Ai-based intelligent bom automated generation system and method

Abstract

An AI-based intelligent BOM automatic generation system according to one embodiment of the present invention comprises: a data collection unit that collects part information, process information, and equipment information from a plurality of data sources; an attribute management unit that preprocesses the collected part information, process information, and equipment information, classifies them by category, and predicts attribute information; a BOM generation unit that generates and provides a BOM based on the predicted attribute information; a BOM verification unit that performs error verification on the generated BOM; and an intelligent autonomous control unit that autonomously controls the operations of the data collection unit, the attribute management unit, the BOM generation unit, and the BOM verification unit.

Description

Technology Field

[0001] The present invention relates to an AI-based intelligent automatic BOM generation system and method, and more specifically, to a system and method for automatically generating a Bill of Materials (BOM) by classifying part, process, and equipment information collected from multiple data sources by part category and predicting attribute information. Background Technology

[0003] A Bill of Materials (BOM) is a document that hierarchically lists all parts, raw materials, quantities, and assembly relationships required to produce a finished product in manufacturing. It plays a pivotal role throughout the entire manufacturing process, including production planning, material purchasing, cost estimation, and quality control. The BOM systematically represents the structure of a product, ranging from the top-level finished product to intermediate assemblies and individual components, and includes the code, name, quantity, and supplier information for each part.

[0004] Traditionally, Bills of Materials (BOMs) have been created manually by design engineers referencing CAD drawings, design documents, and BOMs from similar past products. Engineers complete the BOM document by analyzing product specifications, selecting necessary components, and determining the quantities and assembly relationships of each part. However, this manual method poses a problem in that it requires significant time and manpower for BOM creation, and the workload increases exponentially as product complexity grows. For example, for electronic or mechanical products composed of hundreds or more parts, creating the BOM alone can take anywhere from several days to several weeks.

[0005] Furthermore, manual processes have fundamental limitations, such as frequent human errors including mis-entering part codes, quantity discrepancies, and missing parts. Incorrect part codes lead to the ordering of incorrect components, causing production delays; quantity discrepancies result in material shortages or excess inventory; and missing parts, if discovered during the assembly stage, can trigger urgent orders and production stoppages. These errors have a severe impact on the manufacturing floor, leading to reduced production efficiency, delayed deliveries, and increased costs.

[0006] In particular, creating an optimal Bill of Materials (BOM) that comprehensively considers attribute information such as component cost, lead time, and defect rate is a very difficult task. When multiple substitute components performing the same function exist, the optimal combination must be selected by comparing the cost, lead time, quality, and supply stability of each component. For example, if component A has a lower cost but a longer lead time, and component B has a higher cost but a shorter lead time, the appropriate component must be selected by simultaneously considering the production schedule and cost targets. However, it is practically impossible to select the optimal combination by individually comparing the attributes of each component in a BOM composed of dozens to hundreds of parts.

[0007] To address these issues, research on automatic BOM generation technology utilizing artificial intelligence has recently been underway. Some technologies search historical BOM databases to reference BOMs of similar products, while others use rule-based engines to define part selection rules and generate BOMs accordingly.

[0008] However, conventional technology is limited to simply searching for historical data or applying predefined rules, and thus has limitations in effectively processing complex product structures and various component attributes.

[0009] Therefore, there is a need to develop an improved system capable of efficiently utilizing various data from the manufacturing site and systematically analyzing part attribute information to automatically generate an optimal BOM.

[0010] Various methods are known as related prior art, including rule-based BOM generation, similar BOM search, and PDM / PLM-based Product Structure (BOM) management. However, conventional technologies rely primarily on predefined rules or similar BOM searches, and thus have limitations in providing integrated solutions for combining unstructured documents and historical data from heterogeneous data sources, simultaneously predicting and calculating the reliability of part attributes (cost, lead time, defect rate, etc.), controlling automatic re-collection and regeneration pipelines based on verification results, and recommending BOM candidates that reflect specification constraints and multi-objective optimization. The problem to be solved

[0012] The present invention has been devised to solve the aforementioned problems, and aims to provide an AI-based intelligent automatic BOM generation system and method capable of reducing BOM creation time and preventing human error by automatically collecting part information from multiple data sources, preprocessing the collected part information to classify it by part category, predicting attribute information for each part to automatically generate a BOM, and performing error verification on the generated BOM.

[0014] The problems that the present invention aims to solve are not limited to the problem(s) mentioned above, and other unmentioned problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0016] An AI-based intelligent BOM automatic generation system according to one embodiment of the present invention may include: a data collection unit that collects part information, process information, and equipment information from a plurality of data sources; an attribute management unit that preprocesses the collected part information, process information, and equipment information, classifies them by category, and predicts attribute information; a BOM generation unit that generates and provides a BOM based on the predicted attribute information; a BOM verification unit that performs error verification on the generated BOM; and an intelligent autonomous control unit that autonomously controls the operations of the data collection unit, the attribute management unit, the BOM generation unit, and the BOM verification unit.

[0017] The above data collection unit extracts part information by processing fax technical documents with OCR (Optical Character Recognition) and extracts part information by processing emails with NLP (Natural Language Processing), and can query quotation history, order history, procurement history, after-sales service history, and production history from the MES (Manufacturing Execution System).

[0018] The above BOM verification unit performs error verification on the generated BOM, including the validity of the part code, the appropriateness of the quantity, and the existence of the supplier, and if an error is found, it may correct the BOM or request confirmation from the user terminal.

[0019] The above attribute management unit collects training data including part codes, quantities, and supplier information from a technical document database and MES, classifies multiple parts into multiple groups according to their functional roles based on pre-stored attribute information for each part, calculates a representative attribute value for each group by weighted averaging the attribute information of the parts belonging to each group, trains a multiple output regression model using the representative attribute value for each group as an input variable and total cost, lead time, and expected defect rate as output variables, and when a new BOM creation request is received, inputs it into the trained multiple output regression model to calculate multiple attribute prediction values ​​and prediction reliability.

[0020] The above attribute management unit classifies the plurality of parts into at least one group among a structural parts group, an electrical parts group, and a consumables group, and calculates a representative cost, representative lead time, and representative defect rate for each group by weighting the cost, lead time, and defect rate of the parts belonging to each group by the usage ratio of each part.

[0021] The above attribute management unit can perform data preprocessing on the above part information, including part code normalization, unit unification, and duplicate removal.

[0022] The above attribute management unit can supplement the result by querying the attribute values ​​of identical part codes from the technical document database and MES when missing attribute values ​​are found in part data collected from multiple external systems, and if identical part codes do not exist, it estimates the result by averaging the attribute values ​​of similar parts with the same functional role, and when selecting the similar parts, it can supplement the result by weighting the average of the attribute values ​​of a preset number of parts with the highest similarity scores, considering the part category, specification range, and manufacturer match.

