Rapid quotation method based on parameterized BOM and multi-dimensional cost coupling

Abstract

The invention discloses a rapid quotation method based on parameterized BOM and multi-dimensional cost coupling, and relates to the field of rapid quotation, and the rapid quotation method based on parameterized BOM and multi-dimensional cost coupling comprises the following steps: S1, obtaining data, and verifying a preset rule and a database; s2, labeling to form a data set and storing the data set; s3, preprocessing the data set to extract core associated parameters, and outputting a multi-version list; s4, presetting a scheme and a strategy, and constructing a multi-dimensional cost coupling calculation model; s5, setting an adjustment rule introduction algorithm, and generating a multi-dimensional cost rapid quotation scheme; s6, generating an accounting scheme, and constructing a cost coupling feedback correction model; and S7, performing verification adjustment, generating a report, and outputting the scheme and the report. According to the invention, the self-adaptive parameterized BOM list is generated through the three-dimensional classification labeling mechanism and the four-dimensional dynamic parameter mapping model, and the accuracy of multi-scene customized quotation is improved.

Description

Technical Field

[0001] This invention relates to the field of rapid quotation technology. Specifically, it relates to a rapid quotation method based on the coupling of parameterized BOM and multi-dimensional cost. Background Technology

[0002] In an era of customized transformation in manufacturing and collaborative upgrading of the global supply chain, rapid quotation technology for multi-scenario customization needs, as a core hub connecting customer needs and enterprise production, has been widely applied in key areas such as customized automobile modification, engineering equipment procurement, customized industrial parts, and personalized production of smart hardware. Its timeliness of quotation, accuracy of cost accounting, and flexibility of scenario adaptation are directly related to enterprise order conversion rate, production cost control efficiency, and market competitiveness, becoming a core technological bottleneck that urgently needs to be overcome in the digital transformation of manufacturing.

[0003] The deep integration of parametric BOM technology and lifecycle cost accounting methods, with its advantages of rapid response to customized needs and comprehensive coverage of cost dimensions, has demonstrated groundbreaking value in the field of rapid quotation. By integrating core technical elements such as demand parameter analysis, dynamic BOM generation, multi-stage cost coupling, and algorithm optimization decision-making, it has built an integrated collaborative system of demand-BOM-cost-quotation. This has become a key technical direction for breaking through the limitations of traditional quotation methods and improving the efficiency of quotation in customized scenarios. It is of great significance for improving customer service levels, optimizing resource allocation, and enhancing market competitiveness.

[0004] However, existing rapid quotation methods based on parametric BOMs and multi-dimensional cost coupling mostly rely on static BOM templates for quotation calculation. This lacks dynamic adaptability to customization complexity and cost sensitivity, making it difficult to match the personalized needs of diverse scenarios. Furthermore, cost calculation is often limited to a single production or procurement stage, failing to establish a multi-dimensional coupling mechanism across the entire lifecycle of design and development, procurement, production, operation and maintenance, and disposal. This results in biased and inaccurate cost calculations. Additionally, existing methods often rely on single algorithms for cost optimization, lacking a closed-loop iterative mechanism supported by historical data, and fail to establish a time-varying correlation between parameters and costs. This makes it difficult to cope with the impact of market dynamics such as raw material price fluctuations and exchange rate adjustments, ultimately leading to delayed quotation response, insufficient accuracy, poor personalization adaptability, and an inability to meet the core needs of enterprises. Currently, no effective solutions have been proposed to address these technical problems. Summary of the Invention

[0005] In response to the problems in related technologies, this invention proposes a rapid quotation method based on parameterized BOM and multi-dimensional cost coupling, so as to overcome the above-mentioned technical problems existing in the existing related technologies.

[0006] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows: A rapid quotation method based on the coupling of parameterized BOM and multi-dimensional cost includes the following steps: S1. Obtain product multi-scenario customization requirement data and product multi-scenario customization historical data, preset product scenario parameter bidirectional constraint rules and multimodal historical quotation database, and perform double verification on the customization requirement data based on the product scenario parameter bidirectional constraint rules; S2. The product multi-scenario customization requirement data after double verification is classified and labeled in three dimensions according to scenario type, customization complexity and cost sensitivity to form a product parameter labeling dataset, and then structured and stored in association with the historical quotation database. S3. Standardize and preprocess the product parameter annotation dataset, extract the core related parameters using the parameter BOM mapping algorithm, and preset the adaptive generation rules of BOM hierarchy to build a four-dimensional dynamic parameter mapping model. Input the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list. As a preferred embodiment, the standardization preprocessing of the product parameter annotation dataset, the extraction of core related parameters using a parameter BOM mapping algorithm, the pre-defined adaptive generation rules for BOM hierarchy, the construction of a four-dimensional dynamic parameter mapping model, and the input of the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list include the following steps: S31. Use the 3σ principle to remove outliers in the product parameter annotation dataset, use the K nearest neighbor interpolation method to supplement missing key parameters, use the hash deduplication method to delete duplicate data, and transform the product parameter annotation dataset to the [0,1] interval. Use one-hot encoding to convert non-numerical classification parameters into numerical vectors to form a standardized parameter dataset. S32. Preset the correlation threshold, construct the parameter BOM correlation evaluation matrix using the parameter BOM mapping algorithm, bring the standardized parameter dataset into the parameter BOM correlation evaluation matrix, and filter according to the correlation threshold to obtain the core parameter set. Then, remove redundant parameters through the variance inflation factor to obtain the core correlation parameter set. As a preferred embodiment, the preset correlation threshold is achieved by constructing a parameter BOM correlation evaluation matrix using a parameter BOM mapping algorithm, inputting the standardized parameter dataset into the parameter BOM correlation evaluation matrix, and filtering according to the correlation threshold to obtain a core parameter set. Then, redundant parameters are removed using a variance inflation factor to obtain the core correlation parameter set, which includes the following steps: S321. Perform fitting analysis on the historical data of product customization in multiple scenarios, and preset the correlation threshold based on the fitting analysis results; S322. A standardized parameter and BOM core element correlation evaluation matrix is ​​constructed using the parameter BOM mapping algorithm. The matrix elements in the correlation evaluation matrix are calculated using the Pearson correlation coefficient. The row vectors in the correlation evaluation matrix are the standardized product parameters, and the column vectors in the correlation evaluation matrix are the BOM core elements. S323. Input the standardized parameter dataset into the correlation evaluation matrix to calculate the correlation value, and filter the correlation value to obtain the initial core parameter set. Then, remove the multicollinearity redundancy in the initial core parameter set by the variance inflation factor to obtain the secondary core parameter set. S324. Sort the secondary screening core parameters in the secondary screening core parameter set according to their priority based on parameter size to form a core associated parameter set.

[0007] S33. Define a three-level basic framework consisting of top-level product assembly, mid-level functional components, and bottom-level core parts. Adapt the level of refinement to customized complex quantification values. Add priority labels to cost-sensitive parts. Define the BOM level linkage logic when parameters change through the IF-THEN rule base to form BOM level adaptive generation rules. S34. Based on the adaptive generation rules of BOM hierarchy, using historical data of product customization in multiple scenarios as samples, a four-layer architecture of input layer, feature fusion layer, mapping calculation layer and output layer is built for cross-validation and optimization to obtain dynamic mapping weight matrix. The core associated parameter set is input into the dynamic mapping weight matrix and outputs a multi-version configurable parameterized BOM list.

[0008] S4. Preset multi-dimensional full life cycle cost accounting scheme and cost parameter dynamic mapping strategy. Define the time-varying correspondence and weight coefficients of each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full life cycle cost accounting scheme through the cost parameter dynamic mapping strategy. Based on the time-varying correspondence and weight coefficients, construct a multi-dimensional cost coupling calculation model with the LSTM time series prediction model. As a preferred embodiment, the preset multi-dimensional full lifecycle cost accounting scheme and cost parameter dynamic mapping strategy define the time-varying correspondence and weight coefficients between each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full lifecycle cost accounting scheme through the cost parameter dynamic mapping strategy. The construction of a multi-dimensional cost coupling calculation model based on the time-varying correspondence and weight coefficients and the LSTM time series prediction model includes the following steps: S41. Clearly define the cost accounting dimensions for the five stages of design and development, procurement, production, operation and maintenance and disposal, define quantifiable indicators for each cost accounting dimension, and form a multi-dimensional full life cycle cost accounting solution. S42. Preset a dynamic mapping strategy for cost parameters and construct a time-varying correspondence matrix of parameter costs. The matrix uses the specifications, materials and process parameters in the multi-version configurable parameterized BOM list as row vectors, the life cycle cost accounting items as column vectors, the matrix elements as time-varying correlation coefficients, and set weight coefficients according to cost sensitivity. S43. Based on time-varying correspondence, weight coefficients and LSTM time-series prediction model, construct a multi-dimensional cost coupling calculation model with a four-layer architecture: input layer, feature fusion layer, time-series prediction layer and coupling calculation layer. As a preferred embodiment, the multi-dimensional cost-coupled computation model, which constructs a four-layer architecture of input layer, feature fusion layer, time-series prediction layer, and coupled computation layer based on time-varying correspondence, weight coefficients, and LSTM time-series prediction model, includes the following steps: S431. The input layer receives the specifications, materials and process parameters in the multi-version configurable parameterized BOM list, combines the time-varying correspondence and weight coefficients in the cost parameter dynamic mapping strategy, and outputs a standardized fusion input vector after Min-Max normalization. S432. The feature fusion layer receives the standardized fusion input vector and adopts a weighted concatenation mechanism of parametric features plus time-varying correlation features. It allocates fusion weights according to cost sensitivity and outputs a high-dimensional fusion feature vector through feature dimension alignment and redundancy removal. S433: The time series prediction layer embeds an LSTM time series prediction model, which uses historical cost time series data as the training basis, inputs a standardized fusion input vector for iterative training, learns and analyzes the cost time series change pattern, and outputs the time series trend correction coefficient. S434, the coupling calculation layer combines the parameter cost time-varying correspondence matrix, performs matrix operations to couple the high-dimensional fused feature vector with the time-series trend correction coefficient, generates multi-dimensional full life-cycle cost accounting results, and integrates them according to the hierarchical architecture of input layer, feature fusion layer, time-series prediction layer and coupling calculation layer to form a four-layer architecture multi-dimensional cost coupling calculation model.

[0009] S44. Verify and optimize the multi-dimensional cost coupling calculation model, and output the verified and optimized multi-dimensional cost coupling calculation model.

[0010] S5. Set up dynamic cost adjustment rules, introduce adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm, define multi-objective optimization function and enterprise cost constraints, and perform algorithm fusion and iterative optimization of adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm to obtain cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generate multi-dimensional cost rapid quotation scheme. As a preferred embodiment, the setting of dynamic cost adjustment rules introduces an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm, defines a multi-objective optimization function and enterprise cost constraints, and performs algorithmic fusion and iterative optimization of the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generates a multi-dimensional cost rapid quotation scheme. S51. Clarify the cost adjustment logic and cost response rules based on market factors such as raw material prices, exchange rates, and logistics freight rates, and integrate them to form dynamic cost adjustment rules; S52. Define a multi-objective optimization function with the goals of lowest cost, optimal profit, and strongest price competitiveness, and define enterprise cost constraints based on unit product profit margin, total cost ceiling, and price response time requirements. S53. Adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm are introduced for iterative optimization. The optimal solution is searched by adaptive chaotic sparrow search algorithm, and the optimization result is corrected by Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and integrated to form a multi-dimensional cost rapid quotation scheme.

[0011] As a preferred solution, the method of introducing an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm for iterative optimization, searching for the optimal solution using the adaptive chaotic sparrow search algorithm, and then using the Bayesian fusion algorithm to correct the optimization results, thereby obtaining the cost accounting benchmark value, cost fluctuation range, and trend prediction curve, and integrating them to form a multi-dimensional cost rapid quotation scheme includes the following steps: S531. Guided by the multi-objective optimization function, and combined with the enterprise cost constraints and dynamic cost adjustment rules, determine the optimization objective and boundary of the adaptive chaotic sparrow search algorithm. S532. The output of the multi-dimensional cost coupling calculation model is globally optimized by using the adaptive chaotic sparrow search algorithm, and a set of candidate optimal solutions that meet the enterprise cost constraints is selected. As a preferred embodiment, the step of globally optimizing the output of the multi-dimensional cost coupling calculation model using the adaptive chaotic sparrow search algorithm and selecting a set of candidate optimal solutions that meet the enterprise cost constraints includes the following steps: S5321. Based on the optimization objective and boundary, initialize the core parameters of the adaptive chaotic sparrow search algorithm, and use the output of the multi-dimensional cost coupling calculation model as the initial population. S5322. Perform algorithm iteration to improve population diversity through chaotic perturbation mechanism, combine the division of labor mechanism between sparrow population discoverer and joiner to perform global optimization, and use multi-objective optimization function as fitness function to calculate the fitness value of each individual and sort them. S5323. Based on the fitness value ranking results, the compliance of each cost accounting candidate scheme is checked against the enterprise cost constraints, and schemes that do not meet the constraints are eliminated. Then, the remaining compliant schemes are deduplicated to form a candidate optimal solution set.

