Spark plug manufacturing quality precision traceability method based on whole-process identification mapping

Abstract

This invention provides a method for precise traceability of spark plug manufacturing quality based on full-process identifier mapping, belonging to the technical field of precise traceability of spark plug manufacturing quality. This invention assigns identifier cards to the components and products of the spark plug, stores process data from each stage of production on the identifier cards, obtains basic information of the process equipment and processing data to calculate processing quality coefficients, calculates worker operation quality coefficients based on historical work information, and timestamps the basic information, component process data, and assembly process data of the components assembled into the spark plug to form a product dataset. The product dataset and quality defect types are transformed into a structured data table with the same format, generating production triples and quality triples respectively, and constructing a spark plug quality knowledge graph. The knowledge graph is periodically updated by updating the processing quality coefficient, operation quality coefficient, and triples for each time period.

Description

Technical Field

[0001] This invention relates to the field of precise traceability technology for spark plug manufacturing quality, specifically a method for precise traceability of spark plug manufacturing quality based on full-process identification mapping. Background Technology

[0002] For decades, spark plugs have been a critical component of internal combustion engines, directly impacting a vehicle's power, fuel economy, and environmental performance. Any quality defects can lead to unstable engine operation, shortened lifespan, or even safety accidents. When spark plug quality issues arise, it's necessary to trace the source of the problem to identify its root cause and resolve the safety concerns.

[0003] Implementing quality traceability requires building a data chain across the entire product lifecycle to quickly locate and resolve problems. Through quality traceability, manufacturers can access detailed spark plug records to check raw material characteristics such as ceramic alumina purity, shell hardness, and electrode resistivity, as well as specific process parameters in the assembly process, enabling precise recall and process correction, significantly reducing quality costs. Existing traceability solutions mainly involve hierarchical practices: the most basic is manual paper records, relying on work orders and labels, but data is easily lost and traceability efficiency is low; a more common approach is automated labeling technology such as barcodes or QR codes, combined with scanners and databases, to track component flow; however, this method is often limited to logistics and struggles to integrate real-time process data. Furthermore, traditional methods struggle to link batch defects in ceramic components to the performance of the final product, resulting in incomplete traceability. While existing technologies have made some progress, they still suffer from limitations such as information gaps, slow response times, and insufficient intelligence when facing complex manufacturing processes. New methods are urgently needed to overcome these bottlenecks and achieve seamless monitoring from raw materials to finished products, while simultaneously enabling intelligent queries based on product quality issues to improve overall quality control.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a precise traceability method for spark plug manufacturing quality based on full-process identification mapping, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for precise traceability of spark plug manufacturing quality based on full-process identification mapping, comprising the following steps: Step 1: Assign unique component identifiers and identification cards to the components of the spark plug to be traced, obtain the basic information of each component and record it on the identification card, store the component process data in the identification card during the production process of each component, merge the unique component identifier into a unique product identifier, and store the assembly process data of each assembly process in the identification card. Step 2: Obtain the basic information and processing data of the process equipment in each process step, assign a unique equipment identifier to the process equipment, calculate the processing quality coefficient based on historical processing data, obtain worker identification and work information, and calculate the worker's operation quality coefficient based on historical work information; Step 3: During processing, map the basic information of the equipment, processing quality coefficient and operation quality coefficient to the identification cards of the parts and products, and print the basic information of the components assembled into spark plugs, component process data and assembly process data with timestamps to form a product dataset. Step 4: Obtain the product quality defect type, preprocess the product dataset and quality defect type, convert them into a structured data table with the same format, generate production triples and quality triples respectively, and complete the production triples and quality triples. Step 5: Generate a spark plug quality knowledge graph based on the production triplet and the quality triplet. In subsequent production cycles, dynamically update the processing quality coefficient, operation quality coefficient and triplet corresponding to each time period based on actual quality feedback, and update the knowledge graph synchronously at fixed intervals.

[0007] Furthermore, the components include ceramic components, housing components, and electrode components; Basic information about ceramic components includes: ceramic powder batch number, alumina purity, particle size distribution, and density; Basic information about the shell components includes: bar stock batch number, steel grade, hardness, and tensile strength; Basic information about electrode components includes: electrode metal batch number, resistivity, and alloy composition ratio; The assembly process includes: assembling the shell and ceramic parts, and assembling the electrodes.

[0008] Furthermore, the basic information of the equipment includes operating mode, communication status, and electrical parameters; historical processing data includes yield rate, distribution of key processing data values, and stable operating time of the equipment. The method for calculating the processing quality coefficient is as follows: Historical processing data is divided into time windows, and a walking value is set. Within each time window, the standard deviation of the main processing data values ​​is calculated. The stable operation ratio within the time window is calculated based on the stable operating time of the equipment. The processing quality coefficient is calculated based on the yield rate, the standard deviation of the main processing data values, and the stable operation ratio within the time window using a weighted allocation method.

