Engine part intelligent selection method and system based on multi-dimensional feature vector

Abstract

The application relates to the technical field of engine part data retrieval and inventory management, and discloses an engine part intelligent selection method and system based on a multidimensional feature vector. The method comprises the following steps: receiving part archives, adaptive model records, alternative part mapping relationships and industry dictionaries, generating part feature vector records and index records; analyzing query texts to generate query entity records and candidate filtering conditions; generating candidate matching records, alternative part candidate records and alternative relationship scores based on the candidate filtering conditions and the part feature vector records; reading inventory information to generate inventory state records, and generating recommendation result records and recommendation result versions; performing inventory consistency verification after receiving a confirmation instruction to generate inventory reservation records or candidate rearrangement results. The application forms a continuous record link of query analysis, candidate matching, alternative part processing and inventory verification, and effectively improves the coherence of the selection result and the inventory reservation.

Description

Technical Field

[0001] This invention relates to the field of engine component data retrieval and inventory management technology, and in particular to an intelligent selection method and system for engine components based on multidimensional feature vectors. Background Technology

[0002] In the field of engine component data retrieval and inventory management, existing solutions for linking engine component sales, maintenance spare parts supply, and spare parts inventory selection typically rely on component files, material codes, compatible model records, keyword queries, and inventory information display. While these solutions can complete component queries, inventory quantity retrieval, and candidate result display in typical scenarios, they are prone to limitations when part names, engine model abbreviations, common part names, substitute relationships, and inventory status all change simultaneously. These limitations include difficulty in matching query text with component files, discontinuity between candidate results and compatible models, and inconsistencies between recommended results and inventory versions.

[0003] Existing solutions often rely on manual experience, fixed material codes, common keyword searches, or single inventory quantities for processing. When customer query text contains brand aliases, model abbreviations, English abbreviations, common part names, OE numbers, or supplier codes, and the available inventory, reserved quantity, or inventory version in the inventory information changes, there is a lack of stable connection between the candidate component matching stage and the inventory reservation stage. This can easily lead to incomplete candidate set generation basis, the inability of substitute parts to participate in subsequent inventory processing, and inconsistencies between the inventory status corresponding to the confirmation instruction and the recommended results.

[0004] For the joint processing of query text, component files, compatible model records, alternative part mapping relationships, and inventory information, existing technologies still suffer from insufficient coordination between candidate matching, inventory status recording, recommended result versions, and inventory consistency verification. Therefore, it is necessary to address the issue of generating inventory reservation records or candidate rearrangement results based on query text and component files in the context of engine component selection and inventory reservation scenarios, under real-time changes in inventory status. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an intelligent selection method for engine components based on multi-dimensional feature vectors, comprising:

[0006] S100: Receives component files, compatible model records, substitute part mapping relationships, and industry dictionaries; performs standardized field mapping on component names, physical parameters, compatible models, component categories, and substitute part identifiers; and generates component feature vector records and index records.

[0007] S200: Receive query text, perform word segmentation, entity recognition and synonym normalization based on the industry dictionary, generate query entity records, and generate candidate filtering conditions from model entity and part category entity;

[0008] S300. Based on the candidate filtering conditions and the index records, determine the candidate set, generate candidate matching records based on the query entity records and the component feature vector records, and generate alternative candidate records and alternative relationship scores based on the alternative mapping relationship.

[0009] S400: Based on the candidate matching record and the alternative candidate record, read the inventory information and generate an inventory status record containing available inventory quantity, reserved quantity and inventory version; based on the candidate matching record, the alternative relationship score and the inventory status record, generate a recommendation result record and a recommendation result version.

[0010] S500: Receive confirmation instruction, perform inventory consistency verification based on the recommended result version, the inventory version, and the confirmed quantity, generate inventory reservation record when the verification passes, and generate candidate rearrangement result when the verification fails.

[0011] Further, in S100, the standardized field mapping includes: reading the field name, parameter value, unit, compatible model and part category from the component file; mapping the field name to a standardized field; generating a parameter normalization record for the parameter value and the unit; and writing the parameter normalization record into the component feature vector record.

[0012] Furthermore, the component feature vector record includes physical parameter record, compatible model record, part category record, text alias record, and substitute part identification record; the physical parameter record includes size parameters, material parameters, and interface parameters.

[0013] Furthermore, the industry dictionary includes records of brand aliases, model abbreviations, English abbreviations, common part names, OE numbers, and supplier codes; the synonym normalization includes mapping the brand aliases, model abbreviations, English abbreviations, common part names, OE numbers, and supplier codes in the query text to standardized field records in the query entity records.

[0014] Furthermore, the query entity record includes model entity, part category entity, parameter entity, quantity entity, and special constraint entity; when the model entity or the part category entity is missing, a missing field record is generated; when the query text corresponds to multiple model entities, an ambiguous entity record is generated.

[0015] Further, in S300, determining the candidate set includes: matching the index record with the model entity and part category entity in the candidate filtering conditions to obtain the candidate set; for the parts in the candidate set, calculating the similarity between the query entity record and the part feature vector record, and writing the similarity into the candidate matching record.

[0016] Further, in S300, generating the candidate records of the alternative parts includes: when the number of candidate matching records is less than a preset threshold, the inventory of candidate parts is insufficient, or the highest similarity is lower than the similarity threshold, reading the mapping relationship of the alternative parts; generating the candidate records of the alternative parts and the score of the alternative relationship based on the substitution direction, substitution conditions, and substitution priority.

[0017] Furthermore, in S400, the inventory status record also includes warehouse status and inventory update time; an inventory adequacy is generated based on the available inventory, the reserved inventory, the warehouse status, and the inventory update time; a comprehensive recommendation index is generated based on the candidate matching record, the substitution relationship score, and the inventory adequacy, and the comprehensive recommendation index is written into the recommendation result record.

[0018] Furthermore, in S500, the inventory consistency verification includes: comparing the inventory version corresponding to the recommended result version with the current inventory version, and determining that the current available inventory quantity is not less than the confirmed quantity; when the verification passes, generating the inventory reservation record and writing back the inventory information; when the verification fails, generating an inventory verification failure record, and generating the candidate rearrangement result based on the updated inventory status record.

[0019] Furthermore, an intelligent engine component selection system based on multi-dimensional feature vectors includes: a component file normalization module, a feature vector construction module, a query parsing module, a candidate matching module, an inventory status acquisition module, a recommendation sorting module, and an inventory reservation module; the system is used to implement the method described in any of the above-mentioned embodiments.