[0012] S533. Introduce a Bayesian fusion algorithm, based on historical data of product customization in multiple scenarios, to perform deviation correction and probability verification on candidate optimal solutions, and output cost accounting benchmark value and confidence interval. S534. Combine the cost accounting benchmark value and confidence interval with the cost dynamic adjustment rules to generate the cost fluctuation range, and link it with the trend prediction curve to form a multi-dimensional cost rapid quotation solution.

[0013] S6. Extract historical candidate cost parameter sets from the multimodal historical quotation database, combine and optimize the historical candidate cost parameter sets using the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to generate personalized multi-dimensional cost accounting schemes, and construct a cost coupling feedback correction model based on the weighted summation algorithm, the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm. As a preferred embodiment, the step of extracting a set of historical candidate cost parameters from a multimodal historical pricing database, optimizing the set of historical candidate cost parameters using an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm to generate a personalized multi-dimensional cost accounting scheme, and constructing a cost coupling feedback correction model based on a weighted summation algorithm, an adaptive chaotic sparrow search algorithm, and a Bayesian fusion algorithm includes the following steps: S61. Extract the set of historical candidate cost parameters from the multimodal historical quotation database; S62. The adaptive chaotic sparrow search algorithm is used to globally optimize the historical candidate cost parameter set and select the initial parameter combination that meets the cost accuracy requirements. S63. The initial parameter combination is corrected for deviation and optimized for probability using the Bayesian fusion algorithm. Through the fusion iteration mechanism of optimization correction, the cost parameter combination is obtained, and a personalized multi-dimensional cost accounting scheme is generated based on the cost parameter combination. As a preferred embodiment, the step of using a Bayesian fusion algorithm to correct deviations and optimize probabilistics in the initial parameter combination, obtaining a cost parameter combination through an optimization-correction fusion iteration mechanism, and generating a personalized multi-dimensional cost accounting scheme based on the cost parameter combination includes the following steps: S631. Using historical data of product multi-scenario customization as a historical reference combination, calculate the cost deviation between the initial parameter combination and the historical reference combination. S632. Introduce a Bayesian fusion algorithm, construct a deviation correction model based on historical reference combination and cost deviation value, and perform probabilistic optimization on the initial parameter combination through the deviation correction model to obtain the probabilistically optimized parameter combination. S633. Through the optimization correction fusion iteration mechanism, the combination of probabilistic optimization parameters is fed back to the adaptive chaotic sparrow search algorithm for secondary optimization, and repeated iteratively until the enterprise cost constraint conditions are met, thus obtaining the cost parameter combination. S634. Substitute the cost parameter combination into the multi-dimensional full life cycle cost accounting scheme to generate a suitable personalized multi-dimensional cost accounting scheme.

[0014] S64. Based on the weighted summation algorithm, the optimization capability of the adaptive chaotic sparrow search algorithm and the correction characteristics of the Bayesian fusion algorithm are combined to construct a cost coupling feedback correction model. The input of the cost coupling feedback correction model is defined as the quotation scheme and personalized accounting data, and the output is the deviation correction coefficient.

[0015] S7. Input the personalized multi-dimensional cost accounting scheme into the cost coupling feedback correction model to verify and adjust the multi-dimensional cost rapid quotation scheme, generate a five-dimensional quotation verification report, and output the verified and adjusted multi-dimensional cost rapid quotation scheme and the five-dimensional quotation verification report.

[0016] The beneficial effects of this invention are as follows: 1. This invention acquires multi-scenario customization requirements and historical data for products, constructs a multimodal historical quotation database, and generates an adaptive parameterized BOM list by combining a three-dimensional classification and labeling mechanism and a four-dimensional dynamic parameter mapping model. Relying on a multi-dimensional full lifecycle cost accounting scheme and a time-varying correspondence matrix, it integrates an LSTM time-series prediction model to construct a cost-coupled calculation model. Combined with an adaptive chaotic sparrow search and Bayesian fusion dual-algorithm closed-loop iterative optimization strategy, it achieves accurate and efficient generation of customized quotations for multiple scenarios. This avoids the problems of static BOM generation, cost accounting limited to a single stage, fragmented parameter-cost correlation, and single optimization algorithm in traditional rapid quotation methods. It reduces the risk of large quotation deviations and low customer satisfaction caused by poor quotation-scenario adaptability and one-sided cost accounting. Furthermore, through the deep binding of multi-objective optimization functions with enterprise cost constraints and the real-time response of dynamic cost adjustment rules to market factors, it strengthens the market competitiveness and profit guarantee capability of the quotation scheme, avoids interference from invalid parameters and blind quotations, and improves the accuracy, timeliness, and adaptability of customized quotations for multiple scenarios.

[0017] 2. This invention addresses the pain points of traditional pricing methods—lack of historical data support, lack of closed-loop verification and iteration processes, and insufficient personalized adaptation—by combining a personalized multi-dimensional cost accounting scheme and a cost-coupled feedback correction model with a five-dimensional quotation verification report generation and a scheme optimization mechanism supported by historical data. This reduces repeated debugging costs and the iteration cycle of quotation schemes, improving the reliability and scalability of rapid quotations. Simultaneously, through cross-validation optimization of a four-dimensional dynamic parameter mapping model and a verification and adaptation process of a multi-dimensional cost-coupled calculation model, coupled with a global optimization and deviation correction collaborative mechanism integrating dual algorithms, it ensures the model and data readiness before quotation generation. This avoids quotation failures caused by inaccurate BOM level adaptation, disconnected cost parameter mapping, or one-sided algorithm optimization, and prevents insufficient quotation competitiveness due to delayed market response or lack of personalized adaptation. It guarantees the stability, effectiveness, and market adaptability of quotation data, reducing enterprise quotation risks and operating costs. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a rapid quotation method based on the coupling of parameterized BOM and multi-dimensional cost according to an embodiment of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, the rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to an embodiment of the present invention includes the following steps: S1. Obtain product multi-scenario customization requirement data and product multi-scenario customization historical data, preset product scenario parameter bidirectional constraint rules and multimodal historical quotation database, and perform double verification on the customization requirement data based on the product scenario parameter bidirectional constraint rules; Specifically, acquiring product customization demand data and historical data requires building a multi-channel, multi-dimensional data collection system. Demand data is collected through direct customer surveys (online questionnaires, offline interviews, demand matching meetings), order system data collection (customization parameter selection, personalized remarks), market trend analysis (industry reports, competitor customization cases), and customer feedback platforms. The data focuses on core dimensions such as scenario type, customization complexity, functional requirements, and cost expectations to ensure that the data aligns with actual application scenarios. Historical data is extracted from the company's past customization order archives, quotation records, performance data, and customer reviews. It covers information such as parameter configuration, cost structure, quotation results, and scenario adaptation effects of completed customization projects. At the same time, publicly available customization case data from the industry is integrated to form multi-source data support. After acquisition, both types of data need to be standardized, including data cleaning, deduplication, format unification, and extraction of core fields.

[0023] Based on the acquired multi-source data, pre-set two-way constraint rules for product scenario parameters are established to clarify two-way constraints such as technical feasibility (e.g., the compatibility boundary between customized parameters and existing production processes), cost constraints (e.g., cost upper limit thresholds under different scenarios), delivery cycle (e.g., the longest delivery time corresponding to customization complexity), and performance matching (e.g., the compatibility standard between functional parameters and scenario requirements). This avoids a disconnect between requirements and actual execution capabilities. At the same time, a multimodal historical quotation database is built, integrating structured data (parameter configuration, cost details, quotation amount) and unstructured data (customized solution documents, customer communication records). This supports multi-scenario data retrieval and correlation analysis. During dual verification, the completeness (no missing core parameters) and compliance (meeting technical, cost, and other constraints) of the requirement data are first verified through rules. Then, the adaptability is verified by combining historical data from the database (e.g., the rationality of parameters and the feasibility of expected costs under similar scenarios). Invalid, contradictory, or unexecutable requirement data is eliminated to ensure that the input data is accurate and reliable.

[0024] S2. The product multi-scenario customization requirement data after double verification is classified and labeled in three dimensions according to scenario type, customization complexity and cost sensitivity to form a product parameter labeling dataset, and then structured and stored in association with the historical quotation database. Specifically, the product customization requirements data after double verification need to be categorized and labeled according to three dimensions: "scenario type - customization complexity - cost sensitivity". Scenario type is based on the product application scenario. For example, industrial machinery is categorized as "mining operations / construction / logistics transportation", and consumer electronics are categorized as "home use / business office / outdoor scenarios" to ensure that the labeling matches the actual use scenario. Customization complexity is labeled with a quantitative score, which is divided into three levels: "low-medium-high" based on indicators such as the number of customization parameters, process difficulty, and technological innovation. For example, adjusting only the size is low complexity, while reconstructing the core structure is high complexity. Cost sensitivity is labeled according to the weight of the parameter's impact on the total life cycle cost. For example, key parameters such as materials and core components are labeled as high sensitivity, while non-core parameters such as appearance and color scheme are labeled as low sensitivity. After the three-dimensional labeling is completed, the requirement data and labeling results are integrated to form a product parameter labeling dataset containing "basic parameters - scenario labels - complexity level - sensitivity indicators".

[0025] The structured storage for association requires prior to the structured adaptation of the two databases. The product parameter annotation dataset is designed with standardized fields according to "requirement ID-scenario information-customized parameters-annotation results". The historical quotation database is supplemented with index fields of "scenario tag-complexity matching item-sensitivity associated cost" to achieve data field alignment. During association, the annotation dataset is associated with historical projects of similar scenarios and equal complexity in the historical quotation database through dual keyword matching of "scenario type plus core customized parameters". At the same time, the association similarity score is recorded. The storage adopts a distributed structured database, which partitions the dataset according to scenario type. The associated data is accessed in real time through index mapping, which not only ensures the standardization of data storage, but also ensures that similar historical quotation cases can be quickly traced in the future, providing accurate data association support for parameter BOM mapping and cost accounting.

[0026] S3. Standardize and preprocess the product parameter annotation dataset, extract the core related parameters using the parameter BOM mapping algorithm, and preset the adaptive generation rules of BOM hierarchy to build a four-dimensional dynamic parameter mapping model. Input the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list. In this embodiment of the application, the steps of standardizing and preprocessing the product parameter annotation dataset, extracting core related parameters using a parameter BOM mapping algorithm, pre-setting adaptive generation rules for BOM levels, constructing a four-dimensional dynamic parameter mapping model, and inputting the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list include the following steps: S31. Use the 3σ principle to remove outliers in the product parameter annotation dataset, use the K nearest neighbor interpolation method to supplement missing key parameters, use the hash deduplication method to delete duplicate data, and transform the product parameter annotation dataset to the [0,1] interval. Use one-hot encoding to convert non-numerical classification parameters into numerical vectors to form a standardized parameter dataset. Specifically, the first step in standardized preprocessing is data cleaning. When using the 3σ principle to remove outliers, the mean and standard deviation of each parameter are calculated first. Data exceeding the range of "mean ± 3 times standard deviation" are marked as outliers and removed. This accurately filters extreme data caused by measurement errors and data entry mistakes. For missing key parameters, K-nearest neighbor interpolation is used to fill in the missing data. Using Euclidean distance as an indicator, K samples with the most similar features are matched for the missing data sample. The missing value is calculated based on the parameter mean of the similar samples, ensuring that the supplemented data fits the overall data distribution. Deduplication is achieved using the hash deduplication method. A unique hash value is generated for each data entry. Duplicate data is filtered by comparing the hash values ​​and only one is retained, solving the data redundancy problem caused by repeated data entry.

[0027] After data cleaning, standardization transformation is performed. Min-Max normalization is used to transform all numerical parameters to the range [0, 1]. The formula "(parameter value - minimum value) / (maximum value - minimum value)" is used to eliminate the interference of different parameter magnitudes on subsequent algorithms. For non-numerical classification parameters such as scene type and material, one-hot encoding is used to convert them into numerical vectors. Each classification attribute is decomposed into multiple binary features. For example, "scene type = industrial / civilian" corresponds to two feature vectors. The selected category is marked as 1, and the rest are marked as 0. This preserves classification information and meets the algorithm's requirements for numerical input. The integrated and processed data form a standardized parameter dataset with no anomalies, no missing values, no redundancy, and a uniform format.