[0023] The above system further includes a specification verification unit that refers to a specification database to query usage restriction information and prohibited combination information for each part and generates constraints; and an optimization unit that generates a set of Pareto optimal solutions by executing a multi-objective optimization algorithm that simultaneously optimizes two or more goals based on the predicted attribute information and the generated constraints, wherein the BOM generation unit visualizes and provides trade-off relationships for each BOM candidate included in the set of Pareto optimal solutions and can recommend an optimal BOM according to the user's preference.

[0024] An AI-based intelligent BOM automatic generation method according to another embodiment of the present invention may include: a data collection unit collecting part information, process information, and equipment information from a plurality of data sources; an attribute management unit preprocessing the collected part information, process information, and equipment information, classifying them by category, and predicting attribute information; a BOM generation unit generating a BOM based on the predicted attribute information; and a BOM verification unit performing error verification on the generated BOM.

[0026] Specific details of other embodiments are included in the detailed description and attached drawings. Effects of the invention

[0028] According to the present invention as described above, by automatically collecting part information from multiple data sources, preprocessing the collected part information to classify it by part category, predicting attribute information for each part to automatically generate a BOM, and performing error verification and correction on the generated BOM, the time required for BOM creation and human error can be significantly reduced.

[0029] In addition, by calculating BOM candidates through quantitative consideration of multiple attributes such as part cost, lead time, and defect rate, and reflecting constraints (combinations restricted or prohibited for use, etc.), it is possible to present optimal or near-optimal BOMs suitable for the manufacturing site based on consistent standards.

[0030] In addition, prediction performance can be continuously improved by feeding back user verification and correction results as training data.

[0031] Since the present invention enables the rapid and accurate generation and verification of BOMs used throughout the entire process of design, purchasing, production, quality, and after-sales service in manufacturing, it can be applied to the generation and change management (ECO) of Product BOMs (EBOM), Production BOMs (MBOM), Order BOMs, and Purchase BOMs in various industrial fields such as electronics, machinery, chemicals, automobiles, shipbuilding, and plant engineering, and can be industrially utilized by linking with existing information systems such as MES, ERP, and PLM. Brief explanation of the drawing

[0033] FIG. 1 is a network configuration diagram of an AI-based intelligent BOM automatic generation system according to one embodiment of the present invention. FIG. 2 is an internal configuration diagram of a server according to one embodiment of the present invention. FIG. 3 is an overall flowchart of an AI-based intelligent BOM automatic generation method according to one embodiment of the present invention. FIG. 4 is a flowchart illustrating a component grouping and multiple output regression detail process according to one embodiment of the present invention. FIG. 5 is a flowchart illustrating a detailed process for specification verification and multi-purpose optimization according to one embodiment of the present invention. FIG. 6 is a diagram showing a BOM hierarchy structure according to an embodiment of the present invention and process and equipment information linked thereto. FIG. 7 is a diagram illustrating a result value supplementation process according to one embodiment of the present invention. Specific details for implementing the invention

[0034] The advantages and / or features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Throughout the specification, like reference numerals refer to like components.

[0035] In addition, the preferred embodiments of the present invention described below focus on explaining the functional configurations that must be additionally provided for the present invention, while omitting as much as possible the system functional configurations that are already provided in each system functional configuration or are ordinarily provided in the technical field to which the present invention belongs, in order to efficiently explain the technical components constituting the present invention. If a person skilled in the art to which the present invention belongs is able to easily understand the functions of the components that are conventionally used among the functional configurations that are omitted and not illustrated below, and can also clearly understand the relationship between the components omitted as above and the components added for the present invention.

[0036] Furthermore, in the following description, terms such as "transmission," "communication," "transmission," "reception," and other terms of similar meaning regarding signals or information include not only the direct transfer of signals or information from one component to another but also the transfer through other components. In particular, "transmitting" or "transmitting" a signal or information to a component refers to the final destination of the signal or information and does not imply a direct destination. The same applies to the "reception" of signals or information.

[0038] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0039] FIG. 1 is a network configuration diagram of an AI-based intelligent BOM automatic generation system according to one embodiment of the present invention.

[0040] Referring to FIG. 1, an AI-based intelligent BOM automatic generation system according to one embodiment of the present invention may include a user terminal (110), a communication network (120), a server (130), a technical document database (140), and an MES (150).

[0041] The user terminal (110) performs the role of transmitting a request for BOM generation to the server (130) and receiving the generated BOM from the server (130). The user terminal (110) can be implemented in the form of a smartphone, tablet PC, desktop computer, laptop, etc., and provides an interface that allows the user to input product specifications, target cost, target procurement period, etc.

[0042] As an optional embodiment, the user terminal (110) may be a terminal used by a design engineer or purchasing manager in the manufacturing field. If the user wishes to create a BOM for a new product through the user terminal (110), the user may input product functional requirements, budget constraints, delivery requirements, etc.

[0043] As an optional embodiment, the user terminal (110) may provide a function to input a target attribute range. For example, a target range may be set for each attribute, such as “total cost: 30,000 won or less”, “procurement period: within 14 days”, and “expected defect rate: 0.5% or less”.

[0044] The communication network (120) serves to relay data communication between the user terminal (110) and the server (130). The communication network (120) can be implemented as various types of networks, such as wired internet, wireless internet, and mobile communication networks.

[0045] The server (130) is a core component that receives a BOM creation request and target conditions from a user terminal (110), and autonomously performs processing such as data collection, attribute prediction, BOM creation, and error verification through an intelligent autonomous control unit to create an optimal BOM and provide it to the user terminal (110). In the present invention, the intelligent autonomous control unit refers to an intelligent software system that independently makes decisions based on learned knowledge and algorithms without user intervention.

[0046] As an optional embodiment, the server (130) may be implemented as a cloud server and may receive and process requests simultaneously from multiple user terminals. The server (130) may be connected to a technical document database (140), an MES (150), etc., to collect and manage part information.

[0047] The technical document database (140) is a database in which product item codes, part combination information, process conditions, etc. are stored, and provides an API so that the data collection unit of the server (130) can query the necessary information.

[0048] The MES (150) is a Manufacturing Execution System that manages data such as defect history, process conditions, and batch defect rates that occur during the production process. In this embodiment, the MES (150) may include an After Service (AS) management function and may manage AS requests, repair history, and parts replacement history that occur after product shipment.

[0050] FIG. 2 is an internal configuration diagram of a server according to one embodiment of the present invention.

[0051] Referring to FIG. 2, the server (130) may include a data collection unit (210), an attribute management unit (220), a BOM generation unit (230), a BOM verification unit (240), and an intelligent autonomous control unit (250).

[0052] The data collection unit (210) performs the function of collecting part information from multiple data sources.

