A film thickness intelligent correction method and system based on multi-source data

Abstract

The application provides a film thickness intelligent correction method and system based on multi-source data, and is applied to the technical field of data processing. The application is developed around the intelligent correction of sprayed film thickness. First, multi-dimensional related data such as workpiece identification, spatial position, spraying process parameters and film thickness detection results are collected, and then the data is screened, matched and standard generated according to the data validity and modeling specification setting. Through integrated data processing, low confidence data is removed and a standardized film thickness deviation sample library is constructed, and the information in the library is sorted according to the contribution degree. Combined with process, computing power and other parameters, the parameter recommendation strategy is determined and the related configuration is set, the correlation rule between the sample and the spraying parameter is formed to form a training batch, and finally the parameter recommendation result is generated through multi-source data fusion modeling, which is adapted to the feedforward compensation strategy and applied to the subsequent spraying process, so that the film thickness is accurately controlled.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for intelligent film thickness correction based on multi-source data. Background Technology

[0002] Current film thickness testing methods mostly employ offline sampling or single-dimensional online detection, making the data susceptible to environmental interference and equipment fluctuations, resulting in a large amount of low-confidence data. Furthermore, there is a lack of mechanisms for validating the test data and correlating multi-source data. Key data such as workpiece identification, spatial location, and spraying parameters cannot be accurately linked to the film thickness test results, leading to data fragmentation and difficulty in supporting subsequent modeling and analysis. Traditional spraying parameter adjustments rely on operator experience, failing to establish a quantitative mapping relationship between film thickness deviation and workpiece characteristics and spraying parameters. This prevents dynamic parameter optimization based on workpiece structural differences and spatial changes. Moreover, most methods are reactive, adjusting parameters only after film thickness deviations are detected, resulting in a high rate of film thickness non-compliance in batches of workpieces and poor production efficiency and cost control.

[0003] The poor adaptability of data processing and modeling makes it difficult to meet the needs of industrial scenarios. Existing multi-source data processing technologies do not fully integrate with the characteristics of the spraying process and lack an integrated processing flow for spraying data. Standardization is lacking in steps such as low-confidence data removal, effective data correlation, and deviation sample extraction. The modeling process does not take into account industrial constraints such as the real-time nature of the production process and the upper limit of equipment computing power, resulting in low model iteration efficiency and poor adaptability of parameter recommendations to actual field conditions, making it difficult to apply to continuous spraying production. The compensation control strategy is too simplistic and cannot adapt to complex scenarios. Existing film thickness correction mostly adopts fixed compensation strategies, without considering variables such as the degree of film thickness deviation, differences in workpiece characteristics, and spraying process type, resulting in insufficient compensation accuracy. Furthermore, the lack of a feedforward compensation mechanism makes it impossible to predict deviations and optimize parameters in advance based on historical data and correlation patterns, making it difficult to achieve precise control of film thickness.

[0004] It should be noted that the information disclosed in the background section above is only used 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] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0006] According to one aspect of this application, a method for intelligent film thickness correction based on multi-source data is provided, comprising: acquiring multi-dimensional spraying correlation data, including workpiece identification information, workpiece spatial position data, spraying process parameters, and film thickness detection results; processing and modeling the multi-dimensional spraying correlation data, setting data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards based on film thickness detection data validity requirements and modeling accuracy specifications; performing integrated processing of the multi-dimensional spraying correlation data based on film thickness detection data validity requirements and modeling accuracy specifications, completing low-confidence data removal, effective data correlation modeling, and film thickness deviation sample extraction, constructing a standardized film thickness deviation sample library, reading workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information from the sample library, and sorting the data according to their contribution and correlation to film thickness deviation analysis; and based on the sorted sample data... The parameter recommendation generation strategy is determined by considering the complexity of workpiece coating features, the real-time requirements of production processes, and the upper limit of modeling computing power. Simultaneously, based on the characteristics of film thickness deviation samples, workpiece feature dimensions, and coating process characteristics, parameter recommendation weights, feature matching accuracy, and model iteration optimization frequency are set. Based on the determined parameter recommendation generation strategy and supporting parameters, the correlation between film thickness deviation samples and coating parameters is adaptively mined to form a training batch containing workpiece feature vectors, film thickness deviation labels, and coating parameter attributes. Related data of workpieces of the same type and batch are synchronously matched through an association mechanism. Using the matched training batch as algorithm input, a multi-source data fusion modeling algorithm learns the mapping relationship between film thickness deviation and workpiece features and coating parameters. The recommended coating parameters are dynamically weighted and fused according to workpiece identification, spatial location, and coating parameter features to generate recommended coating parameters. The recommended parameters and feedforward compensation control strategy are adapted based on the degree of film thickness deviation, workpiece feature differences, and coating process requirements, and the recommended parameters are applied to the feedforward compensation control of subsequent coating processes.

[0007] Another aspect of this application is a film thickness intelligent correction system based on multi-source data, the system being configured to execute the above-described film thickness intelligent correction method based on multi-source data by executing the executable instructions.

[0008] According to another aspect of this application, an electronic device includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-described intelligent film thickness correction method based on multi-source data by executing the executable instructions.

[0009] According to another aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a second processor, implements the above-described intelligent film thickness correction method based on multi-source data.

[0010] The beneficial effects of this application are as follows: First, multi-dimensional data such as workpiece identification, spatial location, spraying process parameters, and film thickness detection results are collected. Screening, matching, and generation standards are set according to validity and modeling specifications. Low-confidence data is eliminated through integrated processing, and a standardized film thickness deviation sample library is constructed and sorted. A parameter recommendation strategy is determined by considering process complexity, real-time requirements, and computational constraints. Core configurations such as weights and matching accuracy are set. The correlation between samples and spraying parameters is mined to form training batches. Through multi-source data fusion modeling, the mapping relationship is learned, and dynamically weighted, adaptive spraying parameter recommendation results are generated. These results are adapted to a feedforward compensation strategy and applied to subsequent spraying processes, achieving closed-loop film thickness control.

[0011] This application addresses data fragmentation by employing multi-source data correlation and validity screening to eliminate low-confidence data and construct a standardized sample library. This provides a high-quality data foundation for modeling and improves analytical accuracy. Based on correlation patterns and fusion modeling, it dynamically adapts parameters according to workpiece characteristics and spatial location, eliminating reliance on manual experience and resolving the blindness of traditional parameter adjustments. A feedforward compensation mechanism is adopted, combining the degree of deviation with process requirements to adjust compensation strategies in advance, correcting deviations, reducing the defect rate, and improving product quality consistency. The modeling process balances real-time production requirements with computational power limits, featuring standardized and modular workflows that can flexibly adapt to different painting scenarios, demonstrating strong feasibility and balancing production efficiency with cost control. Attached Figure Description

[0012] Figure 1 This document shows a flowchart illustrating a method for intelligent film thickness correction based on multi-source data, provided in an embodiment of this application.

[0013] Figure 2 A schematic diagram of a film thickness intelligent correction system based on multi-source data provided in an embodiment of this application is shown. Detailed Implementation

[0014] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0015] In one implementation, Figure 1 A schematic flowchart of a film thickness intelligent correction method based on multi-source data according to an embodiment of this application is shown.

[0016] S101, obtain multi-dimensional spraying correlation data.

