Design method of super-large BOM based on flow constraint propagation and thought chain reasoning cascade

Abstract

The application proposes a super-large BOM design method based on flow constraint propagation and thinking chain reasoning cascade, which comprises the following steps: obtaining a target function requirement set; analyzing the target function requirement set, combining a historical structure relationship set, and determining a BOM design initial state through a preset function-structure mapping rule; executing thinking chain reasoning on the BOM design initial state to generate a candidate design state; performing engineering rule judgment and path pruning on the candidate design state based on flow constraint propagation to output a consistent confirmation state that meets the conditions; wherein the engineering rule judgment adopts the execution mode of an engineering rule engine; and performing structure convergence based on the consistent confirmation state and generating a final standardized BOM. The application improves the engineering controllability and delivery efficiency of large equipment BOM design.

Description

Technical Field

[0001] This invention belongs to the field of ultra-large BOM design, and particularly relates to an ultra-large BOM design method based on flow constraint propagation and thought chain reasoning cascade. Background Technology

[0002] As large-scale transportation equipment such as ships, rail transit, automobiles, and aerospace develops towards platformization, serialization, and on-demand customization, the Bill of Materials (BOM) is no longer just an appendix to design drawings within an enterprise, but a core engineering carrier that runs through design, procurement, manufacturing, quality, and service. Large-scale equipment typically features deep hierarchies, numerous components, and many variations. The platform BOM and variant BOMs must maintain structural inheritance while also meeting the local differences arising from different task requirements. Meanwhile, engineering rules such as compatibility, dependency, and mutual exclusion are scattered across platform specifications, interface standards, process constraints, and engineering experience, and are expressed differently in different systems. In current engineering practice, BOM design often adopts a "first assemble a solution, then perform unified verification" approach: engineers select modules or component families based on platform structure and experience to form a relatively complete BOM before conducting consistency checks. Once conflicts are discovered, rework across multiple levels is often required, especially when the number of components is huge and the rule relationships are dense. The delayed exposure of errors can lead to repeated design process fluctuations. Some systems attempt to retrieve auxiliary configurations through predefined rules or similar cases. However, in ultra-large BOM scenarios, the scope of rule triggering is difficult to control, and the impact of local changes is difficult to locate. This results in either overly coarse verification, making it difficult to detect path errors early, or overly comprehensive verification, leading to excessive computational and troubleshooting costs. On the other hand, the lack of a consistent way of carrying out platform structure, rule constraints, and material master data makes it difficult for intermediate choices generated during the design process to form verifiable "incremental changes." Verification and convergence often have to rely on manual experience and centralized verification at the later stage, which is insufficient to meet the requirements of large-scale equipment engineering for controllable design processes and deliverable results.

[0003] Meanwhile, the input requirements of actual projects are increasingly appearing in natural language form. The conversion from requirements to structured configuration items and then to the BOM often relies on manual breakdown and interpretation, which can easily introduce semantic biases. Even with the introduction of intelligent question answering or text understanding tools, the lack of in-process gating by engineering rules and interface constraints often requires a lot of manual review, making it difficult to steadily advance under dynamic constraints. Therefore, there is a need for a method that can progressively advance the design within the platform's structural boundaries, and can promptly execute constraint judgments and path selections for each local expansion. This method should support end-to-end conversion from natural language requirements to a standardized BOM, ensuring that the BOM construction process is within an engineering-controllable solution space from the very beginning, and ultimately outputting a standardized BOM that can be directly used in enterprise systems. Summary of the Invention

[0004] The purpose of this invention is to propose a design method for ultra-large BOMs based on streaming constraint propagation and cascaded thought chain reasoning, thereby solving the above-mentioned problems.

[0005] To achieve the above objectives, this invention provides a method for designing ultra-large BOMs based on streaming constraint propagation and cascading thought chain reasoning, the method comprising the following steps:

[0006] S1: Obtain the natural language requirement set; parse the natural language requirement set and combine it with the historical structural relationship set, and determine the initial state of BOM design through pre-set function-structure mapping rules; wherein, the initial state of BOM design includes a requirement node set, an allowed initial structural relationship set, and a rule scope set;

[0007] S2: Execute the thought chain reasoning in the initial state of the BOM design to generate candidate design states;

[0008] S3: Based on the flow constraint propagation, perform engineering rule determination and path pruning on the candidate design states, and output a consistency confirmation state that meets the conditions; wherein, the engineering rule determination adopts the execution mode of the engineering rule engine;

[0009] S4: Based on the consistency confirmation state, perform structural convergence and generate the final standardized BOM.

[0010] Furthermore, S1 specifically includes:

[0011] Obtain a set of natural language requirements, perform structured extraction using a large language model, and generate a set of requirement nodes;

[0012] Analyze the hierarchical structure from the platform BOM or historical product BOM, and extract the set of historical structural relationships.