[0053] As an optional embodiment, the data collection unit (210) can extract part information by processing a fax technical document using OCR. The technical document received via fax may include part specifications, drawings, quantity information, etc., and the document in image form can be converted into text using OCR technology, and structured information such as part code, part name, and quantity can be extracted by utilizing regular expressions or natural language processing techniques.

[0054] As an optional embodiment, the data collection unit (210) can extract part information by processing emails using NLP. It can analyze emails such as quotations and delivery notices received from suppliers using natural language processing technology and extract information such as part codes, unit prices, lead times, and minimum order quantities.

[0055] As an optional embodiment, the data collection unit (210) can retrieve quotation history, order history, procurement history, after-sales service history, and production history from the MES (150). The MES (150) records the actual cost, lead time, and defect occurrence status of past products by part, and this historical data can be used as training data for predicting part attributes.

[0056] As an optional embodiment, the data collection unit (210) may collect formulation recipe information. For example, it may extract raw material list, target input amount, tolerance (±g / %), and substitute raw material rules from the technical document database (140), collect raw material Lot information, expiration date, and COA (certificate of analysis) data from the MES (150), and collect information on prohibited combinations (non-mixing) rules and pre-mixing (premix) rules.

[0057] As an optional embodiment, the data collection unit (210) may collect process recipe information. For example, it may collect management parameters for each process step (mixing → heating → molding → foaming → cooling, etc.) from the MES (150). The management parameters include temperature zones, line speed, tension, pressure, viscosity, thickness, drying time, foaming conditions, etc., and may also collect information on target values, upper / lower control limits, and upper / lower specification limits for each parameter.

[0058] As an optional embodiment, the data collection unit (210) may collect equipment recipe information. For example, it may collect setting values ​​such as setpoints, PID values, and alarm values ​​for each equipment and zone from a PLC system, and if the equipment specifications differ by line even for the same product, it may separately collect parameter sets for each line.

[0059] As an optional embodiment, the data collection unit (210) may collect parts list and assembly structure information from a design drawing system, collect customer-customized specification information from an order system, collect parts unit price and supplier information from a purchasing system, and collect actual usage information from a material release system. Process performance, equipment operation history, parts defect history, etc. generated during the production process may also be included in the collection targets.

[0060] The attribute management unit (220) performs the function of preprocessing collected part information, classifying it by part category, and predicting attribute information for each part.

[0061] Attribute information refers to itemized data fields extracted from unstructured documents or historical data, such as part name, specification, quantity, unit, supplier, and process / equipment association information.

[0062] As an optional embodiment, the attribute management unit (220) may perform data preprocessing on the collected part information, including part code normalization, unit unification, and duplicate removal. Part code normalization is a process of converting codes referring to the same part into a standard format when they are expressed differently in different data sources, and unit unification is a process of converting costs expressed in various units, such as "won / piece" or "dollar / piece," into a unified unit (e.g., won / piece).

[0063] As an optional embodiment, the attribute management unit (220) may collect learning data including part codes, quantities, and supplier information from the technical document database (140) and the MES (150). For example, as illustrated in FIGS. 6 and 7, data such as part codes, quantities, suppliers, and actual costs, lead times, and defect rates used for each past product may be collected.

[0064] As an optional embodiment, the training data may include part code, quantity, and supplier information (optionally may include process code and equipment code information).

[0065] As an optional embodiment, the attribute management unit (220) may retrieve pre-stored attribute information for each part. The attribute information may include the cost of the part, typical lead time, past defect rate, manufacturer, specifications, etc., and this information may be stored and managed in a part master database.

[0066] As an optional embodiment, the attribute management unit (220) may classify component information into multiple groups according to functional roles. For example, in the case of an electronic product BOM, it may be classified into a structural component group (case, screw, etc.), an electrical component group (MCU, sensor, battery, etc.), and a consumable component group (silicon, wire, etc.), and the attribute information of the components belonging to each group may be weighted averaged to calculate a representative cost, representative lead time, and representative defect rate for each group.

[0067] As an optional embodiment, the attribute management unit (220) learns the collected mixing recipe, process recipe, and equipment recipe information and can automatically predict which parts (raw materials) to use in what ratio, which process conditions (temperature, pressure, speed, etc.) to produce, and which equipment to produce with which setting values ​​when a request to create a new product or semi-product is received.

[0068] As an optional embodiment, the attribute management unit (220) learns combinations of blending recipes, process recipes, and equipment recipes that have been successfully produced for each item from past production history, and when a request to create a BOM for a similar item is received, it can recommend an optimal recipe combination based on the learned combinations. For example, in the case of a chemical product, if the defect rate was low when process conditions of temperature 75°C and pressure 2 bar were applied to a blend of raw material A 30% + raw material B 50%, this combination can be recommended as a priority.

[0069] As an optional embodiment, the attribute management unit (220) may train a multiple output regression model with a representative attribute value for each group as an input variable and total cost, procurement time, and expected defect rate as output variables. The multiple output regression model can simultaneously predict multiple attributes such as total cost, procurement time, and expected defect rate, and can be trained by considering the correlation between each attribute.

[0070] As an optional embodiment, the attribute management unit (220) may input a new BOM generation request into a learned multiple-output regression model to calculate multiple attribute prediction values ​​and prediction confidence. The prediction confidence is a numerical value representing the uncertainty or confidence level for the attribute prediction values, and may be calculated based on at least one of the output probability, error variance, and confidence interval of the prediction model.

[0071] As an optional embodiment, if a missing attribute value is found in the part data collected from a plurality of external systems, the attribute management unit (220) can supplement it by querying the attribute value of the same part code from the technical document database (140) and the MES (150). If the same part code does not exist, it can be estimated by the average of the attribute values ​​of similar parts with the same functional role. When selecting similar parts, the result can be supplemented by weighting the average of the attribute values ​​of a preset number (e.g., top 3) of parts with the highest similarity scores, considering the part category, specification range, and whether the manufacturer matches.

[0072] The BOM generation unit (230) performs the function of generating and providing a BOM based on predicted attribute information.

[0073] As an optional embodiment, the BOM generation unit (230) may generate a BOM document based on attribute information such as the cost, lead time, and expected defect rate of each part predicted by the attribute management unit (220). The BOM document may be hierarchically organized from the top-level product code to intermediate assemblies and individual parts, and for each item, information such as part code, part name, quantity, supplier, estimated cost, and estimated lead time may be included.

[0074] As an optional embodiment, the BOM generation unit (230) may generate a product BOM, production BOM, order BOM, purchase BOM, etc., depending on the application. The product BOM is a standard BOM based on design criteria, the production BOM is a BOM that reflects actual production line conditions, the order BOM is a BOM that reflects customer-customized specifications, and the purchase BOM is a BOM centered on a list of parts required for purchase orders. Each BOM type may optionally include necessary information among a mixing recipe, a process recipe, and an equipment recipe.