[0017] In one implementation, the workpiece identification information serves as unique identification data for the sprayed workpiece and is the core association identifier for achieving multi-source data association modeling. It accurately matches the workpiece's spatial location, spraying parameters, and film thickness detection results, ensuring that various types of data in subsequent analysis correspond to specific workpieces and avoiding data confusion. The collected workpiece identification information must be unique and traceable, and can be collected according to the actual management needs of spraying production. It serves as the basic index for all subsequent data associations. In the association modeling stage, this information will act as the core primary key, binding the other three types of data to specific workpieces, and providing an identification basis for eliminating low-confidence detection data for individual workpieces and generating film thickness deviation samples for individual workpieces.

[0018] Workpiece spatial position data refers to the data related to the spatial geometric position of the workpiece during the spraying process. This is a key physical factor affecting the uniformity of the sprayed film thickness. The acquisition of this type of data must closely match the process layout of the spraying site to accurately reflect the actual position of the workpiece in the spraying station. Since the spatial position of the workpiece directly affects the coating effect of the spraying equipment, thus altering the film thickness detection results, this type of data is an important feature dimension for film thickness deviation analysis. In the subsequent correlation modeling stage, this data will be combined with spraying parameters and film thickness detection results to analyze the correlation between spraying parameters and film thickness under different spatial positions, providing spatial feature basis for generating personalized spraying parameter recommendations for workpieces in different positions.

[0019] Spraying process parameters are the core operating parameters of the spraying equipment during spraying operations. They directly determine the key process data for film thickness formation. The collection of this type of data needs to cover the core execution stages of spraying to accurately reflect the actual operating status of the spraying equipment. Spraying process parameters are the core control object for film thickness correction, and their correlation with film thickness detection results is the core basis for subsequent parameter recommendations. In the subsequent correlation modeling and pattern mining stages, this data will serve as a core independent variable, performing correlation analysis with film thickness deviation samples to uncover the film thickness variation patterns under different combinations of process parameters. Simultaneously, in the final parameter recommendation stage, this type of data is also the core carrier of the model's output recommendation results, providing specific process adjustment basis for feedforward compensation control.

[0020] The film thickness measurement result is the actual measured data of the coating thickness on the workpiece after spraying. It is measured using an online non-contact thickness measurement system and is the core basis for determining whether there is a deviation in film thickness. The acquisition of this type of data must ensure real-time and effective detection, and be synchronized with the workpiece spraying process. The acquired film thickness measurement results are online real-time detection data. Its detection principle is based on thermo-optical technology (ATO) and digital signal processing technology (DSP). The actual coating thickness is obtained by pulse heating of the coating, recording the temperature decay curve, and analyzing it using a patented algorithm. In subsequent stages, this data will serve as a core dependent variable. First, it is used to filter low-confidence detection data. Then, it is combined with workpiece identification, spatial location, and spraying parameters to construct film thickness deviation samples. It also serves as the core data foundation for the model to learn the mapping relationship between film thickness deviation and other features, providing direct numerical basis for judging the degree of film thickness deviation and generating feedforward compensation control strategies.

[0021] The collection of the above four types of data must ensure synchronization, completeness, and accuracy. After collection, the data will be used as the overall multi-dimensional spraying correlation data and input into the subsequent rule setting and data integration processing. All types of data are interconnected and support each other, forming a full-dimensional data system for intelligent film thickness correction, ensuring the scientificity and effectiveness of subsequent modeling analysis and parameter recommendations.

[0022] S102 processes and models multi-dimensional spraying correlation data. Based on the validity requirements of film thickness detection data and the modeling accuracy specifications, it sets data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards.

[0023] In one implementation, based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, the dimensions and correlation logic of setting data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards are integrated to clarify the core constraints and adaptation requirements of each indicator. Based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, the dimensions of the three types of indicators—data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards—are broken down, and the inherent correlation logic between each indicator is analyzed. Finally, the core constraints and actual adaptation requirements of each type of indicator are clarified to ensure that the indicator settings are highly consistent with the business logic of film thickness detection and correlation modeling.

[0024] To address the data confidence screening threshold, the decision dimensions corresponding to the validity of the test data and the accuracy of the modeling data are broken down. Regarding the multi-source data association and matching rules, the association and matching dimensions of four types of data—workpiece identification, spatial location, spraying parameters, and online film thickness results—are broken down. For the deviation sample generation standard, the effective data screening dimensions after removing low-confidence data, the film thickness deviation judgment dimension, and the association dimension between the sample and workpiece features are broken down. Clearly defining the data confidence screening threshold is a prerequisite for multi-source data association and matching; only effective film thickness test data that passes the confidence screening can be associated and matched with data from other dimensions. Effectively integrated data that conforms to the multi-source data association and matching rules is the basis for generating film thickness deviation samples, and the deviation sample generation standard needs to be adapted based on the settings of the first two indicators.

[0025] The core constraints are set around the authenticity of online film thickness detection and the accuracy of multi-source data association modeling. The data confidence screening threshold must be able to effectively identify low-confidence detection data, the multi-source data association matching rules must achieve accurate binding of four types of data with specific workpieces, and the deviation sample generation standard must be able to accurately extract valid samples that reflect the actual film thickness deviation. The adaptation requirements need to be set in combination with the actual detection situation at the spraying site and the actual data requirements for modeling, to ensure that the setting dimensions and association logic of various indicators not only comply with technical specifications, but can also be applied to the actual spraying film thickness modeling process.

[0026] Based on the quality control requirements for constructing film thickness deviation samples, this paper designs standards for indicator setting, clarifies the definition rules of core indicators and auxiliary constraints. Core indicators include data confidence screening thresholds and multi-source data association matching rules, while auxiliary constraints include the allowable range of modeling data validity deviation. Guided by the quality control requirements for film thickness deviation sample construction, and based on the indicator constraints and adaptation requirements defined in the previous step, specific setting standards are formulated for the three types of indicators. At the same time, the scope and judgment rules of core indicators and auxiliary constraints are clearly defined to ensure that the indicator system is clearly prioritized and the focus of quality control is highlighted, laying a standard foundation for the high-quality construction of film thickness deviation samples. Based on the quality requirements for film thickness deviation sample construction, quantitative judgment standards are designed for data confidence screening thresholds, standardized and procedural matching execution standards are designed for multi-source data association matching rules, and specific execution standards for film thickness deviation judgment, sample extraction, and sample annotation are designed for deviation sample generation standards.

[0027] The data confidence screening threshold and multi-source data association matching rules are clearly defined as core indicators because these two types of indicators directly determine the validity and relevance of the data entering the film thickness deviation sample construction stage. They are the core prerequisites for ensuring the quality of film thickness deviation samples. If the core indicators are set unreasonably, it will directly lead to the distortion of deviation samples, and subsequent association modeling and parameter recommendation will lose their data foundation. The definition rule for the data confidence screening threshold is "a quantitative judgment indicator that directly affects the validity of film thickness detection data," and the definition rule for the multi-source data association matching rules is "an execution indicator that directly determines whether multi-dimensional spraying data can be accurately associated."

[0028] The allowable range of deviation in the modeling data is defined as an auxiliary constraint. This constraint is a supplementary limitation to the setting and execution of the core indicators. It is used to regulate the error tolerance of the core indicators in practical applications. Its definition rule is "a reasonable tolerance range for slight data deviations generated during the execution of core indicators in order to ensure the accuracy of modeling". The auxiliary constraint must be compatible with the core indicators. It should not cause the modeling data to be distorted due to the excessive allowable deviation range, nor should it cause the actual execution difficulty to be too high due to the excessively small range, making it unsuitable for the actual data processing scenario on the spraying site.