[0013] Based on the set of demand nodes and the set of historical structural relationships, a function-structure mapping rule is introduced. By matching the rule conditions with the demand nodes in the set of demand nodes, a set of allowed structural relationships that are permitted at the functional level is generated.

[0014] Based on the historical structural relationship set and the allowed structural relationship set, the allowed structural relationships in the initial design state are determined by set intersection operation, and the initial design state of BOM is generated.

[0015] Furthermore, S2 specifically includes:

[0016] The initial state of the BOM design is obtained, the expandable slots in the initial state of the BOM design are identified, and the category identifier associated with the slot is used as the starting point for retrieval. A structured retrieval is performed in the knowledge graph to generate a set of candidate expansion items.

[0017] The candidate extension item set is scored based on predefined scoring rules, and the extension item with the best score is selected and applied to the current state to generate a candidate design state.

[0018] Furthermore, each candidate extension in the candidate extension set carries its minimum structural template and attribute fields for engineering preference evaluation;

[0019] The steps for generating candidate expansion items are as follows:

[0020] Using the standard identifier of the set of requirement nodes in the initial state of the BOM design as a constraint, the module family or component family candidate that can be connected to the current slot is obtained by traversing the graph along the preset relationship. Then, these candidates are matched with the platform module package list or component family list in the enterprise master data at the code level, and are implemented as specific instance identifiers and form corresponding candidate extension items.

[0021] Furthermore, the predefined scoring rules are as follows:

[0022] ;

[0023] in, For candidate expansion terms The set of rule entries that are triggered and determined to be satisfied in the current state is obtained by selecting from the set of allowed initial structural relationships. Match candidate extensions by the target field of the rule entry. After identifying the module / component, execute the rule judgment one by one and summarize the satisfied items; For rule entries Weight parameters; For candidate extensions The set of requirement nodes that can be covered is obtained by using candidate extension items in the requirement-structure mapping table. The module family mapping key is used to look up the set of requirement domain tags, and then compared with... The set of required nodes is intersected and deduplicated, and then summarized. To include candidate expansion items The set of incremental structural relationships generated when the minimum structural template is incorporated into the current state; and Configure parameters for the project.

[0024] Furthermore, the characteristic of S3 is that it specifically includes:

[0025] Calculate the incremental structural relationship set for the candidate design state and the initial BOM design state;

[0026] Based on the incremental structural relationship set, select the rule entries to be triggered from the rule scope set in the initial state of the BOM design, and construct the set of triggered rules;

[0027] For each rule in the triggered rule set, an engineering rule judgment is performed, and the judgment result is combined with the scale constraint on the incremental structural change to prune the candidate design states and summarize them into a consistency confirmation state.

[0028] Furthermore, the step of constructing the triggered rule set specifically comprises:

[0029] From the set of rule scopes, select rule entries that intersect with the design objects involved in the set of incremental structural relationships.

[0030] Furthermore, the determination of the engineering rules adopts the execution method of the engineering rule engine:

[0031] For each triggered rule in the set of triggered rules, check whether the relevant structural relationships in the candidate design state satisfy the two compatible rules and one mutually exclusive rule defined by the current rule. Perform a logical AND operation on the check results of all rules to generate a consistency confirmation state that meets the conditions.

[0032] When performing engineering rule judgments on candidate design states, compatible rules all return true, while mutually exclusive rules return false.

[0033] If the conjunctive result of extracting the structural relationships related to the object of rule action from the candidate design state and comparing them item by item with the rule conditions is false, and thus the consistency confirmation state is false, then the design path of the current candidate design state is pruned.

[0034] If both compatibility rules and one mutual exclusion rule return true and the count value of the incremental structural relationship set is less than or equal to the current project configuration parameter, then the consistency confirmation status is false, and the candidate design status is recorded as the consistency confirmation status.

[0035] Furthermore, the determination result is combined with the scale constraint on the incremental structural change, specifically as follows:

[0036] The current project configuration parameters are designed to limit the maximum number of structural relationships that can be introduced in a single candidate expansion; if the count value of the incremental structural relationship set exceeds the current project configuration parameters, then the candidate design state is determined to not meet the constraints.

[0037] Further, S4 includes:

[0038] The structural relationships in the consistency confirmation status are aligned hierarchically according to the enterprise's established BOM hierarchical template.

[0039] Based on the consistency confirmation status after hierarchical alignment, material master data binding and summary calculation of material quantity are performed. The material master data includes module package number, component family number and unique material code, and the version field, replacement identifier field and effective field are filled in to generate a standardized BOM. The standardized BOM includes at least hierarchical sequence number, parent and child reference, material code, quantity and version fields.