[0020] The key innovations of this invention include:

[0021] (1) Standardize the field mapping of the parts archive, the compatible model record, the alternative parts mapping relationship and the industry dictionary to generate the parts feature vector record and index record. Then, transform the query entity record formed by word segmentation, entity recognition and synonym normalization of the query text into candidate filtering conditions, so that the parts archive field and the query text field form a corresponding processing link at the model entity and part category entity level.

[0022] (2) Based on the candidate filtering conditions and index records, determine the candidate set, generate candidate matching records based on the query entity records and component feature vector records in the candidate set, and transform the substitute part mapping relationship into substitute part candidate records and substitute relationship scores, so that the substitution direction, substitution conditions and substitution priority are used as candidate processing fields to enter the subsequent inventory processing.

[0023] (3) Read inventory information based on candidate matching records and alternative candidate records, generate inventory status records containing available inventory quantity, reserved quantity and inventory version, bind candidate matching records, alternative relationship scores and inventory status records to generate recommended result records and recommended result versions, and then perform inventory consistency verification based on recommended result versions, inventory versions and confirmed quantities to form inventory reserved records or candidate rearrangement results.

[0024] The following are its main beneficial effects:

[0025] (1) To address the inconsistency in name, model and part category expression between query text and parts archives, the continuous generation of standardized field mapping, query entity records and candidate filtering conditions enables the model entity and part category entity in the query text to correspond with the matching model and part category in the index records. The candidate set generation process has clear data sources and field basis.

[0026] (2) To address the problem that substitute parts relationships are difficult to participate in candidate results and inventory processing, the substitute parts mapping relationship is connected to generate substitute parts candidate records and substitute relationship scores through the connection processing of candidate matching records and substitute parts mapping relationship, so that the original candidate parts and substitute parts candidate records can enter the inventory status reading and recommended result record generation process together.

[0027] (3) To address the issue of inconsistency between the recommended result record and the inventory version, the inventory status record, recommended result version and inventory consistency verification are processed continuously to ensure that the inventory version corresponding to the confirmation instruction and the current inventory version are reviewed before the inventory reservation record is generated; when the inventory version or available inventory quantity changes, candidate rearrangement results are generated. Attached Figure Description

[0028] Figure 1 A flowchart illustrating the intelligent selection method for engine components based on multidimensional feature vectors provided in this application embodiment;

[0029] Figure 2 The structural block diagram of the intelligent engine component selection system based on multidimensional feature vectors provided in the embodiments of this application is shown. Detailed Implementation

[0030] Example 1: Refer to Figure 1This is a flowchart illustrating the intelligent selection method for engine components based on multi-dimensional feature vectors provided in this embodiment of the invention. The process may include at least steps S100-S500:

[0031] S100: Receives component files, compatible model records, substitute part mapping relationships, and industry dictionaries; performs standardized field mapping on component names, physical parameters, compatible models, component categories, and substitute part identifiers; and generates component feature vector records and index records.

[0032] S200: Receive query text, perform word segmentation, entity recognition and synonym normalization based on the industry dictionary, generate query entity records, and generate candidate filtering conditions from model entity and part category entity;

[0033] S300. Based on the candidate filtering conditions and the index records, determine the candidate set, generate candidate matching records based on the query entity records and the component feature vector records, and generate alternative candidate records and alternative relationship scores based on the alternative mapping relationship.

[0034] S400: Based on the candidate matching record and the alternative candidate record, read the inventory information and generate an inventory status record containing available inventory quantity, reserved quantity and inventory version; based on the candidate matching record, the alternative relationship score and the inventory status record, generate a recommendation result record and a recommendation result version.

[0035] S500: Receive confirmation instruction, perform inventory consistency verification based on the recommended result version, the inventory version, and the confirmed quantity, generate inventory reservation record when the verification passes, and generate candidate rearrangement result when the verification fails.

[0036] S100 receives component files, compatible model records, alternative part mapping relationships, and industry dictionaries. It performs standardized field mapping on component names, physical parameters, compatible models, component categories, and alternative part identifiers, and generates component feature vector records and index records.

[0037] S100 is executed by the component file normalization module and the feature vector construction module. The component file normalization module is deployed in the server of the engine component selection system, and the server is connected to the database. The component file is a material record stored in the engine component business system, which includes component number, original part name, brand field, specification field, physical parameter field, supplier code field, and inventory system material code field. The compatible model record is a corresponding record between the component and the engine brand, engine series, model number, displacement field, or power segment field. The substitute part mapping relationship is a mapping relationship between the original component and the substitute component, which includes the substitution direction, substitution conditions, substitution priority, and substitute part identifier. The industry dictionary is a name mapping record in the field of engine components, which includes brand alias record, model abbreviation record, English abbreviation record, common part name record, OE number record, and supplier code record, where the OE number is the Original Equipment number, i.e., the original factory matching number.

[0038] In this step, the component file normalization module is activated upon receiving updates to component files, compatible model records, alternative part mapping relationships, or industry dictionary versions. The component files are accessed from the ERP system, warehouse system, supplier tables, or manual maintenance database. The compatible model records are accessed from the manufacturer directory, historical sales records, or maintenance spare parts records. The alternative part mapping relationships are accessed from the supplier alternative list, manual review records, or historical alternative records. The industry dictionary is read from the dictionary table in the database. The component file normalization module reads the field names, parameter values, units, compatible models, and part categories from each component file entry and matches the original field names with standardized fields in the industry dictionary. For example, "cylinder bore," "cylinder diameter," and "Bore" are mapped to the same standardized field; "Cummins," "Cummins," and "Kangji" are mapped to the same brand field; and "piston ring," "ring assembly," and "piston ring" are mapped to the same part category field.

[0039] Specifically, the standardized field mapping includes field name mapping, parameter value records, unit records, and field source records. Field name mapping uses standardized fields from an industry dictionary as the matching basis, writing the original fields from the component archive into the standardized field record. The parameter value record stores the original parameter value, the normalized parameter value, and the field source. The unit record stores the original unit and the normalized unit. The field source record stores the source identifier of the parameter value from the component archive, component specification sheet, compatible model record, or manual maintenance record. For size parameters, in one implementation, the server writes inches and millimeters into the same parameter normalization record and retains the original unit field. For material parameters, the server writes "alloy steel," "cast iron," "rubber," and "composite material" into the material parameter field. For interface parameters, the server writes thread specifications, hole diameter, outer diameter, inner diameter, installation position, or connection type into the interface parameter field.

