Welder skill grading integrated management system based on AI technical analysis

Abstract

The invention discloses a welder skill dynamic hierarchical management system based on AI technical analysis, which comprises a data acquisition and fusion module, a skill hierarchical rule base module, an AI skill evaluation engine and a man-machine interaction and resource configuration interface module. Self-adaptive skill evaluation and level-to-level management of welders from admission to on-duty performance are realized; a real-time evolution skill system which is driven by two wheels of qualification basis and dynamic performance and is strongly associated with QHSE (Quality Health Safety Environment) behaviors is constructed, and accurate matching and optimal configuration of welder resources are realized. According to the system, through modular management, multi-dimensional data collection and analysis means are applied, comprehensive, accurate and dynamic integrated management of each welder is achieved from pre-job check, post-job skill assessment, field performance and the like, welder resource allocation is optimized, and therefore the welding quality and the management efficiency are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of welding technology analysis, and more specifically to an integrated management system for welder skill grading based on AI technology analysis. Background Technology

[0002] In the field of marine engineering welding, welders' qualifications and welding performance are crucial to the safety and reliability of engineering projects. Conventional management methods involve manually tracking, calculating, evaluating, and summarizing the welding process of each welder in real time, which requires significant manpower, resources, and time. This results in low accuracy and timeliness of welding-related data statistics, and makes data maintenance difficult. The main problems are as follows: 1. Incomplete evaluation: For the management of welder qualifications, welder skill assessment is usually based on domestic or international welding standards (such as AWS D1.1 and ASME IX). Welder qualifications are evaluated by referring to welding "standard test pieces". Welders who have obtained welding qualifications are no longer differentiated by skill level. 2. Lack of dynamic adjustment: Evaluation is based solely on relevant welding standards, ignoring the dynamic changes in welders' actual skills. Newly hired but highly skilled welders may be underestimated, while experienced welders whose skills have declined may be overestimated. A single assessment test cannot reflect a welder's true capabilities under different working environments, materials, and process requirements. 3. Neglecting safety practices: Most existing welding assessments do not include safety and quality violations in their scope, resulting in welders not paying enough attention to QHSE (Quality, Health, Safety, Environment), which can easily lead to violations during work and affect the overall quality and safety of the project. 4. Difficulty in adapting to project differences: Different projects have different requirements for welder skills. It is difficult to flexibly adjust according to the special requirements of the project based on relevant welding standards, making it impossible to accurately match suitable welders for the project, which increases the quality risk and cost of the project. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic hierarchical management system for welder skills based on AI technology analysis. Through modular management and the use of multi-dimensional data collection and analysis methods, from pre-employment medical examinations to post-employment skills assessments and on-site performance evaluations, it achieves comprehensive, accurate, and dynamic integrated management of each welder, optimizes welder resource allocation, and thus effectively improves welding quality and management efficiency.

[0004] To achieve the above objectives, the technical solution adopted by this invention is as follows: A dynamic skill grading management system for welders based on AI technology analysis includes a data acquisition and fusion module, a skill grading rule base module, an AI skill assessment engine, and a human-computer interaction and resource allocation interface module. By integrating multi-dimensional data acquisition, artificial intelligence algorithm models, and a dynamic rule engine, it achieves adaptive skill assessment and grading management of welders from entry to on-the-job performance. Construct a real-time evolving skills system driven by both qualification-based and dynamic performance, and strongly correlated with quality, health, safety, environment, and QHSE behaviors, to achieve precise matching and optimized allocation of welder resources.

[0005] Preferably, the data acquisition and fusion module connects to multiple heterogeneous data sources to collect diverse and heterogeneous data from each welder in real time or periodically. Heterogeneous data includes: static qualification data, dynamic process data, and behavioral compliance data; static qualification data includes welder basic information, physical examination qualification information, special operation certificate information, and welder testing qualification information; dynamic process data includes welding task history records, welding workload statistics, welding rework records, and corresponding rework rate calculations; behavioral compliance data includes quality and safety training records and assessment results, and on-site QHSE violation records; The data acquisition and fusion module includes: The welder testing qualification unit is used to input and store the qualification data of welders who pass the standard test piece test. The qualification data includes the unique identifier of the test piece, the test date, the test welding position, the welding method used, the non-destructive testing results, and the corresponding welding procedure qualification number. The welding performance data acquisition unit is used to automatically collect welding process data from the production management system or the welding equipment IoT platform. The welding process data includes the length or number of welds completed by the welder on a regular basis, as well as the identification of rework welds and their reasons confirmed by the quality inspection department, and automatically calculates the periodic and cumulative welding rework rate based on this data. The QHSE behavior recording unit is used to integrate on-site violation reports from the quality and safety inspection system or manual entry. Violations include violations of welding process specifications, failure to use qualified welding materials, failure to wear required personal protective equipment, and unauthorized operations in fire-restricted areas.