[0040] Specifically, the binding of material master data based on the consistency confirmation status after hierarchical alignment includes:

[0041] Using the module package number or component family number carried by the node in the consistency confirmation status as the index key, the unique material code is retrieved and written into the node record; for the same material code that is referenced in multiple places, the same level of detail line merging process under the same parent node is entered, and the quantity value needs to be calculated during merging;

[0042] Specifically, the process of summarizing and calculating the quantity of materials involves:

[0043] Under the same parent node and the same level, count all design instances bound to the same unique material code, and use the count value as the quantity value of the material under that unique material code.

[0044] The beneficial technical effects of the present invention are at least as follows:

[0045] This invention addresses the combinatorial explosion and verification lag problems in ultra-large BOM design, proposing a process-controlled BOM construction approach: At the design initiation stage, requirements, platform historical structure, and rule scope are uniformly encapsulated into a computable initial design state, clearly defining the boundaries of permissible initial structural relationships to prevent subsequent decisions from unfolding haphazardly across the entire space. During the design advancement stage, a chain-like step-by-step decision-making approach is adopted, performing only one local expansion at a time, introducing the instantiation of platform module packages or component families along with their minimum structural templates, ensuring clear incremental changes in candidate states. In the verification stage, incremental filtering and judgment of the rule set are triggered around this incremental change, forming a consistent conclusion through logical conjunction, and eliminating design paths that do not meet rule constraints or expansion scale control at an early stage, thus transforming verification from post-event centralized checking to real-time filtering during the process. In the delivery stage, the structural relationships after consistency confirmation are converged according to the enterprise-level template and encoded, bound, and quantity-summarized with material master data to form a standardized BOM object that can be directly entered into PDM or ERP, ensuring that the effectiveness of the design path is ultimately reflected in the usability of engineering data.

[0046] To adapt to the "natural language requirement-first" input method in engineering sites, this invention extracts natural language requirements in a structured manner using a large language model and aligns them with standard ontology identifiers in the enterprise knowledge graph. This allows requirement nodes to directly participate in platform structural relationship retrieval and rule matching. In the candidate expansion generation and selection stage, the applicability, substitution, and interface compatibility relationships in the knowledge graph are used to constrain the candidate space. Engineering preferences such as cost, performance, delivery time, and supply chain stability are embedded into the candidate ranking in the form of weights, so that candidate generation and multi-objective trade-offs are completed in the same decision-making process. In the consistency determination stage, a streaming constraint propagation triggered by incremental structural changes is adopted, moving the checks of compatibility, dependency, mutual exclusion, and resource conditions forward to after each local expansion, thereby connecting "reasoning generation - engineering verification - structural convergence - standardized delivery" into a single closed-loop process.

[0047] Through the above process, the BOM construction forms a closed loop from the starting boundary, local expansion, incremental determination to engineering convergence. This not only maintains the directionality of design progress in ultra-large structures, but also limits the accumulation of error paths in a timely manner after each expansion, thereby improving the engineering controllability and delivery efficiency of BOM design for large equipment. Attached Figure Description

[0048] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0049] Figure 1 This is a flowchart of the ultra-large BOM design method based on the cascaded flow constraint propagation and thought chain reasoning of the present invention. Detailed Implementation

[0050] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0051] like Figure 1 As shown in the embodiment of the present invention, a method for designing ultra-large BOMs based on streaming constraint propagation and cascading thought chain reasoning is provided. The method includes:

[0052] S1: Obtain the target functional requirement set; parse the target functional requirement set, combine it with the historical structural relationship set, and determine the initial state of BOM design through pre-set function-structure mapping rules; wherein, the initial state of BOM design includes a requirement node set, an allowed initial structural relationship set, and a rule scope set.

[0053] Specifically, this step revolves around how to establish a unified, calculable, and engineering-reasonable design foundation in the initial stage of ultra-large BOM design. Its core objective is to construct the initial design state. This section serves to hold the functional, structural, and rule information required for subsequent design phases. In actual engineering, product requirements typically exist in the form of specifications or task orders, while platform or historical BOMs are stored in a hierarchical structure within the PDM system, and engineering rules are scattered across rule bases or knowledge graphs. If this information is not uniformly organized and aligned at the beginning of the design phase, subsequent design processes will struggle to proceed within the same engineering context. This step, through the structured processing and tailoring of existing engineering data, confines the design space to a range that conforms to both historical engineering practices and current requirement constraints.