[0029] To meet the stability requirements of multi-source data modeling in spraying, dynamic discrimination rules for confidence thresholds and index optimization rules adapted to data correlation are set to ensure that deviation samples meet the requirements of modeling and analysis. Dynamic optimization rules are designed for core indicators. By dynamically adjusting and adapting the data confidence threshold and multi-source data correlation matching rules, the indicators always closely match the actual data conditions at the spraying site, ensuring that the final generated film thickness deviation samples meet the requirements of subsequent correlation modeling and film thickness deviation analysis. Since the film thickness detection process at the spraying site is affected by equipment status, environmental factors, etc., the confidence characteristics of the detection data will change dynamically. Therefore, dynamic discrimination rules are set for the data confidence threshold. These rules can adjust and discriminate the confidence threshold in real time based on the actual film thickness detection data characteristics, equipment operating status, and environmental changes. This avoids low-confidence data not being eliminated or valid data being mistakenly screened due to fixed thresholds, ensuring that the film thickness detection data participating in subsequent processing always has high validity.

[0030] The correlation of multi-dimensional spraying data varies depending on the workpiece type, spraying process, and spatial location. Therefore, rules adapted to data correlation are set for multi-source data correlation matching. These rules can be specifically adjusted based on the actual correlation strength and characteristics of four types of data: workpiece identification, spatial location, spraying parameters, and online film thickness results. Priority is given to ensuring accurate matching of core data with strong correlation, while also considering the integrity of the overall data correlation, ensuring that the correlation results of multi-source data can truly reflect the actual correlation between various data and film thickness deviation. The ultimate goal of setting and implementing the two types of indicator optimization rules is to ensure the effectiveness, correlation, and authenticity of film thickness deviation samples, so that the generated deviation samples can accurately reflect the actual film thickness deviation during the spraying process, which is in line with the stability requirements of multi-source data modeling for spraying.

[0031] The system integrates and processes the indicator constraints, quality control requirements, and indicator optimization rules to generate basic data for membrane thickness modeling configuration, including indicator types, specifications, correlation logic, and optimization strategies. It comprehensively integrates all information defined in previous steps, including data confidence screening thresholds, multi-source data correlation and matching rules, the dimensions for setting deviation sample generation standards, correlation logic, core constraints, adaptation requirements, and specific setting standards; the definition scope and judgment rules for core indicators and auxiliary constraints; and the dynamic optimization rules, optimization objectives, and execution requirements for both types of indicators.

[0032] The integrated basic data must fully include four core elements: indicator type (data confidence screening threshold, multi-source data association matching rules, deviation sample generation standards), setting specifications (specific quantitative standards and implementation standards for various indicators), association logic (the inherent connection relationship between each indicator), and optimization strategy (confidence threshold dynamic identification strategy, association matching rule data association adaptation strategy). These elements are interconnected and support each other to form a complete configuration data system.

[0033] The basic data for film thickness modeling will serve as the unified execution basis for subsequent low-confidence data removal, effective data correlation modeling, and film thickness deviation sample extraction. This will ensure that all subsequent data processing steps have clear and standardized operating procedures, guaranteeing that the processing results of each step are highly consistent with the requirements for the validity of film thickness detection data, the accuracy of modeling, and the stability requirements for multi-source spraying data modeling. Ultimately, this will ensure the overall execution effect and scientific validity of the intelligent film thickness correction method.

[0034] S103, based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, sets data confidence screening thresholds, multi-source data association matching rules, and deviation sample generation standards. It performs integrated processing on multi-dimensional spraying correlation data, completes the removal of low-confidence data, effective data association modeling, and film thickness deviation sample extraction, constructs a standardized film thickness deviation sample library, reads the workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information from the sample library, and sorts them according to the contribution and correlation of the data to the film thickness deviation analysis.

[0035] In one implementation, combining multi-source data fusion processing logic with film thickness modeling data screening requirements, a data validity assessment algorithm and a multi-dimensional data association verification mechanism are introduced to modularly decompose and functionally define the spraying-related data processing steps, generating parameter-marked spraying data processing nodes. Based on both multi-source data fusion processing logic and film thickness modeling data screening requirements, and employing two core technologies—a data validity assessment algorithm and a multi-dimensional data association verification mechanism—the entire processing of spraying-related data is modularly decomposed. After clarifying the core functional positioning of each module, parameter-marked spraying data processing nodes are generated, laying the foundation for the standardization and process-oriented implementation of subsequent data processing.

[0036] A data validity assessment algorithm is introduced, primarily to address the need for removing low-confidence detection data. This algorithm serves as the core criterion for determining the validity of film thickness detection data, directly supporting the screening of film thickness modeling data. A multi-dimensional data association verification mechanism is also introduced, primarily to address the need for association modeling of workpiece identification, spatial location, spraying parameters, and online film thickness results. This mechanism serves as the verification standard for accurate association of these four types of data, ensuring the accuracy of the association modeling. Based on the entire process of spraying association data processing and the application requirements of the aforementioned algorithms and mechanisms, the overall processing is broken down into several functional modules, including a low-confidence data assessment module, a multi-source data association verification module, a valid data integration module, and a film thickness deviation sample initial screening module. Each module undertakes a single and clearly defined processing task, avoiding processing chaos caused by overlapping steps.

[0037] The core functions of each module after decomposition are clearly defined. For example, the low-confidence data identification module is responsible for determining the confidence level of film thickness detection data and marking data to be removed based on the data validity identification algorithm. The multi-source data association verification module is responsible for verifying the association of four types of data and marking abnormal association data based on the association verification mechanism. Based on the functional positioning, exclusive processing parameters are configured for each module, generating spraying data processing nodes with parameter markings. Each node corresponds to a modular processing link, and the parameter markings clarify the processing standards, execution thresholds, and adaptation range of the node to ensure the standardization of node processing.

[0038] By combining deterministic processing rules with a modeling data coupling implementation strategy, the parameter-marked spraying data processing nodes are processed. The processing steps and algorithm functions are matched to generate preliminary spraying data processing infrastructure. The deterministic processing rules establish unified execution criteria for spraying-related data processing, clarifying the processing order of each node, data input / output specifications, and abnormal data handling methods to ensure the standardization of the data processing process. The modeling data coupling implementation strategy clarifies how the processed data from each node is coupled with the film thickness modeling requirements, ensuring that all processing steps revolve around the film thickness modeling data screening requirements and avoiding a disconnect between processing results and modeling needs. Based on the deterministic processing rules, the parameter-marked spraying data processing nodes are sorted, clarifying the sequential execution logic between nodes. For example, after processing the low-confidence data identification node, valid data is input to the multi-source data association verification node. Simultaneously, based on the modeling data coupling implementation strategy, modeling data coupling requirements are configured for each node. For instance, the processing results of the association verification node must meet the requirement of precise coupling between four types of data and workpiece features, adapting to the subsequent formation of film thickness deviation samples.

[0039] Each modular processing step is precisely matched with the specific functions of the data validity identification algorithm and the multi-dimensional data association verification mechanism. For example, the low-confidence data identification step is matched with the confidence judgment function of the data validity identification algorithm, and the multi-source data association verification step is matched with the association verification function of the multi-dimensional data association verification mechanism, ensuring that the functions of the algorithm and mechanism are fully implemented in the corresponding steps. Based on the node sorting and function matching results, the parameter marking, processing logic, and input / output requirements of each node are integrated to generate preliminary spraying data processing framework data. This data provides the basic process and parameter basis for the subsequent data processing framework construction.