[0006] Preferably, the skill grading rule base module pre-stores a welder skill grading system and corresponding grading logic rules based on domain knowledge. The skill level system is divided into three levels: gold, silver, and bronze. The hierarchical logic rules include basic qualification level determination rules based on welding position, welding method, and weldable material range in static qualification data; performance level determination rules based on rework rate threshold and work efficiency benchmark in dynamic process data; and comprehensive rules for handling conflicts and special cases. The comprehensive rules include the principle of choosing the highest level, rules for downgrading due to violations, rules for temporarily suspending qualifications, and rules for demotion. The AI ​​skills assessment engine is communicatively connected to the data acquisition and fusion module and the skills grading rule base module. The machine learning models used in the AI ​​skills assessment engine include: A time-series forecasting model is used to predict the performance trend of welders in the near future based on their historical rework rate and workload data, and the stability of the forecast trend is used as an adjustment factor for dynamic performance scoring. The pattern classification model is used to classify the records of welding rework causes through natural language processing or coding, and to identify whether the rework pattern is systematically related to the welder's personal skill shortcomings. If the correlation exceeds the preset reliability, a negative correction is applied to the welder's dynamic performance score. Preferably, the analysis methods used in the AI ​​skills assessment engine include the following: Step SA1: Receive the multi-dimensional time series data of the target welder after fusion and cleaning; Step SA2: Call the rules in the skill level rule base to parse the static qualification data and generate initial qualification level labels; Step SA3: Use a machine learning model to perform time series analysis and pattern recognition on the dynamic process data, quantitatively evaluate the welder's skill stability and work efficiency trend, and generate a dynamic performance score. Step SA4: Input the behavioral compliance data as a constraint condition to identify the violation pattern and its severity; Step SA5: Combining the initial qualification level label, the dynamic performance score, and the behavioral compliance constraints, an integrated decision model is used to apply the principle of maximizing the highest level or the rule of downgrading for violations to perform weighted or logical operations, and the final comprehensive skill level of the welder in the current assessment cycle is output.

[0007] The implementation methods of the principle of applying the highest authority include the following: Step SF1: Based on the standardized qualification score and the dynamic performance score, independently map the qualification recommendation level and the performance recommendation level respectively; Step SF2: Compare the qualification recommendation level with the performance recommendation level; Step SF3: When the two are inconsistent, select the higher level as the preliminary comprehensive level; Step SF4: The preliminary comprehensive level is downgraded only when the violation downgrade rule or other preset constraints are applied, so as to generate the final comprehensive skill level.

[0008] Preferably, the dynamic update and early warning module is connected to the AI ​​skill assessment engine, and the dynamic update and early warning module is equipped with a dynamic update and early warning method, the method including the following: Step SC1: Automatically trigger the AI ​​skill assessment engine to reassess all registered welders according to the preset cycle; Step SC2: Update the welder's skill level on a monthly basis and store the update results in the welder's skill file; Step SC3: Based on the preset warning threshold, proactively issue warnings for medical examination information, special operation certificates, and welding qualification validity periods that are about to expire; Step SC4: When it is detected that a welder's special operation certificate has expired or there is a serious QHSE violation, the qualification suspension rule or demotion rule will be automatically triggered, and the welder will be marked as demotion or his / her welding task assignment qualification will be suspended in the system. Preferably, the human-computer interaction and resource configuration interface module includes human-computer interaction methods, which include the following: Step SD1: Visually display the welder skill level distribution, individual skill profiles, warning list, and grading criteria to the administrator; Step SD2: Receive project task input, which includes the required welding process, materials, location, and quality requirements; Step SD3: Based on the welder skill level and skill profile output by the AI ​​skill assessment engine, a matching algorithm is used to recommend a list of welders with matching or optimal skill levels for a specific welding task, thereby achieving dynamic matching between welders and projects.

[0009] Preferably, the grading logic rules defined in the skill grading rule base module also include: The quantitative scoring rules for the testing qualification section convert welder test scores, the number and difficulty coefficient of welding positions covered by the qualification, test performance in simulating complex working conditions, and the diversity of weldable materials into standardized qualification scores. The evaluation rules for welding performance set upper limits for monthly cumulative rework rates and benchmark ranges for work efficiency for different skill levels. Specifically, the monthly rework rate threshold for gold-level welders is set below the first preset value, the monthly rework rate threshold for silver-level welders is set below the second preset value, and the monthly rework rate threshold for bronze-level welders is set below the third preset value. The rules for downgrading due to violations should clearly stipulate that when the system identifies a welder committing a QHSE violation of a specific level during the assessment period, regardless of their current qualifications and performance scores, their overall skill level will be forcibly downgraded by one or more levels, and this downgrade status must continue for at least one full assessment period.

[0010] The preferred qualification suspension rules or demotion rules are automatically executed by the dynamic update and early warning module, specifically including: When the system detects that a welder's special operation certificate has expired and has not been renewed within the grace period, it automatically marks the welder's skill level as an assistant and locks all of the welder's task assignment permissions involving welding and thermal cutting. When the system receives a confirmed serious QHSE violation record, it automatically suspends all welding qualifications of the welder in the current project and marks the welder's status as pending retraining and assessment until the required retraining is completed and the assessment is passed. After that, the corresponding qualification and level will be restored by the administrator or by the system based on the assessment results.

[0011] Preferably, the update cycle of the dynamic update and early warning module is monthly. The system is configured to automatically collect, analyze and re-evaluate all welder data from the previous month in the first week of each month, update the level information in the welder skill archive, generate a level change report and a list of welders who meet the standards for each level in the current month. The human-computer interaction and resource allocation interface module also includes a welder screening and comparison function, allowing project managers to input multi-dimensional screening criteria; The multi-dimensional screening criteria include the minimum skill level required, specific welding process qualifications, specific material welding experience, and historical rework rate requirements. Based on these criteria, the system dynamically filters and ranks the most suitable welder candidates from the currently available welder pool, while also providing a comparison view of key performance indicators among the candidates.

[0012] Preferably, the dynamic grading management system for welder skills also includes a feedback closed-loop learning mechanism, including the following: Step SE1: The output of the AI ​​skill assessment engine, namely the actual welding quality result after matching the welder's skill level with the project task, is fed back into the machine learning model as new training data. Step SE2: The system periodically retrains and optimizes the time-series prediction model and pattern classification model using new feedback data to adapt to the impact of different project environments, material properties, or process changes on the evaluation criteria for welder skill performance, thereby achieving continuous iteration and accuracy improvement of the grading model.

[0013] The beneficial effects of this invention are: 1) Improve welding quality: Through strict grading standards, welders are encouraged to improve their skills, reduce rework rates, ensure stable and reliable welding quality, and improve the overall quality of engineering projects.