[0009] Furthermore, the work information includes operation score, pass rate, and number of violations; The technical method for improving the operational quality coefficient is as follows: Obtain daily work information, calculate the operation base based on operation score and pass rate, introduce a penalty factor for the number of violations and perform logarithmic fitting to amplify the impact of the number of violations on product quality, set a calculation period, and use the forgetting coefficient to calculate the operation quality coefficient within the period as the current operation quality coefficient.

[0010] Furthermore, the method for mapping equipment data and operational quality coefficients to identification cards for components and products is as follows: Based on the technological processing relationship between equipment and workers on component parts and overall parts, during the processing, equipment identification, identity information and component identification and product identification are mapped one by one, and equipment data and operation quality coefficients are stored in component and product identification cards.

[0011] Furthermore, the method for preprocessing product data and quality defect types is as follows: Perform integrity and consistency checks on product datasets and quality defect types, identify and handle missing values, outliers, and duplicate records, and then unify heterogeneous data into a standardized format, including mapping unstructured defect description text to a standard defect code library, converting all timestamps to standard ISO format, and ensuring time zone consistency.

[0012] Furthermore, the method for generating production triples is as follows: Based on structured data tables, entity and relation extraction is performed. Component identifiers, product identifiers, and equipment identifiers are extracted as entities, and the corresponding process links are defined as relations. Relevant basic equipment information, processing quality coefficients, and basic information of component components are added as entity attributes. At the same time, component process data and assembly process data are timestamped and mapped to the dynamic attributes of the process relation, thereby constructing triples in the form of "entity-relationship-entity" and "entity-attribute-value". Finally, all such triples are integrated to form a complete set of production triples. The method for generating quality triplet is as follows: In the structured data table, identify entities related to quality, including quality defect types, components, process equipment, and processing workers, and define quality traceability relationships, including causing defects, associating with defects, and being affected as entity attributes. At the same time, extract defect description, severity, and detection time as relation attributes, construct triples of "entity-has defects-defect type" and "entity-cause-defect occurrence", and embed processing quality coefficient and operation quality coefficient as relation attributes to form quality triples.

[0013] Furthermore, the method for completing the triplet is as follows: The extracted entities and relationships are mapped to a low-dimensional continuous vector space. The spatial distance and semantic association scores between entities are calculated using the TransE learning algorithm. On the probabilistic dimension, potential associations that are not directly recorded in the data table but exist logically are predicted and mined.

[0014] Furthermore, the step of generating the spark plug quality knowledge graph is as follows: The previously extracted production triples and quality traceability triples are topologically fused in a unified vector space. Using the product's unique identifier as the core anchor point, the production triple chain describing the manufacturing physical process and the quality triple topology describing the logical evaluation are cross-domain associated. The knowledge graph is then physically stored based on the attribute graph model of the graph database to form a knowledge graph.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention assigns identification cards to the components and products of spark plugs, stores the process data of the production process in the identification cards, obtains the basic information of the process equipment and processing data to calculate the processing quality coefficient, calculates the worker's operation quality coefficient based on historical work information, and timestamps the basic information, component process data and assembly process data of the components assembled into spark plugs to form a product dataset. The product dataset and quality defect types are transformed into a structured data table of the same format, and production triples and quality triples are generated respectively. A spark plug quality knowledge graph is constructed, and the knowledge graph is updated regularly by updating the processing quality coefficient, operation quality coefficient and triples for each time period. This invention constructs a precise traceability system based on full-process identification mapping and knowledge graphs. By assigning a unique identifier to each spark plug and its core components and dynamically associating it with all elements of data such as raw materials, processes, equipment, and personnel, a complete, continuous, and verifiable digital product archive is formed. This enables forward traceability to obtain the entire data chain from ceramic powder batches to the final inspection report with a single click, greatly improving the completeness and efficiency of information retrieval. By integrating dynamic evaluation indicators such as processing quality coefficients and operational quality coefficients with production data, and using knowledge graph technology to perform association mining and completion on the structured production triples and quality triples, intelligent reasoning and rapid attribution of complex quality problems can be achieved. By constructing production triples and quality triples to build a knowledge graph, the originally isolated and static production and quality data are transformed into an interconnected, dynamic, and reasonable knowledge network, realizing full-process transparency and seamless association. This method can locate the relevant information of all its components along the graph relationship chain, forming a complete data chain, and completely solving the problems of information gaps and cumbersome queries in traditional traceability. This technical solution also enables intelligent attribution and impact analysis of quality problems. It can reverse query all affected products based on quality problems, and automatically deduce the most likely root cause by analyzing the process nodes, equipment status and personnel operation records connected to the defects in the analysis graph. This shifts quality control from post-event remediation to pre-event prediction and in-event prevention. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall method flow of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0019] Example: Please see Figure 1 The present invention provides a technical solution: A method for precise traceability of spark plug manufacturing quality based on full-process identification mapping, comprising the following steps: Step 1: Assign unique component identifiers and identification cards to the components of the spark plug to be traced, obtain the basic information of each component and record it on the identification card, store the component process data in the identification card during the production process of each component, merge the unique component identifier into a unique product identifier, and store the assembly process data of each assembly process in the identification card.

