CN-121809295-B - Ultra-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade

CN121809295BCN 121809295 BCN121809295 BCN 121809295BCN-121809295-B

Abstract

The invention provides an ultra-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade connection, which comprises the steps of obtaining a target function requirement set, analyzing the target function requirement set, combining a historical structure relation set, determining a BOM design initial state through a preset function-structure mapping rule, generating a candidate design state in the execution thinking chain reasoning of the BOM design initial state, carrying out engineering rule judgment and path cutting on the candidate design state based on stream constraint propagation, and outputting a consistency confirmation state meeting the condition, wherein the engineering rule judgment adopts an execution mode of an engineering rule engine, and executing structure convergence based on the consistency confirmation state to generate a final standardized BOM. The invention improves engineering controllability and delivery efficiency of BOM design of large equipment.

Inventors

LI ZHI
FAN FENGYUN

Assignees

广东工业大学
上海浙江大学高等研究院

Dates

Publication Date: 20260508
Application Date: 20260309

Claims (8)

1. The ultra-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade is characterized by comprising the following steps: S1, acquiring a natural language demand set, analyzing the natural language demand set, and combining a historical structure relation set to determine a BOM design initial state through a preset function-structure mapping rule, wherein the BOM design initial state comprises a demand node set, an allowed initial structure relation set and a rule scope set; S2, performing thinking chain reasoning in the BOM design initial state to generate a candidate design state; S3, carrying out engineering rule judgment and path cutting on the candidate design states based on stream constraint propagation, and outputting a consistency confirmation state meeting the conditions, wherein the engineering rule judgment adopts an execution mode of an engineering rule engine; S4, performing structure convergence based on the consistency confirmation state and generating a final standardized BOM; the step S2 specifically comprises the following steps: Acquiring the BOM initial design state, identifying an expandable slot in the BOM initial design state, and executing structural search in a knowledge graph by taking a category identifier associated with the slot as a search starting point to generate a candidate expansion item set; scoring the candidate extension item set based on a predefined scoring rule, selecting an extension item with the optimal score to be applied to the current state, and generating a candidate design state; the step S3 specifically comprises the following steps: calculating an incremental structural relation set aiming at the candidate design state and the BOM design initial state; screening rule items needing triggering from the rule scope set in the initial state of the BOM design according to the increment structural relation set, and constructing a triggered rule set; And executing engineering rule judgment on each rule in the triggered rule set, combining judgment results with scale constraint on the incremental structural change, cutting the candidate design states, and summarizing the candidate design states into a consistency confirmation state.
2. The ultra-large BOM design method based on the cascade of stream constraint propagation and thinking chain reasoning according to claim 1, wherein the S1 specifically comprises: acquiring a natural language demand set, and executing structural extraction through a large language model to generate a demand node set; analyzing a hierarchical structure from a platform BOM or a historical product BOM, and extracting to obtain a historical structure relation set; Based on the requirement node set and the history structure relation set, introducing a function-structure mapping rule, and generating an allowed structure relation set allowed at a function level by matching rule conditions with requirement nodes in the requirement node set; and determining the allowed structural relation in the initial design state through set traffic operation based on the historical structural relation set and the allowed structural relation set, and generating the initial BOM state.
3. The method for designing a very large scale BOM based on stream constraint propagation and mental chain reasoning cascade according to claim 1, wherein each candidate extension of the candidate extension set carries its minimum structural template and attribute field for engineering preference evaluation; The candidate expansion item generation steps are as follows: And traversing in the map along a preset relation by taking a standard mark of a demand node set in the initial state of the BOM design as a constraint condition to obtain module group or component group candidates which can be connected with the current slot position, performing coding level matching on the candidates and a platform module package list or a component group list in enterprise owner data, and marking a specific example by the floor and forming a corresponding candidate extension item.
4. The method for designing a very large scale BOM based on a cascade of streaming constraint propagation and mental chain reasoning according to claim 1, wherein the predefined scoring rules are as follows: ; Wherein, the To candidate extension Rule entry set triggered and judged to be satisfied in current state is obtained by allowing initial structural relation set Matching candidate extensions in rule-entry-based action object fields After the module/component identification, rule judgment is carried out item by item, and the satisfaction items are summarized; For rule entries Weight parameters of (2); For candidate extensions The method comprises the steps of obtaining a coverage demand node set by using candidate expansion items in a demand-structure mapping table The module group mapping key of (1) is used for searching the demand domain label set reversely and then is connected with The medium demand node sets are collected, crossed, de-duplicated and summarized; To expand candidate items The minimum structure template of (2) is incorporated into the incremental structure relation set generated when the current state is; and (3) with Parameters are configured for the project.
5. The method for designing a very large scale BOM based on cascading of stream constraint propagation and thinking chain reasoning according to claim 1, wherein the step of constructing the triggered rule set is specifically as follows: and screening out rule entries which have intersections with design objects related to the incremental structural relation set from the rule scope set.
6. The ultra-large scale BOM design method based on stream constraint propagation and thinking chain reasoning cascade according to claim 1, wherein the engineering rule determination adopts an execution mode of an engineering rule engine: checking whether the related structural relation in the candidate design state meets two compatible rules and one mutually exclusive rule defined by the current rule or not according to each triggered rule in the triggered rule set, and performing logical AND operation on checking results of all rules to generate a consistency confirmation state meeting the conditions; Wherein, when engineering rule determination is executed on the candidate design state, the compatible rule is returned to be established, the mutual exclusion rule is returned to be not established, If the combination result of extracting the structural relation related to the rule action object from the candidate design state and comparing the structural relation with the rule condition item by item is false, and further the consistency confirmation state is false, the design path of the current candidate design state is cut; and if the two compatible rules and one mutually exclusive rule are both established in a return mode and the count value of the increment structural relation set is smaller than or equal to the current engineering configuration parameter, the consistency confirmation state is false, and the candidate design state is marked as the consistency confirmation state.
7. The ultra-large scale BOM design method based on stream constraint propagation and thinking chain reasoning cascade according to claim 6, wherein the combining of decision results with scale constraints on the incremental structural changes is specifically: And if the count value of the increment structural relation set exceeds the current engineering configuration parameter, judging that the candidate design state does not meet the constraint.
8. The method for designing a very large scale BOM based on cascade of stream constraint propagation and thinking chain reasoning according to claim 1, wherein S4 comprises: Carrying out level alignment on the structural relation in the consistency confirmation state according to a BOM level template established by an enterprise; Binding and summarizing material main data based on the consistency confirmation state after the level alignment to calculate the quantity of materials, wherein the material main data comprises a module package number, a component family number and a unique material code, and fills a supplementary edition field, a substitute identification field and an effective field to generate a standardized BOM; the binding of the material main data is performed based on the consistency confirmation state after the level alignment, specifically: the module package number or the part family number carried by the corresponding node in the consistency confirmation state is used as an index key, the unique material code is obtained by searching and is written into the corresponding node record; wherein, the total calculated material quantity specifically is: and under the same father node and the same hierarchy, counting all design examples binding the same unique material code, and taking the count value as the quantity value of the material under the unique material code.