[0014] 2) Optimize resource allocation: Clear skill levels help projects allocate welding tasks reasonably according to project needs, improve work efficiency, and reduce costs.

[0015] 3) Strengthen safety and quality awareness: The grading method takes into account quality and safety violations and disciplinary infractions, enhances welders' quality and safety awareness, reduces violations, and ensures production safety.

[0016] 4) Dynamic management: The grading status is updated monthly, which can reflect changes in welders' skills in a timely manner and realize dynamic management of the welder team. Attached Figure Description

[0017] Figure 1 This is a system block diagram of a dynamic grading management system for welder skills based on AI technology analysis.

[0018] Figure 2 This is a graph showing the structural welding performance in 2023 and 2024 after the system was implemented.

[0019] Figure 3 This is a performance curve of pipeline welding in 2023 and 2024 after the system was implemented.

[0020] Figure 4 This is a template showing the design of gold, silver, and bronze badges for welders. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings: Example

[0022] Combination Figures 1 to 4 A dynamic skill grading management system for welders based on AI technology analysis includes a data acquisition and fusion module, a skill grading rule base module, an AI skill assessment engine, and a human-computer interaction and resource allocation interface module. By integrating multi-dimensional data acquisition, artificial intelligence algorithm models, and dynamic rule engines, it achieves adaptive skill assessment and grading management of welders from entry to on-the-job performance. Construct a real-time evolving skills system driven by both qualification-based and dynamic performance, and strongly correlated with quality, health, safety, environment, and QHSE behaviors, to achieve precise matching and optimized allocation of welder resources.

[0023] Example 2, based on the above examples, further discloses the following: The data acquisition and fusion module connects to multiple heterogeneous data sources, collecting diverse and heterogeneous data from each welder in real time or periodically. Heterogeneous data includes: static qualification data, dynamic process data, and behavioral compliance data; static qualification data includes welder basic information, physical examination qualification information, special operation certificate information, and welder testing qualification information; dynamic process data includes welding task history records, welding workload statistics, welding rework records, and corresponding rework rate calculations; behavioral compliance data includes quality and safety training records and assessment results, and on-site QHSE violation records; The data acquisition and fusion module includes: The welder testing qualification unit is used to input and store the qualification data of welders who pass the standard test piece test. The qualification data includes the unique identifier of the test piece, the test date, the test welding position, the welding method used, the non-destructive testing results, and the corresponding welding procedure qualification number. The welding performance data acquisition unit is used to automatically collect welding process data from the production management system or the welding equipment IoT platform. The welding process data includes the length or number of welds completed by the welder on a regular basis, as well as the identification of rework welds and their reasons confirmed by the quality inspection department, and automatically calculates the periodic and cumulative welding rework rate based on this data. The QHSE behavior recording unit is used to integrate on-site violation reports from the quality and safety inspection system or manual entry. Violations include violations of welding process specifications, failure to use qualified welding materials, failure to wear required personal protective equipment, and unauthorized operations in fire-restricted areas.

[0024] The skill grading rule base module pre-stores a welder skill grading system and corresponding grading logic rules based on domain knowledge. The skill grading system has three levels: gold, silver, and bronze. The grading logic rules include basic qualification grading rules based on welding position, welding method, and weldable material range in static qualification data; performance grading rules based on rework rate threshold and work efficiency benchmark in dynamic process data; and comprehensive rules for handling conflicts and special situations. The comprehensive rules include the principle of choosing the highest grade, rules for downgrading due to violations, rules for temporary suspension of qualifications, and rules for demotion.

[0025] The AI ​​skills assessment engine communicates with the data acquisition and fusion module and the skills grading rule base module. The machine learning models used in the AI ​​skills assessment engine include: a time series prediction model, which predicts the welder's performance trend in the near future based on the welder's historical rework rate data and workload data, and uses the stability of the predicted trend as an adjustment factor for the dynamic performance score; and a pattern classification model, which performs natural language processing or encoding classification on the records of welding rework reasons, identifies whether the rework pattern is systematically related to the welder's personal skill shortcomings, and applies a negative correction to the welder's dynamic performance score if the correlation exceeds the preset reliability.

[0026] Example 3: Based on the above examples, this example further discloses the following: The analytical methods used in the AI ​​skills assessment engine include the following: Step SA1: Receive the multi-dimensional time series data of the target welder after fusion and cleaning; Step SA2: Call the rules in the skill level rule base to parse the static qualification data and generate initial qualification level labels; Step SA3: Use a machine learning model to perform time series analysis and pattern recognition on the dynamic process data, quantitatively evaluate the welder's skill stability and work efficiency trend, and generate a dynamic performance score. Step SA4: Input the behavioral compliance data as a constraint condition to identify the violation pattern and its severity; Step SA5: Combining the initial qualification level label, the dynamic performance score, and the behavioral compliance constraints, an integrated decision model is used to apply the principle of maximizing the highest level or the rule of downgrading for violations to perform weighted or logical operations, and the final comprehensive skill level of the welder in the current evaluation period is output.

[0027] The implementation methods of the principle of applying the highest authority include the following: Step SF1: Based on the standardized qualification score and the dynamic performance score, independently map the qualification recommendation level and the performance recommendation level respectively; Step SF2: Compare the qualification recommendation level with the performance recommendation level; Step SF3: When the two are inconsistent, select the higher level as the preliminary comprehensive level; Step SF4: The preliminary comprehensive level is downgraded only when the violation downgrade rule or other preset constraints are applied, so as to generate the final comprehensive skill level.