[0075] The BOM verification unit (240) performs the function of performing error verification on the generated BOM.

[0076] As an optional embodiment, the BOM verification unit (240) may perform error verification on the generated BOM, including part code validity, quantity appropriateness, and supplier existence. Part code validity verification is a process of verifying whether the part code listed in the BOM actually exists in the part master database, and quantity appropriateness verification is a process of verifying whether the quantity of each part is within a reasonable range (e.g., not negative, not excessively large). Supplier existence verification is a process of verifying whether a designated supplier can actually supply the part.

[0077] As an optional embodiment, the BOM verification unit (240) may automatically correct the BOM or request confirmation from the user terminal (110) if an error is found. For example, if a part code is determined to be a typo, it may be automatically corrected to a similar part code, and if the quantity is excessively large, confirmation may be requested from the user.

[0078] The intelligent autonomous control unit (250) performs the function of autonomously controlling the operation of the data collection unit, attribute management unit, BOM generation unit and BOM verification unit.

[0079] As an optional embodiment, the intelligent autonomous control unit (250) can perform a pipeline control function. Pipeline control refers to a function that automatically coordinates and controls the execution order, conditions, and data flow between multiple modules. The intelligent autonomous control unit (250) analyzes a BOM creation request and controls each step of data collection, attribute extraction, BOM creation, and BOM verification to be executed according to an appropriate order and conditions.

[0080] For example, when a user inputs a request such as "Generate a BOM for new product A," the intelligent autonomous control unit (250) can automatically pipeline control a series of processes including (1) instructing a data collection unit to collect necessary data from design documents, technical specifications, and past similar product BOMs, (2) controlling an attribute management unit to extract part attributes and relationship information from the collected data, (3) coordinating a BOM generation unit to generate a BOM tree structure based on the extracted attribute information, (4) instructing a BOM verification unit to verify the consistency and rule compliance of the generated BOM, and (5) providing the final BOM to the user terminal (110).

[0081] As an optional embodiment, the intelligent autonomous control unit (250) may dynamically determine the next step based on the execution results of each module. For example, if the BOM verification unit discovers an error in the omission of essential parts during the verification process, the intelligent autonomous control unit (250) may instruct the data collection unit to supplementally collect missing part information from additional data sources (e.g., supplier catalog, engineering change history), control the attribute management unit to re-extract part attributes including the supplemented data, and coordinate the BOM generation unit to regenerate the BOM with the supplemented information.

[0082] As an optional embodiment, the intelligent autonomous control unit (250) may autonomously determine an appropriate response measure when an exception occurs. For example, if the attribute information of a specific part is incomplete, the intelligent autonomous control unit (250) may select the most suitable response measure according to the situation, among (1) replacing it with the attribute information of a similar part, (2) requesting additional information from the user, and (3) displaying the part as an optional item.

[0083] As an optional embodiment, the intelligent autonomous control unit (250) can optimize the performance of the entire system by learning repetitive usage patterns. For example, if it learns a pattern in which the same combination of parts is always used in a specific product family, the intelligent autonomous control unit (250) can shorten the overall processing time by calling the data collection unit in advance to collect relevant data in advance when a request for BOM generation for that product family is made.

[0084] As an optional embodiment, the intelligent autonomous control unit (250) may be implemented based on a Large Language Model (LLM) and can understand the intent of a user request and establish an appropriate module execution plan. The intelligent autonomous control unit (250) may interpret intermediate results returned from each module, determine the execution conditions for the next step, and convert and provide the final BOM into a form that is easy for the user to understand.

[0086] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0087] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0089] FIG. 3 is an overall flowchart of an AI-based intelligent BOM automatic generation method according to one embodiment of the present invention.

[0090] Referring to FIG. 3, in one embodiment, the data collection unit (210) can collect part information from a plurality of data sources (S310).

[0091] As an optional embodiment, the data collection unit (210) can extract part information by processing a fax technical document using OCR. The technical document received via fax may include part specifications, drawings, quantity information, etc., and the document in image form can be converted into text using OCR technology, and structured information such as part code, part name, and quantity can be extracted by utilizing regular expressions or natural language processing techniques.

[0092] As an optional embodiment, the data collection unit (210) can extract part information by processing emails using NLP. It can analyze emails such as quotations and delivery notices received from suppliers using natural language processing technology and extract information such as part codes, unit prices, lead times, and minimum order quantities.

[0093] As an optional embodiment, the data collection unit (210) can extract item codes, part combination information, and process conditions from the technical document database (140). For example, for item code "P001", information on the composition ratio of "Part A 30%, Part B 50%, Part C 20%" and process information on "assembly process, inspection process" can be extracted.

[0094] As an optional embodiment, the data collection unit (210) can retrieve quotation history, order history, procurement history, after-sales service history, and production history from the MES (150) using the extracted item code. For example, for item code "P001", actual data such as "2025-01-10, cost: 25,000 won, lead time: 10 days, defect rate: 0.3%" can be retrieved.

[0095] As an optional embodiment, the data collection unit (210) may additionally collect after-service (AS) management data from the MES (150). The after-service management data may include the history of after-service requests, repair details, parts replacement history, and after-service processing time that occurred after product shipment. For example, data such as "Item P001, after-service occurrence date: 2025-01-25, repair details: battery replacement, processing time: 2 hours" may be collected.

[0096] The attribute management unit (220) can preprocess the collected part information, classify it by part category, and predict attribute information for each part (S320).

[0097] As an optional embodiment, the attribute management unit (220) may first perform data preprocessing on the collected part information, including part code normalization, unit unification, and duplicate removal. Part code normalization is a process of converting codes referring to the same part from different data sources (e.g., "MCU-001A", "MCU_001A", "MCU001A") into a standard format (e.g., "MCU-001A"). Unit unification is a process of converting costs from various units such as "won / piece", "dollar / piece", or "yen / piece" into a unified unit (e.g., won / piece), and duplicate removal is a process of integrating the same part information into one when it is collected redundantly from multiple data sources.

[0098] As an optional embodiment, the attribute management unit (220) may classify preprocessed part information into structural part groups, electrical part groups, consumable part groups, etc., according to their functional roles. For example, in the case of a Smart Tumbler product, the Stainless Steel Body may be classified as a structural part group, the Temperature Sensor, MCU, Bluetooth Module, LED Display Panel, and Battery as electrical part groups, and the Silicon Lid as a consumable part group.

[0099] As an optional embodiment, the attribute management unit (220) can calculate a representative cost, a representative procurement period, and a representative defect rate for each group by weighting the attribute information of the parts belonging to each group by the usage ratio of each part. For example, if an MCU (cost 5,000 won, quantity 1) and a sensor (cost 3,000 won, quantity 1) belong to the automotive parts group, the representative cost of the automotive parts group can be calculated as (5,000×1 + 3,000×1) / (1+1) = 4,000 won.