[0028] S32. Preset the correlation threshold, construct the parameter BOM correlation evaluation matrix using the parameter BOM mapping algorithm, bring the standardized parameter dataset into the parameter BOM correlation evaluation matrix, and filter according to the correlation threshold to obtain the core parameter set. Then, remove redundant parameters through the variance inflation factor to obtain the core correlation parameter set. In this embodiment of the application, the preset correlation threshold is constructed by using a parameter BOM mapping algorithm to build a parameter BOM correlation evaluation matrix, inputting the standardized parameter dataset into the parameter BOM correlation evaluation matrix, and filtering according to the correlation threshold to obtain a core parameter set. Then, redundant parameters are removed using the variance inflation factor to obtain the core correlation parameter set, which includes the following steps: S321. Perform fitting analysis on the historical data of product customization in multiple scenarios, and preset the correlation threshold based on the fitting analysis results; Specifically, the core of fitting analysis of historical data on product customization across multiple scenarios is to uncover the correlation patterns between customization parameters and target variables such as BOM elements and cost composition. First, core dimensions are selected from the historical data, retaining key fields such as scenario type, customization parameters (specifications, materials, etc.), core BOM elements, and cost data. After cleaning and removing outliers and missing values, data subsets are divided according to scenario type to ensure the analysis's relevance. The fitting model is selected to suit the data characteristics; linear correlations (such as size and material usage) use linear regression, while non-linear correlations (such as process complexity and processing cost) use multinomial regression. A model is constructed with customization parameters as independent variables and correlation targets (such as the importance of BOM elements) as dependent variables. The model is fitted using a training set and validated using a validation set to optimize it. The correlation strength is assessed using R² goodness-of-fit and residual analysis to clarify the influence weight and correlation trend of each parameter on the target variable, providing data support for setting the correlation threshold.

[0029] The preset correlation threshold should be based on the fitting analysis results and dynamically adjusted according to actual business needs. First, extract the correlation data of the parameters output by the fitting model, and statistically analyze the distribution characteristics of the correlation in different scenarios (such as mean, median, and peak interval). If the correlation is normally distributed, use "mean + 1 standard deviation" as the initial threshold. If it is skewed, use the lower limit of the high-frequency correlation interval as the benchmark, and then refine the adjustment according to the scenario type. For highly customized and complex scenarios, relax the threshold (e.g., reduce it by 10%-15%) to avoid missing potential correlation parameters. For standardized scenarios, increase the threshold to eliminate weakly correlated and redundant data. At the same time, refer to the actual effect of parameter screening in historical quotations. If the threshold is too high and core parameters are missing, or too low and invalid data is introduced, fine-tune it based on business experience, and finally determine the correlation threshold that balances correlation and practicality.

[0030] S322. A standardized parameter and BOM core element correlation evaluation matrix is ​​constructed using the parameter BOM mapping algorithm. The matrix elements in the correlation evaluation matrix are calculated using the Pearson correlation coefficient. The row vectors in the correlation evaluation matrix are the standardized product parameters, and the column vectors in the correlation evaluation matrix are the BOM core elements. Specifically, the matrix first clarifies the definitions of row and column vectors: the row vectors fully cover the standardized product parameters, including core parameters such as specifications, dimensions, material types, process requirements, and functional configurations after normalization or coding conversion; the column vectors focus on the core elements of the BOM, covering structural elements such as top-level product assemblies, mid-level functional components, and bottom-level core parts, as well as key attributes such as cost weights, process adaptability, and supply chain stability corresponding to each element, ensuring that the matrix can fully represent the correlation dimensions between parameters and the BOM system.

[0031] Subsequently, a matrix framework was built based on the parameter BOM mapping algorithm. An empty matrix was constructed according to the rule of "one standardized parameter per row and one core BOM element per column" to clarify the matrix dimensions and coverage and avoid missing key parameters or BOM elements. The core step is to calculate the matrix elements using the Pearson correlation coefficient: based on the historical data of product customization in multiple scenarios, the numerical vectors of standardized parameters and the feature data of the corresponding core BOM elements (such as cost ratio, process matching degree and other quantitative indicators) in each historical record are extracted, and substituted into the Pearson correlation coefficient formula to calculate the linear correlation between each pair, and obtain the coefficient value in the interval [-1, 1]. The closer the absolute value of the coefficient is to 1, the stronger the correlation between the standardized parameter and the core BOM element. The closer it is to 0, the weaker the correlation. The calculated correlation coefficients are filled into the corresponding positions in the matrix one by one, and finally a structured correlation evaluation matrix is ​​formed.

[0032] S323. Input the standardized parameter dataset into the correlation evaluation matrix to calculate the correlation value, and filter the correlation value to obtain the initial core parameter set. Then, remove the multicollinearity redundancy in the initial core parameter set by the variance inflation factor to obtain the secondary core parameter set. Specifically, when bringing the standardized parameter dataset into the correlation evaluation matrix, the standardized parameters of each record in the dataset are precisely matched with the matrix row vectors according to the rule of "one data point corresponds to a set of parameter vectors". The correlation values ​​between each parameter in each data point and the corresponding core elements of the BOM are calculated by extracting or weighting the matrix elements, thus forming a parameter-correlation correspondence table for each data point.

[0033] Subsequently, a screening process is performed based on a preset correlation threshold, retaining parameters with correlation values ​​higher than the threshold. For highly customized and complex scenarios, the threshold can be appropriately lowered to avoid missing potential key parameters, while for standardized scenarios, the threshold is increased to remove weakly correlated parameters. All parameters that meet the conditions are summarized and deduplicated to obtain the initial core parameter set. In order to eliminate multicollinearity redundancy, the variance inflation factor (VIF) of the initial screening set needs to be calculated: a multiple linear regression model is constructed with each parameter in the initial screening set as the independent variable and the correlation strength of the core elements of the BOM as the dependent variable. The VIF value of each parameter is calculated using a formula. Usually, VIF > 10 is used as the criterion for the existence of severe collinearity. For parameter groups with collinearity, parameters with high VIF values ​​and relatively low correlation are removed first, based on the priority of correlation values. The VIF values ​​of the remaining parameters are calculated repeatedly until all parameters have a VIF ≤ 10, finally forming a secondary core parameter set without multicollinearity redundancy.

[0034] S324. Sort the secondary screening core parameters in the secondary screening core parameter set according to their priority based on parameter size to form a core associated parameter set.

[0035] Specifically, the additional attributes of the secondary screening parameters are first sorted out: the absolute value of the Pearson correlation coefficient (correlation strength index) of each parameter in the correlation evaluation matrix, the cost sensitivity level marked in the early stage (high / medium / low corresponding to 3 / 2 / 1 points), and the weight of the parameter's influence on the core function of the product (such as the weight of core performance parameter 0.4, auxiliary parameter 0.1).

[0036] Subsequently, a weighted scoring model was constructed, with correlation strength (50%), cost sensitivity (30%), and functional impact weight (20%) as core indicators. After quantifying the attributes of each parameter, a comprehensive score was calculated. The higher the score, the greater the importance of the parameter. After scoring, the parameters were sorted in descending order of score. If parameters with the same score were encountered, the frequency of their use in historical customized data was further referenced (higher frequency takes priority). After sorting, the priority level of the parameters was marked (P1 is the highest to P3 is the lowest). Weakly correlated parameters with too low scores at the end of the sort were removed, and finally, a core correlation parameter set with "strong correlation, high business value, and low redundancy" was formed.

[0037] S33. Define a three-level basic framework consisting of top-level product assembly, mid-level functional components, and bottom-level core parts. Adapt the level of refinement to customized complex quantification values. Add priority labels to cost-sensitive parts. Define the BOM level linkage logic when parameters change through the IF-THEN rule base to form BOM level adaptive generation rules. Specifically, when defining the three-tiered basic framework of BOM, the top-level product assembly is the complete product unit to be delivered in the end, which clarifies the overall functions and performance indicators of the product, such as the "whole-house intelligent control terminal" of smart devices and the "hydraulic excavator" of construction machinery; the middle-level functional components are broken down according to core functions and are modular units that support the top-level assembly, such as the "main control module / communication module" of the terminal and the "hydraulic system / walking system" of the excavator; the bottom-level core parts are the key components of the middle-level components, focusing on the indivisible core components or basic components, such as chips, sensors, hydraulic pumps, etc.

[0038] Customized complex quantification values ​​are assigned based on indicators such as the number of parameters and process difficulty (1-10 points). In high-scoring (7-10 points) scenarios, mid-level components need to be refined to sub-modules and bottom-level parts need to have clear model specifications. In low-scoring (1-3 points) scenarios, mid-level classification can be simplified and similar bottom-level parts can be merged. For cost-sensitive parts (such as core chips and customized molds), priority labels "A / B / C" (A being the highest) are added based on cost ratio. Linkage logic is defined based on the IF-THEN rule base, such as "IF top-level product power parameters increase (>10%), THEN mid-level power module model is upgraded and the bottom-level core motor priority is adjusted to A" and "IF bottom-level part material is changed to titanium alloy, THEN related mid-level component process parameters are updated synchronously". The integrated framework definition, complexity adaptation, priority label and linkage logic form the BOM level adaptive generation rules.

[0039] S34. Based on the adaptive generation rules of BOM hierarchy, using historical data of product customization in multiple scenarios as samples, a four-layer architecture of input layer, feature fusion layer, mapping calculation layer and output layer is built for cross-validation and optimization to obtain dynamic mapping weight matrix. The core associated parameter set is input into the dynamic mapping weight matrix and outputs a multi-version configurable parameterized BOM list.

[0040] Specifically, the process begins by selecting valid samples from historical data on product customization across multiple scenarios. Outliers are cleaned and removed, and missing information is supplemented. Complete data containing core correlation parameters, customized complexity metrics, cost sensitivity indicators, and corresponding BOM structures is retained as the foundational sample set for model training and optimization. Subsequently, a four-layer architecture is constructed: the input layer receives core correlation parameters and auxiliary features such as customized complexity and cost sensitivity from the sample set. The feature fusion layer, according to the BOM hierarchy adaptive rules, weights and fuses parameter features with hierarchy adaptation requirements and cost priority features, outputting a high-dimensional fused feature vector through dimension alignment and redundancy removal. The mapping calculation layer embeds BOM hierarchy linkage logic, learns the mapping relationship between core parameters and elements at each level of the BOM based on historical samples, and iteratively optimizes the mapping coefficients using a gradient descent algorithm. The output layer initially generates a candidate parameterized BOM list.

[0041] To improve mapping accuracy, K-fold cross-validation is used to optimize the four-layer architecture. By adjusting parameters such as feature fusion weights and linkage logic thresholds, the deviation between the BOM list and the historical actual BOM is minimized, ultimately resulting in a dynamic mapping weight matrix (matrix elements represent the correlation strength between core parameters and elements at each level of the BOM). The core correlation parameter set is input into this matrix, and combined with customized complexity quantification values ​​and cost sensitivity priorities for different scenarios, the BOM level refinement is automatically adapted to generate multiple versions of configurable parameterized BOM lists (such as a BOM version refined to sub-modules for high complexity scenarios, and a BOM version marked with A-level priority parts for high cost sensitivity scenarios), achieving dynamic adaptation between parameters and BOM structure.

[0042] S4. Preset multi-dimensional full life cycle cost accounting scheme and cost parameter dynamic mapping strategy. Define the time-varying correspondence and weight coefficients of each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full life cycle cost accounting scheme through the cost parameter dynamic mapping strategy. Based on the time-varying correspondence and weight coefficients, construct a multi-dimensional cost coupling calculation model with the LSTM time series prediction model. In this embodiment of the application, the preset multi-dimensional full life cycle cost accounting scheme and cost parameter dynamic mapping strategy define the time-varying correspondence and weight coefficients between each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full life cycle cost accounting scheme through the cost parameter dynamic mapping strategy, and construct a multi-dimensional cost coupling calculation model based on the time-varying correspondence and weight coefficients and the LSTM time series prediction model, including the following steps: S41. Clearly define the cost accounting dimensions for the five stages of design and development, procurement, production, operation and maintenance and disposal, define quantifiable indicators for each cost accounting dimension, and form a multi-dimensional full life cycle cost accounting solution. Specifically, clarifying the cost accounting dimensions for the five stages of the entire life cycle requires focusing on the core cost components of each stage, defining indicators based on quantifiable targets, and ultimately forming a systematic accounting scheme. The design and development stage focuses on "controllable early investment," and the accounting dimensions include design fees, depreciation of R&D equipment, testing and verification costs, and patent application fees. Quantifiable indicators are set as design cycle (days), R&D personnel man-hours (person-hours), number of tests (times), and unit design cost (yuan / item). The procurement stage revolves around "supply chain cost optimization," and the dimensions cover raw material procurement costs, supplier service costs, and transportation and warehousing costs. The indicators are refined as raw material unit price (yuan / unit), procurement batch (unit), transportation loss rate (%), and supplier on-time delivery rate (%).