[0040] The component feature vector record is generated by the feature vector construction module based on standardized field records. This component feature vector record is not a single text field, but rather stores physical parameter records, compatible model records, part category records, text alias records, and substitute part identification records in record form. The physical parameter records include dimensional parameters, material parameters, and interface parameters; the compatible model records include engine brand, engine series, model entity, and compatibility range; the part category records include standard part categories and part classification levels; the text alias records include the original part name, brand alias, model abbreviation, English abbreviation, part common name, OE number, and supplier code; the substitute part identification record includes the substitute part group identifier, substitution direction, substitution conditions, and substitution priority. The feature vector construction module writes these multiple records into the data record corresponding to the same component number and assigns a record version to that data record.

[0041] The index records are established by the feature vector construction module after generating component feature vector records. These index records include an engine model index, a part category index, an OE number index, a supplier code index, and a substitute part group index. The engine model index uses the engine brand, engine series, and engine entity from the engine model record as the index key; the part category index uses the standard part category as the index key; the OE number index and supplier code index are used to receive the number fields appearing in subsequent query text; and the substitute part group index is used to read the substitute part mapping relationship in the S300. The index records are bound to the component feature vector records through the component number and record version, and the database stores this binding relationship.

[0042] In one implementation, when a field is missing in the component file, the component file normalization module generates a field missing marker and writes the read fields into the standardized field record. The field missing marker includes the missing field name, the source of the missing record, and the processing status. For cases where multiple parameter values ​​exist for the same component number, the server generates a parameter conflict marker based on the field source record and the record update time, while retaining the original parameter value. Before manual review is completed, the feature vector construction module writes the parameter conflict marker into the status field of the component feature vector record. For cases where the compatible model record and the part category record are inconsistent, the server generates an adaptation conflict record and marks the component number as pending review in the index record.

[0043] In one implementation, the piston ring file maintained by the operating entity has the original part name "6BT Ring Set," the brand field "Cummins," and the specification fields "102mm, Cast Iron, Set." The compatible model record includes "Cummins 6BT." After reading the file, the component file normalization module maps "Cummins" to the standard brand field, writes "6BT" into the model entity, maps "Ring Set" to the part category field "Piston Ring," writes "102mm" into the size parameter, and writes "Cast Iron" into the material parameter. The feature vector construction module generates the corresponding component feature vector record based on this standardized field record and establishes an index record of "6BT—Piston Ring" in the model index and part category index. This index record is called by the candidate filtering conditions in S300, and the physical parameter record, compatible model record, and substitute part identification record in the component feature vector record enter the candidate matching process.

[0044] After S100 is completed, the server writes the standardized field records, component feature vector records, and index records into the database, and records the industry dictionary version, component file version, and record generation time; the component feature vector records and index records serve as inputs for S300 to determine the candidate set and generate candidate matching records, and the industry dictionary serves as inputs for S200 to perform word segmentation, entity recognition, and synonym normalization on the query text.

[0045] S200: Receive query text, perform word segmentation, entity recognition, and synonym normalization based on the industry dictionary, generate query entity records, and generate candidate filtering conditions from the model entity and part category entity.

[0046] S200 is executed by the query parsing module, which is deployed on the server of the engine parts selection system and connected to the sales terminal and the database. The query text is the text content entered by the sales terminal, the customer's online inquiry portal, or the remarks on the repair work order, including brand aliases, model abbreviations, common names of parts, English abbreviations, OE numbers, supplier codes, parameter descriptions, quantity descriptions, or special constraint descriptions. The query parsing module starts when it receives a query text submission event and reads the industry dictionary, the record version of the parts feature vector record, and the index version of the index record stored in S100; the industry dictionary serves as the basis for entity recognition and synonym normalization in this step.

[0047] The word segmentation refers to the query parsing module dividing the query text into fragments according to brand alias records, model abbreviation records, English abbreviation records, part common names records, OE number records, and supplier code records in the industry dictionary. The entity recognition refers to matching the segmented text fragments into model entities, part category entities, parameter entities, quantity entities, or special constraint entities. The synonym normalization refers to mapping the brand aliases, model abbreviations, English abbreviations, part common names, OE numbers, and supplier codes in the query text to standardized field records in the query entity records. When performing synonym normalization, the query parsing module retains the original text fragments, standardized fields, dictionary version, and matching source, generating query entity records.

[0048] The query entity record includes engine type entity, part category entity, parameter entity, quantity entity, and special constraint entity. The engine type entity consists of a brand field, an engine series field, and an engine number field; the part category entity consists of a standard part category and a part classification level; the parameter entity consists of a parameter name, parameter value, and unit; the quantity entity consists of the requested quantity and unit; and the special constraint entity consists of brand requirements, original equipment manufacturer (OEM) requirements, replacement part acceptance status, or warehouse restriction fields. In one implementation, the query text is "Cummins 6BT piston rings 2 sets". The query parsing module normalizes "Cummins" to a brand field, identifies "6BT" as an engine type entity, "piston rings" as a part category entity, and "2 sets" as a quantity entity, and writes these fields into the same query entity record.

[0049] Candidate filtering conditions are generated by the query parsing module based on the model entity and part category entity in the query entity record. These candidate filtering conditions include model conditions, part category conditions, number conditions, and parameter conditions. Model conditions are derived from the engine brand, engine series, and model entity in the query entity record; part category conditions are derived from the standard part category field; number conditions are derived from the OE number or supplier code; and parameter conditions are derived from size parameters, material parameters, or interface parameters. In this step, the model entity and part category entity serve as the main fields for candidate filtering conditions. Parameter entities, quantity entities, and special constraint entities are written into the query entity record and used as input fields for S300 matching and S400 inventory processing.

[0050] When no machine type entity is identified in the query text, the query parsing module generates a missing field record, which includes the missing field name, the original query text, the session identifier, and the dictionary version. When no part category entity is identified in the query text, the query parsing module generates a corresponding missing field record and marks the candidate filter conditions as pending completion. For cases where the same query text corresponds to multiple machine type entities, the query parsing module generates an ambiguous entity record; this ambiguous entity record includes multiple machine type entities, the dictionary source for each machine type entity, the original text fragment, and the sorting status. In one implementation, when "6B" in the query text simultaneously corresponds to multiple engine series, the query parsing module writes multiple machine type entities into the ambiguous entity record and marks the candidate filter conditions as grouped candidate states.