[0054] Furthermore, during implementation, the natural language requirements will be set. As input, structured extraction is performed through a large language model to generate a set of requirement nodes. The specific approach involves segmenting the requirement text into sentences and feeding it into the model. The model is then constrained to output a structured result that conforms to the enterprise's requirement slot template. This structured result includes at least functional domain labels, key constraint phrases, and referential information related to platform modules. Subsequently, the extracted terms are aligned with the ontology nodes of the enterprise knowledge graph. Existing standard identifiers in the graph are used to replace free expressions in the text, thereby obtaining a set of requirement nodes carrying standard identifiers. For example, when the text contains expressions such as "electric drive" and "low temperature adaptation", the model outputs the corresponding functional domain labels and constraint phrases, and merges them into the existing power system domain and environmental adaptation domain node identifiers of the enterprise through the knowledge graph, so that subsequent steps can directly perform relationship retrieval and rule matching based on standard identifiers.

[0055] Subsequently, the platform BOM or historical product BOM is exported from the PDM system, its hierarchical structure is parsed, and the set of historical structural relationships is extracted. This set consists of the hierarchical relationships between modules, submodules, and components, reflecting the structural organization methods that the company has actually used in its existing products.

[0056] Furthermore, in obtaining the set of demand nodes Relationship with historical structure Next, function-structure mapping rules from the engineering rule base are introduced. These rules typically exist in the form of "condition-permitted structures," describing which modules or structures are usually used to implement a certain type of function in an engineering context. This is achieved by mapping rule conditions to a set of requirement nodes. By matching the requirement nodes in the data, a set of permissible structural relationships that are allowed at the functional level can be generated. Based on this, the permissible structural relationships in the initial design state are determined through set intersection operations, and their calculation form is as follows:

[0057] ;

[0058] in, This represents the set of historical structural relationships extracted from the historical BOM. This represents the set of allowed structural relationships obtained by combining requirement nodes with function-structure mapping rules. This represents the initial design state of the BOM (Bill of Materials) that is ultimately incorporated into the initial design state. This calculation unifies historical feasibility and demand rationality into the same structural set, ensuring that the initial design space is subject to the dual constraints of engineering practice and demand constraints from the very beginning.

[0059] To better understand the above process, an engineering scenario can be used as an example. Assume the data is extracted from the platform's BOM. It contains three structural relationships: system nodes Includes modules Module Includes component families System nodes Includes modules In the current demand, The set of requirement nodes obtained through parsing This indicates that the system needs to satisfy a certain dynamic functional domain, and according to the function-structure mapping rules in the rule base, this functional domain is allowed a set of structural relationships. Include" Include "and" Include ", and also allows " Includes modules ".Will and After substituting the above relation trimming process, the result is for" Include "and" Include This result demonstrates that the initial design state only retains structural relationships that both appeared in the historical platform and match the current functional requirements, thereby avoiding the introduction of historically unverified or mismatched structural combinations at the design starting stage.

[0060] Furthermore, in determining Next, the scope of the engineering constraint rules is limited, retaining only rule entries whose constraint objects appear in requirement nodes or permitted structural relationships. Finally, the set of requirement nodes is... Allow initial set of structural relations Together with the scope of the trimmed rules, they are encapsulated into the initial state of the BOM design. This state exists in the form of a structured engineering object, which can be directly read and expanded by subsequent steps, providing a well-defined, content-constrained, and engineering-reasonable starting point for the step-by-step design of a very large BOM.

[0061] S2: Execute the thought chain reasoning in the initial state of the BOM design to generate candidate design states.

[0062] Specifically, in the initial design stage The set of required nodes and the set of allowed initial structural relationships have been given. and the set of rule scopes Building upon this foundation, this step revolves around "advancing a local decision within the constrained structural boundary" to generate candidate design states. During implementation, first allow an initial set of structural relationships. The system locates expandable slots and uses the category identifier associated with those slots as the starting point for retrieval. It then performs a structured retrieval within the knowledge graph to generate a set of candidate expansion items. The specific approach is to... medium demand node set The standard identifier serves as a constraint. The system iterates through the graph along relationships such as "applicable," "substitutable," and "interface compatible" to obtain candidate module families or component families that can be connected to the current slot. These candidates are then matched at the code level with the platform module package list or component family list in the enterprise master data, resulting in specific instance identifiers and forming... Each It also carries its minimum structure template (from the mandatory sub-module relationships and mandatory interface relationships in the platform module package template table) and attribute fields for engineering preference evaluation (from the master data fields or the project configuration table for the classification fields of indicators such as cost, performance, and supply chain stability), thereby ensuring that the candidate options not only meet the connectability of the knowledge graph constraint topology, but can also be directly mapped to the deliverable engineering coding object.