[0100] As an optional embodiment, the attribute management unit (220) can input component configuration information and representative attribute values ​​by group into a pre-trained multiple-output regression model to calculate attribute prediction values ​​such as total cost, procurement time, and expected defect rate. The multiple-output regression model can predict multiple attributes simultaneously and can reflect the correlation between each attribute (e.g., the tendency for the defect rate to decrease as the cost increases) in the learning process. Along with the prediction values, the prediction reliability is calculated so that the user can judge the reliability of the prediction results.

[0101] The BOM generation unit (230) can generate a BOM based on predicted attribute information (S330).

[0102] As an optional embodiment, the BOM generation unit (230) may generate a BOM document based on attribute information such as the cost, lead time, and expected defect rate of each part predicted by the attribute management unit (220). The BOM document may be hierarchically organized from the top-level product code to intermediate assemblies and individual parts, and for each item, information such as part code, part name, quantity, supplier, estimated cost, and estimated lead time may be included.

[0103] As an optional embodiment, the BOM generation unit (230) may represent the BOM in a hierarchical structure as shown in FIG. 6. At level 01 (top level), the finished product "Smart Tumbler Model X" is located, at level 02, the main assemblies "Stainless Steel Body Assembly", "Electronics Assembly", "Lid Assembly", and "Hardware" are located, and at level 03, the individual parts constituting each assembly are located.

[0104] In this specification, "BOM" may be used to encompass not only a manufacturing BOM (MBOM) as a list of parts, but also a process BOM containing process step information linked to the manufacturing BOM and an equipment BOM containing equipment resource information corresponding to the process.

[0105] The BOM verification unit (240) can perform error verification on the generated BOM (S340).

[0106] As an optional embodiment, the BOM verification unit (240) can verify the validity of the part code for the generated BOM. Part code validity verification is a process of confirming whether the part code listed in the BOM actually exists in the part master database. For example, if a part code named "MCU-999Z" is listed in the BOM but the corresponding code does not exist in the part master database, it may be determined as an error.

[0107] As an optional embodiment, the BOM verification unit (240) can verify the quantity appropriateness. Quantity appropriateness verification is a process of checking whether the quantity of each part is within a reasonable range. For example, if the quantity of screws is negative or the quantity of batteries is excessively large, such as more than 100, it may be determined as an error.

[0108] As an optional embodiment, the BOM verification unit (240) may verify the existence of a supplier. Verifying the existence of a supplier is a process of confirming whether a designated supplier can actually supply the corresponding part. For example, if "Supplier A" is designated for a specific part, but it is confirmed in the supplier database that Supplier A does not handle the part, it may be determined as an error.

[0109] As an optional embodiment, the BOM verification unit (240) may automatically correct the BOM when an error is found or request verification from the user terminal (110). As an example of automatic correction, if a part code is determined to be a typo (e.g., incorrectly written as "MCU-001A" as "MCU-001A"), it may be automatically corrected to the part code with the highest similarity (e.g., "MCU-001A"). As an example of a user verification request, if the quantity is excessively large, a message such as "The battery quantity is set to 100. Please check if the quantity is correct" may be displayed.

[0110] As an optional embodiment, the overall operation flow of the intelligent autonomous control unit is as follows. When a user inputs product specifications and target constraints (e.g., total cost of 30,000 won or less, procurement period within 14 days), the intelligent autonomous control unit autonomously searches for BOM information of similar products from the technical document database (140) and MES (150) (S310), preprocesses the collected data, groups parts, and predicts attributes (S320). Subsequently, it generates a BOM based on the predicted attribute information (S330) and performs error verification on the generated BOM (S340). The user can review the presented BOM and modify it if necessary, and can continuously improve the attribute prediction model by providing feedback on cost, procurement period, and defect rate data generated after actual production. In this series of processes, the user simply sets the goal, and the intelligent autonomous control unit autonomously performs the entire process of data collection, preprocessing, prediction, BOM generation, and verification.

[0112] FIG. 4 is a flowchart illustrating a component grouping and multiple output regression detail process according to one embodiment of the present invention.

[0113] Referring to FIG. 4, in one embodiment, the attribute management unit (220) can collect learning data from the technical document database (140) and the MES (150) (S410).

[0114] As an optional embodiment, the training data may include part codes, quantities, and supplier information. As illustrated in FIG. 6, actual data such as part codes used for each past product (e.g., MCU-001A, Sensor-002B), quantities (e.g., 1 MCU, 1 Sensor), suppliers (e.g., Supplier A, Supplier B), and actual costs incurred (e.g., 25,000 won), lead times (e.g., 10 days), and defect rates (e.g., 0.3%) may be collected.

[0115] As an optional embodiment, the training data may include hundreds to thousands of historical BOM records, and the more data there is, the better the prediction accuracy of the machine learning model may be. For example, in the field of electronics, at least 500 historical BOM records may be used as training data.

[0116] The attribute management unit (220) can query and group part attributes (S420).

[0117] As an optional embodiment, the attribute management unit (220) can query attribute information that is pre-stored in the component master database for each component. The attribute information may include cost, lead time, past defect rate, manufacturer, specifications, weight, size, power consumption, etc. For example, for MCU-001A, information such as "cost: 5,000 won, lead time: 7 days, past defect rate: 0.2%, manufacturer: ABC Electronics, specifications: 5V 32bit, weight: 5g" can be queried.

[0118] As an optional embodiment, the attribute management unit (220) may classify a plurality of parts into a plurality of groups according to their functional roles. The functional role refers to the main function that a part performs within a product, and, for example, in the case of an electronic product, it may be classified into parts that provide structural support (structural parts group), parts that perform electrical functions (electrical parts group), consumable materials (consumables group), etc.

[0119] As an optional embodiment, for a Smart Tumbler product, the structural parts group may include a Stainless Steel Body and a Sensor Connection Wire, the electrical parts group may include a Temperature Sensor, a Micro Controller Unit (MCU), a Bluetooth Module, an LED Display Panel, and a Battery, the consumables group may include a Silicone Lid, and the hardware group may include a Screw.

[0120] As an optional embodiment, the attribute management unit (220) may calculate a representative attribute value for each group by weighting the attribute information of the parts belonging to each group. When weighting the average, the usage ratio (quantity or weight ratio) of each part may be used as a weight. For example, if an MCU (cost 5,000 won, quantity 1, procurement period 7 days, defect rate 0.2%) and a sensor (cost 3,000 won, quantity 1, procurement period 10 days, defect rate 0.5%) belong to the electrical equipment group, the representative cost of the electrical equipment group may be calculated as (5,000×1 + 3,000×1) / (1+1) = 4,000 won, the representative procurement period as (7×1 + 10×1) / (1+1) = 8.5 days, and the representative defect rate as (0.2×1 + 0.5×1) / (1+1) = 0.35%.