[0043] The production stage focuses on "precise process costs," with accounting dimensions including depreciation of production equipment, direct labor costs, energy consumption costs, and scrap losses. The indicators are unit product man-hours (hours / unit), energy consumption (kWh / unit), scrap rate (%), and production line utilization rate (%). The operation and maintenance stage aims for "controllable service costs," with dimensions including repair parts costs, labor service costs, and remote operation and maintenance costs. The indicators are set as average number of repairs per year (times / unit), cost per repair (yuan), and operation and maintenance response time (hours). The scrapping stage revolves around "environmental protection and recycling value," calculating recycling and disposal fees, environmental treatment fees, and material residual value. The indicators are recycling rate (%), environmental treatment cost (yuan / unit), and residual value revenue (yuan / unit). By integrating the five stages of dimensions and indicators, clarifying data collection rules, accounting formulas, and weight allocation (such as higher weights for the production and procurement stages), a multi-dimensional, full life-cycle cost accounting solution is formed.

[0044] S42. Preset a dynamic mapping strategy for cost parameters and construct a time-varying correspondence matrix of parameter costs. The matrix uses the specifications, materials and process parameters in the multi-version configurable parameterized BOM list as row vectors, the life cycle cost accounting items as column vectors, the matrix elements as time-varying correlation coefficients, and set weight coefficients according to cost sensitivity. Specifically, the preset cost parameter dynamic mapping strategy needs to revolve around the three-dimensional correlation of "parameter-cost-time". First, the data collection and update mechanism should be clarified: integrate customized historical cost data of products in multiple scenarios, market dynamic data (raw material prices, exchange rates, freight rate fluctuations) and cost feedback data of each stage of the entire life cycle, establish a real-time updated basic database, and define the calculation cycle and adjustment threshold of the time-varying correlation coefficient (such as monthly updates, and immediate adjustment triggered when the fluctuation exceeds 5%) to ensure that the strategy adapts to changes in the market and time dimensions.

[0045] Based on this, a time-varying correlation matrix of parameter costs is constructed: the row vectors accurately extract the core parameters from the multi-version configurable parameterized BOM list, covering three key parameters: specifications (size, power, precision, etc.), materials (steel type, alloy type, plastic grade, etc.), and processes (stamping, welding, 3D printing, etc.); the column vectors correspond to the specific cost accounting items in the five stages of the entire life cycle, such as the test fees in the design and development stage, the raw material costs in the procurement stage, the energy consumption costs in the production stage, the maintenance costs in the operation and maintenance stage, and the environmental treatment costs in the scrapping stage, etc., while the matrix elements are time-varying correlation coefficients. Through regression analysis, time series prediction and other methods, the influence of each BOM parameter on the corresponding cost accounting item at different time nodes is quantified (the coefficient value range is [0, 1], the closer to 1, the stronger the influence). For example, when the material is upgraded from ordinary steel to high-strength alloy, the correlation coefficient of procurement cost and processing cost increases simultaneously.

[0046] Finally, weighting coefficients are set according to cost sensitivity: parameters previously marked as highly sensitive (such as core materials and key processes) are assigned higher weights (0.6-1.0), medium-sensitive parameters are assigned medium weights (0.3-0.6), and low-sensitive parameters are assigned low weights (0-0.3). The total weight coefficient is 1. By weighting calculation, the impact of key parameters on cost accounting is strengthened, and a complete dynamic mapping system of "parameter-time-varying correlation-weight adaptation" is finally formed.

[0047] S43. Based on time-varying correspondence, weight coefficients and LSTM time-series prediction model, construct a multi-dimensional cost coupling calculation model with a four-layer architecture: input layer, feature fusion layer, time-series prediction layer and coupling calculation layer. In this embodiment of the application, the multi-dimensional cost coupling calculation model, which constructs a four-layer architecture of input layer, feature fusion layer, time series prediction layer and coupling calculation layer based on time-varying correspondence, weight coefficients and LSTM time series prediction model, includes the following steps: S431. The input layer receives the specifications, materials and process parameters in the multi-version configurable parameterized BOM list, combines the time-varying correspondence and weight coefficients in the cost parameter dynamic mapping strategy, and outputs a standardized fusion input vector after Min-Max normalization. Specifically, the system first accurately receives the core parameters from multiple versions of configurable parameterized BOM lists, covering specifications (quantitative parameters such as size, power, and accuracy level), materials (categorized parameters such as steel type and alloy type, which have been converted into numerical vectors through unique thermal encoding), and processes (process types such as stamping, welding, and 3D printing, and their corresponding process parameters), ensuring that the parameters cover key cost-related dimensions.

[0048] Subsequently, based on the dynamic mapping strategy of cost parameters, the row vectors of the time-varying correspondence matrix of the received parameters are precisely matched with the parameters. The time-varying correlation coefficients of each parameter and each cost accounting item in the whole life cycle are extracted. Then, the time-varying correlation coefficients are weighted according to the preset weight coefficients (0.6-1.0 for high cost sensitivity parameters, 0.3-0.6 for medium sensitivity parameters, and 0-0.3 for low sensitivity parameters) to strengthen the influence weight of key parameters on cost accounting. In order to eliminate the interference of different parameter scale differences on subsequent model calculations, the weighted time-varying correlation feature vector is subjected to Min-Max normalization processing. All feature values ​​are mapped to the interval [0, 1] by the formula "(feature value - minimum value) / (maximum value - minimum value)". Finally, the standardized weighted features corresponding to specifications, materials and processes are integrated to form a standardized fusion input vector with unified dimensions and standardized values, providing high-quality data support for further processing of the feature fusion layer.

[0049] S432. The feature fusion layer receives the standardized fusion input vector and adopts a weighted concatenation mechanism of parametric features plus time-varying correlation features. It allocates fusion weights according to cost sensitivity and outputs a high-dimensional fusion feature vector through feature dimension alignment and redundancy removal. Specifically, after receiving the standardized fusion input vector from the input layer, the feature fusion layer first performs feature splitting on the vector, clearly separating two types of core features: one type is the parameter features of the inherent attributes of the product (including standardized numerical vectors of specifications, materials, and processes), and the other type is the time-varying correlation features that reflect the dynamic correlation of costs (time-varying correlation coefficient vectors after weighting), laying the foundation for weighted splicing.

[0050] Subsequently, a weighted concatenation mechanism of "parameter features plus time-varying correlation features" is adopted, and the fusion weights are allocated according to cost sensitivity: high cost sensitivity parameter features and time-varying correlation features of the same parameter are assigned a high fusion weight of 0.6-1.0, medium sensitivity features are assigned a medium weight of 0.3-0.6, and low sensitivity features are assigned a low weight of 0-0.3. The sum of the weight ratios of the two types of features for the same parameter is 1, ensuring that the key features dominate after fusion. Before concatenation, feature dimension alignment is performed. Through zero-padding or feature dimensionality reduction techniques, the dimensions of parameter features and time-varying correlation features are made completely consistent to avoid fusion deviation caused by dimension mismatch. Then, variance screening and mutual information analysis methods are used to remove redundancy, deleting invalid features with variance less than a preset threshold and highly correlated features with excessively high mutual information values ​​(greater than 0.8), reducing the interference of data redundancy on subsequent model calculations. Finally, the two types of features after weight allocation, dimension alignment and redundancy removal are concatenated and integrated to output a high-dimensional fusion feature vector with unified dimensions, prominent key features and no redundant information.

[0051] S433: The time series prediction layer embeds an LSTM time series prediction model, which uses historical cost time series data as the training basis, inputs a standardized fusion input vector for iterative training, learns and analyzes the cost time series change pattern, and outputs the time series trend correction coefficient. Specifically, when embedding the LSTM time series prediction model into the time series prediction layer, the training data is first preprocessed: historical cost time series data for each stage of the entire life cycle is extracted from the multimodal historical pricing database (sequences are divided into monthly / quarterly units according to the time dimension), covering core time series information such as raw material price fluctuations, changes in production energy consumption, and operation and maintenance cost trends. After removing outliers and filling in missing data through interpolation, the data is precisely aligned with the standardized fusion input vector output by the input layer according to the time dimension and divided into a training set (accounting for 70%-80%) and a validation set (accounting for 20%-30%) to ensure the continuity of data time series and feature matching degree.

[0052] Subsequently, an LSTM model architecture adapted to the cost prediction scenario was constructed: 2-3 hidden layers were set (each containing 64-128 hidden units), and a dropout layer was embedded (dropout rate set to 0.2-0.3) to avoid overfitting. The input gate, forget gate, and output gate were adapted to the needs of capturing short-term fluctuations, remembering long-term trends, and selecting key features in cost time series data, respectively. During the model training phase, the standardized fused input vector was input into the LSTM model, with mean squared error (MSE) as the loss function, and the Adam optimizer was used to dynamically adjust the learning rate (initial learning rate 0.001-0.01). Through multiple rounds of iterative training, the model gradually learns the time-series changes in costs, including seasonal fluctuations in raw material prices, cost reduction trends caused by process upgrades, and sudden fluctuations caused by changes in market supply and demand. At the same time, the model's prediction accuracy is monitored in real time using a validation set. Training stops when the loss function value converges to a preset threshold (e.g., MSE < 0.01). The finally trained LSTM model outputs a quantized time-series trend correction coefficient (range 0.8-1.2) based on the standardized fused input vector. A coefficient greater than 1 indicates an upward trend in costs, less than 1 indicates a downward trend in costs, and close to 1 indicates stable costs.

[0053] S434, the coupling calculation layer combines the parameter cost time-varying correspondence matrix, performs matrix operations to couple the high-dimensional fused feature vector with the time-series trend correction coefficient, generates multi-dimensional full life-cycle cost accounting results, and integrates them according to the hierarchical architecture of input layer, feature fusion layer, time-series prediction layer and coupling calculation layer to form a four-layer architecture multi-dimensional cost coupling calculation model.

[0054] Specifically, the high-dimensional fusion feature vector (including standardized features of specifications, materials, and processes, as well as time-varying correlation weights) output by the feature fusion layer is first converted into a column vector form. This vector is then multiplied element-wise with the time-series trend correction coefficient (matching corresponding coefficients according to the cost accounting item dimension) output by the time-series prediction layer to achieve time-series trend correction of the feature vector. Next, the corrected feature vector is multiplied with the parameter cost time-varying correspondence matrix. By weighted summation of matrix elements, the cost contribution of each BOM parameter in each stage of the entire life cycle (design and development, procurement, production, etc.) is quantified, generating a multi-dimensional full life cycle cost accounting result that includes cost details for each stage, total cost summary, and cost fluctuation warning.

[0055] During architecture integration, four logical connections are clearly defined: the input layer provides standardized parameters and core BOM data as the model data entry point; the feature fusion layer performs feature purification and weighted fusion on the input data to output high-dimensional features; the time series prediction layer trains an LSTM model based on historical time series data to provide time series correction basis for cost accounting; the coupled calculation layer takes over the previous output, completes cost coupling calculation and outputs the results, and deeply integrates the four functional modules and data flow logic by defining data interface standards for each layer and optimizing the transmission efficiency of features and coefficients, ultimately forming a multi-dimensional cost coupling calculation model with a closed loop of "input-fusion-prediction-computation", ensuring that cost accounting has both accuracy, dynamism and full life cycle coverage.

[0056] S44. Verify and optimize the multi-dimensional cost coupling calculation model, and output the verified and optimized multi-dimensional cost coupling calculation model.

[0057] Specifically, the first step is to build a comprehensive validation dataset: select historical project data (accounting for 30%) from the multimodal historical quotation database, covering different scenario types, customization complexity, and cost sensitivity, and combine it with real cost data from 3-5 actual customized projects. The dataset is then split along the entire "parameter-BOM-cost" chain to form a three-dimensional validation data system that includes training set, validation set, and test set, ensuring data coverage and authenticity.