[0020] This step, by assigning a unique identifier to each core component (ceramic, housing, electrode) and equipping it with an accompanying digital identification card, is equivalent to establishing a dynamically evolving "holographic file" for each part from its inception. From the basic attributes of raw material batches to real-time process parameters for each step, everything is natively collected and bound, ensuring the granularity and reliability of traceability data. By integrating component identifiers into product identifiers, a physical assembly relationship chain from part to product is naturally formed, allowing subsequent traceability to unfold naturally along this chain. This solves the problem of information gaps, and process data is written to the production line in real time, avoiding the delays and errors of manual entry afterward. This step provides high-quality, highly relevant source data for subsequent steps, forming the physical and data foundation for building a reliable "production triplet."

[0021] In this embodiment, the components include a ceramic component, a housing component, and an electrode component; Basic information about ceramic components includes: ceramic powder batch number, alumina purity, particle size distribution, and density; Basic information about the shell components includes: bar stock batch number, steel grade, hardness, and tensile strength; Basic information about electrode components includes: electrode metal batch number, resistivity, and alloy composition ratio; The assembly process includes: assembling the shell and ceramic parts, and assembling the electrodes.

[0022] The identification card is an RFID electronic tag attached to the component carrier or the product itself, carrying a unique digital identifier and serving as a medium for dynamic data storage and interaction. For the component manufacturing process, the stored data includes: stamping pressure, sintering temperature curves, holding time, and dimensional accuracy and crack detection results after firing for ceramic components during forming and sintering; plastic deformation, cutting speed and feed rate, surface roughness, heat treatment temperature and quenching medium parameters, and real-time hardness detection values ​​for shell components during cold heading, machining, and heat treatment; and alloy powder mixing uniformity, pressing density, welding current and time, and weld penetration data for electrode components during metal powder preparation, pressing, and welding. For the assembly process, the stored data includes: the type and amount of encapsulating adhesive used in the assembly of the shell and ceramic parts, the drying temperature and time curves; the relative position accuracy of the center electrode and side electrodes, gap setting values, screw-in torque or welding energy parameters, and the overall sealing test results after assembly in the electrode assembly process. These real-time, multi-dimensional process data are continuously written into the identification cards, which together form a complete and verifiable digital manufacturing archive for each component and the final product.

[0023] Step 2: Obtain the basic information and processing data of the process equipment in each process step, assign a unique equipment identifier to the process equipment, calculate the processing quality coefficient based on historical processing data, obtain worker identification and work information, and calculate the worker's operation quality coefficient based on historical work information.

[0024] This step introduces a quantitative assessment and integration of dynamic capabilities and states during the production process into the traceability system. It transforms equipment and personnel factors, previously considered static background elements, into key, traceable, and analyzable variables that directly impact product quality. Existing traceability technologies often neglect equipment and personnel records, yet their status significantly influences product quality. This step defines and calculates "processing quality coefficients" and "operational quality coefficients," achieving a leap from "recording participating entities" to "assessing the contribution status of participating entities." Based on historical yield rates and operational stability data of the equipment, as well as information such as workers' first-pass yields and violation records, a dynamic and quantitative capability coefficient is generated through an algorithmic model. This ensures that each production activity is not only linked to the executor but also includes the executor's current capability "confidence" or "risk level." In the subsequent knowledge graph, an abnormal process parameter event occurring on equipment with a low "processing quality coefficient" has a significantly higher probability of being identified as the root cause. Similarly, an assembly defect strongly correlated with a worker experiencing a sudden drop in the "operational quality coefficient" can directly point to human factors. Therefore, the coefficients provided in this step are the core bridge for intelligently associating and reasoning about the production triplet and the quality triplet, enabling the entire traceability method to understand the "causal weights" behind the data, thereby achieving a qualitative change from simple traceability to intelligent diagnosis and predictive control.

[0025] In this embodiment, the basic information of the device includes operating mode, communication status, and electrical parameters; historical processing data includes yield rate, distribution of key processing data values, and stable operating time of the device. The method for calculating the processing quality coefficient is as follows: Historical processing data is divided into time windows, and a walking value is set. Within each time window, the standard deviation of the main processing data values ​​is calculated. The stable operation ratio within the time window is calculated based on the stable operating time of the equipment. The processing quality coefficient is calculated based on the yield rate, the standard deviation of the main processing data values, and the stable operation ratio within the time window using a weighted allocation method.