[0040] Based on the requirements of the film thickness modeling data processing workflow, algorithm function information, and data processing setup specifications, the preliminary spraying data processing setup data is systematically organized to complete the construction of a basic framework for integrated processing of multi-dimensional correlated spraying data. The film thickness modeling data processing workflow requirements clearly define the core processing objectives that the framework must adapt to: low-confidence data removal, effective data correlation modeling, and film thickness deviation sample extraction. All workflow designs within the framework must revolve around these objectives. The algorithm function information clarifies the functional boundaries and application scenarios of the data validity identification algorithm and the multi-dimensional data correlation verification mechanism within the framework, ensuring deep integration of algorithm functions with the framework workflow. The data processing setup specifications generation rules clearly define the framework's structural design, node connections, parameter configuration, and other specifications, ensuring the framework's scientific validity and feasibility.

[0041] Based on the above three criteria, the preliminary data was organized in multiple dimensions, including sorting out the parameter markings of each processing node to ensure that the parameters are compatible with the screening requirements of film thickness modeling data; verifying the execution logic between nodes to ensure that the process from data input to the initial screening of deviation samples is smooth and uninterrupted; matching the node processing functions with the algorithm functions to ensure that the algorithm is applied accurately and without deviation within the framework; and supplementing the abnormal fault tolerance configuration of data processing to ensure that the framework can cope with various abnormal data situations at the spraying site.

[0042] Based on the systematically organized data, a well-structured, logically coherent, and fully functional integrated framework for processing multi-dimensional related data in spraying is established. The framework starts with "multi-dimensional spraying related data input" and ends with "preliminary extraction of film thickness deviation samples." It includes core process layers such as low-confidence data identification, multi-source data correlation verification, effective data correlation modeling, and deviation sample extraction. Each process layer corresponds to a processing node with pre-defined parameters and functional matching. Algorithm interfaces and parameter adjustment entry points are also reserved to ensure the framework's flexibility and adaptability, enabling integrated processing of four types of data: workpiece identification, spatial location, spraying parameters, and online film thickness results.

[0043] Based on the correlation and logical relationship of the spraying data processing steps, core data processing nodes are automatically generated. Anomalies are flagged and structural adjustments are made based on a node algorithm adaptability detection mechanism. Based on the inherent correlation of the entire spraying data processing process, key steps that play a decisive role in the final data quality are accurately identified from the modular processing steps and transformed into core data processing nodes. This ensures that the core nodes can directly control the core quality indicators of data processing. The identification of core nodes uses "impact of data processing effect" as the core judgment criterion, focusing on three core objectives: data validity, relevance, and model adaptability. Among these steps, the core step of determining the confidence level of film thickness detection data is the "first hurdle" for data quality, directly determining whether subsequent data has modeling value, and requiring accurate identification of low-confidence data; the core step of accurately associating four types of multi-source data is to achieve "data-workpiece" binding, ensuring a one-to-one correspondence between workpiece identification, spatial location, spraying parameters, and film thickness detection results; the core step of effective data association modeling is the key to transforming discrete data into usable modeling data, requiring the completion of structured integration of data and construction of association logic; and the core step of accurately extracting film thickness deviation samples is to generate modeling training data, requiring the accurate selection of effective samples that can reflect the film thickness deviation pattern.

[0044] After transforming these key steps into core nodes for data processing, each core node has a clear functional positioning and processing standards: the confidence level judgment node is responsible for outputting data confidence level judgment results based on the data validity identification algorithm; the multi-source data accurate association node is responsible for completing the binding of four types of data based on the multi-dimensional data association verification mechanism; the effective data modeling node is responsible for converting the associated effective data into structured modeling data; and the deviation sample extraction node is responsible for screening and outputting membrane thickness deviation samples according to the set standards. These core nodes, as the core units of the data processing framework, directly determine the accuracy of low-confidence data removal (error tolerance ≤ ±3%), the accuracy of multi-source data association modeling (association accuracy ≥ 98%), and the effectiveness of membrane thickness deviation sample extraction (effective sample ratio ≥ 90%).

[0045] A node algorithm compatibility testing mechanism is introduced, conducting full-process compatibility testing across four dimensions: functional matching, parameter matching, efficiency matching, and integration matching between core nodes and supporting algorithms. This ensures that the algorithms can accurately support the implementation of core node functions. The core verification checks whether the core functions of the supporting algorithms can meet the processing requirements of the core nodes. For example, for the core node for determining the confidence level of film thickness detection data, it checks whether the data validity identification algorithm has the function of quantitatively determining confidence based on key indicators such as the fitting degree of the temperature decay curve and the difference in heat storage coefficient, and whether it can accurately output three levels of determination results: "high confidence," "medium confidence," and "low confidence." For the node for accurate association of four types of multi-source data, it checks whether the multi-dimensional data association verification mechanism has the function of multi-data primary key association based on workpiece identifiers, and whether it can handle association logic under abnormal scenarios such as missing data and inconsistent formats.

[0046] The key focus is on verifying the consistency between the algorithm execution parameters and the parameter labels of the core nodes. The parameter labels of the core nodes clearly define key configurations such as processing thresholds and accuracy requirements. For example, the confidence core judgment node sets a parameter label of "high confidence data fit ≥ 0.9," and it is necessary to check whether the confidence judgment threshold of the data validity screening algorithm is consistent with this label. The multi-source data accurate association node sets a time synchronization parameter of "association deviation allowable range ≤ ±5ms," and it is necessary to check whether the time synchronization threshold of the multi-dimensional data association verification mechanism meets this requirement to avoid distortion of processing results due to parameter mismatch. The focus is also on the adaptability of algorithm processing efficiency to the real-time requirements of spraying production. Based on the real-time requirements of the production process, the core nodes must meet specific processing efficiency standards, such as a single workpiece data processing time ≤ 1 second. The detection mechanism needs to quantify and statistically analyze the processing time of the supporting algorithms, such as the time for the data validity screening algorithm to process a single film thickness detection data, and the time for the multi-dimensional data association verification mechanism to complete the association of four types of data for a single workpiece, ensuring that the algorithm processing efficiency can support the real-time requirements of the core nodes and avoiding disruption to the production line rhythm due to processing delays. The algorithm connections between core nodes should be checked for smoothness to avoid functional conflicts or logical gaps. For example, can the high-confidence data output by the core confidence determination node be directly read and processed by the algorithm of the multi-source data accurate association node? Can the structured data output by the effective data association modeling node be seamlessly integrated into the algorithm screening logic of the membrane thickness deviation sample extraction node? This ensures that the data format and output definition are consistent between algorithms, without any gaps in connection.

[0047] Through a node algorithm adaptability detection mechanism, various abnormal matching situations are accurately marked. Combined with film thickness modeling data screening requirements and actual spraying production, targeted adjustments are made to achieve a high degree of adaptability between core nodes and algorithms. Detected anomalies are categorized, marked, and their root causes analyzed. For example, when marking an anomaly as "algorithm execution threshold and core node parameter marking mismatch," it is determined whether the algorithm's default threshold is not synchronized with the updated parameters of the core node, or whether the algorithm's threshold adjustment range cannot cover the core node parameter requirements. When marking an anomaly as "algorithm processing efficiency does not meet real-time requirements," it is analyzed whether the algorithm's computational complexity is too high, or whether lightweight optimization for industrial scenarios has not been performed. When marking an anomaly as "disconnection in algorithm connection between core nodes," it is located whether the data format is incompatible, or whether the correlation logic is inconsistent.

[0048] Adjust the parameter labels of the core nodes or optimize the algorithm execution parameters. If the confidence threshold of the data validity screening algorithm does not match the "high confidence fit ≥ 0.9" set by the core node, optimize the algorithm threshold calculation logic to adapt it to the parameter requirements of the core node; or fine-tune the parameter thresholds within the allowable range of the core node parameter labels to ensure compatibility with the algorithm function. Perform lightweight optimization of the algorithm, such as simplifying the complex calculation steps of the data validity screening algorithm while retaining the core judgment indicators; or adopt parallel computing logic to improve the processing speed of the multi-source data association verification mechanism and ensure that the processing time of a single task is controlled within the real-time requirements.