[0121] The attribute management unit (220) can calculate representative attribute values ​​for each group (S430).

[0122] As an optional embodiment, the attribute management unit (220) can calculate the representative cost, representative lead time, and representative defect rate for each group by weighting the cost, lead time, and defect rate of the parts belonging to each group by the usage ratio of each part, as described in S420.

[0123] As an optional embodiment, representative attribute values ​​by group enable more stable predictions than attribute values ​​of individual parts. While the attributes of individual parts may change frequently due to factors such as changes in suppliers or fluctuations in market conditions, representative attribute values ​​at the group level are average values ​​of multiple parts, thereby having the effect of mitigating the variability of individual parts.

[0124] The attribute management unit (220) can train a multiple output regression model (S440).

[0125] As an optional embodiment, the attribute management unit (220) may configure a multiple output regression model with a part code, quantity, and representative attribute value by group as input variables, and a total cost, lead time, and expected defect rate as output variables. The input variables are in the form, for example, "Representative cost of structure group: 10,000 won, Representative lead time of structure group: 5 days, Representative cost of electrical group: 4,000 won, Representative lead time of electrical group: 8.5 days, ...", and the output variables are in the form, "Total cost: 25,000 won, Total lead time: 10 days, Expected defect rate: 0.3%".

[0126] As an optional embodiment, the multi-output regression model can simultaneously predict multiple attributes with a single model and can incorporate correlations between each attribute into the learning process. For example, correlations such as a tendency for the defect rate to decrease as higher-cost parts are used, or a tendency for costs to be higher as parts with shorter lead times are used, can be learned.

[0127] As an optional embodiment, the multi-output regression model can be implemented using algorithms such as a Multi-layer Neural Network, Random Forest, Gradient Boosting, XGBoost, and LightGBM. The selection of the algorithm may be determined by considering the size of the training data, the number of input variables, and the required prediction speed.

[0128] As an optional embodiment, the attribute management unit (220) may divide the collected training data into training data (e.g., 70%) and validation data (e.g., 30%), train a model with the training data, and then evaluate the performance of the model with the validation data. Model performance may be measured by indicators such as mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R²). For example, if the R² value for total cost prediction is 0.85 or higher, it may be judged as excellent prediction performance.

[0129] As an optional embodiment, the model training process of S410 to S440 may be performed once during the initial system setup or retrained at regular intervals (e.g., once a month, once a quarter). During retraining, newly collected actual data is added to the training data, thereby enabling improved prediction accuracy that reflects the latest market conditions (e.g., fluctuations in raw material prices, changes in the supply chain).

[0130] As an optional embodiment, when a new BOM creation request is received, the attribute management unit (220) inputs new part configuration information into a learned multiple-output regression model to calculate multiple attribute prediction values ​​and prediction reliability. Prediction reliability is an indicator representing the uncertainty of the prediction values ​​output by the model; for example, when using a Bayesian neural network, a confidence interval can be calculated through the probability distribution of the prediction values.

[0131] As an optional embodiment, if the predicted attribute value falls outside the target range set by the user, the attribute management unit (220) may search for past BOMs that satisfy the attribute criteria while having a component configuration greater than or equal to a preset similarity criterion from the training data and provide them to the user terminal (110). For example, if the predicted total cost is 35,000 won and exceeds the target cost of 30,000 won, three past BOMs that had a component configuration similarity of 0.8 or higher and an actual cost of 30,000 won or less may be searched for and provided in the form of "similar product examples." Similarity may be calculated as cosine similarity or Euclidean distance-based similarity between component configuration vectors.

[0132] As an optional embodiment, search accuracy can be improved by additionally considering meta-information such as part category, product use, and target market when searching for similar BOMs. For example, past BOMs corresponding to the "household" target market within the same "electronics" category can be searched preferentially.

[0134] FIG. 5 is a flowchart illustrating a detailed process for specification verification and multi-purpose optimization according to one embodiment of the present invention.

[0135] Referring to FIG. 5, in one embodiment, the specification verification unit can collect regulatory data from a specification database (S510).

[0136] As an optional example, the standard verification unit may look up information on usage restrictions for each part by referring to standard databases such as RoHS (Restriction of Hazardous Substances), REACH (Registration, Evaluation, Authorization and Restriction of Chemicals), and the Occupational Safety and Health Act. For example, information such as "maximum lead (Pb) content of 0.1%", "prohibition of use of certain phthalate compounds", and "prohibition of use of mercury (Hg)" may be collected.

[0137] As an optional embodiment, the specification verification unit may query prohibited combination information. A prohibited combination refers to a combination where the combination of specific parts and parts or parts and processes is not permitted according to specifications, regulations, safety standards, or manufacturing process constraints. For example, combination restriction information such as "safety verification required when using a combination of a lithium battery and a specific metal case" or "prohibition of simultaneous use of a specific plastic material and a specific solvent" may be collected.

[0138] As an optional embodiment, the standards database may be updated periodically, and the latest regulatory information may be automatically reflected upon amendment of laws. For example, the database may be updated by monitoring announcements of law amendments from relevant agencies, such as the Ministry of Environment and the Ministry of Trade, Industry and Energy.

[0139] The specification verification unit and the optimization unit can merge user goals and regulatory conditions (S520).

[0140] As an optional embodiment, the optimization unit may integrate constraints entered by the user and constraints provided by the specification verification unit. User-entered constraints may include cost constraints (e.g., total cost of 30,000 won or less), lead time constraints (e.g., lead time within 14 days), quality constraints (e.g., defect rate of 0.5% or less), and component preference conditions (e.g., preferential use of components from a specific manufacturer).

[0141] As an optional embodiment, the optimization unit may convert the integrated constraints into a mathematical form. For example, they may be expressed as inequality or logical constraints such as "total cost ≤ 30,000 won", "procurement lead ≤ 14 days", "defect rate ≤ 0.5%", "part X lead (Pb) content ≤ 0.1%", "prohibition of using combination of part A + part B".

[0142] As an optional embodiment, the optimization unit may verify whether there is a conflict between constraints. If user input constraints conflict with regulatory constraints, a warning message may be provided to the user terminal (110), and the constraints may be modified. For example, if the user inputs "Part X must be used" but the regulatory database confirms "Part X is prohibited," a message such as "Part X is prohibited from use due to RoHS regulations. Please select an alternative part" may be displayed.

[0143] The optimization unit can perform a multi-objective optimization algorithm (S530).

[0144] As an optional embodiment, the optimization unit may simultaneously optimize two or more objectives, including total cost and multiple quality indicators. For example, multiple objectives may be set, such as "Objective 1: Minimize total cost," "Objective 2: Minimize lead time," "Objective 3: Minimize defect rate," and "Objective 4: Maximize environmental friendliness."