[0026] Basic equipment information includes operating mode, communication status, and electrical parameters. This information directly affects the real-time working status of the equipment and the reliability of the processing. The operating mode determines whether the equipment is in the preset optimal processing state. Abnormal modes may lead to process deviations. The communication status ensures the continuity of production data acquisition and command issuance. Interruptions can cause information black holes and distort traceability. The stability of electrical parameters such as voltage and current is directly related to the accuracy of the equipment's power output. Fluctuations in these parameters will be immediately reflected in key quality characteristics such as processing dimensions or strength. Therefore, basic information is the primary basis for assessing equipment health and real-time risks.

[0027] Traditional methods for calculating processing quality coefficients using a single index cannot fully capture the dynamic and complex nature of equipment performance. This method analyzes historical data by dividing time windows, integrating three dimensions: yield rate, standard deviation of key processing data values, and ratio of stable equipment operation. A weighted allocation method is used to balance the influence of each dimension, generating a quantitative index that comprehensively reflects the "capability level" of the equipment within a specific time period. This transforms processing quality from a vague qualitative assessment to a precise quantitative one, making equipment status evaluation more scientific and objective. Furthermore, the temporal changes in the coefficient provide early warnings of performance degradation trends. In quality traceability, this coefficient, as a key attribute embedded in a knowledge graph, directly links equipment status and product quality defects, significantly improving the accuracy and efficiency of root cause analysis. This, in turn, drives the transformation of the manufacturing process from passive response to proactive, intelligent control.

[0028] In this embodiment, the work information includes operation score, pass rate, and number of violations; The technical method for improving the operational quality coefficient is as follows: Obtain daily work information, calculate the operation base based on operation score and pass rate, introduce a penalty factor for the number of violations and perform logarithmic fitting to amplify the impact of the number of violations on product quality, set a calculation period, and use the forgetting coefficient to calculate the operation quality coefficient within the period as the current operation quality coefficient.

[0029] The operational scores in the work information reflect the worker's compliance with procedures and operational standards. The first-pass yield directly reflects the worker's work efficiency and the ability to produce qualified output in one go for the process they are responsible for. The number of violations is a direct risk signal that may cause quality fluctuations and safety accidents. These three factors together constitute a multi-dimensional dynamic profile for assessing the worker's impact on product quality. This method combines scores and first-pass yield to calculate the operational baseline, establishing an operational benchmark. Subsequently, a penalty factor is introduced for the number of violations, and logarithmic fitting is performed. The core consideration is that the logarithmic function can significantly amplify the impact difference between first-time violations and repeated violations, providing a moderate warning for occasional mistakes, while repetitive and systematic violations will cause the coefficient to deteriorate sharply. This is more in line with the modeling principle of "low probability, high impact" events in quality risk management. At the same time, by setting the calculation period and forgetting coefficient, the coefficient can dynamically reflect the worker's recent operational level, emphasizing the higher correlation between recent performance and current quality risk, and also giving workers the possibility of improving their behavior to increase the coefficient, reflecting the management's continuous improvement orientation. In subsequent knowledge graph construction and quality traceability, this coefficient, as a key relational attribute, can accurately and strongly correlate specific personnel status with the quality of products produced within a specific time period. When performing defect root cause analysis, this method can quickly identify product batches produced during periods of abnormally low operational quality coefficients, thereby efficiently pinpointing quality fluctuations caused by human factors. This shifts personnel management from subjective experience-based judgment to objective data-driven approaches, ultimately providing measurable and predictable intelligent control measures for preventing batch quality problems caused by human factors.

[0030] Step 3: During processing, map the basic information of the equipment, processing quality coefficient and operation quality coefficient to the identification cards of the parts and products, and print the basic information of the components assembled into spark plugs, component process data and assembly process data with timestamps to form a product dataset.

[0031] In this embodiment, the method for mapping equipment data and operational quality coefficients to identification cards for components and products is as follows: Based on the technological processing relationship between equipment and workers on component parts and overall parts, during the processing, equipment identification, identity information and component identification and product identification are mapped one by one, and equipment data and operation quality coefficients are stored in component and product identification cards.

[0032] This step achieves real-time dynamic binding and accompanying growth of data across all elements of "people, machines, materials, and methods" during the manufacturing process. Existing traceability technologies, such as barcode-based technologies, typically only perform point-to-point scanning at the beginning or end of a process. The recorded relationships are static and retroactive. Equipment status data (such as real-time electrical parameters) and personnel capability data (such as dynamic operation coefficients) are stored independently in different systems, with weak or even disconnected connections to specific products and components. This results in traceability only knowing "a certain equipment produced a certain batch," but not being able to answer "what was the precise state of the equipment and the real-time skill level of the operator when producing this specific component." This step, by mapping equipment identification, worker identity, component identification, and product identification one-to-one at the moment of processing, and immediately writing the basic equipment information, the processing quality coefficient representing the long-term health of the equipment, and the operation quality coefficient representing the current state of the worker into the "accompanying" identification card of the component and product, establishes a unique dynamic file containing full-dimensional process context for each smallest manufacturing unit (component) and the final product.