[0028] The dynamic update and early warning module is connected to the AI ​​skill assessment engine. The dynamic update and early warning module is equipped with a dynamic update and early warning method, which includes the following: Step SC1: Automatically trigger the AI ​​skill assessment engine to reassess all registered welders according to the preset cycle; Step SC2: Update the welder's skill level on a monthly basis and store the update results in the welder's skill file; Step SC3: Based on the preset warning threshold, proactively issue warnings for medical examination information, special operation certificates, and welding qualification validity periods that are about to expire; Step SC4: When it is detected that a welder's special operation certificate has expired or there is a serious QHSE violation, the qualification suspension rule or demotion rule will be automatically triggered, and the welder will be marked as demotion or his / her welding task assignment qualification will be suspended in the system.

[0029] Example 4: Based on the above examples, this example further discloses the following: The human-computer interaction and resource configuration interface module includes human-computer interaction methods, which are as follows: Step SD1: Visually display the welder skill level distribution, individual skill profiles, warning list, and grading criteria to the administrator; Step SD2: Receive project task input, which includes the required welding process, materials, location, and quality requirements; Step SD3: Based on the welder skill level and skill profile output by the AI ​​skill assessment engine, a matching algorithm is used to recommend a list of welders with matching or optimal skill levels for a specific welding task, thereby achieving dynamic matching between welders and projects.

[0030] The skill grading rule base module also defines the following grading logic rules: The quantitative scoring rules for the testing qualification section convert welder test scores, the number and difficulty coefficient of welding positions covered by the qualification, test performance in simulating complex working conditions, and the diversity of weldable materials into standardized qualification scores. The evaluation rules for welding performance set upper limits for monthly cumulative rework rates and benchmark ranges for work efficiency for different skill levels. Specifically, the monthly rework rate threshold for gold-level welders is set below the first preset value, the monthly rework rate threshold for silver-level welders is set below the second preset value, and the monthly rework rate threshold for bronze-level welders is set below the third preset value. The rules for downgrading due to violations should clearly stipulate that when the system identifies a welder committing a QHSE violation of a specific level during the assessment period, regardless of their current qualifications and performance scores, their overall skill level will be forcibly downgraded by one or more levels, and this downgrade status must continue for at least one full assessment period.

[0031] The rules for temporary suspension of qualifications or demotion are automatically executed by the dynamic update and early warning module, specifically including: When the system detects that a welder's special operation certificate has expired and has not been renewed within the grace period, it automatically marks the welder's skill level as an assistant and locks all of the welder's task assignment permissions involving welding and thermal cutting. When the system receives a confirmed serious QHSE violation record, it automatically suspends all welding qualifications of the welder in the current project and marks the welder's status as pending retraining and assessment until the required retraining is completed and the assessment is passed. After that, the corresponding qualification and level will be restored by the administrator or by the system based on the assessment results.

[0032] Example 5: Based on the above examples, this example further discloses the following: The dynamic update and early warning module is updated monthly. The system is configured to automatically collect, analyze and reassess all welder data from the previous month in the first week of each month, update the level information in the welder skill archive, generate a level change report and a list of welders who meet the standards for each level in the current month. The human-computer interaction and resource allocation interface module also includes a welder screening and comparison function, allowing project managers to input multi-dimensional screening criteria; The multi-dimensional screening criteria include the minimum skill level required, specific welding process qualifications, specific material welding experience, and historical rework rate requirements. Based on these criteria, the system dynamically filters and ranks the most suitable welder candidates from the currently available welder pool, while also providing a comparison view of key performance indicators among the candidates.

[0033] The dynamic skill grading management system for welders also includes a feedback closed-loop learning mechanism, including the following: Step SE1: The output of the AI ​​skill assessment engine, namely the actual welding quality result after matching the welder's skill level with the project task, is fed back into the machine learning model as new training data. Step SE2: The system periodically retrains and optimizes the time-series prediction model and pattern classification model using new feedback data to adapt to the impact of different project environments, material properties, or process changes on the evaluation criteria for welder skill performance, thereby achieving continuous iteration and accuracy improvement of the grading model.

[0034] Example 6: Based on the above examples, this example further discloses the following: 1.1 Company Overview (Example): The main business of a certain marine equipment company includes the construction of high-end equipment such as deep-sea drilling platforms, offshore wind power installation vessels, and large marine modules. Welding is its core production process, and its quality directly affects the structural safety, service life, and operational reliability of the equipment.

[0035] The company has a number of welders, and its welding processes cover a variety of methods such as SAW (submerged arc welding), GMAW (gas metal arc welding), SMAW (manual arc welding), and FCAW (flux-cored wire arc welding), involving a variety of materials such as carbon steel, high-strength steel, stainless steel, and nickel-based alloys. The construction locations include horizontal, horizontal, vertical, overhead and various complex pipe joints.

[0036] 1.2 Challenges faced before implementation: Before the introduction of this system, the company's welder management mainly relied on traditional, static qualification certificate management and the experience-based judgment of team leaders, which had the following prominent problems: Skills assessment is one-sided: once a welder's grade is determined based on an initial exam, it often remains unchanged for a long time, failing to reflect fluctuations in performance, skill progress, or regression in actual production.

[0037] Experience-based manpower allocation: When project managers assign critical welding tasks, they often rely on personal impressions or welders' reputations, lacking data support. This may lead to a mismatch between skills and tasks, creating potential quality problems or causing efficiency losses.

[0038] Quality attribution is vague: when welding rework occurs, the cause analysis often stays at the process or material level, with insufficient systematic identification of the welder's individual skill shortcomings, and training lacks specificity.

[0039] QHSE management is passive: safety violations are decoupled from the allocation of welder skill resources, failing to form a strong correlation constraint between "safety-skills-job", and the early warning mechanism is lagging behind.

[0040] Low management efficiency: The expiration and renewal reminders for qualification certificates rely on manual ledgers, which are prone to oversight; the welding skill files are not updated in a timely manner, making it impossible to grasp the overall picture of the company's human resources skills in real time.