[0063] Furthermore, the ranking of candidate expansion items adopts the linear weighted scalarization concept from classic multi-objective decision-making (derived from the weighted sum method commonly used in multi-objective optimization and engineering decision-making), and makes targeted modifications for the BOM design scenario: "rule constrainability" and "demand anchoring" are used as benefit items, and "structural deviation" is used as a penalty item, so that each expansion introduces candidates to positions that can be constrained by more established rules and aligned with more demand nodes, while suppressing excessive structural increments introduced by the expansion. Its scoring function uses set cardinality (from set theory). As an operable quantity, the calculation formula is as follows:

[0064] ;

[0065] in, For the extension item The set of rule entries that are triggered and determined to be satisfied in the current state is obtained by... Matching the target field of the rule entry After identifying the module / component, execute the rule judgment one by one and summarize the satisfied items; For rule entries The weight parameters are derived from the preference weight mapping of objectives such as cost, performance, and supply chain stability in the enterprise project configuration table. In engineering implementation, these preferences can be mapped to several "evaluation rules" and assigned higher weights. By assigning basic weights to hard engineering rules, both hard constraint orientation and multi-objective preference orientation can be reflected in the same scoring framework. for The set of requirement nodes that can be covered is obtained by referring to the requirement-structure mapping table. The module family mapping key is used to look up the set of requirement domain tags, and then compared with... The set of required nodes is intersected and deduplicated, and then summarized. To be The minimum structural template is incorporated into the incremental structural relationship set generated when the current state is reached. This set is obtained by reading the platform module package list. The bound set of "mandatory submodule relationships / mandatory interface relationships" templates, minus... The existing homogeneous relationships are used to obtain an incremental set; and The project configuration parameters are derived from the enterprise design strategy configuration file and can be set according to platform or project stage. Both sides of the equation are scalars obtained by linearly combining discrete set cardinality; all terms involved in the calculation are count quantities and do not involve physical quantities, therefore there is no issue of dimensional inconsistency. The modification of the weighted sum method lies in... As a direct penalty for structural deviation, the engineering constraint of "single-step expansion to maintain locality" is explicitly encoded into the scoring function, so that the scoring result has an inhibitory effect on the combinatorial expansion of ultra-large BOMs.

[0066] Furthermore, after obtaining the scores for each candidate extension item, the extension item with the highest score is selected and applied. Forming candidate design states This selection process corresponds to the maximization selection in classical optimization (derived from discrete optimization). The definition was implemented and engineered using BOM state updates: Extended items are written into slots and their minimum structure templates are merged using deterministic application operators, thereby generating candidate states that can be directly processed by subsequent consistency checks. The relationship is expressed as follows:

[0067] ;

[0068] in, This is the optimal extension term; For deterministic state update operations, the implementation method is as follows: Write the instance identifier to the selected slot, The minimum structural template relation set and Perform a union set operation to obtain the updated set of relations, while retaining the original set. The set of demand nodes and the set of rule scopes make A local specification is completed within the same demand and rule context. The logical relationship between the two equations above is as follows: the first equation gives a score for a single candidate extension item, and the second equation performs discrete maximization selection and state update based on the score. Together, they constitute the process from... arrive The single-step advancement process.

[0069] To illustrate the calculation process, an engineering-style substitution example is given. Assume that a certain expansion slot corresponds to three candidate expansion items. Determined and summarized from the rule base ; obtained from demand-structure mapping matching ; Obtained by subtracting existing relationships from the minimum structural template of the platform module package Retrieve configuration parameters Substituting into the scoring formula, we get , , ,therefore Then execute :Will Write the instance identifier to the slot and put The minimum structural template relation set is incorporated into A new set of relationships is formed, thereby obtaining candidate design states. In engineering implementation, the three cardinal values ​​of each candidate extension term are typically used. Values, scoring results, and final selection The record is a calculation trajectory, and the trajectory content comes directly from the intermediate results of rule retrieval count, mapping matching count and template difference count, which makes it easy to verify the consistency between candidate selection behavior and data source when iterating platform products.

[0070] S3: Based on the flow constraint propagation, perform engineering rule determination and path pruning on the candidate design state, and output the consistency confirmation state that meets the conditions; wherein, the engineering rule determination adopts the execution mode of the engineering rule engine.

[0071] Specifically, this step follows the candidate design state generated in step two. The engineering constraints of the new relationships introduced by a single structural expansion are determined, and the design path is pruned accordingly. Because... Is It is obtained through a deterministic expansion within the bounded structural boundary, and its relative to The changes can be clearly defined as a set of newly added or specified structural relationships, denoted as... In engineering practice, these new relationships typically correspond to "which new module instances have been added," "which interface connections have been completed," and "which component families have been introduced." These changes are most likely to trigger compatibility, dependency, or mutual exclusion rules. This step focuses on processing this incremental set, ensuring that constraint propagation strictly revolves around "the design changes that have just occurred," thus directly connecting with the step-by-step expansion logic in step two.