[0033] Step 4: Obtain the product quality defect type, preprocess the product dataset and quality defect type, convert them into a structured data table with the same format, generate production triples and quality triples respectively, and complete the production triples and quality triples.

[0034] Existing technologies for quality analysis typically rely on fixed queries or simple statistics on relational databases. The relationships between data are implicit and rigid, making it difficult to automatically discover complex connections across stages and entities. For example, traditional methods cannot automatically link seemingly independent events such as "an increase in the standard deviation of electrode resistivity over a certain period" and "a higher ignition voltage in the final product" into a complete causal chain. This step, however, generates "production triples" and "quality triples" and uses vector completion technology to uncover potential relationships. It explicitly and semantically defines and stores the complex relationships between manufacturing entities (parts, equipment, workers), activities (processing, assembly), and results (quality defects), forming a computable and scalable knowledge network.

[0035] This step, through preprocessing the dataset, ensures the quality and consistency of all input knowledge, laying the foundation for reliable analysis. Then, by constructing production triples and quality triples separately, it systematically establishes the logical graphs of the physical manufacturing process (production triples) and the causal relationships of quality (quality triples), providing a complete navigation map for bidirectional traceability. Through triple completion based on algorithms such as TransE, it is possible to infer "hidden" relationships in the data that are not explicitly recorded but are highly likely to exist, greatly enhancing the completeness and reasoning ability of the knowledge graph.

[0036] In this embodiment, the method for preprocessing product data and quality defect types is as follows: Perform integrity and consistency checks on product datasets and quality defect types, identify and handle missing values, outliers, and duplicate records, and then unify heterogeneous data into a standardized format, including mapping unstructured defect description text to a standard defect code library, converting all timestamps to standard ISO format, and ensuring time zone consistency.

[0037] Preprocessing product data and quality defect types is an indispensable foundational step in building a reliable knowledge graph. Raw data collected directly from the manufacturing site is inherently heterogeneous and of varying quality. For example, defect descriptions might come from the verbal notes of quality inspectors (e.g., "the ceramic has cracks"), while equipment timestamps might use local time zones. Directly using this raw data for association analysis will lead to semantic confusion, temporal misalignment, and association errors, thereby compromising the accuracy and reliability of the subsequent knowledge graph. Preprocessing transforms "heterogeneous data" into "clean knowledge." Through integrity and consistency checks, it eliminates traceability gaps caused by missing or contradictory data at the source. By handling missing and outlier values, it corrects noise and errors in data collection, improving the overall quality and representativeness of the dataset. By mapping unstructured text to standard defect codes and unifying time formats, it completely resolves semantic ambiguity and spatiotemporal misalignment issues.

[0038] Integrity checks set non-null constraints on key fields in the data table (such as component unique identifiers, process step codes, and inspection times). By scanning records, entries lacking critical information are automatically marked, and real-time alarms or supplementary work orders are triggered according to preset strategies to ensure a seamless data chain. Consistency checks verify the rationality of data relationships at the business logic level. For example, it verifies that the timestamp of a component assembly must be later than the manufacturing completion time of all its constituent components to prevent logical fallacies of time reversal; it checks whether the physical properties (such as hardness values) of the same batch of raw materials are consistent within a reasonable error range across different records; and it verifies whether the coding rules for product identification conform to the established standard format to prevent information silos caused by data entry errors. By writing and running such business rule scripts or using data quality tools for automated auditing, the logical consistency of the data is ensured at the source.

[0039] In this embodiment, the method for generating production triples is as follows: Based on structured data tables, entity and relation extraction is performed. Component identifiers, product identifiers, and equipment identifiers are extracted as entities, and the corresponding process links are defined as relations. Relevant basic equipment information, processing quality coefficients, and basic information of component components are added as entity attributes. At the same time, component process data and assembly process data are timestamped and mapped to the dynamic attributes of the process relation, thereby constructing triples in the form of "entity-relationship-entity" and "entity-attribute-value". Finally, all such triples are integrated to form a complete set of production triples. The method for generating quality triplet is as follows: In the structured data table, identify entities related to quality, including quality defect types, components, process equipment, and processing workers, and define quality traceability relationships, including causing defects, associating with defects, and being affected as entity attributes. At the same time, extract defect description, severity, and detection time as relation attributes, construct triples of "entity-has defects-defect type" and "entity-cause-defect occurrence", and embed processing quality coefficient and operation quality coefficient as relation attributes to form quality triples.