[0051] In one implementation, the query parsing module receives customer input "Kangji 6BT ring, 102, in stock"; the query parsing module maps "Kangji" to a brand field based on an industry dictionary, writes "6BT" into the machine model entity, maps "ring" to the part category entity "piston ring", uses "102" as a size parameter entity, and uses "in stock" as an inventory constraint field in a special constraint entity; then, a query entity record is generated, and candidate filtering conditions are generated from the machine model entity "6BT" and the part category entity "piston ring". These candidate filtering conditions are matched against the index record in S300, and the size parameter entity in the query entity record enters the candidate matching record generation process.

[0052] In another implementation, when the query text contains an OE number or supplier code, the query parsing module first writes the OE number or supplier code into the number field of the query entity record. If the number field matches an OE number record or supplier code record in the industry dictionary, the query parsing module converts the number field into a part category entity or component number candidate and writes this number condition into the candidate filter conditions. If the number field has an incomplete format or cannot be matched, the query parsing module generates a missing field record or a number mismatch record and saves the original number field to the query entity record.

[0053] After S200 is completed, the query parsing module writes the query entity records, candidate filter conditions, missing field records, and ambiguous entity records to the database. The query entity records and candidate filter conditions serve as input for S300 to determine the candidate set and generate candidate matching records, while the missing field records and ambiguous entity records serve as status fields when S300 generates the candidate set. The query entity records are also bound to a sales-side session identifier; subsequently, when S400 generates recommendation result records, the server writes this session identifier into the recommendation result version.

[0054] S300. Based on the candidate filtering conditions and the index records, determine the candidate set, generate candidate matching records based on the query entity records and the component feature vector records, and generate alternative candidate records and alternative relationship scores based on the alternative mapping relationship.

[0055] S300 is executed by the candidate matching module, which receives the query entity record and candidate filtering conditions generated by S200, and reads the index record, component feature vector record and substitute part mapping relationship generated by S100; the candidate set is the set of component numbers matched from the index record according to the candidate filtering conditions; the candidate matching record is the matching result record between each component in the candidate set and the query entity record, including component number, matching field, similarity, matching status and record version; the substitute part candidate record is the candidate record generated based on the substitute part mapping relationship, including the original candidate component number, substitute component number, substitution direction, substitution condition and substitution priority.

[0056] The candidate matching module first matches index records based on the model entity and part category entity in the candidate filtering conditions. Specifically, when the candidate filtering conditions include model entities and part category entities, the candidate matching module reads the model index and part category index respectively, and takes the part number that matches both the model entity and the part category entity as the candidate set; when the candidate filtering conditions also include OE number or supplier code, the candidate matching module reads the OE number index or supplier code index, and writes the part number obtained by number matching into the candidate set; when the candidate filtering conditions include parameter entities, the candidate matching module reads the physical parameter record in the corresponding part feature vector record in the candidate set, and uses the parameter entity as the input for subsequent similarity calculation.

[0057] When generating candidate matching records, the candidate matching module reads the model entity, part category entity, parameter entity, quantity entity, and special constraint entity from the query entity record. The candidate matching module also reads the part feature vector record corresponding to each part number in the candidate set and performs field matching in the physical parameter record, compatible model record, part category record, text alias record, and substitute part identifier record. For the model entity, the candidate matching module matches the model entity in the query entity record with the compatible model record in the part feature vector record; for the part category entity, the candidate matching module matches the standard part category with the part category record; for the parameter entity, the candidate matching module matches the parameter name, parameter value, and unit with the physical parameter record; for the text alias, the candidate matching module matches the query text fragment with the text alias record.

[0058] The similarity score is written to the candidate matching record. The calculation of similarity is not limited to a single model; in one implementation, the candidate matching module calculates similarity according to field matching methods, generating field matching values ​​for machine model entity matching, part category entity matching, parameter entity matching, and text alias matching respectively, and then writes multiple field matching values ​​to the candidate matching record; in another implementation, the candidate matching module performs vector calculations on the query entity record and the part feature vector record, and writes the calculation results to the similarity field. Regardless of whether field matching or vector calculation is used, the candidate matching record stores the matching field, similarity score, part number, and record version. Subsequently, S400 reads inventory information using the part number in the candidate matching record.

[0059] The candidate records for replacement parts are generated by the candidate matching module based on the replacement part mapping relationship. In one implementation, when the number of candidate matching records is less than a preset threshold, the candidate matching module reads the replacement part group identifier in the replacement part mapping relationship and writes the part numbers that meet the replacement conditions within the same replacement part group into the candidate record for replacement parts. In another implementation, when the inventory of candidate parts is insufficient, the candidate matching module reads the replacement part identifier corresponding to the candidate part and reads the replacement part number according to the replacement direction and replacement conditions. In yet another implementation, when the highest similarity is lower than the similarity threshold, the candidate matching module reads the replacement part mapping relationship and writes the replacement part number with higher replacement priority into the candidate record for replacement parts. The preset threshold and similarity threshold are stored in the configuration records in the database, and the server reads the corresponding version during calculation.

[0060] The substitution relationship score is generated based on substitution direction, substitution conditions, and substitution priority. Substitution direction distinguishes between unidirectional and bidirectional substitution; substitution conditions record compatible models, parameter ranges, brand requirements, or part category requirements; substitution priority records the sorting field among multiple substitute parts. The candidate matching module first reads the substitution direction field from the substitution mapping relationship, then matches the substitution conditions with the model entity, part category entity, and parameter entity in the query entity record. If the substitution conditions match the fields in the query entity record, the candidate matching module reads the substitution priority and generates a substitution relationship score; if the substitution conditions do not match the fields in the query entity record, the candidate matching module generates a non-matching substitution record and does not write the substitution part number into the substitution candidate record.

[0061] For cases where S200 generates records with missing fields or ambiguous entity records, S300 processes them according to the status field. When the candidate filtering condition is a state to be supplemented, the candidate matching module reads the identified part category entities or number fields from the query entity records and generates a candidate set with fewer limiting fields, while writing low-confidence records into the candidate matching records. When the candidate filtering condition is a grouped candidate state, the candidate matching module matches index records with multiple model entities respectively, generates multiple candidate sets, and writes the corresponding model entity identifier into the candidate matching records. The above records enter the generation process of recommendation result records in S400, which facilitates the saving of candidate sources and matching status in the recommendation result records.