[0041] 1.3 Implementation of Overall Objectives: The deployment of an "AI-based dynamic skill grading management system for welders" aims to build a new paradigm for intelligent management of welder resources that is data-driven, real-time sensing, and dynamically optimized. Specific objectives include: Improve welding quality: By accurately matching welder skills with task requirements, reduce rework rate and increase first-pass yield from the source.

[0042] Optimize project efficiency: Achieve scientific scheduling and optimized allocation of welding resources, shorten critical path duration, and improve overall project execution efficiency.

[0043] Strengthen QHSE management: directly embed safety and quality behaviors into the skills evaluation system, establish a proactive early warning and constraint mechanism, and reduce operational risks.

[0044] Stimulate talent vitality: Establish a fair, transparent, and dynamic skill value measurement system to guide welders to proactively improve their skills and standardize their behavior.

[0045] Achieve digital management: Create digital assets of welder skills for the enterprise, providing decision support for human resource planning and training system optimization.

[0046] II. System Deployment and Core Module Implementation: 2.1 First Phase: Deployment and Integration of the Data Acquisition and Fusion Module: This stage forms the cornerstone of system operation, with the goal of breaking down data silos and achieving automated or semi-automated collection of welder data across all dimensions.

[0047] Static qualification data integration, with the following sources: the company's existing HR system, training management system, and special equipment operator database.

[0048] Implementation: Develop a data interface to periodically synchronize welder basic information, medical examination report validity period, welder's certificate (special operation certificate) items and validity period. The welder testing qualification unit is a special module, entered through a dedicated port in the process quality department. For example, if welder "Zhang San" passes a process qualification test for "50mm thick EH36 high-strength steel, 2G horizontal welding position, SAW welding," the test date, specimen number, RT (radiological inspection) result (e.g., Level I qualified), and the corresponding WPS (Welding Procedure Specification) number PQR-2024-001 are fully recorded, forming his skill tag.

[0049] Dynamic process data acquisition, interface source: the company's MES (Manufacturing Execution System), WMS (Welding Material Management System), and IoT platform for key welding equipment (such as digital welding machines).

[0050] Implementation method: The welding performance data acquisition unit automatically retrieves the daily "weld number," "weld length," or "number of welds" completed by welders from the work reporting module of the MES. When a weld is determined to require rework by NDT (non-destructive testing) or VT (visual inspection), the inspector marks it as "rework" in the MES and selects or fills in the reason (such as "porosity," "lack of fusion," "undercut," etc.). The system automatically links to the performing welder and calculates their "individual monthly rework rate" (reworked weld length / total weld length) in real time.

[0051] IoT welding machine data (such as welding parameter stability and arc time) are input as auxiliary reference data for subsequent in-depth analysis by AI models.

[0052] Integrate behavioral compliance data, connecting to: the company's QHSE management system, on-site safety inspection app, and training and assessment system.

[0053] Implementation: The QHSE behavior recording unit automatically captures all safety and quality training records and assessment scores of welders. Violation reports submitted by on-site safety officers or quality inspectors via mobile devices (such as "not wearing a protective mask", "unauthorized modification of welding current", "failure to obtain a hot work permit in a fire-prone area") are pushed to this unit in real time after confirmation and are associated with the specific welder.

[0054] 2.2 Second Phase: Localization Configuration of the Skill Level Rule Base Module: Based on the company's actual situation, the system's preset rules were parameterized to form enterprise standards.

[0055] Skill level system definition: Adopting the three-tier system of "Gold, Silver, and Bronze," and clearly defining its corporate connotation: Top-tier welders: Core skill experts for the company. Capable of undertaking the most complex and critical welding tasks (such as deep-sea platform nodes and high-pressure pipelines), and are the main force in tackling process challenges and mentoring apprentices.

[0056] Silver Medal Welder: A key skilled worker. Capable of independently and efficiently completing most routine and challenging welding tasks with consistent and reliable quality.

[0057] Bronze-level welder: A qualified skilled worker. Capable of performing general welding tasks and required to perform some important operations under supervision.

[0058] Quantization of hierarchical logic rules (example): Basic qualification scoring rules: The difficulty coefficient is set by combining welding position (1G to 6G), welding method (SMAW, GMAW, etc.), and material type (carbon steel, stainless steel, etc.). Welders receive corresponding points for each test qualification they obtain. The cumulative score maps to the "recommended qualification level".

[0059] Dynamic performance rules: Rework rate thresholds: The monthly rework rate is set at ≤0.5% for gold-level welders; ≤1.5% for silver-level welders; and ≤3.5% for bronze-level welders (specific values ​​are determined based on historical company data). The welder skill grading rules are shown in Table 1 below: Table 1. Structural Welder Skill Classification Rules:

[0060] Work efficiency benchmark: Based on the standard working hours set for different processes, the length of welds that meet the standard for effective arc time per welder unit is counted and used as the efficiency score.

[0061] Comprehensive rules: The principle of choosing the highest standard: If a welder's quality score reaches the silver medal level, but their performance score has consistently reached the gold medal level for the past six months, then the initial level is gold medal.

[0062] Rule for downgrading due to violations: If a "serious violation" occurs (such as causing a quality accident or serious safety hazard), the overall rating for that period will be forcibly downgraded by one level (Gold to Silver, Silver to Bronze) and will be maintained for at least one evaluation period (one month).

[0063] Qualification Suspension / Demotion Rules: If a certificate expires and is not renewed, the system will automatically mark the worker's status as "auxiliary worker," allowing them to perform non-welding tasks such as grinding and cleaning. In the event of a major safety violation, the status will be marked as "awaiting retraining," and all welding task assignments will be suspended.

[0064] 2.3 The third stage is the training and launch of the AI ​​skills assessment engine, which is the construction of the system's "intelligent brain".