[0040] This step involves constructing production and quality triplets, aiming to achieve more precise traceability and analysis through professional division of labor. The construction of the production triplets strictly follows the logic of the flow of "materials" and the execution of "processes," structuring entities such as components, products, equipment, and workers, as well as the processing and assembly relationships between them, and attaching specific process parameters and timestamps. The production triplets form a set that purely describes the fundamental question of "how a product is manufactured step by step," providing an indisputable, fine-grained data chain down to each process for forward traceability, ensuring the objectivity and completeness of the traceability basis.

[0041] The quality triplet is used for the characterization, attribution, and impact analysis of quality problems. It extracts information from manufacturing facts, focusing on identifying and defining the specific entity of "defect," and establishing specialized relationships such as "caused," "associated," and "affected." It dynamically links defects to potential risk sources such as people, machines, materials, methods, and environment, embedding quantitative risk indicators such as processing quality coefficients and operational quality coefficients as relational attributes. Therefore, the quality triplet can directly describe the two key questions: "What might cause a defect?" and "What scope will a problem affect?" providing a dedicated reasoning graph for efficient reverse traceability and intelligent root cause analysis. The production triplet provides all the facts needed for traceability, while the quality triplet provides the logic and weights needed for problem diagnosis. This separation design makes the methodology clear, ensuring the purity of manufacturing data records while giving quality analysis professionalism and depth. When the two are associated through unified product and component identifiers, deep intelligent diagnosis can be achieved while maintaining accurate traceability.

[0042] In this embodiment, the method for completing the triplet is as follows: The extracted entities and relationships are mapped to a low-dimensional continuous vector space. The spatial distance and semantic association scores between entities are calculated using the TransE learning algorithm. On the probabilistic dimension, potential associations that are not directly recorded in the data table but exist logically are predicted and mined.

[0043] Data completion is an essential step to overcome the inherent incompleteness of data collection at the manufacturing site, thereby significantly improving the reasoning ability and practical value of knowledge graphs. Sensor failures, human omissions, or system interface limitations inevitably lead to some key associations not being recorded, resulting in "information gaps" in the initial graph. If these gaps are not filled, intelligent traceability and root cause analysis based on the graph will fail at these breakpoints, making it impossible to discover complex, cross-stage, hidden causal chains.

[0044] The TransE learning algorithm is a classic model for knowledge graph representation learning. Its core idea is to map entities and relations in the knowledge graph to the same low-dimensional continuous vector space and assume a simple vector operation principle: for a correct triple (head entity h, relation r, tail entity t), the translation relationship h + r ≈ t should be satisfied in the vector space. That is, the vector of the head entity plus the vector of the relation should be as close as possible to the vector of the tail entity. TransE represents entities (such as equipment, defects) and relations (such as causing, processing) as vectors, so that the method can evaluate the probability of the triple being true (i.e., semantic association score) by calculating the spatial distance between h + r and the candidate tail entity t.

[0045] When the system encounters a potential association that is not recorded (e.g., suspecting that equipment E caused defect D), it can retrieve the corresponding vector and calculate a score. If the score is high (close), the potential association is determined to be highly likely on a probabilistic dimension, and thus added to the knowledge graph as a complete triple. This method enables the system to break through the boundaries of explicit data and automatically discover deep, cross-stage implicit logic such as "Although it is not directly recorded that equipment A processes part B, the processing pattern vector of equipment A is highly similar to the vectors of all equipment that have processed part B, so it is inferred that A may be involved" or "The pattern vector of defect C is close to the vectors of raw material D and process E, so it is inferred that D and E jointly caused C." This greatly enhances the completeness and intelligent reasoning ability of the knowledge graph and is key to achieving accurate and forward-looking quality traceability.

[0046] Step 5: Generate a spark plug quality knowledge graph based on the production triplet and the quality triplet. In subsequent production cycles, dynamically update the processing quality coefficient, operation quality coefficient and triplet corresponding to each time period based on actual quality feedback, and update the knowledge graph synchronously at fixed intervals.

[0047] Even if existing quality management or manufacturing execution systems integrate data, they often manifest as fixed reports or isolated traceability paths. Data from each stage cannot automatically discover deep cross-domain connections, nor do they have the ability to evolve based on new feedback.

[0048] This step involves constructing and continuously updating a dynamically evolving spark plug quality knowledge graph. It integrates production triplets describing the physical reality ("what material was manufactured by what process and what equipment") with quality triplets describing logical attribution ("what defect caused what reason and affected what product"), using the product's unique identifier as an anchor point, within a unified vector space. This fundamentally solves the core challenge of accurate traceability in manufacturing quality: how to efficiently, accurately, and intelligently achieve reverse localization "from effect to cause" and forward tracing "from cause to effect" from massive, heterogeneous manufacturing data. The mechanism lies in the fact that this knowledge graph is not a static database, but an intelligent reasoning engine that uses a unified vector space as its underlying logic, the product's unique identifier as its global anchor point, and deeply integrates the physical manufacturing chain and the quality causal network. For forward traceability, when a spark plug product identifier is input, it is positioned as the core anchor entity in the graph. Then, it traverses along a series of time-series relationships consisting of production triplets, such as "product - assembled by... - component", "component - processed by... - equipment", and "equipment - operated by... worker". This accurately presents the raw material batch, process parameters, equipment status, and operators of each link, achieving completeness and penetration of traceability information.