[0049] Unify the algorithm data interfaces and association logic between core nodes. For example, standardize the output data format of the confidence core judgment node and the multi-source data association node, and unify data field definitions and encoding rules; reconstruct the algorithm connection logic between core nodes, and add a data format conversion module to ensure seamless data flow. Supplement algorithm function modules or adjust the functional positioning of core nodes. If the multi-dimensional data association verification mechanism lacks data missing scenario completion function, an interpolation completion module based on historical data can be added; if the data validity identification algorithm cannot cover confidence judgment under special working conditions, the core algorithm logic can be optimized, and working condition identification and corresponding judgment rules can be added.

[0050] Following the pre-defined data processing setup rules, multi-source data processing and parsing algorithms are integrated with the data processing framework and core processing nodes. This achieves algorithmic coupling for low-confidence data removal, effective data association modeling, and film thickness deviation sample extraction. A standardized film thickness deviation sample library is constructed, and workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information are read from the sample library. The data is then sorted according to its contribution and correlation to the film thickness deviation analysis. Following the pre-defined data processing setup rules, multi-source data processing and parsing algorithms (including data validity identification algorithms, multi-dimensional data association verification mechanisms, and deviation sample extraction algorithms) are fully integrated into the algorithm interface of the integrated processing framework. Precise associations are established with ordinary processing nodes and core data processing nodes within the framework. Dedicated algorithm processing logic and execution parameters are configured for each node, achieving deep integration of algorithms with the framework and nodes, ensuring efficient and accurate implementation within the framework.

[0051] By integrating algorithms, frameworks, and nodes, this system achieves algorithmic coupling for three core functions: low-confidence data removal, effective data association modeling, and film thickness deviation sample extraction. Specifically, a unified algorithmic system and framework process enable the coordinated execution of these three functions: First, a data validity screening algorithm accurately removes low-confidence film thickness detection data, retaining only valid data. Second, a multi-dimensional data association verification mechanism precisely associates valid film thickness detection data with workpiece identification, spatial location, and spraying parameters, completing association modeling of valid data and meeting association modeling requirements. Finally, a deviation sample extraction algorithm extracts film thickness deviation samples from the effective data after association modeling, forming a preliminary film thickness deviation sample set. The extracted film thickness deviation sample set after algorithm coupling undergoes standardization processing, including unified sample labeling (labeling core information such as workpiece identification, spatial location, spraying parameters, and film thickness deviation values), sample quality grading (classifying sample quality levels based on data association and validity), and structured sample storage. Ultimately, a standardized film thickness deviation sample library is constructed, ensuring the standardization, structure, and traceability of data within the library.

[0052] Core data such as workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information are accurately retrieved from a standardized film thickness deviation sample library. Based on the actual needs of film thickness deviation analysis, the contribution and correlation of various data to the analysis are comprehensively ranked. For example, data with high contribution and strong correlation to film thickness deviation analysis, such as spraying parameters and film thickness deviation values, are ranked first, while auxiliary data such as sample quality assessment information are ranked later. This ensures that core data can be obtained first when formulating parameter recommendation strategies and mining correlation patterns, thereby improving the processing efficiency and accuracy of subsequent steps.

[0053] S104, based on the sorted sample data, determines the parameter recommendation generation strategy by combining the complexity of workpiece spraying features, the real-time requirements of production process and the upper limit of modeling computing power. At the same time, it sets the parameter recommendation weight, feature matching accuracy and model iteration optimization frequency according to the film thickness deviation sample features, workpiece feature dimensions and spraying process characteristics.

[0054] In one implementation, multi-dimensional constraint analysis technology is used to integrate and process the process adaptation requirements, generating workpiece coating feature complexity assessment results, production process real-time requirement indicators, and modeling computing power upper limit threshold data. First, the full-dimensional requirements for coating process adaptation are systematically integrated, breaking down fragmented requirement statements and breaking them down into three independent yet comprehensive core dimensions based on "core factors influencing parameter recommendations": workpiece characteristics, production execution, and modeling implementation. Each dimension undergoes in-depth analysis focusing on "whether parameter recommendations can be accurately implemented and whether they are suitable for the actual scenario," completing the transformation from qualitative requirements to quantitative data through quantitative evaluation and threshold definition.

[0055] In the workpiece coating dimension, the focus is on the impact of the workpiece's own characteristics on parameter recommendations. Key analyses include workpiece structural complexity (e.g., presence of irregular structures, hollow areas, etc.), the number of coating surfaces (single-sided coating, multi-sided coating, etc.), and coating requirements (coating material, target thickness range, uniformity standards, etc.). By quantitatively scoring these characteristics (e.g., structural complexity rated on a 1-5 scale) and classifying them into levels (simple, medium, complex), a workpiece coating feature complexity assessment result is generated, intuitively reflecting the differentiated requirements of different workpieces for coating parameters. In the production process dimension, focusing on the execution efficiency and rhythm requirements of the coating production line, key indicators such as coating cycle time (number of workpieces coated per unit time), workpiece turnover speed (workpiece dwell time at the coating station, turnover interval), and parameter adjustment response efficiency (production line's response speed to changes in coating parameters) are analyzed. These indicators are then transformed into real-time requirement indicators such as parameter recommendation generation speed (e.g., requiring the generation of recommended parameters for a single workpiece within 1 second) and update frequency (e.g., updating the recommended parameter model every 10 workpieces), ensuring that parameter recommendations keep pace with the production line and do not affect production efficiency.

[0056] In terms of computing power for modeling, we consider the hardware support capabilities for parameter recommendation modeling, and analyze hardware indicators such as computing speed (the amount of data that can be processed per second), data processing capacity (the scale of modeling data that can be carried in a single run), and model training efficiency (the time to complete a model training run). By setting thresholds, we generate computing power limit thresholds such as the upper limit of modeling data volume (e.g., a maximum of 100,000 sample data points can be used in a single modeling run), the iteration speed threshold (e.g., a single model iteration should not exceed 5 minutes), and the upper limit of computational complexity (e.g., avoid using deep learning networks with more than 10 layers). This ensures that parameter recommendation modeling can be efficiently implemented under existing hardware conditions and avoids modeling failure or recommendation delays due to insufficient computing power.

[0057] The parameter recommendation generation requirements are integrated with process constraint standards, sample feature association rules, and process characteristic adaptation requirements to establish a precise matching relationship between strategy formulation and multi-dimensional constraints, generating an adaptive parameter recommendation generation strategy. The four core elements of this integration each have a clear role: the parameter recommendation generation requirements are the core guideline, generating spraying parameter recommendations that can accurately correct film thickness deviations based on film thickness deviation samples and workpiece characteristics; process constraint standards are the hard boundaries, namely, the three types of constraint data generated in the preceding steps: workpiece spraying feature complexity, production process real-time performance, and modeling computing power limits; sample feature association rules are the data basis, namely, the inherent correlation between film thickness deviation sample features, workpiece characteristics, and spraying parameters (such as the positive correlation between film thickness deviation and spraying pressure for a specific workpiece structure); and process characteristic adaptation requirements are the scenario adaptation basis, namely, the specific requirements for parameter recommendations based on the characteristic differences of different spraying processes (such as powder coating and paint coating) (such as paint atomization requirements and drying speed).

