. For example, 95%. The preset confidence level reflects the reliability of the estimation results. In industry practice, 95% is often chosen as the default value, which means that there is a 95% probability that the actual cost change will fall within the calculated range.

[0135] To calculate the upper bound of cost fluctuation and the lower realm The module solves for the cumulative distribution function of this distribution function, such that the cumulative distribution from the left side of the distribution to... The probability is Accumulated from the left side of the distribution to The probability is These two values and This constitutes the range of target automotive component cost fluctuations at a given confidence level.

[0136] The specific formula is expressed as follows:

[0137]

[0138]

[0139]

[0140] in: The distribution function representing the confidence interval of the cost is the change in total cost. The probability density function. represents the kernel density estimation function, which estimates the probability density by smoothing out the influence of each data point, where It is the width parameter of the kernel function. It is the type of kernel function, for example, Gaussian kernel. Representing Monte Carlo simulation The set of simulated values ​​for the total cost change obtained this time. This represents the inverse function of the cumulative distribution function, i.e., the quantile function. This represents the preset confidence level, usually expressed as a decimal, for example... represent The confidence level. This represents the lower bound of cost fluctuations, expressed as a percentage. This represents the upper limit of cost fluctuations, expressed as a percentage.

[0141] In one embodiment of the present invention, the monitoring and command output module is used to perform the following operations:

[0142] Compare the upper bound of the total cost change rate or cost fluctuation with the preset safety deviation threshold; if it exceeds the safety deviation threshold, search the preset strategy rule base according to the event type in the standardized supply chain risk signal flow; match the response strategy corresponding to the event type, and automatically generate procurement strategy suggestions including price lock-in agreement suggestions, safety stock adjustment suggestions, or alternative path switching suggestions.

[0143] First, the obtained dynamic cost increments are summarized with the generated cost confidence interval distribution function to produce a final cost report containing point value estimates and interval estimates. The software component of the report generator receives the dynamic cost increments and the upper bound of cost fluctuations. and the lower world and the cost confidence interval distribution function The report generator integrates this information into a structured JSON object or a visual HTML page, forming the final cost report. This report is a structured data or visual report that integrates three core elements: quantitative forecasting, uncertainty quantification, and the time dimension, clearly showing the expected cost changes of the target automotive component. For example, the report content might include: "Forecast object: power battery casing; Expected cost increment: 0.884%; 95% confidence interval: [0.52%, 1.35%]; Expected effective date: 2025-11-13; Driving events: rising copper prices, port delays."

[0144] Secondly, the final cost report is compared with preset safety control standards. If the deviation exceeds the preset safety threshold, procurement strategy recommendations are automatically generated based on the event types in the standardized supply chain risk signal flow. The software component of the risk and strategy decision-making module receives the final cost report, as well as the company's internally preset procurement contract terms and cost control strategy database. The safety threshold is an internally set percentage limit for acceptable cost fluctuations, and its value is determined based on the importance of components, cost structure, and the company's risk appetite.

[0145] The decision-making module compares the estimated total cost change rate point value and upper limit of fluctuation in the final cost report with the preset safe percentage threshold for cost fluctuation. In this example, the point value estimate is 0.884%. If the preset safe threshold is 1.0%, although the point value is not exceeded, the upper limit of 1.35% exceeds the threshold, and the system identifies it as a potential risk and triggers an alert. The decision-making module then queries the generated and stored standardized supply chain risk signal stream to find the core driving event causing this cost fluctuation. Based on the event type and the predefined strategy rule base, the decision-making module automatically generates one or more specific procurement strategy recommendations. It should be noted that the strategy rule base is an IF-THEN rule set built based on domain expert knowledge and historical experience, which maps specific supply chain event types to a set of recommended response strategies. For example, "IF event type = significant exchange rate fluctuation AND impact > 5% THEN recommends initiating hedging tools."

[0146] Finally, the final cost report and procurement strategy recommendations are sent to mobile devices, enabling closed-loop auxiliary management of dynamic costs for automotive parts. The final cost report and procurement strategy recommendations are received by a software component of the notification and push service. This service first formats this information into a message format suitable for mobile devices, for example, via email, enterprise instant messaging tools, or a dedicated mobile application. The message is sent to the mobile devices of pre-registered procurement managers, cost analysts, or senior management personnel via a secure transmission protocol. Upon receiving the push notification, the application on the mobile device alerts the user with a pop-up, vibration, or sound, and displays a core summary of the final cost report and detailed procurement strategy recommendations. Users can directly view cost fluctuation details on their mobile devices and make quick decisions or initiate the next internal approval process based on the system's strategy recommendations, thus achieving real-time, closed-loop auxiliary management from risk perception to decision support.