[0072] Furthermore, in specific implementation, the first step is to base it on the incremental structural relationship set. From step one, following the initial state of BOM design The set of encapsulated rule scopes is used to filter out the rule entries that need to be triggered. This filtering method is derived from the incremental constraint evaluation concept in the Constraint Satisfaction Problem (CSP), which states that when a variable changes, only the constraints related to that variable are re-evaluated. This application applies this concept to the BOM design context by matching the target objects of rule entries with the structural relationships to construct the set of triggered rules. Its expression is as follows:

[0073] ;

[0074] in, The set of rules determined in step one, with data derived from rule entries exported from the enterprise rule base or knowledge graph; For step two in execution The set of incremental structural relationships introduced at that time; This represents the deterministic process of resolving the target identifier from rule entries or structural relationships. Specifically, it involves reading the "target field" of the rule entry (such as module family identifier, interface type identifier, component family identifier) ​​and the endpoint identifier of the structural relationship record to form a deduplicated set. This formula belongs to set filtering operations, and the participating objects are all discrete identifier sets, without involving any physical quantities or continuous values, so there is no problem of inconsistent dimensions.

[0075] Furthermore, after obtaining Then, engineering rule judgment is performed on each rule, and the judgment results are summarized into an overall consistency conclusion. The rule judgment process adopts the conventional execution method of the engineering rule engine: for compatible rules, a search is performed... Check whether newly added modules or interface combinations appear in the compatibility table; for dependency rules, check if the closure condition "if a prerequisite relationship exists, a conclusion relationship must also exist" is satisfied; for mutual exclusion rules, check whether the newly added structure is compatible with... The existing structures form prohibited combinations. The result of each rule is a Boolean value. To ensure that path pruning not only reflects rule consistency but also embodies the engineering requirement of "maintaining locality in a single expansion" in ultra-large BOM design, this application introduces a constraint on the expansion scale based on the classic conjunction decision. This constraint originates from the concept of "single change scale control" in engineering change management, and is used to limit the number of structural relationships introduced in a single design expansion, thereby keeping subsequent problem identification and engineering adjustments controllable. After combining rule decisions and expansion scale constraints, the consistency decision is expressed as follows:

[0076] ;

[0077] in, Indicates the rule The specific determination process is implemented by starting from... Extract the structural relationships related to the objects affected by the rules and compare them item by item with the rule conditions; This is a logical conjunction operation, derived from the definition of conjunction in propositional logic; The count value for incremental structural relation entries is obtained directly through... The result is obtained by deduplication and counting. These are engineering configuration parameters, derived from enterprise design strategy configuration files or project configuration tables, used to limit the maximum number of structural relationships allowed to be introduced in a single candidate expansion. Both sides of this equation are logical judgment results, composed of Boolean checks and counting threshold comparisons, and do not involve physical quantities, therefore maintaining consistent dimensions.

[0078] The logical relationship between the two formulas above is as follows: the first formula constructs the set of rules to be triggered, clarifying "which rules need to be checked in this expansion"; the second formula performs specific judgments on this set of rules and superimposes expansion scale constraints, thus giving a final judgment on whether to retain the design path. To illustrate the operability of this calculation process, an engineering example can be used. Suppose that an expansion introduces... It includes two newly added relationships, with configuration parameters taken as follows: If so, then the condition for scaling up is met.

[0079] Filtered from the rule scope using set matching. It contains three rules: two are compatibility rules and one is a mutual exclusion rule. When executing rule determination, if all compatible rules return true and all mutually exclusive rules return false, then... The conjunctive result is false, therefore If false, the design path is pruned in this step. Alternatively, if all three rules return true and... ,but If true, the system will Recorded as consistency confirmation status In engineering implementation, The construction results, the judgment results of each rule, The count value and the final consistency determination conclusion can be saved as a calculation record for subsequent design review or problem backtracking.

[0080] Through this step, the candidate design status... Design paths that do not meet engineering rules or expansion scale constraints are promptly pruned, while paths that do meet the conditions are confirmed. This process then proceeds to subsequent steps, ensuring that the entire ultra-large BOM design process maintains engineering controllability and consistency of rules while gradually expanding.

[0081] S4: Based on the consistency confirmation state, perform structural convergence and generate the final standardized BOM.