[0058] The evaluation results of the complexity of workpiece spraying features are matched with the sample feature association rules and process characteristic adaptation requirements. For example, workpieces with complex structures correspond to more refined sample feature association rules, and the painting process corresponds to parameter recommendation logic adapted to the atomization characteristics of the coating. The real-time requirements of the production process are matched with the requirements of parameter recommendation generation speed and update frequency. For example, high-cycle production lines correspond to faster parameter recommendation generation speed. The upper limit threshold data of modeling computing power is matched with the computational complexity of parameter recommendation modeling. For example, low computing power equipment corresponds to lightweight modeling logic to avoid complex calculations.

[0059] Based on the above matching relationship, an adaptive parameter recommendation generation strategy is generated: This strategy clarifies the principles of differentiated parameter recommendation (generating exclusive recommended parameters for workpieces of different complexity and different process types), rapid generation (prioritizing parameter recommendation speed to adapt to the real-time requirements of the production line), and lightweight modeling (controlling the computational complexity of modeling to adapt to the upper limit of computing power). Simultaneously, the strategy's applicable scope (e.g., applicable to multi-variety, medium-cycle production lines in small and medium-sized painting enterprises), update mechanism (e.g., updating the strategy quarterly based on new sample data), and adjustment logic (e.g., prioritizing adjustments to parameter recommendation generation speed-related configurations when the production line cycle time changes) are defined to ensure the strategy can flexibly adapt to different scenarios and changing needs.

[0060] Based on the parameter recommendation generation strategy, core configuration requirements are extracted. Using film thickness deviation sample features as the input dimension, workpiece features as the core parameters, and spraying process characteristics as the judgment basis, the scientific setting of parameter recommendation weights, feature matching accuracy, and model iteration optimization frequency is achieved. First, core configuration requirements are extracted from the adaptability parameter recommendation generation strategy, focusing on "how to achieve the strategy goal through parameter configuration." The setting direction and constraints of three core parameters—parameter recommendation weights, feature matching accuracy, and model iteration optimization frequency—are clarified. For example, the strategy differentiation principle requires that weight allocation reflect the differences in feature importance, and the rapid generation principle requires that the model iteration optimization frequency not be too high, affecting efficiency. The parameter recommendation weights are set using film thickness deviation sample features as the input dimension and workpiece features as the core parameters. The influence of different sample features and workpiece features on film thickness deviation is analyzed, and weights are allocated accordingly. For example, sample features corresponding to core spraying parameters such as spraying pressure and spraying distance have a higher impact on film thickness deviation and are given a weight of 30%-40%; workpiece features such as workpiece structural complexity and number of sprayed surfaces are given a weight of 20%-30%; other auxiliary features (such as ambient temperature) are given a lower weight (below 10%) to ensure that the weight allocation is tilted towards features that have a significant impact on film thickness deviation.

[0061] The feature matching accuracy is set based on the characteristics of the spraying process, combined with the complexity of the workpiece features and process requirements. This involves setting a matching accuracy threshold and error tolerance between the film thickness deviation sample features and the workpiece features. For example, for complex workpiece structures or processes requiring high film thickness uniformity (such as precision instrument casing spraying), a higher matching accuracy threshold (e.g., feature matching similarity ≥ 95%) is set, with an error tolerance controlled within ±1%. For simple workpiece structures or ordinary processes (such as general hardware spraying), the matching accuracy threshold can be appropriately lowered (e.g., feature matching similarity ≥ 85%), with an error tolerance relaxed to ±3%, balancing matching accuracy and processing efficiency.

[0062] The frequency of model iteration optimization needs to be set in conjunction with process stability and parameter adjustment frequency, and by setting iteration time periods or data volume trigger thresholds and trigger conditions. For example, for production lines with high process stability and low parameter adjustment frequency (such as large-scale production of a single product), a longer iteration time period (such as iterating the model once a month) should be set; for production lines with low process stability and frequent parameter adjustments (such as small-batch production of multiple products), a shorter iteration time period (such as iterating once a week) should be set, or a data volume trigger threshold should be set (such as triggering iteration when 5,000 new sample data are added) to ensure that the model can adapt to process changes and data updates in a timely manner and maintain recommendation accuracy.

[0063] Based on the recommended adaptability strength and process execution priority requirements, the core configuration parameters are quantitatively calibrated and optimized. The adaptability weights, feature matching ranges, and iteration update cycles of various parameters are clarified to generate parameter configuration standards that meet the requirements of the spraying process. The core basis for adjustment focuses on "adaptability" and "priority": the recommended adaptability strength requirement means that the core configuration parameters must be highly compatible with the workpiece spraying characteristics, production process requirements, and modeling computing power conditions to avoid parameters being out of sync with the scenario; the process execution priority requirement means that, based on the core demands of the spraying production site, the priority of different process indicators is clarified (e.g., some production lines prioritize ensuring film thickness correction accuracy, while others prioritize ensuring production efficiency), ensuring that parameter adjustments are tilted towards core demands.

[0064] In the calibration of parameter recommendation weights, the weight ratio of each feature dimension is adjusted according to the process execution priority. For example, if the production line prioritizes film thickness correction accuracy, the weight of features related to spraying parameters is further increased (e.g., from 30% to 40%). If production efficiency is prioritized, the weight of easily identifiable features in the workpiece features is appropriately increased (e.g., the weight of the number of sprayed surfaces is increased from 20% to 25%), while the weight ratio of complex features is reduced to speed up weight calculation. In defining the range of feature matching accuracy, the upper and lower limits of feature matching accuracy are adjusted according to the workpiece complexity and adaptation requirements for different scenarios. For example, for ultra-complex structural workpieces (such as aerospace parts), the upper limit of matching accuracy is increased to 98%, and the lower limit is set to 90%; for simple structural workpieces, the upper limit of accuracy is set to 90%, and the lower limit is reduced to 80%. At the same time, the accuracy adaptation range of different process types is clearly defined (e.g., the accuracy range of powder coating process is 85%-95%, and the accuracy range of painting process is 90%-98%) to ensure that the accuracy requirements are adapted to the scenario.

[0065] In optimizing the frequency of model iterations, the iteration threshold and update magnitude are adjusted according to the real-time requirements and stability of the production process. For example, if a high-cycle production line requires absolutely real-time parameter recommendations, the iteration cycle can be optimized from weekly to bi-weekly, but the update magnitude of a single iteration can be increased (e.g., the model parameter adjustment ratio is increased from 5% to 10%). If the process stability is poor, the data volume trigger threshold can be reduced from 5000 to 3000, while controlling the update magnitude of a single iteration (not exceeding 8%) to avoid fluctuations in recommended parameters due to frequent and large iterations. Finally, the optimized core configuration parameters are standardized and solidified, clarifying the specific values ​​of various parameters (e.g., weight allocation ratio, accuracy threshold, iteration cycle), applicable scope (e.g., applicable to which workpieces, processes, and production lines), adjustment boundaries (e.g., the maximum adjustable range of parameters), and execution specifications (e.g., the approval process and effective time for parameter adjustments). This forms a standardized parameter configuration standard, providing a unified and clear execution basis for subsequent parameter recommendation modeling and correlation pattern mining, ensuring the standardization of the parameter recommendation process and the consistency of the results.

[0066] S105, based on the determined parameters, recommends generation strategies and supporting parameters, adaptively mines the correlation between film thickness deviation samples and spraying parameters, and forms a training batch containing workpiece feature vectors, film thickness deviation labels and spraying parameter attributes. The correlation data of workpieces of the same type and batch are synchronously matched through the correlation mechanism.