[0049] For reverse tracing, for example, when a quality defect such as "unstable ignition voltage" is discovered, the method first activates all "cause" and "association" relationships directly connected to the defect entity in the quality triplet network, starting from the defect entity, to directly lock the suspected problematic parts or processes. Then, by utilizing the cross-domain fusion capability of the graph, through vector space calculation (such as comparing the vector similarity between the defect pattern and historical cases), it can uncover potential root causes that are not explicitly stated in the records but have strong semantic associations (for example, the temperature control pattern vector of a certain oven in a specific period is highly similar to the defect pattern vector in this case). At the same time, along the "defect-impact-product" relationship chain, all affected product batches and ranges can be expanded with one click.

[0050] Each new quality feedback (such as online inspection data or new types of quality defects) triggers the recalculation and updating of the processing quality coefficients and worker operation quality coefficients for the relevant equipment. The time periods for dynamically updating the processing quality coefficients, operation quality coefficients, and triples for each time period can be determined through expert consultation based on the product's production and sales cycles. These updated coefficients are written back to the graph as new attributes or relationship weights, thereby dynamically adjusting the strength and risk probability of connections between entities. This allows the graph to assess the real-time risk status of each node in the current production chain based on the latest coefficients and historical patterns, providing predictive warnings for potential defects. This enables closed-loop management of source tracing from post-investigation to pre-investigation prevention, ultimately achieving precise location of quality problems, in-depth root cause analysis, and continuous optimization of the production process.

[0051] In this embodiment, the step of generating the spark plug quality knowledge graph is as follows: The previously extracted production triples and quality traceability triples are topologically fused in a unified vector space. Using the product's unique identifier as the core anchor point, the production triple chain describing the manufacturing physical process and the quality triple topology describing the logical evaluation are cross-domain associated. The knowledge graph is then physically stored based on the attribute graph model of the graph database to form a knowledge graph.

[0052] Through knowledge representation learning, such as the TransE algorithm, all entities and relations in the production and quality triples are mapped to a unified low-dimensional continuous vector space, enabling heterogeneous entities such as "ceramic parts" and "insulation defects" to obtain computable and comparable mathematical representations. Subsequently, the core topology fusion is performed. This process uses the "product unique identifier" as the global anchor and fusion center. The product entity node is located in the vector space. First, the physical chain of "component-process-equipment" described by the associated production triple is reconstructed and attached according to vector similarity and relational logic to form a manufacturing fact subtree rooted at the product. Next, the defects and traceability relationships (such as "caused" and "associated") of the product or its components and equipment in the quality triple network are accurately attached to the corresponding nodes of the subtree. By calculating the semantic distance in the vector space, high-confidence implicit inference relationships are automatically created (for example, inferring equipment that is not directly recorded but has a highly similar vector pattern as a potential defect source). This achieves a deep cross-domain association between the physical manufacturing chain and the logical causal network, forming a cohesive and inferable complete graph at the logical layer.

[0053] Finally, physical storage is performed based on the attribute graph model of the graph database. The attribute graph model is a data model used to represent and store complex relational networks in the real world. It is the underlying core structure of the graph database (Neo4j), and its design intuitively reflects the essence of "entities" and their "relationships". This model consists of three basic elements: nodes, relationships, and attributes. That is, this logical graph is instantiated: each entity (such as a spark plug or a piece of equipment) is created as a "node" with a type label, the relationships between entities (such as "processed at" or "caused") are created as directed "edges", and all basic information (such as material batch number) is stored as "node attributes" and process context data (such as processing timestamps and real-time parameters) is stored as "edge attributes". In this way, a dynamically related knowledge network is persistently stored in graph databases such as Neo4j, forming a physical intelligent quality traceability method that supports precise bidirectional traceability and allows for complex queries and traversals at the millisecond level.