[0058] Subsequently, multi-dimensional validation metrics were established: in addition to general algorithm metrics such as mean squared error (MSE) and mean absolute error (MAE), new business scenario metrics were added (such as cost accounting deviation rate at each stage, total cost prediction accuracy, and time series trend fit), constructing a two-dimensional evaluation system of "algorithm accuracy plus business adaptation." The validation method combines K-fold cross-validation (K=5-10) with time series cross-validation to verify the model's generalization ability while ensuring the trend capture accuracy of the time series prediction layer. Simultaneously, through comparative validation with real-world cases, the model's output cost accounting results were compared one by one with real project cost data to pinpoint the sources of deviation. During the optimization phase, iterative adjustments were made to address the issues exposed during validation: if the time series trend... If the difference is large, adjust the number of hidden layer units, dropout rate, and learning rate of the LSTM model; if the cost coupling accuracy is insufficient, correct the time-varying correlation coefficient and weight allocation of the time-varying correspondence matrix of parameters and costs; if feature redundancy leads to model inefficiency, optimize the dimension alignment and redundancy removal threshold of the feature fusion layer. After multiple iterations (3-5 rounds) until the validation metrics meet the standards (e.g., MSE < 0.005, cost deviation rate < 3%), solidify the model structure, parameter configuration, and data processing flow, output the validated and optimized multi-dimensional cost coupling calculation model, and simultaneously attach a model validation report (including metric compliance status, optimization logs, and applicable scenario descriptions) to ensure that the model can be directly used for customized cost accounting in multiple scenarios.

[0059] S5. Set up dynamic cost adjustment rules, introduce adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm, define multi-objective optimization function and enterprise cost constraints, and perform algorithm fusion and iterative optimization of adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm to obtain cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generate multi-dimensional cost rapid quotation scheme. In this embodiment of the application, the setting of dynamic cost adjustment rules introduces an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm, defines a multi-objective optimization function and enterprise cost constraints, and performs algorithm fusion and iterative optimization of the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generate a multi-dimensional cost rapid quotation scheme: S51. Clarify the cost adjustment logic and cost response rules based on market factors such as raw material prices, exchange rates, and logistics freight rates, and integrate them to form dynamic cost adjustment rules; Specifically, the core logic of cost adjustment is first established: using the benchmark cost output by the multi-dimensional cost coupling calculation model as the anchor point, two scenarios are divided into "benchmark periodic adjustment" and "emergency immediate adjustment", and the adjustment trigger threshold is clarified (such as initiating periodic adjustment when the benchmark cost deviation rate exceeds 5% and initiating immediate adjustment when the daily fluctuation of market factors exceeds 8%). The responsible entities for the entire process from data collection and impact assessment to adjustment implementation are also standardized.

[0060] Next, the cost response rules for the three major market factors are refined: Regarding raw material prices, they are graded according to material sensitivity (core metals are grade 1, and general plastics are grade 2). When the price fluctuation of grade 1 materials exceeds 5%, the procurement and production costs are adjusted synchronously according to the correlation coefficient in the parameter cost time-varying correspondence matrix. If the inventory exceeds 30 days, the impact is amortized in batches. Regarding exchange rates, for major settlement currencies such as the US dollar and the euro, an adjustment is triggered when the exchange rate fluctuation exceeds 3%. For export products, the exchange rate cost is calculated based on the settlement cycle weighted average, while for imported raw materials, the procurement cost is directly linked and updated in real time. Regarding logistics freight rates, differentiating between road and sea transportation modes, when the freight rate index fluctuates by more than 10%, the logistics cost is adjusted according to the formula "weight × distance × fluctuation coefficient" based on the customized product's transportation weight, distance, and transportation-sensitive parameters in the BOM list.

[0061] Finally, a rule system is formed by integrating the adjustment logic and response rules of each factor into the cost management system, constructing a standardized module of "trigger condition - response formula - adjustment range - execution time", and simultaneously establishing a dynamic rule update mechanism (verifying the rationality of thresholds monthly in conjunction with market data) to ensure that cost adjustments not only accurately match market changes, but also meet the customized cost accounting needs of multiple scenarios.

[0062] S52. Define a multi-objective optimization function with the goals of lowest cost, optimal profit, and strongest price competitiveness, and define enterprise cost constraints based on unit product profit margin, total cost ceiling, and price response time requirements. Specifically, defining a multi-objective optimization function requires first quantifying and establishing a synergistic relationship between the three objectives of "lowest cost, optimal profit, and strongest price competitiveness" to avoid objective conflicts. The first objective is to minimize the total cost (C) throughout the entire product lifecycle, which is quantified as the sum of costs in five stages, including design, procurement, and production. The second objective is to maximize the profit per unit product (P), which is the difference between the quoted price (Q) and the total cost. The third objective is to maximize price competitiveness (R), which is measured by the deviation rate (|Q-Q0| / Q0) between the quoted price and the average price of similar products in the industry. The lower the deviation rate, the stronger the competitiveness.

[0063] Subsequently, dynamic weight coefficients (ω1, ω2, ω3, and ω1+ω2+ω3=1) are introduced. These weights are adjusted in conjunction with the company's strategy, with ω3 being higher during the growth phase (e.g., 0.4) and ω2 being higher during the maturity phase (e.g., 0.4). This ultimately forms the optimization function: Minimize(ω1×C-ω2×P+ω3×(|Q-Q0| / Q0)). By converting the sign, the multiple objectives are unified into a minimization problem. The company's cost constraints must closely align with actual operational requirements: a minimum threshold is set for the unit product profit margin (e.g., not less than 15%), i.e., (QC) / Q≥r (r is the minimum profit margin per unit of output). Minimum profit margin); Total cost ceiling constraints clearly stipulate that the total cost of a single batch or unit product must not exceed a preset value (C≤Cmax, where Cmax is determined in conjunction with production capacity and financial status); Quotation response time constraints stipulate the maximum time limit from receiving demand to outputting quotation (e.g., within 24 hours), and convert it into model calculation efficiency constraints (algorithm iteration time≤t, where t is 70% of the response time, with reserved review time). Constraints must be linked with the optimization function to ensure that the optimization process meets both the company's profitability and operational requirements, and achieves a dynamic balance of the three objectives, providing a quantitative basis for quotation decisions.

[0064] S53. Adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm are introduced for iterative optimization. The optimal solution is searched by adaptive chaotic sparrow search algorithm, and the optimization result is corrected by Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and integrated to form a multi-dimensional cost rapid quotation scheme.

[0065] In this embodiment of the application, the process of introducing an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm for iterative optimization, searching for the optimal solution using the adaptive chaotic sparrow search algorithm, and then using the Bayesian fusion algorithm to correct the optimization results, to obtain the cost accounting benchmark value, cost fluctuation range, and trend prediction curve, and integrating them to form a multi-dimensional cost rapid quotation scheme includes the following steps: S531. Guided by the multi-objective optimization function, and combined with the enterprise cost constraints and dynamic cost adjustment rules, determine the optimization objective and boundary of the adaptive chaotic sparrow search algorithm. Specifically, when determining the optimization objectives and boundaries of the adaptive chaotic sparrow search algorithm based on the multi-objective optimization function, the core demands of the function are first transformed into quantifiable objectives that the algorithm can execute. Then, the optimization range is defined by combining constraints and adjustment rules. The optimization objectives are directly anchored to the three core objectives of "lowest cost, optimal profit, and strongest price competitiveness". The minimization objective of "ω1×C-ω2×P+ω3×(|Q-Q0| / Q0)" in the multi-objective optimization function is taken as the core optimization direction of the algorithm, and simultaneously decomposed into three sub-objectives: first, the total cost of the entire life cycle (C) is lower than the preset upper limit Cmax; second, the unit product profit (P) is not lower than the threshold corresponding to the minimum profit rate r; and third, the deviation rate between the price (Q) and the industry average price (Q0) is controlled within ±5% to ensure competitiveness. The algorithm balances the conflicts of sub-objectives through the chaos mechanism to avoid the overall benefits being damaged due to the optimization of a single objective.

[0066] Determining the optimization boundary requires dual anchoring constraints and rules: the hard boundary is set based on enterprise cost constraints, transforming the quotation response time into an upper limit for the number of algorithm iterations (e.g., convergence within 200 iterations), and the unit product profit margin constraint into the boundary of the difference between "quotation and total cost" (QC≥Q×r), with the upper limit of total cost directly serving as the cost parameter boundary (C≤Cmax); the dynamic boundary is generated in conjunction with dynamic cost adjustment rules, such as raw material price fluctuations corresponding to a ±10% fluctuation range for procurement cost parameters, exchange rate fluctuations corresponding to a ±8% boundary for imported component costs, and logistics freight rate fluctuations corresponding to a ±12% range for transportation costs. The algorithm updates the boundary range in real time through an adaptive mechanism, while using the time-varying correlation coefficient in the parameter cost time-varying correspondence matrix as a supplement to the boundary constraints, ensuring that the correlation logic between parameters and costs in the optimization process conforms to reality, ultimately forming an algorithm optimization framework that is "clear in objectives, controllable in boundaries, and dynamically adaptable".

[0067] S532. The output of the multi-dimensional cost coupling calculation model is globally optimized by using the adaptive chaotic sparrow search algorithm, and a set of candidate optimal solutions that meet the enterprise cost constraints is selected. In this embodiment of the application, the step of globally optimizing the output of the multi-dimensional cost coupling calculation model using the adaptive chaotic sparrow search algorithm and screening the candidate optimal solution set that meets the enterprise cost constraints includes the following steps: S5321. Based on the optimization objective and boundary, initialize the core parameters of the adaptive chaotic sparrow search algorithm, and use the output of the multi-dimensional cost coupling calculation model as the initial population. Specifically, when initializing the core parameters of the adaptive chaotic sparrow search algorithm based on the optimization objectives (lowest cost, best profit, strongest price competitiveness) and boundaries (cost upper limit, profit margin threshold, etc.), the optimization objectives should adhere to the principles of "adaptation to objectives, controllable boundaries, and efficient optimization". At the same time, the output of the multi-dimensional cost coupling calculation model should be transformed into a high-quality initial population.

[0068] Regarding the initialization of core parameters, the population size is set to 50-80 individuals based on the complexity of the optimization target. Each individual corresponds to a set of cost-quotation schemes to ensure coverage of multiple scenarios with high / low customization complexity and different cost sensitivities. The chaos factor is set to 0.3-0.6 and adjusted in conjunction with the dynamics of the optimization boundary (such as the fluctuation range of raw material prices) to enhance population diversity and avoid local optima. The initial value of the adaptive weight is allocated according to the target priority. The weight of the quotation competitiveness is set to 0.4 during the growth stage and the profit weight is adjusted to 0.4 during the maturity stage, and the weight is dynamically updated with iteration. The number of iterations is set to 200-300 times based on the quotation response time constraint to ensure that the optimization is completed within the 24-hour response time limit.

[0069] The initial population construction is based on the output of the multi-dimensional cost coupling calculation model: extract key data such as the full life cycle cost, unit profit, and quotation amount corresponding to the multiple versions of BOM output by the model, encapsulate each "cost composition-profit value-quotation scheme" into a sparrow individual, and the individual characteristics cover the cost details of five stages such as procurement and production, as well as the deviation rate of quotation from the industry average price. At the same time, the population is screened according to the enterprise constraints, and invalid individuals with profit margins below 15% and costs exceeding the upper limit are removed. Derivative schemes that meet the boundaries (such as cost variants that adjust material parameters) are added to finally form an initial population that is reasonably sized, comprehensive, and compliant with constraints.

[0070] S5322. Perform algorithm iteration to improve population diversity through chaotic perturbation mechanism, combine the division of labor mechanism between sparrow population discoverer and joiner to perform global optimization, and use multi-objective optimization function as fitness function to calculate the fitness value of each individual and sort them. Specifically, the first step is to activate the chaotic perturbation mechanism: a chaotic sequence is generated based on the Logistic chaotic mapping formula, and the parameters of 30%-40% of the individuals with poor fitness values ​​in the current population are perturbed. Within the optimization boundary, their characteristic parameters such as cost composition and bidding scheme are slightly and randomly adjusted. This not only avoids the population from getting stuck in local optima, but also improves the population diversity through the randomness of the chaotic sequence, ensuring that the algorithm can explore more potential optimal solutions.