[0062] In one implementation, the candidate filtering condition output by S200 is "engine type entity = 6BT, part category entity = piston ring", and the query entity record also includes the parameter entity "102mm". The candidate matching module reads the index record to obtain multiple piston ring part numbers and reads the corresponding part feature vector record one by one. For parts with a cylinder diameter of 102mm, an engine type record containing 6BT, and a part category record of piston ring, the candidate matching module generates candidate matching records with high similarity. If one of the candidate parts has no available inventory or the number of candidate matching records is less than a preset threshold, the candidate matching module reads the alternative part mapping relationship, generates alternative part candidate records based on the substitution direction and substitution conditions, and writes the substitution relationship score. The part numbers in the candidate matching records and alternative part candidate records are used as index fields for reading inventory information in S400.

[0063] After S300 is completed, the candidate matching module writes the candidate matching record, the alternative candidate record, the substitution relationship score, the low confidence record, and the alternative unmatched record into the database. The candidate matching record and the alternative candidate record are used as inputs for S400 to read inventory information, the substitution relationship score is used as inputs for S400 to generate recommendation result records and comprehensive recommendation index, and the low confidence record is written into the status field of the recommendation result record.

[0064] S400: Based on the candidate matching record and the alternative candidate record, read the inventory information and generate an inventory status record containing available inventory quantity, reserved quantity and inventory version. Based on the candidate matching record, the alternative relationship score and the inventory status record, generate a recommendation result record and a recommendation result version.

[0065] S400 is executed by the inventory status acquisition module and the recommendation sorting module; the inventory status acquisition module receives the candidate matching records and alternative part candidate records generated by S300, and reads the inventory information based on the part number therein; the inventory information comes from the warehousing system, ERP system or inventory database, and includes warehouse identifier, current quantity, reserved quantity, available inventory quantity, warehouse status, inventory version and inventory update time; the inventory status record is the data record formed after binding the inventory information with the candidate matching records in this step, and includes part number, warehouse identifier, available inventory quantity, reserved quantity, inventory version, inventory update time, warehouse status and inventory read time.

[0066] The inventory status acquisition module reads inventory information one by one according to the component numbers in the candidate matching records. For alternative component candidate records, the inventory status acquisition module reads the inventory information corresponding to the alternative component number and writes both the original candidate component number and the alternative component number into the inventory status record. The available inventory quantity is generated from the existing quantity and reserved quantity in the inventory information. When the inventory system directly provides the available inventory quantity, the inventory status acquisition module reads and records this field; when the inventory system only provides the existing quantity and reserved quantity, the inventory status acquisition module reads both fields and generates the available inventory quantity. The inventory version is used to identify the update time of the inventory information or the inventory record version. The inventory status acquisition module writes the inventory version into the inventory status record and binds the inventory status record to the record version of the candidate matching record.

[0067] In one implementation, the inventory information includes inventory records from multiple warehouses. The inventory status acquisition module reads the inventory information under multiple warehouse identifiers according to the part number and generates inventory status records for each. The warehouse status records whether the warehouse is open for outbound shipments, whether there is an inventory count, and whether there is a frozen status. The inventory update time records the time when the inventory information is generated or synchronized. If the inventory interface returns a failure, the inventory status acquisition module generates an inventory interface exception record and marks the inventory status of the corresponding part number as unconfirmable. If the inventory version is missing, the inventory status acquisition module generates an inventory version missing record and retains candidate matching records, waiting for S500 or the next round of inventory reading for re-verification.

[0068] The recommendation ranking module generates recommendation result records based on candidate matching records, substitution relationship scores, and inventory status records. Each recommendation result record includes the query entity record identifier, component number, candidate source, similarity, substitution relationship score, inventory adequacy, comprehensive recommendation index, inventory version, recommendation result version, and recommendation status. Inventory adequacy is generated based on available inventory, reserved quantity, warehouse status, and inventory update time. In one implementation, when available inventory is greater than or equal to the quantity entity in the query entity record and the warehouse status is "ready to ship," the inventory adequacy is written to a "reservable" status; when available inventory is less than the confirmed quantity or the warehouse status is "frozen," the inventory adequacy is written to a "non-reservable" status; when the inventory interface is abnormal or the inventory version is missing, the inventory adequacy is written to a "pending confirmation" status.

[0069] The comprehensive recommendation index is generated by the recommendation ranking module based on candidate matching records, substitution relationship scores, and inventory adequacy. The similarity in candidate matching records reflects the matching between the query entity record and the component feature vector record; the substitution relationship score reflects the substitution direction, substitution conditions, and substitution priority in the candidate substitution records; and inventory adequacy reflects the available inventory, reserved inventory, warehouse status, and inventory update time in the inventory status record. The recommendation ranking module writes these fields into the comprehensive recommendation index according to the configuration records and then writes the comprehensive recommendation index into the recommendation result record. In another implementation, the recommendation ranking module does not output a single value but instead writes similarity, substitution relationship scores, and inventory adequacy as multiple ranking fields into the recommendation result record; the sales terminal reads the recommendation result record and displays the candidate component list according to the ranking fields.

[0070] The recommended result version is generated by the recommendation sorting module when generating the recommended result record. This recommended result version is bound to the query entity record identifier, candidate matching record version, alternative candidate record version, and inventory version. The recommended result version does not replace the inventory version; instead, it records the data combination used when generating this recommended result. Subsequently, when S500 receives a confirmation command, the inventory reservation module reads the recommended result version and performs inventory consistency verification based on the inventory version in the recommended result version and the current inventory version. For low-confidence records, records with missing fields, or ambiguous entity records, the recommendation sorting module writes the corresponding status into the recommended result record and retains its binding relationship with the query entity record.

[0071] In one implementation, the candidate matching record generated in S300 includes component A and component B, and the candidate alternative part record includes component C. After reading the inventory information of the three components, the inventory status acquisition module finds that component A has an available inventory of 3 in warehouse 1, a reserved quantity of 1, and an inventory version of V12; component B has an available inventory of 0 and an inventory version of V18; and component C has an available inventory of 5 and an inventory version of V21. The recommendation ranking module writes component A, component B, and component C into the recommendation result record, writes component B into the non-reservable status, writes component C into the alternative part candidate source field, and generates a comprehensive recommendation index based on the candidate matching record, the substitution relationship score, and the inventory adequacy. The recommendation result version record records the inventory versions V12, V18, and V21 read this time, and uses them as input for inventory consistency verification in S500.

[0072] After S400 is completed, the inventory status acquisition module writes the inventory status record to the database, and the recommendation sorting module writes the recommendation result record and recommendation result version to the database and outputs them to the sales terminal. The part number, inventory version, available inventory quantity, comprehensive recommendation index and recommendation result version in the recommendation result record enter S500. When the sales terminal receives the confirmation instruction, it submits the confirmation instruction, confirmation quantity and recommendation result version to the inventory reservation module.