[0147] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0148] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An intelligent dynamic cost estimation system for automotive parts that integrates supply chain fluctuations, characterized in that: include: The data acquisition and alignment module is used to collect raw data from multi-source heterogeneous supply chains, perform timestamp alignment processing, and generate a multimodal supply chain raw dataset. The semantic association analysis module is used to perform semantic association analysis on the original dataset of the multimodal supply chain and quantify and generate a standardized supply chain risk signal stream that includes event confidence factors. The causal graph construction module is used to parse the bill of materials of the target automotive parts, extract core nodes and configure transmission parameters to construct a dynamic tensor causal graph. The graph dynamic adjustment module uses the event confidence factor in the standardized supply chain risk signal flow to perform real-time parameter space adjustment on the dynamic tensor causal graph. The cost transfer calculation module is used to drive the modulated dynamic tensor causal graph to perform numerical transfer and calculate the dynamic cost increment of the target automotive parts. The uncertainty quantification module uses Monte Carlo simulation to iteratively sample dynamic cost increments and fits the distribution function of cost confidence intervals. The monitoring and instruction output module is used to summarize dynamic cost increments and cost confidence interval distribution functions to generate cost reports, trigger early warnings based on preset thresholds, and output procurement strategy suggestions.

2. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The data acquisition and alignment module is used to perform the following steps: Real-time data on raw material futures prices, energy market indices, and real-time freight rates from logistics platforms can be captured using application programming interfaces (APIs). Unstructured text data is obtained through web crawling, and structured tuples containing time, location, and event type are extracted using natural language processing techniques. Real-time freight rate data and structured tuples are uniformly mapped to a preset standard time axis, forming a multi-dimensional data structure indexed by the time axis.

3. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The semantic association analysis module is used to perform the following steps: The original dataset of the multimodal supply chain is input into a pre-defined semantic network model to identify the causal relationships between data of different dimensions and cluster them into primary supply chain event tuples. A multimodal cross-validation engine is used to compare the information consistency of primary supply chain event tuples under different confidence sources, and an information consistency index is generated. By combining the preset information source authority, information consistency index, and historical event correlation, an event confidence factor is calculated, which includes the confidence probability of the event occurring, the potential impact range, and the timeliness of the information.

4. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The causal graph construction module is used to perform the following steps: The types of raw materials, processing steps, logistics routes, and energy consumption quotas in the raw material list are used as the core nodes of the dynamic tensor causal graph. A nonlinear cost transmission function is established between core nodes. The nonlinear cost transmission function includes marginal transmission coefficients determined by a piecewise function. Configure time delay parameters that characterize the physical transmission cycle among core nodes, and configure elastic buffer factors that quantify the ability of each node to absorb cost shocks.

5. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The dynamic adjustment module for the spectrum is used to perform the following steps: Based on the entity types involved in the primary supply chain event tuples, match the corresponding core nodes in the dynamic tensor causal graph; If the confidence probability of an event is greater than a preset threshold, the direct causal path in the dynamic tensor causal graph is forcibly activated, and the random perturbation parameter of the nonlinear cost transmission function is narrowed proportionally. If the confidence probability of an event is not greater than a preset threshold, activate alternative paths or elastic nodes in the dynamic tensor causal graph and proportionally increase the random perturbation parameters.

6. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The cost propagation calculation module is used to perform the following operations: Starting from the source node of the dynamic tensor causal graph, local cost changes are propagated downstream layer by layer, and the local cost changes are attenuated according to the elastic buffer factor. By using the real-time marginal transfer coefficient operator and combining the value proportion weight of each core node in the bill of materials, the impact of the source disturbance after being transmitted through the activation path is weighted and aggregated to generate the total cost change rate. By accumulating time delay parameters along the activation path in the dynamic tensor causal graph, the specific time point at which the total cost change rate reaches the target automotive component finished product is determined.

7. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The uncertainty quantification module is used to perform the following operations: Establish a probabilistic model of random variables based on the modulated random perturbation parameters and the potential impact amplitude; Based on a probabilistic model of random variables, multiple iterative simulations were performed, and the frequency distribution of the rate of change of the simulated total cost generated in each iteration was statistically analyzed. The frequency distribution is smoothly fitted using the kernel density estimation method, and the upper and lower bounds of cost fluctuations are calculated based on the preset confidence level.

8. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The monitoring and command output module is used to perform the following operations: Compare the upper bound of the total cost change rate or cost fluctuation with the preset safety deviation threshold; If the safety deviation threshold is exceeded, the preset strategy rule base is retrieved based on the event type in the standardized supply chain risk signal flow; Match the response strategy to the event type and automatically generate procurement strategy suggestions, including suggestions for price lock-in agreements, safety stock adjustments, or alternative path switching.

9. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, The generated cost report and procurement strategy recommendations are sent to the mobile terminal via a secure transmission protocol, and the internal approval process is initiated based on the decision instructions fed back by the mobile terminal.

10. The intelligent cost estimation system for automotive parts based on supply chain fluctuations as described in claim 1, characterized in that, Standardized supply chain risk signal flow is used as a control input and coupled with dynamic tensor causal graph. Event confidence factors are mapped to graph control parameters to adjust the activation weights of causal edges and the range of random perturbation parameters in the dynamic tensor causal graph.

Images