[0071] Subsequently, global optimization is performed based on the discoverer-joiner division of labor mechanism of the sparrow population: 20%-30% of high-quality individuals are selected as discoverers based on their fitness values, responsible for exploring new optimization areas and updating their positions based on cost dynamic adjustment rules (such as raw material price fluctuation ranges) to find better fitness values; the remaining individuals act as joiners, updating their own states by imitating the positions of discoverers or using random search strategies. Some joiners (10%-15%) will deviate from the discoverer's trajectory to conduct local exploration, balancing the efficiency of global search and local development. During the iteration process, a multi-objective optimization function is used as the fitness function. The "cost-profit-price competitiveness" parameters corresponding to each sparrow individual are substituted into the function to calculate a quantified fitness value. This value directly reflects the degree to which the individual fits the goal of "lowest cost, best profit, and strongest price competitiveness". Finally, individuals are sorted from smallest to largest fitness value (because the optimization function is a minimization objective), and the top 50% of high-quality individuals are retained to enter the next round of iteration, while individuals with poor fitness values ​​are eliminated, providing a better population foundation for subsequent iterations and gradually approaching the global optimum.

[0072] S5323. Based on the fitness value ranking results, the compliance of each cost accounting candidate scheme is checked against the enterprise cost constraints, and schemes that do not meet the constraints are eliminated. Then, the remaining compliant schemes are deduplicated to form a candidate optimal solution set.

[0073] Specifically, the first step is to conduct compliance verification on all ranked cost accounting candidate schemes. The core of the verification is to strictly adhere to the enterprise's cost constraints: verify whether the unit product profit margin of each scheme reaches the preset minimum threshold (e.g., not less than 15%), whether the total cost of the entire life cycle is controlled within the preset upper limit, and whether the quotation response time meets the 24-hour time limit requirement. At the same time, in conjunction with the dynamic cost adjustment rules, the cost adaptability under the fluctuations of market factors such as raw material prices and exchange rates is further verified (e.g., the cost after fluctuation adjustment does not exceed the dynamic boundary), to ensure that the schemes not only meet static constraints but also adapt to dynamic market changes.

[0074] For solutions found during the verification process that fail to meet profit margin standards, exceed total cost limits, have timeouts, or fail to adapt dynamic costs, they are directly marked as non-compliant and removed to prevent invalid solutions from entering subsequent stages. After the compliant solutions are screened, a deduplication strategy of "core parameter hash similarity comparison" is adopted: key features of each solution (such as total cost, quoted amount, core BOM cost ratio, and profit value) are extracted to generate unique hash values. At the same time, the cosine similarity of key features between solutions is calculated. If the hash values ​​are duplicated or the similarity is higher than 95%, it is determined to be a duplicate solution, and only the one with the better fitness value is retained. Finally, the solutions that have passed the compliance verification and have no duplicates are integrated to form a candidate optimal solution set that covers different customized scenarios, has high fitness values, and combines compliance and diversity.

[0075] S533. Introduce a Bayesian fusion algorithm, based on historical data of product customization in multiple scenarios, to perform deviation correction and probability verification on candidate optimal solutions, and output cost accounting benchmark value and confidence interval. Specifically, firstly, project data that is highly similar to the current customized scenario (such as specification type, material grade, and process difficulty) is extracted from the historical database. The corresponding cost accounting results, actual implementation deviations, and market environment parameters are then selected. The probability distribution of historical cost deviations is fitted by kernel density estimation and used as prior information for Bayesian fusion (such as assuming that the deviation follows a normal prior distribution).

[0076] Subsequently, each scheme in the candidate optimal solution set is used as an observation. Combining the prior distribution and the likelihood function (constructed based on the mapping relationship between "candidate solution-actual cost" in historical data), the posterior probability distribution of the cost accounting results of each scheme is calculated using Bayes' formula to correct the bias. For accounting biases that frequently occur in historical data (such as underestimation of raw material costs and omission of operation and maintenance costs), the cost composition of the candidate solutions is adjusted through the posterior distribution to reduce systematic errors. Then, probability verification is performed: the confidence level of each corrected scheme is calculated based on the posterior distribution, a 95% confidence level is set, and the confidence interval of the cost accounting results is obtained through integration. The narrower the interval, the higher the reliability of the result. Finally, the schemes are sorted according to the size of the posterior probability, and the corrected result with the largest posterior probability is selected as the cost accounting benchmark value. Its corresponding 95% confidence interval is output simultaneously to form an accounting result with "clear benchmark value and controllable fluctuation range".

[0077] S534. Combine the cost accounting benchmark value and confidence interval with the cost dynamic adjustment rules to generate the cost fluctuation range, and link it with the trend prediction curve to form a multi-dimensional cost rapid quotation solution.

[0078] Specifically, using the cost accounting benchmark as the core, the initial cost fluctuation range is first defined by combining it with a 95% confidence interval (upper limit = benchmark value + upper deviation of confidence interval, lower limit = benchmark value - lower deviation of confidence interval), clarifying the basic fluctuation boundary of cost accounting. Then, dynamic cost adjustment rules are embedded to adjust the fluctuation range in real time for the three core market factors of raw material prices, exchange rates, and logistics freight rates, according to preset response thresholds (such as raw material price fluctuations exceeding 5% and exchange rate fluctuations exceeding 3%): if raw material prices rise by 8%, the procurement cost and the upper limit of the fluctuation range are adjusted upward according to the correlation coefficient in the time-varying relationship matrix of parameter costs; if logistics freight rates fall by 10%, the transportation cost and the lower limit of the fluctuation range are adjusted downward simultaneously to ensure that the fluctuation range dynamically adapts to market changes.

[0079] Based on this, the cost trend prediction curve output by the associated time-series prediction layer is superimposed on the curve with the dynamically corrected fluctuation range, intuitively presenting the cost change trend (such as steady increase or stable fluctuation) for the next 1-6 months. The trend inflection points and corresponding fluctuation extremes are marked, and finally integrated into a multi-dimensional cost rapid quotation solution: The solution includes a benchmark quotation (based on the cost accounting benchmark value), a dynamic fluctuation range (including upper and lower limits after market factor correction), a trend prediction explanation (including curve interpretation and risk warning), a full life cycle cost detail (broken down into five stages such as design and procurement), and a quotation validity period (set according to the stability of the fluctuation range, such as 7-15 days). At the same time, it supports fine-tuning of customized scenario parameters (such as material changes or specification adjustments) to update the quotation in real time, which not only ensures the accuracy and timeliness of the quotation, but also provides customers with a clear cost structure and trend reference.

[0080] S6. Extract historical candidate cost parameter sets from the multimodal historical quotation database, combine and optimize the historical candidate cost parameter sets using the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to generate personalized multi-dimensional cost accounting schemes, and construct a cost coupling feedback correction model based on the weighted summation algorithm, the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm. In this embodiment of the application, the step of extracting a set of historical candidate cost parameters from a multimodal historical pricing database, optimizing the set of historical candidate cost parameters by combining an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm to generate a personalized multi-dimensional cost accounting scheme, and constructing a cost coupling feedback correction model based on a weighted summation algorithm, an adaptive chaotic sparrow search algorithm, and a Bayesian fusion algorithm includes the following steps: S61. Extract the set of historical candidate cost parameters from the multimodal historical quotation database; Specifically, the extraction scope is first defined: historical quotation projects are selected based on product category (e.g., mechanical equipment, electronic equipment), customization complexity level (high / medium / low), and cost sensitivity type (core materials / general processes, etc.), focusing on data from the past 3-5 years to ensure timeliness, while simultaneously excluding abnormal projects (e.g., records with a data missing rate exceeding 30% or significant differences in customization scenarios). Then, preprocessing is performed according to the characteristics of multimodal data: for structured data (e.g., BOM list, cost detail table), quantitative parameters such as specifications (size, power), materials (model codes), and processes (processing methods) are directly extracted, and outliers (e.g., cost data exceeding 3σ) are removed through data cleaning, and missing items are filled by interpolation; for unstructured data (e.g., technical agreements, quotation documentation), NLP technology (keyword extraction, entity recognition) is used to extract hidden cost parameters (e.g., customized process requirements, special operation and maintenance clauses), and they are converted into standardized numerical vectors; for time-series data (e.g., historical raw material prices, exchange rate fluctuation records), time-series parameter fragments matching the cycle of the quotation project are extracted by aligning them with the time dimension.

[0081] The core extraction stage focuses on the cost-related dimensions throughout the entire lifecycle, covering key parameters in five stages: design and development (number of tests, R&D hours), procurement (unit price of raw materials, procurement batch size), production (energy consumption, scrap rate), operation and maintenance (number of repairs, spare parts cost), and disposal (recycling rate, environmental protection costs). It also links these parameters to the corresponding parameter attributes in the three-level BOM framework (top-level assembly, middle-level components, bottom-level parts) to ensure that parameters are directly linked to cost accounting items. After extraction, standardization is performed: unified parameter units (e.g., size converted to millimeters, cost unified to RMB), coding rules (materials use national standard codes, processes use industry-unified classifications), variance screening (retaining valid parameters with variance > 0.1), and correlation analysis (removing highly redundant parameters with mutual information > 0.8). Finally, a structured historical candidate cost parameter set is formed, containing attributes such as parameter name, value range, corresponding cost stage, and related BOM elements.

[0082] S62. The adaptive chaotic sparrow search algorithm is used to globally optimize the historical candidate cost parameter set and select the initial parameter combination that meets the cost accuracy requirements. Specifically, the historical candidate cost parameter set is first preprocessed: core parameters directly related to the entire life cycle cost (such as specifications, materials, processes and cost drivers at each stage) are retained. By standardizing and unifying units and coding rules, outliers are eliminated and missing items are filled by interpolation using the 3σ criterion. Then, correlation analysis is performed (redundant parameters with mutual information > 0.8 are eliminated) to compress the optimization space and improve parameter quality.

[0083] Then, the core parameters of the algorithm are initialized: the population size is set to 50-80 (each individual corresponds to a set of parameter combinations), the chaos factor is set to 0.3-0.6 to balance diversity and convergence, and the number of iterations is set to 200-300 times (to adapt to the quotation response time constraint). The range of preprocessed parameter values ​​is used as the optimization boundary to ensure that the optimization does not deviate from the actual business scenario. The optimization stage relies on the algorithm's division of labor mechanism and chaotic perturbation: 20%-30% of high-quality individuals are selected as discoverers to explore low-bias parameter regions based on historical cost accuracy data; the remaining individuals act as joiners to imitate the discoverers or explore locally. Chaotic perturbation is applied to the 30%-40% of individuals with poor fitness to avoid getting trapped in local optima. The fitness function is constructed with cost accuracy as the core, and the quantitative indicator is the mean absolute error (MAE) between the historical cost accounting value and the actual cost corresponding to the parameter combination. The smaller the error, the better the fitness. During the iteration process, the fitness of individuals is calculated and sorted in each round, and the top 50% of high-quality individuals are retained to enter the next round until the iteration converges (fitness change < 0.001 for 10 consecutive generations). Finally, the selection is carried out according to cost and accuracy requirements: an error threshold is set (e.g., MAE ≤ 3%), the parameter combination with the best fitness that meets the threshold is extracted, and after deduplication (removing duplicate combinations with similarity > 95%), an initial parameter combination with low bias and high correlation is formed.

[0084] S63. The initial parameter combination is corrected for deviation and optimized for probability using the Bayesian fusion algorithm. Through the fusion iteration mechanism of optimization correction, the cost parameter combination is obtained, and a personalized multi-dimensional cost accounting scheme is generated based on the cost parameter combination. In this embodiment of the application, the step of using a Bayesian fusion algorithm to correct the deviation and optimize the probability of the initial parameter combination, obtaining the cost parameter combination through an optimization and correction fusion iterative mechanism, and generating a personalized multi-dimensional cost accounting scheme based on the cost parameter combination includes the following steps: S631. Using historical data of product multi-scenario customization as a historical reference combination, calculate the cost deviation between the initial parameter combination and the historical reference combination. Specifically, the first step is to screen historical reference combinations: based on the core attributes of the initial parameter combination (product category, customization complexity level, cost sensitivity type, key BOM parameters), the cosine similarity algorithm or Euclidean distance model is used to match and select the top 20-30 highly similar historical parameter combinations as reference samples from the multimodal historical quotation database. Abnormal samples with a data missing rate of more than 10% or a large deviation between actual cost and calculated cost (>10%) are removed to ensure the effectiveness of the reference combination.