[0073] S500: Receive confirmation instruction, perform inventory consistency verification based on the recommended result version, the inventory version, and the confirmed quantity, generate inventory reservation record when the verification passes, and generate candidate rearrangement result when the verification fails.

[0074] S500 is executed by the inventory reservation module, which receives a confirmation instruction submitted by the sales terminal or business system. The confirmation instruction includes the recommended result version, the confirmed part number, the confirmed quantity, the sales terminal session identifier, and the confirmation time. The recommended result version comes from the recommended result record generated by S400. The inventory version comes from the inventory status record generated by S400 and is compared by the inventory reservation module after reading the current inventory information during confirmation. The confirmed quantity comes from the quantity entity in the query entity record or the quantity field entered by the sales terminal during confirmation.

[0075] Upon receiving the confirmation command, the inventory reservation module first reads the recommendation result record corresponding to the recommended version, and then reads the inventory status record saved by S400 based on the confirmed part number in the recommendation result record. Subsequently, the inventory reservation module reads the current inventory information from the inventory system, obtaining the current inventory version and the current available inventory quantity. The inventory consistency verification includes two processes: first, comparing the inventory version corresponding to the recommended version with the current inventory version; second, determining that the current available inventory quantity is not less than the confirmed quantity. The inventory reservation module writes the results of these two processes into the inventory verification result and binds the inventory verification result to the confirmation command.

[0076] When the inventory version corresponding to the recommended result version matches the current inventory version, and the current available inventory quantity is not less than the confirmed quantity, the inventory reservation module generates an inventory reservation record. The inventory reservation record includes a reservation record number, confirmed component number, confirmed quantity, warehouse identifier, inventory version, recommended result version, sales session identifier, reservation status, and reservation time. After generating the inventory reservation record, the inventory reservation module writes back inventory information to the inventory system. The written-back content includes the reservation record number, reservation quantity, inventory deduction field, and inventory version update field. After the inventory system completes the write-back, the inventory reservation module saves the inventory write-back record to the database and binds the inventory reservation record to the query entity record, candidate matching record, and recommended result record.

[0077] When the inventory version corresponding to the recommended result version is inconsistent with the current inventory version, the inventory reservation module generates an inventory verification failure record. This record includes the confirmation instruction, the recommended result version, the inventory version at the time of recommendation, the current inventory version, the confirmation quantity, and the failure status. The inventory reservation module then reads the updated inventory status record and regenerates the candidate re-ranking results based on it. The candidate re-ranking results include the original recommended part number, the updated available inventory quantity, the updated inventory version, the candidate sorting field, and the re-ranking time. If the current available inventory quantity is less than the confirmed quantity, the inventory reservation module also generates an inventory verification failure record and generates candidate re-ranking results based on the updated inventory status record.

[0078] In one implementation, the sales terminal submits a confirmation command for component A in the recommended result record, confirming a quantity of 2 and the inventory version bound to the recommended result version as V12. Upon confirmation, the inventory reservation module reads the current inventory information from the inventory system, finding that the current inventory version is still V12 and the current available inventory quantity is 3. The inventory reservation module generates an inventory verification result indicating success and generates an inventory reservation record, recording the confirmed component number, confirmed quantity, warehouse identifier, inventory version, and recommended result version. The inventory reservation module then writes the inventory reservation record back to the inventory system and saves the inventory write-back record. This inventory reservation record is subsequently read by the order processing system or the warehousing system.

[0079] In another implementation, when the sales terminal submits a confirmation command, the inventory reservation module reads that the current inventory version is V13 and the recommended result version saves the inventory version as V12. The inventory reservation module generates an inventory verification failure record and reads the inventory status record corresponding to V13. If the updated inventory status record shows that the available inventory of the original component A is insufficient, the inventory reservation module generates a candidate re-ranking result based on the recommended result record in S400 and the updated inventory status record, and ranks the candidate component or substitute component candidate record with the reservable status in the first position. If there is a substitute component candidate record, the inventory reservation module writes the substitute component number, substitution relationship score and updated inventory status record in the substitute component candidate record into the candidate re-ranking result.

[0080] For cases where the confirmation instruction lacks a confirmation quantity, the recommended result version is nonexistent, or the confirmed part number is inconsistent with the recommended result record, the inventory reservation module generates a confirmation instruction exception record but does not generate an inventory reservation record. For cases where the inventory system write-back times out, the inventory reservation module writes the inventory reservation record to the pending write-back state and generates an inventory write-back exception record. When the next inventory write-back task is triggered, the inventory reservation module reads the inventory reservation record in the pending write-back state and resubmits the write-back content to the inventory system. For cases where the inventory interface returns a failure, the inventory reservation module generates an inventory interface exception record and marks the candidate reordering result as an inventory pending confirmation state.

[0081] In one implementation, the inventory reservation module also writes inventory reservation records, inventory verification failure records, candidate rearrangement results, and inventory write-back records into a processing record under the same sales session identifier. The processing record stores the correspondence between query text, query entity records, candidate matching records, recommended result versions, inventory versions, and confirmation instructions. When inventory information is subsequently updated or the sales end resubmits a confirmation instruction, the inventory reservation module reads this processing record and re-executes the inventory consistency verification according to the current inventory version. After S500 is completed, the system generates an inventory reservation record or candidate rearrangement result; the inventory reservation record enters the inventory system, order processing system, or warehousing system, and the candidate rearrangement result is returned to the sales terminal as input for the next confirmation instruction.

[0082] Example 2: Figure 2 A structural block diagram of an intelligent engine component selection system based on multi-dimensional feature vectors according to an embodiment of the present invention is shown. Figure 2 As shown, the structure may include:

[0083] The component file normalization module 01 is used to receive component files, compatible model records, substitute part mapping relationships, and an industry dictionary. It performs standardized field mapping on part names, physical parameters, compatible models, part categories, and substitute part identifiers, generating standardized field records. Specifically, the component file normalization module receives input objects from component files, compatible model records, substitute part mapping relationships, and an industry dictionary. It reads the part names, physical parameters, compatible models, part categories, and substitute part identifiers from the component files. It matches the part names with brand alias records, model abbreviation records, English abbreviation records, common part names, OE number records, and supplier code records from the industry dictionary. It performs field name mapping, parameter value recording, and unit recording for the size parameters, material parameters, and interface parameters in the physical parameters. It writes the compatible model and the part category into the model entity field and part category field, respectively. It records the substitution direction, substitution conditions, and substitution priority in the substitute part identifier and the substitute part mapping relationship. The component file normalization module performs field reading upon receiving updates to component files, compatible model records, substitute part mapping relationships, or industry dictionaries. When field names are missing, parameter values ​​are missing, or the compatible model and part category are inconsistent, the original fields, missing statuses, and conflict statuses are written to a standardized field record. This standardized field record includes the component number, standard part name, standard part category, physical parameters, compatible model, substitute part identifier, and field source. The component file normalization module provides this standardized field record to the feature vector construction module and retains the correspondence between the standardized field record and the component files, compatible model records, substitute part mapping relationships, and industry dictionaries.