[0145] As an optional embodiment, the optimization unit may use evolutionary algorithms such as NSGA-II (Non-dominated Sorting Genetic Algorithm II), NSGA-III, and MOEA / D (Multi-Objective Evolutionary Algorithm based on Decomposition) as multi-objective optimization algorithms. These algorithms can search for a set of Pareto optimal solutions by considering trade-offs between multiple objectives.

[0146] As an optional embodiment, the optimization unit may generate a set of Pareto optimal solutions. A Pareto optimal solution is a solution in which the performance of one objective cannot be improved without sacrificing the performance of another objective, and can represent various balance points between cost and quality. For example, various options may be included, such as "BOM Candidate 1: Cost 25,000 won, Lead time 12 days, Defect rate 0.3%", "BOM Candidate 2: Cost 28,000 won, Lead time 10 days, Defect rate 0.2%", and "BOM Candidate 3: Cost 30,000 won, Lead time 8 days, Defect rate 0.1%".

[0147] As an optional embodiment, the optimization unit may provide a plurality of BOM candidates included in the Pareto optimal solution set to the user terminal (110). Each BOM candidate may include information such as a list of part codes, quantity, estimated cost, estimated lead time, estimated defect rate, and regulatory compliance.

[0148] As an optional embodiment, the optimization unit may visualize and provide the trade-off relationship between total cost and quality indicators for each BOM candidate. In a 2D graph, the x-axis can be set as total cost and the y-axis as the quality total score (weighted sum of the inverse of the defect rate + the inverse of the lead time, etc.), and each BOM candidate can be marked as a point to visually represent the Pareto Front. When using a 3D graph, the x-axis can be set as cost, the y-axis as lead time, and the z-axis as the defect rate to represent it in three dimensions.

[0149] As an optional embodiment, the optimization unit may recommend an optimal BOM based on the user's preference direction. If the user inputs a "cost-first" preference, the BOM with the lowest cost among the Pareto optimal solution set is recommended; if the user inputs a "quality-first" preference, the BOM with the lowest defect rate is recommended; and if the user inputs a "balanced" preference, the BOM located at the midpoint between cost and quality is recommended. When making recommendations, scores such as "Cost-first: 95 points," "Quality-first: 85 points," and "Balanced-first: 90 points" may be displayed for each BOM candidate to assist the user in making a selection.

[0150] As an optional embodiment, the optimization unit may provide personalized recommendations by learning the past selection history of users. For example, if a specific user has primarily selected cost-priority BOMs in the past, cost-priority BOMs may be presented to that user as a default recommendation.

[0152] FIG. 6 is a diagram showing a BOM hierarchy structure according to an embodiment of the present invention and process and equipment information linked thereto.

[0153] Referring to Fig. 6, the BOM can be hierarchically organized from the top-level finished product to intermediate assemblies and individual parts, and optionally, process (routing) stages and equipment (resource) information of the corresponding product can be linked together.

[0154] As an optional embodiment, FIG. 6 illustrates the BOM hierarchy of the Smart Tumbler Model X, along with examples of process (routing) steps for producing the BOM and equipment (resources) used in each process. The BOM hierarchy is generally represented by level numbers such as 01, 02, and 03, where a smaller number indicates a higher level and a larger number indicates a lower level.

[0155] As an optional embodiment, the finished product "Smart Tumbler Model X" is located at level 01 (top-most part). This refers to a finished product that is ultimately sold to a customer.

[0156] As an optional embodiment, the major assemblies constituting the finished product are located at Level 02 (upper component). In the case of the Smart Tumbler Model X, "Stainless Steel Body Assembly", "Electronics Assembly", "Lid Assembly", and "Hardware" correspond to Level 02. Each assembly refers to an intermediate component made by assembling multiple parts.

[0157] As an optional embodiment, individual components constituting each assembly are located at Level 03 (sub-component). The Stainless Steel Body Assembly consists of a "Stainless Steel Body" and a "Sensor Connection Wire (Quantity: 1)". The Electronics Assembly consists of a "Temperature Sensor", a "Micro Controller Unit (MCU)", a "Bluetooth Module", an "LED Display Panel", and a "Battery". The Lid Assembly consists of a "Silicone Lid", and the Hardware consists of "Screws (Quantity: 4)".

[0158] As an optional embodiment, additional attribute information may be associated with each part in addition to the part code, part name, and quantity. For example, for an MCU, detailed information such as "Part Code: MCU-001A, Part Name: Micro Controller Unit, Quantity: 1, Supplier: ABC Electronics, Unit Price: 5,000 won, Lead Time: 7 days, Past Failure Rate: 0.2%, Specification: 5V 32bit, Weight: 5g, RoHS Compliance: Yes" may be stored.

[0159] As an optional embodiment, the BOM hierarchy may reflect the flow of the manufacturing process. For example, manufacturing may proceed in the order in which the Stainless Steel Body Assembly is first assembled by attaching the Sensor Connection Wire to the Stainless Steel Body, the Electronics Assembly is assembled by soldering the MCU, sensor, battery, etc., onto the PCB, and finally all assemblies are combined to create a finished product.

[0160] As an optional embodiment, the BOM hierarchy can also be utilized for inventory management and cost calculation. By managing inventory at each level, the inventory status at the assembly level can be identified, and the costs of higher-level assemblies and finished products can be calculated by summing the costs of lower-level parts.

[0162] FIG. 7 is a diagram illustrating a result value supplementation process according to one embodiment of the present invention.

[0163] Referring to FIG. 7, the attribute management unit (220) can automatically supplement missing attribute values ​​found in the part data.

[0164] As an optional embodiment, FIG. 7 illustrates a process for supplementing results in a situation where lead time information for MCU components is missing. Missing values ​​may occur if specific attribute information is missing during the data collection process or if historical data is unavailable due to new components.

[0165] As an optional embodiment, the attribute management unit (220) first queries the attribute value of the same part code "MCU-001A" from the technical document database (140) and the MES (150). As a result of the query, it is confirmed that there is no procurement period information for the corresponding code.

[0166] As an optional embodiment, if information cannot be found with the same part code, the attribute management unit (220) searches for similar parts with the same functional role. The search conditions include part category (e.g., "electrical parts"), specification range (e.g., "5V, 32bit"), and manufacturer (e.g., "same or similar manufacturer"). As a result of the search, MCU-002B (similarity: 0.92, lead time: 7 days), MCU-003C (similarity: 0.88, lead time: 10 days), and MCU-004D (similarity: 0.85, lead time: 5 days) are found.