[0065] Model training and initialization: Data preparation: Extract historical performance data, rework records and corresponding task characteristics (process, material, location) of all welders from the past two years, and clean and label them.

[0066] Time-series forecasting models (such as LSTM) use historical data to train the model and learn the changing patterns of each welder's rework rate and workload. After deployment, the model can predict the welder's performance trend for the next month based on data from the most recent 3-6 months. If the trend is upward and stable, a positive adjustment factor is applied to their dynamic performance score; if the trend is downward or highly volatile, a negative adjustment factor is applied.

[0067] Pattern classification models (such as NLP-based text classification or encoding classification): The model is trained to identify the text indicating the reason for rework. For example, the system finds that in welder "Li Si's" recent rework records, the frequency of "incomplete fusion" is significantly higher than the average level of other welders in the same process, and it is strongly correlated with a specific position (such as overhead welding). The model will then determine that he has a skill deficiency of "poor fusion in overhead welding" and apply a continuous negative correction to his dynamic performance score until the pattern disappears.

[0068] Online evaluation process (executing claim steps SA1-SA5): At the beginning of each month, the engine automatically performs the following for each welder: SA1: Retrieve all fusion data for this welder from the previous month.

[0069] SA2: Analyze its static qualifications and generate a "recommended qualification level" (e.g., silver medal).

[0070] SA3: Analyzes dynamic data using time series models and pattern classification models, and outputs a "dynamic performance score" (converted to an equivalent level, such as gold medal trend).

[0071] SA4: Verification of compliance data; no serious violations found.

[0072] SA5: Applying the "highest standard principle," the qualification recommendation (silver) and performance recommendation (gold) are compared, with gold selected as the initial overall level. Due to the absence of any violations, the final overall skill level is rated gold. This result will be used for task matching this month.

[0073] 2.4. The fourth stage involves the application of dynamic updates, early warning systems, and human-computer interaction modules: This phase ensures the system operates dynamically and interacts with managers and project managers.

[0074] Dynamic update and early warning module (executing steps SC1-SC4 of claims): The system is set to automatically trigger a global reassessment (SC1) on the 5th of each month, updating all welder files (SC2).

[0075] At the same time, the system automatically scans the certificate validity period and sends alerts (SC3) to the individual and HR for medical examination and welding certificates that expire within 30 days.

[0076] If the system detects that welder “Wang Wu”’s welder’s certificate expired last month, it will immediately trigger the “qualification suspension rule” and mark him as an “auxiliary worker” in the system, and all his welding task qualifications will be locked by the system (SC4).

[0077] Human-computer interaction and resource configuration interface module (executing steps SD1-SD3 of claims and extended functions): Management Dashboard (SD1): Displays a pie chart of welder skill levels across the company, skill heatmaps for each workshop, and an alert list to the Production Manager and HR Director. Clicking on any welder allows viewing their "skill profile": qualification radar chart, performance trend curve, historical violation records, and AI-generated analysis of skill strengths and weaknesses.

[0078] Intelligent Task Dispatch (SD2, SD3): When a project manager creates a task titled "SAW Welding of S355 Steel Thick Plates at 4G Position on a Platform in the South China Sea," the system manager inputs the process, materials, location, and quality requirements into the system interface. The system immediately and automatically recommends a list of welders from the currently available (excluding those on leave or demoted) pool who simultaneously meet the criteria of "possessing SAW qualifications," "having experience with S355 steel," "passing the 4G position test," and currently holding a gold or silver skill level. The list is then sorted by matching degree (e.g., historical rework rate, experience value for similar tasks).

[0079] Advanced Filtering and Comparison (Extended Functionality): Project managers can further filter welders who have "no porosity-related repair records in the past six months." The system will list candidates and provide a comparative view of their key indicators (such as total weld length, average repair rate, and certification status) to assist in the final decision.

[0080] Example 7: Based on the above examples, this example further discloses the following: Complete application example: Taking a critical closure joint welding project of a 100,000-ton ship hull module as an example.

[0081] Project background and tasks: The company's 100,000-ton ship hull project has entered the main structure assembly stage. There is a full-penetration K-type node weld, made of DH36 high-strength steel, welded at position 6G (pipeline inclined fixation), using FCAW-G (flux-cored wire gas shielded welding) process. This weld is a fatigue critical point with extremely high quality requirements, necessitating 100% UT (ultrasonic testing).

[0082] System-enabled task execution process: The project welding engineer creates a task in the system, specifying the above technical requirements.

[0083] The system uses a matching algorithm from the human-computer interaction and resource configuration interface module to quickly select welders who are currently in the factory but have not been assigned critical tasks.

[0084] Screening Logic: ① Static Qualifications: Must hold test qualification records covering "DH36 steel, 6G position, FCAW"; ② Current Dynamic Rating: "Gold Medal" welders are preferred; ③ Behavioral Compliance: No recent QHSE violation records; ④ AI Performance Trend: Recent performance is stable or improving.

[0085] System recommendation results: Welder Zhao Liu (Gold Medal) is the top priority recommendation. His skill profile shows that he possesses the qualification for this test, has an average rework rate of only 0.3% over the past 12 months, and the AI ​​model evaluates his "stable downhill welding skills." Welder Qian Qi (Silver Medal) is also recommended as an alternative.

[0086] Task execution and process monitoring: The project manager adopted the system's suggestion and assigned Zhao Liu as the chief welder, and Qian Qi as the assistant / alternate welder.

[0087] During the welding process, the data (voltage, current, wire feed speed) from the IoT welding machine used by Zhao Liu remained stable in real time and were all within the process window. No violations were found during the on-site safety inspection.

[0088] Results feedback and closed-loop learning: After the weld was completed, it passed the UT test on the first attempt and was rated as Grade I.