[0084] Feature vector construction module 02, connected to the component file normalization module, is used to generate component feature vector records and index records based on the standardized field records. Specifically, the feature vector construction module receives standardized field records from the component file normalization module, reads the component number, standard part name, physical parameters, compatible model, part category, text alias, and substitute part identifier from the standardized field records; writes the physical parameters into the physical parameter record, the compatible model into the compatible model record, the part category into the part category record, the text alias into the text alias record, and the substitute part identifier into the substitute part identifier record; and generates component feature vector records based on the above records. The feature vector construction module establishes index records according to the compatible model, the part category, the OE number, the supplier code, and the substitute part identifier; the index records include a model index, a part category index, an OE number index, a supplier code index, and a substitute part group index. When a missing or conflicting state exists in the standardized field record, the feature vector construction module writes the missing or conflicting state into the state field of the component feature vector record; the index record maintains consistency with the component number and record version of the component feature vector record. The feature vector construction module provides the industry dictionary reading results to the query parsing module and the component feature vector record and the index record to the candidate matching module; the physical parameter record, compatible model record, part category record, text alias record, and substitute part identifier record in the component feature vector record serve as input objects for the candidate matching module.

[0085] The query parsing module 03, connected to the feature vector construction module, receives query text, performs word segmentation, entity recognition, and synonym normalization based on the industry dictionary, generates query entity records, and generates candidate filtering conditions from the model entity and part category entity. Specifically, the query parsing module receives query text and calls the brand alias record, model abbreviation record, English abbreviation record, part common name record, OE number record, and supplier code record from the industry dictionary; it performs word segmentation on the query text, matches the segmented text fragments with the fields in the industry dictionary, and writes the matched brand field, model field, part category field, parameter field, quantity field, and special constraint field into the query entity record. The query parsing module normalizes the synonyms in the query text, mapping the brand alias, model abbreviation, English abbreviation, part common name, OE number, and supplier code to standardized field records in the query entity record; the query entity record includes model entity, part category entity, parameter entity, quantity entity, and special constraint entity. The query parsing module generates candidate filtering conditions based on the model entity and the part category entity; when the model entity or the part category entity is missing, a missing field record is generated; when the query text corresponds to multiple model entities, an ambiguous entity record is generated. The query parsing module provides the query entity record, the candidate filtering conditions, the missing field record, and the ambiguous entity record to the candidate matching module; the model entity and part category entity in the candidate filtering conditions serve as input fields for the candidate matching module to match the index record.

[0086] The candidate matching module 04, connected to the query parsing module, is used to determine a candidate set based on the candidate filtering conditions and the index records, generate candidate matching records based on the query entity records and the component feature vector records, and generate alternative component candidate records and alternative relationship scores based on the alternative component mapping relationship. Specifically, the candidate matching module receives candidate filtering conditions and query entity records from the query parsing module, and reads index records and component feature vector records from the feature vector construction module; it matches the index records with the model entity and part category entity in the candidate filtering conditions to obtain a candidate set. The candidate matching module reads the component feature vector records corresponding to each component in the candidate set, and matches the model entity, part category entity, parameter entity, quantity entity, and special constraint entity in the query entity records with the matching model record, part category record, physical parameter record, text alias record, and alternative component identifier record in the component feature vector records, respectively; it writes the matching fields, similarity, component number, and record version into the candidate matching record. The candidate matching module reads the alternative part mapping relationship when the number of candidate matching records is less than a preset threshold, the inventory of candidate parts is insufficient, or the highest similarity is lower than the similarity threshold; it generates alternative part candidate records and an alternative relationship score based on the substitution direction, substitution conditions, and substitution priority. The candidate matching module generates unmatched alternative part records for parts that do not meet the substitution conditions and saves these unmatched alternative part records corresponding to the query entity records. The candidate matching module provides the candidate matching records and the alternative part candidate records to the inventory status acquisition module and provides the alternative relationship score to the recommendation and sorting module; the part numbers in the candidate matching records and the alternative part candidate records serve as input fields for reading inventory information.

[0087] The inventory status acquisition module 05, connected to the candidate matching module, is used to read inventory information based on the candidate matching record and the alternative part candidate record, and generate an inventory status record containing available inventory quantity, reserved quantity, and inventory version. Specifically, the inventory status acquisition module receives candidate matching records and alternative part candidate records from the candidate matching module, reads the part numbers in the candidate matching records and the alternative part candidate records, reads inventory information based on the part numbers, and writes the warehouse identifier, available inventory quantity, reserved quantity, warehouse status, inventory version, and inventory update time from the inventory information into the inventory status record. When reading inventory information corresponding to multiple warehouses, the inventory status acquisition module generates inventory status records according to the part number and warehouse identifier respectively; when the inventory information lacks an inventory version, an inventory version missing record is generated; when the inventory information reading fails, an inventory interface exception record is generated; when the available inventory quantity is less than the quantity entity in the query entity record, an insufficient inventory status is written into the inventory status record. The inventory status record corresponds to the candidate matching record and the alternative part candidate record through the part number; the inventory status acquisition module provides the inventory status record to the recommendation sorting module, and uses the available inventory quantity, reserved quantity, inventory version, warehouse status and inventory update time in the inventory status record as input fields for generating recommendation result records.