[0085] Then, parameter alignment is performed: the parameter dimensions of the initial parameters and historical reference combinations are unified (missing parameters are supplemented with industry averages or medians), unit standards (such as dimensions are unified to millimeters, and cost units are unified to RMB), and coding rules (materials are coded according to national standards, and processes are coded according to industry classification standards) to eliminate accounting deviations caused by inconsistent parameter specifications. Next, a unified accounting standard is adopted to calculate costs: based on the core logic of the multi-dimensional full life cycle cost accounting scheme, the theoretical accounting cost of the initial parameter combination (covering the cost summary of five stages including design, procurement, and production) is calculated separately, as well as the historical actual cost corresponding to each historical reference combination (prioritizing the use of real cost data from implemented projects, and using verified historical accounting costs when no real data is available), ensuring that the accounting scope and calculation rules of the two are completely consistent. In the deviation quantification stage, the deviation value of a single reference combination is first calculated, including the absolute deviation (|initial parameter combination accounting cost - historical reference combination actual cost|) and the relative deviation (absolute deviation / historical reference combination actual cost × 100%). Then, the overall average relative deviation is calculated by weighted average (assigning weights according to the similarity between the historical reference combination and the initial combination). At the same time, the maximum deviation, minimum deviation, and deviation standard deviation are statistically analyzed, and finally, a complete cost deviation value result containing individual deviation details and overall deviation statistics is formed.

[0086] S632. Introduce a Bayesian fusion algorithm, construct a deviation correction model based on historical reference combination and cost deviation value, and perform probabilistic optimization on the initial parameter combination through the deviation correction model to obtain the probabilistically optimized parameter combination. Specifically, the first step is to associate the selected highly similar historical reference combinations (including their parameter attributes and full lifecycle cost data) with the corresponding cost deviation values ​​(absolute deviation and relative deviation) to construct a prior information base for the Bayesian model. The probability distribution of historical cost deviations (such as normal distribution and Gamma distribution) is fitted by kernel density estimation, and the mapping relationship between the parameters and deviations of the historical reference combinations is transformed into a prior probability density function to clarify the prior probability of deviation under different parameter combinations. Then, a likelihood function is constructed: based on the parameter similarity between the initial parameter combination and the historical reference combination, the probability of observing cost deviation of the initial parameter combination in the current scenario is quantified. That is, the mapping relationship between "parameter difference degree - deviation size" is trained through historical data so that the likelihood function can reflect the reasonable range of deviation of the initial parameter combination.

[0087] Based on Bayes' theorem, the prior distribution and likelihood function are fused to calculate the posterior probability distribution. This distribution characterizes the corrected probability features of the initial parameter combination after considering historical deviation information. The peak of the posterior distribution corresponds to the parameter adjustment direction with the smallest deviation, and the confidence interval corresponds to the reasonable fluctuation range of the parameters. In the probabilistic optimization stage, with the goal of "maximizing the posterior probability and minimizing the deviation," the parameter adjustment amount with the highest probability density in the posterior distribution is searched within the value boundary of the initial parameter combination. The core dimensions of the initial parameter combination (such as material cost coefficient and process energy consumption parameters) are fine-tuned. At the same time, a cross-validation mechanism (such as K-fold cross-validation) is introduced to verify the stability of the model, and abnormal adjustment schemes with a posterior probability lower than 90% are eliminated to ensure that the optimization process conforms to the actual business constraints. Finally, a probabilistically optimized parameter combination is output. This combination retains the core features of the initial parameter combination and reduces cost accounting errors through probabilistic correction of historical deviations. Its corresponding posterior confidence interval can intuitively reflect the reliability of the parameter combination.

[0088] S633. Through the optimization correction fusion iteration mechanism, the combination of probabilistic optimization parameters is fed back to the adaptive chaotic sparrow search algorithm for secondary optimization, and repeated iteratively until the enterprise cost constraint conditions are met, thus obtaining the cost parameter combination. Specifically, the cost parameter combination is optimized through an iterative optimization and correction mechanism. The core is to construct a closed-loop iterative process of "secondary optimization - deviation correction - constraint verification." First, the probabilistic optimization parameter combination output by the Bayesian algorithm is used as the core seed, supplemented with a small number of randomly generated parameter variants. These together constitute the initial population for the secondary optimization of the adaptive chaotic sparrow search algorithm. This inherits the high precision of the previous probabilistic optimization while maintaining population diversity through random variants, avoiding getting trapped in local optima. During the secondary optimization stage, the core parameters of the algorithm are dynamically adjusted: the chaos factor is adaptively adjusted according to the iteration process (initially 0.5, decreasing by 0.05 to 0.2 every 10 generations), enhancing the ability for global exploration in the early stages and local convergence in the later stages; adaptive weights are tilted towards enterprise cost constraints (e.g., the profit margin constraint weight is increased to 0.4), making the optimization direction more aligned with compliance requirements.

[0089] After each round of optimization, the output parameter combination is input again into the Bayesian bias correction model. This model updates the posterior probability distribution by combining historical reference combinations with the latest cost bias value, completing a secondary bias correction and forming a fusion mechanism of "algorithm optimization - Bayesian correction." During the iteration process, enterprise cost constraint verification is performed simultaneously: the unit product profit margin (≥15%), total lifecycle cost (≤ preset upper limit), and quotation response time (≤24 hours) corresponding to each parameter combination are checked one by one. If the constraints are not met, the parameter combination is fed back to the algorithm for re-optimization; if the constraints are met, it is recorded as a candidate solution. An iteration termination condition is set: iteration stops when three consecutive generations of candidate solutions meet the constraints and the fitness value fluctuation is ≤0.001, or the number of iterations reaches the upper limit of 300. Finally, from the candidate solutions that meet the conditions, the solution with the best fitness (lowest cost, best profit, and strongest price competitiveness) is selected as the final cost parameter combination, ensuring that it has both optimization performance and compliance.

[0090] S634. Substitute the cost parameter combination into the multi-dimensional full life cycle cost accounting scheme to generate a suitable personalized multi-dimensional cost accounting scheme.

[0091] Specifically, the cost parameter combination is first standardized and analyzed to extract core parameters such as specifications, material grade, and process type, as well as cost driving factors at each stage (such as purchase batch, production energy consumption coefficient, and maintenance frequency). According to the "parameter-cost accounting item" mapping rule, the parameters are accurately matched to the five-stage accounting modules of design and development, procurement, production, maintenance, and scrapping to ensure that the input and accounting dimensions correspond one-to-one.

[0092] Subsequently, personalized calculations are performed according to the accounting logic of each stage: in the design and development stage, the cost of testing and labor hours is calculated based on customized complexity parameters (such as the quantity of special technical requirements); in the procurement stage, the actual procurement cost is calculated by associating raw material prices, exchange rate parameters, and procurement batches, and adding dynamic adjustment rules; in the production stage, the manufacturing cost is calculated based on process parameters, energy consumption coefficient, and scrap rate; in the operation and maintenance stage, the cost of repair and parts replacement is predicted based on product usage scenario parameters (such as usage frequency and environmental conditions); and in the scrapping stage, the terminal cost is calculated based on recycling rate and environmental treatment standard parameters. A personalized adaptation mechanism is integrated into the accounting process: the weight of design and development costs is strengthened for high-complexity customized scenarios, exchange rate fluctuation correction is focused on export products, and the allocation of operation and maintenance costs is refined for long-life cycle products. Finally, the results of the five stages of accounting are integrated to form a multi-dimensional personalized solution, which includes the total cost of the entire life cycle, cost details of each stage, unit product cost, analysis of key cost drivers, cost fluctuation warning range, and profit margin calculation. This solution not only fully covers the cost composition of the entire life cycle but also accurately matches customized needs, providing customized and implementable accounting basis for subsequent pricing decisions and cost control.

[0093] S64. Based on the weighted summation algorithm, the optimization capability of the adaptive chaotic sparrow search algorithm and the correction characteristics of the Bayesian fusion algorithm are combined to construct a cost coupling feedback correction model. The input of the cost coupling feedback correction model is defined as the quotation scheme and personalized accounting data, and the output is the deviation correction coefficient.

[0094] Specifically, based on the weighted summation algorithm, a closed-loop cost-coupled feedback correction model is constructed by hierarchically fusing the global optimization capability of the adaptive chaotic sparrow search algorithm with the probabilistic correction characteristics of the Bayesian fusion algorithm. The standardized boundaries of the model input and output are clarified. First, the core role of the weighted summation algorithm is established. Dynamic weights are set according to cost accounting accuracy (weight 0.4), optimization efficiency (weight 0.3), and deviation correction reliability (weight 0.3). Feature extraction of input data and results of multiple algorithms are integrated to form a unified computing framework.

[0095] During the fusion process, the adaptive chaotic sparrow search algorithm performs global optimization on the input pricing scheme (including pricing amount, cost breakdown structure, and profit expectation target) and personalized accounting data (cost details for the five stages of the entire lifecycle, customized parameter combinations, and market factor correlation data). It uncovers the optimal correlation between cost components and parameters, filters key cost drivers (such as core material price coefficients and process energy consumption ratios), and outputs the initial cost deviation optimization direction. The Bayesian fusion algorithm, on the other hand, relies on customized historical data from multiple product scenarios to construct a prior probability distribution of cost deviations. Combined with the likelihood information of the current input data (such as the similarity between the pricing scheme and similar historical projects, and the credibility of personalized accounting data), it calculates the posterior probability distribution and performs probability correction on the initial deviation direction obtained through optimization, reducing systematic and random errors.

[0096] During model execution, the core feature vectors of the input data are first integrated using a weighted summation algorithm, followed by an adaptive chaotic sparrow search algorithm to output initial correction suggestions. Then, a Bayesian fusion algorithm is used to complete the deviation calibration. Finally, the optimization results and correction results are integrated using a weighted summation algorithm to output multi-dimensional deviation correction coefficients. These coefficients cover cost correction coefficients for five stages, including design, R&D, and procurement, as well as comprehensive total cost correction coefficients and profit matching correction coefficients. This forms a coupled feedback mechanism of "input-optimization-correction-output," ensuring that the correction coefficients are both tailored to personalized accounting scenarios and statistically reliable.

[0097] S7. Input the personalized multi-dimensional cost accounting scheme into the cost coupling feedback correction model to verify and adjust the multi-dimensional cost rapid quotation scheme, generate a five-dimensional quotation verification report, and output the verified and adjusted multi-dimensional cost rapid quotation scheme and the five-dimensional quotation verification report.

[0098] Specifically, the personalized multi-dimensional cost accounting scheme (including cost details for the five stages of the entire life cycle, customized parameter combinations, and market factor related data) and the initial multi-dimensional cost rapid quotation scheme (benchmark quotation, dynamic fluctuation range, trend prediction curve, etc.) are jointly input into the cost coupling feedback correction model to start the verification and adjustment process.

[0099] The model leverages the chaotic sparrow search optimization capabilities integrated with the weighted summation algorithm and the Bayesian fusion correction characteristics. It first extracts the core features of the two types of input data (such as the deviation between the calculated cost and the price, and the matching degree between customized parameters and market factors). Then, it precisely adjusts the pricing scheme through deviation correction coefficients: for cost deviations at each stage, it optimizes the cost breakdown structure using five-stage specific correction coefficients; it adjusts the benchmark price based on the overall total cost correction coefficient; it updates the dynamic fluctuation range boundary by combining the market adaptation correction coefficient; and it calibrates the profit margin by referring to the profit adaptation correction coefficient. This ensures that the adjusted scheme not only fits the personalized calculation results but also meets the goals of lowest cost, optimal profit, and strongest pricing competitiveness. A five-dimensional quotation verification report is constructed simultaneously, covering cost accuracy (calculation deviation rate, effectiveness of correction coefficient), profit compliance (unit product profit margin, total cost compliance), quotation competitiveness (deviation rate from industry average price, market acceptance forecast), market adaptability (adaptability to raw material / exchange rate / freight rate fluctuations), and response timeliness (quote generation time, adjustment efficiency). Each dimension clearly defines verification indicators, compliance thresholds, and actual test results. After multiple rounds of verification and adjustment (until the core indicators in the report meet the standards, such as cost deviation rate <3%, profit margin ≥15%), the final output is a verified and adjusted multi-dimensional cost rapid quotation solution (including updated benchmark quotation, dynamic fluctuation range, trend prediction curve, and quotation validity period) and a five-dimensional quotation verification report. The former ensures accuracy, timeliness, and personalization, while the latter provides comprehensive verification support for quotation decisions.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid quotation method based on parametric BOM and multi-dimensional cost coupling, characterized in that, Includes the following steps: S1. Obtain product multi-scenario customization requirement data and product multi-scenario customization historical data, preset product scenario parameter bidirectional constraint rules and multimodal historical quotation database, and perform double verification on the customization requirement data based on the product scenario parameter bidirectional constraint rules; S2. The product multi-scenario customization requirement data after double verification is classified and labeled in three dimensions according to scenario type, customization complexity and cost sensitivity to form a product parameter labeling dataset, and then structured and stored in association with the historical quotation database. S3. Standardize and preprocess the product parameter annotation dataset, use the parameter BOM mapping algorithm to extract the core related parameters, and preset the adaptive generation rules of BOM hierarchy to build a four-dimensional dynamic parameter mapping model. Input the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list. S4. Preset multi-dimensional full life cycle cost accounting scheme and cost parameter dynamic mapping strategy. Define the time-varying correspondence and weight coefficients of each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full life cycle cost accounting scheme through the cost parameter dynamic mapping strategy. Based on the time-varying correspondence and weight coefficients, construct a multi-dimensional cost coupling calculation model with the LSTM time series prediction model. S5. Set up dynamic cost adjustment rules, introduce adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm, define multi-objective optimization function and enterprise cost constraints, and perform algorithm fusion and iterative optimization of adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm to obtain cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generate multi-dimensional cost rapid quotation scheme. S6. Extract historical candidate cost parameter sets from the multimodal historical quotation database, combine and optimize the historical candidate cost parameter sets using the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to generate personalized multi-dimensional cost accounting schemes, and construct a cost coupling feedback correction model based on the weighted summation algorithm, the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm. S7. Input the personalized multi-dimensional cost accounting scheme into the cost coupling feedback correction model to verify and adjust the multi-dimensional cost rapid quotation scheme, generate a five-dimensional quotation verification report, and output the verified and adjusted multi-dimensional cost rapid quotation scheme and the five-dimensional quotation verification report.