[0089] The quality inspector confirmed that the weld was "qualified on the first attempt" in the MES, and the data was automatically transmitted back to the data acquisition and fusion module, becoming a positive performance record for Zhao Liu.

[0090] The successful match in this task (gold medal welder + high-difficulty task = high-quality result) is used as a positive sample and, together with the task features and welder features, is fed back into the model training library of the AI ​​skill assessment engine.

[0091] The system periodically (e.g., quarterly) uses this newly generated feedback data to retrain the time series prediction model and the pattern classification model, making the model's judgment on "which characteristics of welders perform better under which working conditions" more and more accurate, thus realizing the continuous self-optimization of the classification rules.

[0092] Traditional model: Engineers may choose an experienced welder based on memory or team recommendation. However, this experienced welder may be more skilled in flat welding than 6G welding, or their main qualifications may be in SMAW rather than FCAW, which may lead to a smooth welding process, rework, delays of 3-5 days, and additional costs.

[0093] Through a year of system implementation and operation, the company has achieved significant results in welder management and engineering efficiency: Welding quality has significantly improved: the company's overall first-pass yield rate has increased from 97.2% before implementation to 98.8%, and the rework rate has decreased by 35% year-on-year. After targeted training was provided to individual welders whose skill deficiencies were identified by AI, their individual rework rates decreased by more than 50%.

[0094] Continuous optimization of engineering efficiency: Delays caused by personnel skill mismatches in welding tasks on the critical path have been virtually eliminated. Project manager assignment decision time has been reduced by an average of 70%, and the accuracy of average project welding duration estimates has improved by 25%.

[0095] QHSE risks were effectively reduced: the system automatically blocked welders with expired certificates or those in a demotion state from performing welding operations, and no compliance incidents caused by this occurred throughout the year. Welders' awareness of proactively paying attention to the validity of their certificates and standardizing their operating procedures has significantly increased, and the number of recordable QHSE violations decreased by 28% compared to the previous period.

[0096] A virtuous cycle in the skills ecosystem: Dynamic grading results are published monthly, forming a transparent incentive mechanism of "performance improvement, grade improvement, more opportunities, and increased income." Welders have shifted from "I have to learn" to "I want to learn," with a 40% increase in the number of people actively applying for skills expansion tests. Companies have a clear understanding of their gold and silver medal welder reserves, providing them with the talent pool to undertake more challenging projects.

[0097] The company has accumulated valuable big data assets on welding skills. The Human Resources Department can plan recruitment and training based on skill distribution; the Process Department can optimize process design based on the actual ability distribution of welders; and the company's leadership can gain a direct understanding of the current status and trends of the company's core manufacturing capabilities.

[0098] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A dynamic grading management system for welder skills based on AI technology analysis, characterized in that, It includes a data acquisition and fusion module, a skill grading rule base module, an AI skill assessment engine, and a human-computer interaction and resource allocation interface module. By integrating multi-dimensional data acquisition, artificial intelligence algorithm models, and dynamic rule engines, it enables adaptive skill assessment and grading management of welders from entry to on-the-job performance. Construct a real-time evolving skills system driven by both qualification-based and dynamic performance, and strongly correlated with quality, health, safety, environment, and QHSE behaviors, to achieve precise matching and optimized allocation of welder resources.

2. The dynamic hierarchical management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The data acquisition and fusion module connects to multiple heterogeneous data sources, collecting diverse and heterogeneous data from each welder in real time or periodically. Heterogeneous data includes: static qualification data, dynamic process data, and behavioral compliance data; static qualification data includes welder basic information, physical examination qualification information, special operation certificate information, and welder testing qualification information; dynamic process data includes welding task history records, welding workload statistics, welding rework records, and corresponding rework rate calculations; behavioral compliance data includes quality and safety training records and assessment results, and on-site QHSE violation records; The data acquisition and fusion module includes: The welder testing qualification unit is used to input and store the qualification data of welders who pass the standard test piece test. The qualification data includes the unique identifier of the test piece, the test date, the test welding position, the welding method used, the non-destructive testing results, and the corresponding welding procedure qualification number. The welding performance data acquisition unit is used to automatically collect welding process data from the production management system or the welding equipment IoT platform. The welding process data includes the length or number of welds completed by the welder on a regular basis, as well as the identification of rework welds and their reasons confirmed by the quality inspection department, and automatically calculates the periodic and cumulative welding rework rate based on this data. The QHSE behavior recording unit is used to integrate on-site violation reports from the quality and safety inspection system or manual entry. Violations include violations of welding process specifications, failure to use qualified welding materials, failure to wear required personal protective equipment, and unauthorized operations in fire-restricted areas.

3. The dynamic grading management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The skill grading rule base module pre-stores a welder skill grading system and corresponding grading logic rules based on domain knowledge. The skill level system is divided into three levels: gold, silver, and bronze. The hierarchical logic rules include basic qualification level determination rules based on welding position, welding method, and weldable material range in static qualification data; performance level determination rules based on rework rate threshold and work efficiency benchmark in dynamic process data; and comprehensive rules for handling conflicts and special cases. The comprehensive rules include the principle of choosing the highest level, rules for downgrading due to violations, rules for temporarily suspending qualifications, and rules for demotion. The AI ​​skills assessment engine is communicatively connected to the data acquisition and fusion module and the skills grading rule base module. The machine learning models used in the AI ​​skills assessment engine include: A time-series forecasting model is used to predict the performance trend of welders in the near future based on their historical rework rate and workload data, and the stability of the forecast trend is used as an adjustment factor for dynamic performance scoring. The pattern classification model is used to classify the records of welding rework causes through natural language processing or coding, and to identify whether the rework pattern is systematically related to the welder's personal skill deficiencies. If the correlation exceeds the preset reliability, a negative correction is applied to the welder's dynamic performance score.