[0088] The recommendation ranking module 06, connected to the inventory status acquisition module, is used to generate recommendation result records and recommendation result versions based on the candidate matching records, the substitution relationship scores, and the inventory status records. Specifically, the recommendation ranking module receives candidate matching records and substitution relationship scores from the candidate matching module, and receives inventory status records from the inventory status acquisition module; it reads the component number, similarity, and matching field from the candidate matching records, the substitution direction, substitution conditions, and substitution priority from the substitution relationship scores, and the available inventory quantity, reserved quantity, inventory version, warehouse status, and inventory update time from the inventory status records. The recommendation ranking module generates an inventory adequacy level based on the available inventory quantity, reserved quantity, warehouse status, and inventory update time; it generates a comprehensive recommendation index based on the candidate matching records, substitution relationship scores, and inventory adequacy level, and writes the comprehensive recommendation index into the recommendation result record. The recommended result record includes the query entity record identifier, component number, candidate source, similarity, substitution relationship score, inventory adequacy, comprehensive recommendation index, inventory version, and recommendation status. The recommendation sorting module generates a recommended result version when generating the recommended result record and saves the recommended result version in correspondence with the query entity record, the candidate matching record, the candidate substitution part, and the inventory status record. The recommendation sorting module provides the recommended result record and the recommended result version to the inventory reservation module; the inventory version, component number, and recommendation status in the recommended result version serve as input fields for the inventory reservation module to perform inventory consistency verification.

[0089] The inventory reservation module 07, connected to the recommendation and sorting module, is used to receive confirmation instructions, perform inventory consistency verification based on the recommended result version, the inventory version in the inventory status record, and the confirmed quantity, and generate an inventory reservation record or candidate rearrangement result. Specifically, the inventory reservation module receives the recommended result record and recommended result version from the recommendation and sorting module, and receives confirmation instructions; the confirmation instructions include the confirmed part number, confirmed quantity, recommended result version, and confirmation time. The inventory reservation module reads the recommended result record corresponding to the recommended result version and reads the inventory version in the recommended result record; simultaneously, it reads the current inventory information based on the confirmed part number and obtains the current inventory version and the current available inventory quantity. The inventory reservation module compares the inventory version corresponding to the recommended result version with the current inventory version and determines whether the current available inventory quantity is not less than the confirmed quantity; when the comparison results are consistent and the current available inventory quantity is not less than the confirmed quantity, an inventory reservation record is generated. The inventory reservation record includes a reservation record number, confirmed part number, confirmed quantity, warehouse identifier, inventory version, recommended result version, reservation status, and reservation time, and the inventory reservation record is written into the inventory information. When the comparison results are inconsistent or the current available inventory is less than the confirmed quantity, the inventory reservation module generates an inventory verification failure record; it generates candidate re-ranking results based on the updated inventory status record and returns the candidate re-ranking results to the corresponding recommendation result record of the recommendation sorting module. The inventory reservation module stores the correspondence between the confirmation instruction, the inventory reservation record, the inventory verification failure record, and the candidate re-ranking results.

Claims

1. A method for intelligent selection of engine components based on multi-dimensional feature vectors, characterized in that, include: S100: Receives component files, compatible model records, substitute part mapping relationships, and industry dictionaries; performs standardized field mapping on component names, physical parameters, compatible models, component categories, and substitute part identifiers; and generates component feature vector records and index records. S200: Receive query text, perform word segmentation, entity recognition and synonym normalization based on the industry dictionary, generate query entity records, and generate candidate filtering conditions from model entity and part category entity; S300. Based on the candidate filtering conditions and the index records, determine the candidate set, generate candidate matching records based on the query entity records and the component feature vector records, and generate alternative candidate records and alternative relationship scores based on the alternative mapping relationship. S400: Based on the candidate matching record and the alternative candidate record, read the inventory information and generate an inventory status record containing available inventory quantity, reserved quantity and inventory version; based on the candidate matching record, the alternative relationship score and the inventory status record, generate a recommendation result record and a recommendation result version. S500: Receive confirmation instruction, perform inventory consistency verification based on the recommended result version, the inventory version, and the confirmed quantity, generate inventory reservation record when the verification passes, and generate candidate rearrangement result when the verification fails.

2. The method according to claim 1, characterized in that, In S100, the standardized field mapping includes: reading the field name, parameter value, unit, compatible model and part category from the component file; mapping the field name to a standardized field; generating parameter normalization records for the parameter value and the unit; and writing the parameter normalization records into the component feature vector record.

3. The method according to claim 1, characterized in that, The component feature vector record includes physical parameter record, compatible model record, part category record, text alias record, and substitute part identification record; the physical parameter record includes size parameter, material parameter, and interface parameter.

4. The method according to claim 1, characterized in that, The industry dictionary includes records of brand aliases, model abbreviations, English abbreviations, common part names, OE numbers, and supplier codes; the synonym normalization includes mapping the brand aliases, model abbreviations, English abbreviations, common part names, OE numbers, and supplier codes in the query text to standardized field records in the query entity records.

5. The method according to claim 1, characterized in that, The query entity records include model entity, part category entity, parameter entity, quantity entity, and special constraint entity; when the model entity or the part category entity is missing, a missing field record is generated; when the query text corresponds to multiple model entities, an ambiguous entity record is generated.

6. The method according to claim 1, characterized in that, In S300, determining the candidate set includes: matching the index records with the model entity and part category entity in the candidate filtering conditions to obtain the candidate set; for the parts in the candidate set, calculating the similarity between the query entity record and the part feature vector record, and writing the similarity into the candidate matching record.

7. The method according to claim 1, characterized in that, In S300, generating alternative candidate records includes: when the number of candidate matching records is less than a preset threshold, the inventory of candidate parts is insufficient, or the highest similarity is lower than the similarity threshold, reading the alternative mapping relationship; and generating the alternative candidate records and the alternative relationship score based on the substitution direction, substitution conditions, and substitution priority.

8. The method according to claim 1, characterized in that, In S400, the inventory status record includes warehouse status and inventory update time; inventory adequacy is generated based on the available inventory, the reserved inventory, the warehouse status, and the inventory update time; A comprehensive recommendation index is generated based on the candidate matching records, the substitution relationship score, and the inventory adequacy, and the comprehensive recommendation index is written into the recommendation result record.

9. The method according to claim 1, characterized in that, In S500, the inventory consistency verification includes: comparing the inventory version corresponding to the recommended result version with the current inventory version, and determining that the current available inventory quantity is not less than the confirmed quantity; when the verification passes, generating the inventory reservation record and writing back the inventory information; when the verification fails, generating an inventory verification failure record, and generating the candidate rearrangement result based on the updated inventory status record.

10. An intelligent selection system for engine components based on multi-dimensional feature vectors, characterized in that, include: The module includes a parts file normalization module, a feature vector construction module, a query parsing module, a candidate matching module, an inventory status acquisition module, a recommendation and sorting module, and an inventory reservation module. The system is used to implement the method described in any one of claims 1-9.

Images