[0067] In one implementation, the correlation characteristics between the parameter recommendation generation strategy and film thickness deviation samples and spraying parameters are analyzed through pattern mining and feature extraction to generate sample deviation features and spraying parameter correlation features. Combining the differentiated and precise recommendation requirements in the parameter recommendation generation strategy, the correlation characteristics between the parameter recommendation generation strategy and film thickness deviation samples and spraying parameters are mined. The focus is on analyzing the distribution patterns of film thickness deviation samples under different parameter recommendation weights and feature matching accuracies (e.g., the concentration range of film thickness deviation under specific weight configurations), and the impact of spraying parameter adjustments on film thickness deviation (e.g., the positive and negative correlation between spraying pressure changes and film thickness deviation), ensuring that the mining direction is highly aligned with the parameter recommendation objectives.

[0068] Based on the correlation characteristics discovered, two types of core features are extracted. First, sample deviation features are extracted, focusing on the key attributes of film thickness deviation samples, including the magnitude of the deviation, the type of deviation distribution, and the correspondence between the deviation and workpiece features, intuitively reflecting the core characteristic dimensions of film thickness deviation. Second, spraying parameter correlation features are extracted, focusing on the correlation logic between spraying parameters and film thickness deviation, extracting features such as the value range, adjustment range, and combination mode of core parameters like spraying pressure, spraying distance, and spraying speed, clarifying the influence path of spraying parameters on film thickness deviation. Both types of feature extraction serve subsequent correlation modeling, ensuring that the features possess strong discriminative power and modeling adaptability.

[0069] The sample deviation features and spraying parameter correlation features are integrated and analyzed for dimensionality to generate a multi-dimensional film thickness parameter correlation feature set. The sample deviation features and spraying parameter correlation features are integrated from both data and business logic dimensions. In the data dimension, the quantification standards, value ranges, and data formats of the features are unified to ensure comparability between features. In the business logic dimension, features of corresponding dimensions are bound and integrated according to the correlation chain of "workpiece features - film thickness deviation - spraying parameters" (e.g., binding sample deviation features corresponding to a specific workpiece structure with suitable spraying parameter correlation features), forming a logically coherent feature combination. Correlation analysis and causal inference methods are used to deeply analyze the correlation between the integrated features. The correlation strength (e.g., strong, moderate, weak) between sample deviation features and spraying parameter correlation features is verified, and core correlation feature pairs are identified (e.g., a strong correlation between excessive film thickness deviation and insufficient spraying pressure). At the same time, redundant features with no or weak correlation are eliminated to avoid feature redundancy affecting modeling efficiency.

[0070] Based on the integration and analysis results, a multi-dimensional film thickness parameter correlation feature set is generated. This feature set is based on "workpiece features, film thickness deviation features, and spraying parameter correlation features as the control direction," covering multi-dimensional and strongly correlated feature combinations. It not only retains the core information of various features but also reflects the inherent business logic between features, providing comprehensive and accurate feature input for subsequent model construction.

[0071] Based on a multi-dimensional film thickness parameter association feature set, a strategy adaptation model is used to specifically construct sample mining rules, parameter association logic, and feature matching mechanisms, generating standardized training batches containing workpiece feature vectors, film thickness deviation labels, and spraying parameter attributes. The model adopts a modular design with parameter recommendation generation strategies as the adaptation target, comprising three core modules: a sample mining rule module, a parameter association logic module, and a feature matching mechanism module. The sample mining rule module defines the selection criteria for effective training samples (e.g., retaining only samples corresponding to feature combinations with a strong correlation greater than a set threshold); the parameter association logic module solidifies the correlation between film thickness deviation and spraying parameters (e.g., clarifying the spraying parameter adjustment direction corresponding to different deviation ranges); and the feature matching mechanism module achieves accurate matching of workpiece features, deviation features, and spraying parameter features (e.g., grouping and matching the three types of features according to the workpiece feature dimension). These three modules are interconnected; the model's input is the multi-dimensional film thickness parameter association feature set, and the output is the feature processing result that meets the modeling requirements.

[0072] Through strategy adaptation model processing, standardized training batches containing core elements are generated. Each sample in the training batch includes a complete workpiece feature vector (such as quantitative vectors of workpiece structural complexity and number of sprayed surfaces), film thickness deviation labels (such as explicit labels such as "deviation exceeds standard" and "deviation is acceptable"), and spraying parameter attributes (such as specific values ​​and combination modes of spraying parameters), ensuring that each training sample has a complete input-output correspondence. Simultaneously, the training batches are standardized, unifying the sample format and dividing the sample ratio (such as the ratio of training set to validation set), ensuring that the training batches have good modeling adaptability and can be directly used in the subsequent multi-source data fusion modeling training process.

[0073] The standardized training batch and the associated data of similar workpieces in the same batch are synchronously matched through an association mechanism to complete the full-process data integration and matching for mining the correlation patterns between film thickness deviation samples and spraying parameters. A workpiece-based association mechanism is adopted, using the workpiece identifier as the core association key, to accurately match the samples in the standardized training batch with the associated data (including workpiece spatial location data, complete spraying process parameter records, and full-process film thickness detection data) of similar workpieces in the same batch. This association mechanism ensures that each sample in the training batch corresponds to specific workpiece production data, achieving a deep binding between training data and actual production scenarios.

[0074] The matching operation is performed according to the principle of "grouping by type and aligning by batch". First, workpieces are grouped by type (e.g., workpieces with the same structural complexity are grouped together). Then, within the same type, data is aligned by production batch to ensure that the matched training data reflects the common patterns and individual differences of workpieces of the same type and batch. During the matching process, the completeness and consistency of the data are simultaneously verified, and abnormal samples with missing data or logical contradictions are removed to ensure the reliability of the matching results.

[0075] After matching, integrated data with both modeling and practical application value is formed, realizing full-process data adaptation for mining the correlation between film thickness deviation samples and spraying parameters. This integrated data includes both the core elements of standardized training samples and complete correlation data from actual production, providing a training foundation closer to actual production for subsequent multi-source data fusion modeling, ensuring that the model training results can effectively adapt to actual spraying production scenarios.

[0076] S106 uses the completed training batch as the algorithm input. Through a multi-source data fusion modeling algorithm, it learns the mapping relationship between film thickness deviation and workpiece features and spraying parameters. It dynamically weights and fuses the workpiece identification, spatial position, and spraying parameter features to generate recommended spraying parameters. It then adapts the recommended parameters and feedforward compensation control strategy by combining the degree of film thickness deviation, workpiece feature differences, and spraying process requirements. Finally, it applies the recommended parameters to the feedforward compensation control of the subsequent spraying process.

[0077] In one implementation, the learning logic, feature mapping dimension, and weight allocation rules of the multi-source data fusion modeling algorithm are trained and patterns are extracted to generate a film thickness deviation feature vector, a workpiece feature association dimension, and a coating parameter mapping rule, forming the basic information for parameter recommendation modeling. The algorithm's three core elements—learning logic, feature mapping dimension, and weight allocation rules—are clearly defined. The learning logic revolves around the causal relationship between "film thickness deviation - workpiece features - coating parameters," with "minimizing film thickness deviation" as the optimization objective. The feature mapping dimension covers the numerical characteristics of film thickness deviation, the structural and coating characteristics of the workpiece, and the process characteristics of the coating parameters, ensuring that the mapping dimension comprehensively covers the core influencing factors. The weight allocation rule prioritizes core process parameters and key workpiece features based on the degree of influence of each feature on film thickness deviation.