[0082] Specifically, this step involves confirming the consistency status. As input, this completes structural convergence and final BOM generation, transforming the design state, which has been trimmed through constraint propagation in previous steps, into a BOM object usable by the enterprise engineering system. In engineering practice, The BOM contains specific module instances, component family instances, and their structural relationships. These relationships already satisfy the check results for triggering rules within the rule's scope. Therefore, this step revolves around "hierarchical template alignment, material master data binding, quantity summarization, and version field population," ultimately outputting a standardized BOM result. During implementation, first read... The system sets structural relationships, grouped and sorted according to the enterprise's established BOM hierarchical template. The hierarchical template typically originates from the enterprise's PDM BOM type configuration or platform specification documents, defining the hierarchical order of system level, subsystem level, module level, and component level, as well as the allowed parent-child relationship types. For each structural relationship record, its level is determined based on the category field of its endpoint node and the hierarchical template rules. Under the same parent node, child nodes at the same level are sorted according to template sequence number or lexicographical order, thus obtaining a stable hierarchical expansion order.

[0083] Furthermore, after completing the hierarchical alignment, for Each node performs material master data binding. The material master data table is exported from the material master data interface of PDM or ERP, containing a mapping between module package number, component family number, and unique material code. The binding process uses a common primary key lookup in engineering: using the module package number or component family number carried by the node as the index key, the unique material code is retrieved and written to the node record. For the same material code referenced in multiple places, the same-level detail row merging process under the same parent node is entered, and the quantity field needs to be calculated during merging. The quantity calculation comes from the classic method of discrete counting and summarizing (set / multiset counting, grouping aggregation). This application uses it for BOM level detail generation, and obtains the quantity value by summarizing the occurrence frequency of structural relationships or expanded templates. Its calculation form is as follows:

[0084] ;

[0085] in, Indicates a material code; This indicates that in the hierarchical expansion results, it belongs to the material code. The instance appears as a record set, which is obtained by binding the material code to the node and then using the parent node identifier and the material code. Used as a grouping key, it summarizes instance rows within the same group; This indicates the material code under this group. The quantity value. This formula is a discrete count summation, not involving continuous physical quantities. Both sides are counted quantities, so the dimensions are consistent. After completing the quantity summary, the hierarchical information, parent-child references, material codes, and quantity fields are written into the BOM detail row structure, and the version field, replacement identifier field, and effective field are supplemented according to the enterprise BOM template; the values ​​of these fields are derived from the enterprise template default values ​​or... The instance attribute fields carried in the template are populated by writing them field by field using a template mapping rule table.

[0086] Furthermore, to illustrate the feasibility of the above quantitative calculations, an engineering calculation example is given. Assume... In the hierarchical expansion results, three component-level instance records appeared under the same parent node, and their bound material codes were respectively , , For material codes... Corresponding It contains two instance records, which, when substituted into the counting formula, yields... For material codes Corresponding Contains one instance record, and gets Based on this, two BOM detail lines are generated, each recording the material code as follows. The quantity is 2, and the material code is The quantity is 1, and its parent node reference and hierarchical position are preserved. In engineering implementation, "hierarchical expansion order, master data binding result, group key, ... The number of records and the final quantity "Save it as a one-time generation record for use as a basis for tracing the formation of detailed lines during subsequent process reviews or BOM verification."

[0087] After completing hierarchy alignment, master data binding, and quantity aggregation, the final BOM result is generated. . The output format follows the enterprise BOM template definition, and at least includes hierarchical sequence number, parent-child reference, material code, quantity and version fields. It can also be exported as a structured exchange format according to PDM / ERP interface requirements for system integration. This step involves... The deterministic mapping and counting of the confirmed structural relationships in the system enables structural convergence and engineering implementation from the consistent confirmation state to the standardized BOM result.

[0088] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0089] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or units may be electrical, mechanical, or other forms.

[0090] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0091] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for designing ultra-large BOMs based on streaming constraint propagation and cascading thought chain reasoning, characterized in that: The method includes: S1: Obtain the natural language requirement set; parse the natural language requirement set and combine it with the historical structural relationship set, and determine the initial state of BOM design through pre-set function-structure mapping rules; wherein, the initial state of BOM design includes a requirement node set, an allowed initial structural relationship set, and a rule scope set; S2: Execute the thought chain reasoning in the initial state of the BOM design to generate candidate design states; S3: Based on the flow constraint propagation, perform engineering rule determination and path pruning on the candidate design states, and output a consistency confirmation state that meets the conditions; wherein, the engineering rule determination adopts the execution mode of the engineering rule engine; S4: Based on the consistency confirmation state, perform structural convergence and generate the final standardized BOM; S2 specifically includes: The initial state of the BOM design is obtained, the expandable slots in the initial state of the BOM design are identified, and the category identifier associated with the slot is used as the starting point for retrieval. A structured retrieval is performed in the knowledge graph to generate a set of candidate expansion items. The candidate extension item set is scored based on predefined scoring rules, and the extension item with the best score is selected and applied to the current state to generate a candidate design state. S3 specifically includes: Calculate the incremental structural relationship set for the candidate design state and the initial BOM design state; Based on the incremental structural relationship set, select the rule entries to be triggered from the rule scope set in the initial state of the BOM design, and construct the set of triggered rules; For each rule in the triggered rule set, an engineering rule judgment is performed, and the judgment result is combined with the scale constraint on the incremental structural change to prune the candidate design states and summarize them into a consistency confirmation state.

2. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, S1 specifically includes: Obtain a set of natural language requirements, perform structured extraction using a large language model, and generate a set of requirement nodes; Analyze the hierarchical structure from the platform BOM or historical product BOM, and extract the set of historical structural relationships. Based on the set of demand nodes and the set of historical structural relationships, a function-structure mapping rule is introduced. By matching the rule conditions with the demand nodes in the set of demand nodes, a set of allowed structural relationships that are permitted at the functional level is generated. Based on the historical structural relationship set and the allowed structural relationship set, the allowed structural relationships in the initial design state are determined by set intersection operation, and the initial design state of BOM is generated.

3. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, Each candidate extension in the candidate extension set carries its minimum structure template and attribute fields for engineering preference evaluation; The steps for generating candidate expansion items are as follows: Using the standard identifier of the set of requirement nodes in the initial state of the BOM design as a constraint, the module family or component family candidate that can be connected to the current slot is obtained by traversing the graph along the preset relationship. Then, these candidates are matched with the platform module package list or component family list in the enterprise master data at the code level, and are implemented as specific instance identifiers and form corresponding candidate extension items.

4. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, The predefined scoring rules are as follows: ； in, For candidate expansion terms The set of rule entries that are triggered and determined to be satisfied in the current state is obtained by selecting from the set of allowed initial structural relationships. Match candidate extensions by the target field of the rule entry. After identifying the module / component, execute the rule judgment one by one and summarize the satisfied items; For rule entries Weight parameters; For candidate extensions The set of requirement nodes that can be covered is obtained by using candidate extension items in the requirement-structure mapping table. The module family mapping key is used to look up the set of requirement domain tags, and then compared with... The set of required nodes is intersected and deduplicated, and then summarized. To include candidate expansion items The set of incremental structural relationships generated when the minimum structural template is incorporated into the current state; and Configure parameters for the project.

5. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, The specific steps for constructing the set of rules to be triggered are as follows: From the set of rule scopes, select rule entries that intersect with the design objects involved in the set of incremental structural relationships.

6. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, The engineering rule determination adopts the execution method of the engineering rule engine: For each triggered rule in the set of triggered rules, check whether the relevant structural relationships in the candidate design state satisfy the two compatible rules and one mutually exclusive rule defined by the current rule. Perform a logical AND operation on the check results of all rules to generate a consistency confirmation state that meets the conditions. When performing engineering rule judgments on candidate design states, compatible rules all return true, while mutually exclusive rules return false. If the conjunctive result of extracting the structural relationships related to the object of rule action from the candidate design state and comparing them item by item with the rule conditions is false, and thus the consistency confirmation state is false, then the design path of the current candidate design state is pruned. If both compatibility rules and one mutual exclusion rule return true and the count value of the incremental structural relationship set is less than or equal to the current project configuration parameter, then the consistency confirmation status is false, and the candidate design status is recorded as the consistency confirmation status.

7. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 6, characterized in that, The determination result is combined with the scale constraint on the incremental structural change, specifically as follows: The current project configuration parameters are designed to limit the maximum number of structural relationships that can be introduced in a single candidate expansion; if the count value of the incremental structural relationship set exceeds the current project configuration parameters, then the candidate design state is determined to not meet the constraints.

8. The ultra-large BOM design method based on streaming constraint propagation and cascaded thought chain reasoning as described in claim 1, characterized in that, The S4 includes: The structural relationships in the consistency confirmation status are aligned hierarchically according to the enterprise's established BOM hierarchical template. Based on the consistency confirmation status after hierarchical alignment, material master data binding and summary calculation of material quantity are performed. The material master data includes module package number, component family number and unique material code, and the version field, replacement identifier field and effective field are filled in to generate a standardized BOM. The standardized BOM includes at least hierarchical sequence number, parent and child reference, material code, quantity and version fields. Specifically, the binding of material master data based on the consistency confirmation status after hierarchical alignment includes: Using the module package number or component family number carried by the node corresponding to the consistency confirmation status as the index key, the unique material code is retrieved and written to the corresponding node record; for the same material code that is referenced in multiple places, the same level of detail line merging process under the same parent node is entered, and the quantity value needs to be calculated during merging. Specifically, the process of summarizing and calculating the quantity of materials involves: Under the same parent node and the same level, count all design instances bound to the same unique material code, and use the count value as the quantity value of the material under that unique material code.

Images