2. The rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 1, characterized in that, The process of standardizing and preprocessing the product parameter annotation dataset, extracting core related parameters using a parameter BOM mapping algorithm, pre-setting adaptive BOM hierarchy generation rules, constructing a four-dimensional dynamic parameter mapping model, and inputting the core related parameters into the four-dimensional dynamic parameter mapping model to output a multi-version configurable parameterized BOM list includes the following steps: S31. Use the 3σ principle to remove outliers in the product parameter annotation dataset, use the K nearest neighbor interpolation method to supplement missing key parameters, use the hash deduplication method to delete duplicate data, and transform the product parameter annotation dataset to the [0,1] interval. Use one-hot encoding to convert non-numerical classification parameters into numerical vectors to form a standardized parameter dataset. S32. Preset the correlation threshold, construct the parameter BOM correlation evaluation matrix using the parameter BOM mapping algorithm, bring the standardized parameter dataset into the parameter BOM correlation evaluation matrix, and filter according to the correlation threshold to obtain the core parameter set. Then, remove redundant parameters through the variance inflation factor to obtain the core correlation parameter set. S33. Define a three-level basic framework consisting of top-level product assembly, mid-level functional components, and bottom-level core parts. Adapt the level of refinement to customized complex quantification values. Add priority labels to cost-sensitive parts. Define the BOM level linkage logic when parameters change through the IF-THEN rule base to form BOM level adaptive generation rules. S34. Based on the adaptive generation rules of BOM hierarchy, using historical data of product customization in multiple scenarios as samples, a four-layer architecture of input layer, feature fusion layer, mapping calculation layer and output layer is built for cross-validation and optimization to obtain dynamic mapping weight matrix. The core associated parameter set is input into the dynamic mapping weight matrix and outputs a multi-version configurable parameterized BOM list.

3. The rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 1, characterized in that, The preset multi-dimensional full lifecycle cost accounting scheme and cost parameter dynamic mapping strategy define the time-varying correspondence and weight coefficients between each parameter in the multi-version configurable parameterized BOM list and the multi-dimensional full lifecycle cost accounting scheme through the cost parameter dynamic mapping strategy. Based on the time-varying correspondence and weight coefficients, a multi-dimensional cost coupling calculation model is constructed with the LSTM time series prediction model, including the following steps: S41. Clearly define the cost accounting dimensions for the five stages of design and development, procurement, production, operation and maintenance and disposal, define quantifiable indicators for each cost accounting dimension, and form a multi-dimensional full life cycle cost accounting solution. S42. Preset a dynamic mapping strategy for cost parameters and construct a time-varying correspondence matrix of parameter costs. The matrix uses the specifications, materials and process parameters in the multi-version configurable parameterized BOM list as row vectors, the life cycle cost accounting items as column vectors, the matrix elements as time-varying correlation coefficients, and set weight coefficients according to cost sensitivity. S43. Based on time-varying correspondence, weight coefficients and LSTM time-series prediction model, construct a multi-dimensional cost coupling calculation model with a four-layer architecture: input layer, feature fusion layer, time-series prediction layer and coupling calculation layer. S44. Verify and optimize the multi-dimensional cost coupling calculation model, and output the verified and optimized multi-dimensional cost coupling calculation model.

4. The rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 1, characterized in that, The proposed dynamic cost adjustment rules introduce an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm, define a multi-objective optimization function and enterprise cost constraints, and perform algorithmic fusion and iterative optimization of the adaptive chaotic sparrow search algorithm and the Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and generate a multi-dimensional rapid cost quotation scheme. S51. Clarify the cost adjustment logic and cost response rules based on market factors such as raw material prices, exchange rates, and logistics freight rates, and integrate them to form dynamic cost adjustment rules; S52. Define a multi-objective optimization function with the goals of lowest cost, optimal profit, and strongest price competitiveness, and define enterprise cost constraints based on unit product profit margin, total cost ceiling, and price response time requirements. S53. Adaptive chaotic sparrow search algorithm and Bayesian fusion algorithm are introduced for iterative optimization. The optimal solution is searched by adaptive chaotic sparrow search algorithm, and the optimization result is corrected by Bayesian fusion algorithm to obtain the cost accounting benchmark value, cost fluctuation range and trend prediction curve, and integrated to form a multi-dimensional cost rapid quotation scheme.

5. The rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 1, characterized in that, The process of extracting historical candidate cost parameter sets from a multimodal historical pricing database, optimizing these sets using an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm to generate a personalized multi-dimensional cost accounting scheme, and constructing a cost coupling feedback correction model based on a weighted summation algorithm, an adaptive chaotic sparrow search algorithm, and a Bayesian fusion algorithm includes the following steps: S61. Extract the set of historical candidate cost parameters from the multimodal historical quotation database; S62. The adaptive chaotic sparrow search algorithm is used to globally optimize the historical candidate cost parameter set and select the initial parameter combination that meets the cost accuracy requirements. S63. The initial parameter combination is corrected for deviation and optimized for probability using the Bayesian fusion algorithm. Through the fusion iteration mechanism of optimization correction, the cost parameter combination is obtained, and a personalized multi-dimensional cost accounting scheme is generated based on the cost parameter combination. S64. Based on the weighted summation algorithm, the optimization capability of the adaptive chaotic sparrow search algorithm and the correction characteristics of the Bayesian fusion algorithm are combined to construct a cost coupling feedback correction model. The input of the cost coupling feedback correction model is defined as the quotation scheme and personalized accounting data, and the output is the deviation correction coefficient.

6. The rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 2, characterized in that, The preset correlation threshold is determined by constructing a parameter BOM correlation evaluation matrix using a parameter BOM mapping algorithm. The standardized parameter dataset is then input into the parameter BOM correlation evaluation matrix, and filtered according to the correlation threshold to obtain a core parameter set. Redundant parameters are then removed using a variance inflation factor to obtain the core correlation parameter set. The process includes the following steps: S321. Perform fitting analysis on the historical data of product customization in multiple scenarios, and preset the correlation threshold based on the fitting analysis results; S322. A standardized parameter and BOM core element correlation evaluation matrix is ​​constructed using the parameter BOM mapping algorithm. The matrix elements in the correlation evaluation matrix are calculated using the Pearson correlation coefficient. The row vectors in the correlation evaluation matrix are the standardized product parameters, and the column vectors in the correlation evaluation matrix are the BOM core elements. S323. Input the standardized parameter dataset into the correlation evaluation matrix to calculate the correlation value, and filter the correlation value to obtain the initial core parameter set. Then, remove the multicollinearity redundancy in the initial core parameter set by the variance inflation factor to obtain the secondary core parameter set. S324. Sort the secondary screening core parameters in the secondary screening core parameter set according to their priority based on parameter size to form a core associated parameter set.

7. A rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 3, characterized in that, The multi-dimensional cost-coupled computation model, which constructs a four-layer architecture (input layer, feature fusion layer, time-series prediction layer, and coupled computation layer) based on time-varying correspondence, weight coefficients, and LSTM time-series prediction model, includes the following steps: S431. The input layer receives the specifications, materials and process parameters in the multi-version configurable parameterized BOM list, combines the time-varying correspondence and weight coefficients in the cost parameter dynamic mapping strategy, and outputs a standardized fusion input vector after Min-Max normalization. S432. The feature fusion layer receives the standardized fusion input vector and adopts a weighted concatenation mechanism of parametric features plus time-varying correlation features. It allocates fusion weights according to cost sensitivity and outputs a high-dimensional fusion feature vector through feature dimension alignment and redundancy removal. S433: The time series prediction layer embeds an LSTM time series prediction model, which uses historical cost time series data as the training basis, inputs a standardized fusion input vector for iterative training, learns and analyzes the cost time series change pattern, and outputs the time series trend correction coefficient. S434, the coupling calculation layer combines the parameter cost time-varying correspondence matrix, performs matrix operations to couple the high-dimensional fused feature vector with the time-series trend correction coefficient, generates multi-dimensional full life-cycle cost accounting results, and integrates them according to the hierarchical architecture of input layer, feature fusion layer, time-series prediction layer and coupling calculation layer to form a four-layer architecture multi-dimensional cost coupling calculation model.

8. A rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 4, characterized in that, The process of introducing an adaptive chaotic sparrow search algorithm and a Bayesian fusion algorithm for iterative optimization, searching for the optimal solution using the adaptive chaotic sparrow search algorithm, and then refining the optimization results using the Bayesian fusion algorithm, yields the cost accounting baseline, cost fluctuation range, and trend prediction curve. This integrated approach forms a multi-dimensional rapid cost quotation scheme, comprising the following steps: S531. Guided by the multi-objective optimization function, and combined with the enterprise cost constraints and dynamic cost adjustment rules, determine the optimization objective and boundary of the adaptive chaotic sparrow search algorithm. S532. The output of the multi-dimensional cost coupling calculation model is globally optimized by using the adaptive chaotic sparrow search algorithm, and a set of candidate optimal solutions that meet the enterprise cost constraints is selected. S533. Introduce a Bayesian fusion algorithm, based on historical data of product customization in multiple scenarios, to perform deviation correction and probability verification on candidate optimal solutions, and output cost accounting benchmark value and confidence interval. S534. Combine the cost accounting benchmark value and confidence interval with the cost dynamic adjustment rules to generate the cost fluctuation range, and link it with the trend prediction curve to form a multi-dimensional cost rapid quotation solution.

9. A rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 5, characterized in that, The process of using a Bayesian fusion algorithm to correct deviations and optimize probabilistics in the initial parameter combination, obtaining a cost parameter combination through an iterative fusion mechanism of optimization correction, and generating a personalized multi-dimensional cost accounting scheme based on the cost parameter combination includes the following steps: S631. Using historical data of product multi-scenario customization as a historical reference combination, calculate the cost deviation between the initial parameter combination and the historical reference combination. S632. Introduce a Bayesian fusion algorithm, construct a deviation correction model based on historical reference combination and cost deviation value, and perform probabilistic optimization on the initial parameter combination through the deviation correction model to obtain the probabilistically optimized parameter combination. S633. Through the optimization correction fusion iteration mechanism, the combination of probabilistic optimization parameters is fed back to the adaptive chaotic sparrow search algorithm for secondary optimization, and repeated iteratively until the enterprise cost constraint conditions are met, thus obtaining the cost parameter combination. S634. Substitute the cost parameter combination into the multi-dimensional full life cycle cost accounting scheme to generate a suitable personalized multi-dimensional cost accounting scheme.

10. A rapid quotation method based on parameterized BOM and multi-dimensional cost coupling according to claim 8, characterized in that, The process of globally optimizing the output of the multi-dimensional cost coupling calculation model using the adaptive chaotic sparrow search algorithm and selecting a set of candidate optimal solutions that meet the enterprise cost constraints includes the following steps: S5321. Based on the optimization objective and boundary, initialize the core parameters of the adaptive chaotic sparrow search algorithm, and use the output of the multi-dimensional cost coupling calculation model as the initial population. S5322. Perform algorithm iteration to improve population diversity through chaotic perturbation mechanism, combine the division of labor mechanism between sparrow population discoverer and joiner to perform global optimization, and use multi-objective optimization function as fitness function to calculate the fitness value of each individual and sort them. S5323. Based on the fitness value ranking results, the compliance of each cost accounting candidate scheme is checked against the enterprise cost constraints, and schemes that do not meet the constraints are eliminated. Then, the remaining compliant schemes are deduplicated to form a candidate optimal solution set.

Images