Description

Ultra-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade Technical Field The invention belongs to the field of ultra-large BOM design, and particularly relates to an ultra-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade. Background As large-scale carrying equipment such as ships, rail transit, automobiles, aerospace and the like develop to the direction of platform, serialization and on-demand customization, BOMs are no longer just an affiliated list of design drawings in enterprises, but are core engineering carriers throughout design, purchase, manufacture, quality and service. The large-scale equipment has the characteristics of deep hierarchy, multiple parts and multiple variants, the structure inheritance between the platform BOM and the variant BOM is required to be maintained, the local difference caused by different task demands is required to be met, and meanwhile, engineering rules such as compatibility, dependence, mutual exclusion and the like are dispersed in platform specifications, interface standards, process constraints and engineering experience and are expressed in different modes in different systems. In the prior engineering practice, the BOM design adopts a working mode of 'first combining out scheme and then uniformly checking', namely engineers select modules or component families based on platform structures and experiences, and then carry out consistency check after forming a more complete BOM, and once conflicts are found, the cross-level reworking is often required, and particularly when the number of components is huge and the rule relation is dense, the design process can be repeatedly oscillated due to error exposure lag. Some systems attempt to retrieve auxiliary configurations via predefined rules or similar cases, but in very large BOM scenarios, the rule trigger ranges are difficult to control and the local change effects are difficult to locate, resulting in verification that is either too coarse, difficult to discover path errors early, or too full, computationally and debug costs are prohibitive. On the other hand, a consistent bearing mode is lacking among the platform structure, the rule constraint and the material main data, the intermediate selection generated in the design process is difficult to form checkable increment change, verification and convergence are often forced to depend on manual experience and later concentrated verification, and the requirements of large equipment engineering on the controllability of the design process and the deliverability of results are difficult to meet. Meanwhile, the demand input of actual projects is increasingly in a natural language form, the conversion from the demand to the structural configuration item to the BOM is often dependent on manual disassembly and interpretation, semantic deviation is easy to introduce, and even if intelligent question-answering or text understanding tools are introduced, a large amount of manual review is often required due to the lack of in-process gating of engineering rules and interface constraint, so that stable propulsion under dynamic constraint conditions is difficult. Therefore, there is a need for a method that can advance the design gradually within the platform structure boundaries, perform constraint decisions and path filtering in time around each local extension, and support end-to-end conversion from natural language requirements to standardized BOMs, so that the BOM construction process is within an engineering-controllable solution space from the start point, and the final output can be directly used for standardized BOMs of enterprise systems. Disclosure of Invention The invention aims to provide a super-large BOM design method based on stream constraint propagation and thinking chain reasoning cascade connection, which solves the problems. In order to achieve the above purpose, the present invention provides a method for designing a very large BOM based on stream constraint propagation and mental chain reasoning cascade, comprising the steps of: S1, acquiring a natural language demand set, analyzing the natural language demand set, and combining a historical structure relation set to determine a BOM design initial state through a preset function-structure mapping rule, wherein the BOM design initial state comprises a demand node set, an allowed initial structure relation set and a rule scope set; S2, performing thinking chain reasoning in the BOM design initial state to generate a candidate design state; S3, carrying out engineering rule judgment and path cutting on the candidate design states based on stream constraint propagation, and outputting a consistency confirmation state meeting the conditions, wherein the engineering rule judgment adopts an execution mode of an engineering rule engine; and S4, performing structural convergence based on the consistency confirmation