[0167] As an optional embodiment, the similarity score may be calculated by comprehensively considering the component's specifications, manufacturer, price range, frequency of past use, functional role, etc. The method for calculating similarity is as follows. First, similarity is calculated individually for each attribute. For example, specification similarity is calculated as the ratio of matching specification items such as voltage (5V) and bit count (32-bit), and manufacturer similarity can be set as 1.0 for the same manufacturer, 0.8 for similar manufacturers, and 0.5 for different manufacturers. Next, the final similarity is calculated by weighting the similarity for each attribute. For example, weights such as "Similarity = Specification Similarity × 0.5 + Manufacturer Similarity × 0.3 + Price Range Similarity × 0.2" may be applied.

[0168] As an optional embodiment, the attribute management unit (220) estimates the result by weighting the procurement time of the top 3 parts by similarity scores. The estimated procurement time is calculated as = (7×0.92 + 10×0.88 + 5×0.85) / (0.92 + 0.88 + 0.85) = (6.44 + 8.8 + 4.25) / 2.65 = 19.49 / 2.65 = 7.35 days.

[0169] As an optional embodiment, the number of similar parts used when calculating the weighted average can be pre-set. For example, it can be set to the top 3, top 5, etc., and it is generally known that 3 to 5 is appropriate. If the number of similar parts is too small, the reliability of the estimate is low, and if it is too large, parts with low similarity are included, which may reduce the accuracy of the estimate.

[0170] As an optional embodiment, the lead time for MCU components is supplemented as 7.35 days (estimated), and the mark “(estimated)” or “*” may be included to indicate that it is an estimate.

[0171] As an optional embodiment, the attribute management unit (220) may notify the user terminal (110) of the result supplementation result. For example, it may display a message such as "The procurement time information for MCU-001A was missing, so it was estimated using the average value of three similar parts. Estimated procurement time: 7.35 days, Estimated reliability: 0.88".

[0172] As an optional embodiment, the attribute management unit (220) can process multiple attributes simultaneously when supplementing the result value. For example, if both the procurement period and the defect rate of MCU-001A are missing, the procurement period and the defect rate can be estimated simultaneously by utilizing the same similar part search results.

[0173] As an optional embodiment, the attribute management unit (220) can update the estimated value to the actual value when actual data is collected after supplementing the result value. For example, if MCU-001A is actually ordered and the procurement period is confirmed to be 8 days, the estimated value of 7.35 days can be replaced with the actual value of 8 days, and this can be added to the training data to improve future prediction accuracy.

[0175] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0176] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0177] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0179] 110: User terminal 120: Communication network 130: Server 140: Technical Document Database 150: MES 210: Data Collection Unit 220: Attribute Management Department 230: BOM Generation Section 240: BOM Verification Section 250: Intelligent Autonomous Control Unit

Claims

Claim 1 A data collection unit that collects part information, process information, and equipment information from multiple data sources; an attribute management unit that preprocesses the collected part information, process information, and equipment information, classifies them by part category, and predicts attribute information for each part; a BOM generation unit that generates and provides a BOM based on the predicted attribute information; and a BOM verification unit that performs error verification on the generated BOM. The AI-based intelligent BOM automatic generation system comprises an intelligent autonomous control unit that autonomously controls the operations of the data collection unit, the attribute management unit, the BOM generation unit, and the BOM verification unit, wherein the attribute management unit collects learning data including part codes, quantities, and supplier information from a technical document database and an MES, classifies multiple parts into multiple groups according to their functional roles based on attribute information stored in advance for each part, calculates a representative attribute value for each group by weighting the attribute information of the parts belonging to each group, trains a multiple output regression model with the representative attribute value for each group as an input variable and total cost, lead time, and expected defect rate as output variables, and when a new BOM generation request is received, inputs it into the trained multiple output regression model to calculate multiple attribute prediction values ​​and prediction reliability. Claim 2 The AI-based intelligent BOM automatic generation system according to claim 1, wherein the data collection unit processes a fax technical document using OCR to extract part information and processes an email using NLP to extract part information, and queries quotation history, order history, procurement history, after-sales service history, and production history from an MES (Manufacturing Execution System); and the BOM verification unit performs error verification on the generated BOM, including part code validity, quantity appropriateness, and supplier existence, and if an error is found, corrects the BOM or requests confirmation from a user terminal. Claim 3 delete Claim 4 An AI-based intelligent BOM automatic generation system according to claim 1, wherein the attribute management unit classifies the plurality of parts into at least one group among a structural part group, an electrical part group, and a consumable part group, and calculates a representative cost, representative lead time, and representative defect rate for each group by weighting the cost, lead time, and defect rate of the parts belonging to each group by the usage ratio of each part. Claim 5 An AI-based intelligent BOM automatic generation system according to claim 1, wherein the attribute management unit performs data preprocessing on the part information, including part code normalization, unit unification, and duplicate removal. Claim 6 The AI-based intelligent BOM automatic generation system according to claim 1, wherein the attribute management unit supplements the result by querying the attribute value of the same part code from the technical document database and MES when a missing attribute value is found in part data collected from multiple external systems, and when no same part code exists, estimates the result by the average of the attribute values ​​of similar parts with the same functional role, and supplements the result by weighted averaging the attribute values ​​of a preset number of parts with the highest similarity score, considering the part category, specification range, and manufacturer match when selecting the similar parts. Claim 7 An AI-based intelligent automatic BOM generation system according to claim 1, further comprising: a specification verification unit that references a specification database to query usage restriction information and prohibited combination information for each part and generates constraints; and an optimization unit that executes a multi-objective optimization algorithm to simultaneously optimize two or more goals based on the predicted attribute information and the generated constraints to generate a set of Pareto optimal solutions, wherein the BOM generation unit visualizes and provides trade-off relationships for each BOM candidate included in the set of Pareto optimal solutions and recommends an optimal BOM according to the user's preference direction. Claim 8 A data collection unit collects part information, process information, and equipment information from multiple data sources; an attribute management unit preprocesses the collected part information, process information, and equipment information and classifies them by category; The above-mentioned attribute management unit collects training data including part code, quantity, and supplier information from a technical document database and an MES; The above-mentioned attribute management unit classifies multiple parts into multiple groups according to their functional roles based on attribute information stored in advance for each part, and calculates a representative attribute value for each group by weighting the attribute information of the parts belonging to each group; The above-mentioned attribute management unit trains a multiple output regression model using the representative attribute value for each group as an input variable and the total cost, procurement period, and expected defect rate as output variables; The above attribute management unit, upon receiving a new BOM generation request, inputs it into the above-mentioned learned multiple-output regression model to calculate multiple attribute prediction values ​​and prediction reliability; AI-based intelligent automatic BOM generation method characterized by comprising: a BOM generation unit, a step of generating a BOM based on the predicted attribute information; and a BOM verification unit, a step of performing error verification on the generated BOM. Claim 9 A computer-readable recording medium having a program for executing the method of paragraph 8 on a computer.

Images