[0054] All the above formulas use dimensionless numerical values ​​for calculation, and the numerical values ​​substituted into the formulas are all in the International System of Units (SI). The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0055] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0056] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0057] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for precise traceability of spark plug manufacturing quality based on full-process identification mapping, characterized in that, The specific steps include: Step 1: Assign unique component identifiers and identification cards to the components of the spark plug to be traced, obtain the basic information of each component and record it on the identification card, store the component process data in the identification card during the production process of each component, merge the unique component identifier into a unique product identifier, and store the assembly process data of each assembly process in the identification card. Step 2: Obtain the basic information and processing data of the process equipment in each process step, assign a unique equipment identifier to the process equipment, calculate the processing quality coefficient based on historical processing data, obtain worker identification and work information, and calculate the worker's operation quality coefficient based on historical work information; Step 3: During processing, map the basic information of the equipment, processing quality coefficient and operation quality coefficient to the identification cards of the parts and products, and print the basic information of the components assembled into spark plugs, component process data and assembly process data with timestamps to form a product dataset. Step 4: Obtain the product quality defect type, preprocess the product dataset and quality defect type, convert them into a structured data table with the same format, generate production triples and quality triples respectively, and complete the production triples and quality triples. Step 5: Generate a spark plug quality knowledge graph based on the production triplet and the quality triplet. In subsequent production cycles, dynamically update the processing quality coefficient, operation quality coefficient and triplet corresponding to each time period based on actual quality feedback, and update the knowledge graph synchronously at fixed intervals.

2. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping as described in claim 1, characterized in that: The components include ceramic components, housing components, and electrode components; Basic information about ceramic components includes: ceramic powder batch number, alumina purity, particle size distribution, and density; Basic information about the shell components includes: bar stock batch number, steel grade, hardness, and tensile strength; Basic information about electrode components includes: electrode metal batch number, resistivity, and alloy composition ratio; The assembly process includes: assembling the shell and ceramic parts, and assembling the electrodes.

3. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping as described in claim 1, characterized in that: The basic information of the equipment includes operating mode, communication status, and electrical parameters. Historical processing data includes yield rate, distribution of key processing data values, and stable operating time of the equipment. The method for calculating the processing quality coefficient is as follows: Historical processing data is divided into time windows, and a walking value is set. Within each time window, the standard deviation of the main processing data values ​​is calculated. The stable operation ratio within the time window is calculated based on the stable operating time of the equipment. The processing quality coefficient is calculated based on the yield rate, the standard deviation of the main processing data values, and the stable operation ratio within the time window using a weighted allocation method.

4. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping according to claim 1, characterized in that: The work information includes operation score, pass rate, and number of violations; The technical method for improving the operational quality coefficient is as follows: Obtain daily work information, calculate the operation base based on operation score and pass rate, introduce a penalty factor for the number of violations and perform logarithmic fitting to amplify the impact of the number of violations on product quality, set a calculation period, and use the forgetting coefficient to calculate the operation quality coefficient within the period as the current operation quality coefficient.

5. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping according to claim 1, characterized in that: The method for mapping equipment data and operational quality coefficients to component and product identification cards is as follows: Based on the technological processing relationship between equipment and workers on component parts and overall parts, during the processing, equipment identification, identity information and component identification and product identification are mapped one by one, and equipment data and operation quality coefficients are stored in component and product identification cards.

6. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping according to claim 1, characterized in that: The method for preprocessing product data and quality defect types is as follows: Perform integrity and consistency checks on product datasets and quality defect types, identify and handle missing values, outliers, and duplicate records, and then unify heterogeneous data into a standardized format, including mapping unstructured defect description text to a standard defect code library, converting all timestamps to standard ISO format, and ensuring time zone consistency.

7. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping according to claim 1, characterized in that: The method for generating production triples is as follows: Based on structured data tables, entity and relation extraction is performed. Component identifiers, product identifiers, and equipment identifiers are extracted as entities, and the corresponding process links are defined as relations. Relevant basic equipment information, processing quality coefficients, and basic information of component components are added as entity attributes. At the same time, component process data and assembly process data are timestamped and mapped to the dynamic attributes of the process relation, thereby constructing triples in the form of "entity-relationship-entity" and "entity-attribute-value". Finally, all such triples are integrated to form a complete set of production triples. The method for generating quality triplet is as follows: In the structured data table, identify entities related to quality, including quality defect types, components, process equipment, and processing workers, and define quality traceability relationships, including causing defects, associating with defects, and being affected as entity attributes. At the same time, extract defect description, severity, and detection time as relation attributes, construct triples of "entity-has defects-defect type" and "entity-cause-defect occurrence", and embed processing quality coefficient and operation quality coefficient as relation attributes to form quality triples.

8. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping as described in claim 1, characterized in that: The method for completing the triplet is as follows: The extracted entities and relationships are mapped to a low-dimensional continuous vector space. The spatial distance and semantic association scores between entities are calculated using the TransE learning algorithm. On the probabilistic dimension, potential associations that are not directly recorded in the data table but exist logically are predicted and mined.

9. The method for precise traceability of spark plug manufacturing quality based on full-process identification mapping according to claim 1, characterized in that: The steps for generating the spark plug quality knowledge graph are as follows: The previously extracted production triples and quality traceability triples are topologically fused in a unified vector space. Using the product's unique identifier as the core anchor point, the production triple chain describing the manufacturing physical process and the quality triple topology describing the logical evaluation are cross-domain associated. The knowledge graph is then physically stored based on the attribute graph model of the graph database to form a knowledge graph.

Images