4. The dynamic hierarchical management system for welder skills based on AI technology analysis according to claim 3, characterized in that, The analytical methods used in the AI ​​skills assessment engine include the following: Step SA1: Receive the multi-dimensional time series data of the target welder after fusion and cleaning; Step SA2: Call the rules in the skill level rule base to parse the static qualification data and generate initial qualification level labels; Step SA3: Use a machine learning model to perform time series analysis and pattern recognition on the dynamic process data, quantitatively evaluate the welder's skill stability and work efficiency trend, and generate a dynamic performance score. Step SA4: Input the behavioral compliance data as a constraint condition to identify the violation pattern and its severity; Step SA5: Combining the initial qualification level label, the dynamic performance score, and the behavioral compliance constraints, an integrated decision model is used to apply the principle of maximizing the highest level or the rule of downgrading for violations to perform weighted or logical operations, and the final comprehensive skill level of the welder in the current assessment cycle is output. The implementation methods of the principle of applying the highest authority include the following: Step SF1: Based on the standardized qualification score and the dynamic performance score, independently map the qualification recommendation level and the performance recommendation level respectively; Step SF2: Compare the qualification recommendation level with the performance recommendation level; Step SF3: When the two are inconsistent, select the higher level as the preliminary comprehensive level; Step SF4: The preliminary comprehensive level is downgraded only when the violation downgrade rule or other preset constraints are applied, so as to generate the final comprehensive skill level.

5. The dynamic grading management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The dynamic update and early warning module is connected to the AI ​​skill assessment engine. The dynamic update and early warning module is equipped with a dynamic update and early warning method, which includes the following: Step SC1: Automatically trigger the AI ​​skill assessment engine to reassess all registered welders according to the preset cycle; Step SC2: Update the welder's skill level on a monthly basis and store the update results in the welder's skill file; Step SC3: Based on the preset warning threshold, proactively issue warnings for medical examination information, special operation certificates, and welding qualification validity periods that are about to expire; Step SC4: When it is detected that a welder's special operation certificate has expired or there is a serious QHSE violation, the qualification suspension rule or demotion rule will be automatically triggered, and the welder will be marked as demotion or his / her welding task assignment qualification will be suspended in the system.

6. The dynamic hierarchical management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The human-computer interaction and resource configuration interface module includes human-computer interaction methods, which are as follows: Step SD1: Visually display the welder skill level distribution, individual skill profiles, warning list, and grading criteria to the administrator; Step SD2: Receive project task input, which includes the required welding process, materials, location, and quality requirements; Step SD3: Based on the welder skill level and skill profile output by the AI ​​skill assessment engine, a matching algorithm is used to recommend a list of welders with matching or optimal skill levels for a specific welding task, thereby achieving dynamic matching between welders and projects.

7. The dynamic grading management system for welder skills based on AI technology analysis according to claim 3, characterized in that, The skill grading rule base module also defines the following grading logic rules: The quantitative scoring rules for the testing qualification section convert welder test scores, the number and difficulty coefficient of welding positions covered by the qualification, test performance in simulating complex working conditions, and the diversity of weldable materials into standardized qualification scores. The evaluation rules for welding performance set upper limits for monthly cumulative rework rates and benchmark ranges for work efficiency for different skill levels. Specifically, the monthly rework rate threshold for gold-level welders is set below the first preset value, the monthly rework rate threshold for silver-level welders is set below the second preset value, and the monthly rework rate threshold for bronze-level welders is set below the third preset value. The rules for downgrading due to violations should clearly stipulate that when the system identifies a welder committing a QHSE violation of a specific level during the assessment period, regardless of their current qualifications and performance scores, their overall skill level will be forcibly downgraded by one or more levels, and this downgrade status must continue for at least one full assessment period.

8. The dynamic grading management system for welder skills based on AI technology analysis according to claim 5, characterized in that, The rules for temporary suspension of qualifications or demotion are automatically executed by the dynamic update and early warning module, specifically including: When the system detects that a welder's special operation certificate has expired and has not been renewed within the grace period, it automatically marks the welder's skill level as an assistant and locks all of the welder's task assignment permissions involving welding and thermal cutting. When the system receives a confirmed serious QHSE violation record, it automatically suspends all welding qualifications of the welder in the current project and marks the welder's status as pending retraining and assessment until the required retraining is completed and the assessment is passed. After that, the corresponding qualification and level will be restored by the administrator or by the system based on the assessment results.

9. The dynamic grading management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The update cycle of the dynamic update and early warning module is monthly. The system is configured to automatically collect, analyze and re-evaluate all welder data from the previous month in the first week of each month, update the level information in the welder skill archive, generate a level change report and a list of welders who meet the standards for each level in the current month. The human-computer interaction and resource allocation interface module also includes a welder screening and comparison function, allowing project managers to input multi-dimensional screening criteria; The multi-dimensional screening criteria include the minimum skill level required, specific welding process qualifications, specific material welding experience, and historical rework rate requirements. Based on these criteria, the system dynamically filters and ranks the most suitable welder candidates from the currently available welder pool, while also providing a comparison view of key performance indicators among the candidates.

10. The dynamic grading management system for welder skills based on AI technology analysis according to claim 1, characterized in that, The dynamic skill grading management system for welders also includes a feedback loop learning mechanism, including the following: Step SE1: The output of the AI ​​skill assessment engine, namely the actual welding quality result after matching the welder's skill level with the project task, is fed back into the machine learning model as new training data. Step SE2: The system periodically retrains and optimizes the time-series prediction model and pattern classification model using new feedback data to adapt to the impact of different project environments, material properties, or process changes on the evaluation criteria for welder skill performance, thereby achieving continuous iteration and accuracy improvement of the grading model.

Images