[0078] Using standardized training batches as input data, the process follows a three-step workflow: data preprocessing, model training, and pattern extraction. In the data preprocessing stage, the workpiece feature vectors, film thickness deviation labels, and spraying parameter attributes in the training batch are normalized and standardized to ensure a consistent data format. In the model training stage, an iterative training mode is adopted, improving the model's fitting accuracy of the mapping relationship between film thickness deviation and feature variables by adjusting weight allocation rules and optimizing feature mapping logic. In the pattern extraction stage, based on the trained model, core information such as the quantitative correlation patterns between film thickness deviation and various features, and the sensitivity patterns of spraying parameter adjustments are extracted.

[0079] Through training and pattern extraction, three types of core foundational information are generated. The film thickness deviation feature vector is a quantitative expression of key attributes of film thickness deviation, such as vector data for dimensions like deviation magnitude, distribution type, and rate of change. The workpiece feature association dimension clarifies the correlation strength and influence between different workpiece features (structural complexity, number of sprayed surfaces, etc.) and film thickness deviation. The spraying parameter mapping rules define the quantitative mapping relationship between spraying parameters (pressure, distance, speed, etc.) and film thickness deviation, such as the change in film thickness deviation corresponding to a 1 MPa change in spraying pressure. These three types of information together form the basis of parameter recommendation modeling, ensuring that the modeling process has clear rules to follow.

[0080] The algorithm is designed and quantified to define the weighted fusion logic, dynamic adjustment range, and recommendation accuracy target for workpiece identification, spatial location, and spraying parameter features. This generates feature weighting coefficients, parameter adjustment boundaries, and recommendation result achievement thresholds, forming constraint information for parameter recommendation generation. Focusing on three core dimensions—workpiece identification, spatial location, and spraying parameter features—a dynamic weighted fusion logic is designed to ensure that recommended parameters can adapt to the personalized needs of different workpieces and locations. The fusion logic uses workpiece identification as the primary key, incorporating the influence of spatial location on spraying effect and the process priority of spraying parameter features into the weighted calculation, achieving precise recommendations of "one parameter per workpiece."

[0081] Key constraints related to the fusion logic are quantitatively defined. Feature weighting coefficients are set according to spatial importance and process priority of spraying parameters. For example, spatial features corresponding to key spraying areas of the workpiece are assigned higher weighting coefficients, and feature coefficients corresponding to core spraying parameters (such as pressure) are higher than those of auxiliary parameters. Parameter adjustment boundaries are set based on the physical operating limits of the spraying equipment and process safety requirements, clearly defining the upper and lower limits of adjustment for each spraying parameter (e.g., the spraying pressure adjustment range is 0.3-0.8MPa). The recommended result attainment threshold defines the accuracy requirements of parameter recommendations. For example, the film thickness deviation after application of recommended parameters must be controlled within ±2μm to ensure the practicality of the recommended results. The above quantitatively defined constraints are integrated into standardized parameter recommendation generation constraint information, clarifying the weighting rules, adjustment range, and accuracy targets of parameter recommendations, defining clear execution boundaries for the parameter recommendation process, and avoiding recommendations that exceed equipment capabilities or process requirements.

[0082] The process involves demand decomposition and process formulation for the adaptation analysis dimensions, strategy matching rules, and compensation control logic of film thickness deviation, workpiece feature differences, and spraying process requirements. This generates deviation adaptation judgment conditions, process matching benchmarks, and feedforward compensation execution specifications, forming compensation strategy adaptation execution information. The three core adaptation dimensions—film thickness deviation, workpiece feature differences, and spraying process requirements—are decomposed into three levels: slight deviation, moderate deviation, and severe deviation, with clear judgment criteria for each level. Workpiece feature differences are further decomposed into three categories—simple, medium, and complex—based on structural complexity and coating requirements, defining the core feature indicators for each category. Spraying process requirements are decomposed into different process scenarios based on spraying type (paint, powder coating), coating material, etc., clarifying the core control objectives for each scenario.

[0083] Based on the requirements breakdown, two core elements are defined. The strategy matching rules clarify the compensation strategies corresponding to different combinations of adaptation dimensions. For example, a severe deviation combined with a complex workpiece and a painting process corresponds to a compensation strategy of "significantly adjusting core parameters + slightly optimizing auxiliary parameters." The feedforward compensation control process is formulated according to the logic of "deviation judgment - strategy matching - parameter output - execution feedback," ensuring the compensation process is streamlined and standardized. For instance, the deviation level is first determined through film thickness detection results, then the corresponding compensation strategy is matched, recommended parameters are output, and the execution effect is fed back. The requirements breakdown results, matching rules, and control process are integrated to generate three types of core execution information. The deviation adaptation judgment criteria are quantitative judgment standards for different deviation levels; the process matching benchmark clarifies the compensation strategy benchmark templates corresponding to each process scenario; and the feedforward compensation execution specification defines the execution steps, responsibility nodes, and feedback mechanisms of the compensation process. These three types of information together constitute the compensation strategy adaptation execution information, ensuring that feedforward compensation control has clear execution basis and process specifications.

[0084] This system integrates basic information for parameter recommendation modeling, constraint information for parameter recommendation generation, and execution information for compensation strategy adaptation. It performs end-to-end collaborative processing on feature mapping of film thickness deviation, weighted recommendation of spraying parameters, and strategy adaptation of feedforward compensation. This generates spraying parameter recommendations that are both adaptable and accurate, which are then applied to feedforward compensation control in subsequent spraying processes. An integration strategy of "dimensional alignment - logical association - redundancy removal" is employed to systematically fuse these three types of information. In the dimensional alignment stage, the data formats and quantification standards of various information types are unified to ensure the relevance of information. In the logical association stage, based on business logic, the mapping rules in the basic modeling information are established to correlate the adjustment boundaries in the constraint information and the matching rules in the compensation information. For example, the spraying parameter mapping rules are bound to the parameter adjustment boundaries to ensure that the mapping results are within a reasonable adjustment range. In the redundancy removal stage, duplicate and conflicting information is deleted, while core valid data is retained, improving the efficiency of collaborative processing.

[0085] The optimization process revolves around three core stages: film thickness deviation feature mapping, weighted recommendation of spraying parameters, and adaptation of feedforward compensation strategies. The feature mapping stage, based on modeling information, transforms the actual film thickness deviation into a feature vector and matches it with the corresponding workpiece feature association dimensions and spraying parameter mapping rules. The weighted recommendation stage combines weighting coefficients and adjustment boundaries from the constraint information to dynamically weight and calculate the spraying parameters, generating initial recommendation results. The compensation adaptation stage, based on compensation execution information, matches corresponding compensation strategies according to deviation level, workpiece type, and process scenario, optimizing and adjusting the initial recommendation results to ensure the results adapt to feedforward compensation requirements.

[0086] Collaborative processing generates recommended spraying parameters that are both adaptable and accurate. These parameters include specific values, adjustment ranges, and execution priorities. The recommended results are synchronized to the industrial control system of the spraying equipment via a data interface. The control system, following feedforward compensation specifications, automatically adjusts equipment operating parameters during subsequent spraying processes to proactively correct film thickness deviations. Simultaneously, a feedback mechanism for execution effectiveness is established, synchronizing subsequent film thickness detection results back to the modeling stage for dynamic model optimization, forming a closed-loop management system of "recommendation-execution-feedback-optimization."

[0087] In one implementation, such as Figure 2 As shown, this application also provides a film thickness intelligent correction system based on multi-source data, comprising:

[0088] The multi-dimensional spraying correlation data acquisition module 201 is used to acquire multi-dimensional spraying correlation data throughout the entire spraying process, including workpiece identification information, workpiece spatial position data, spraying process parameters, and film thickness detection results.

[0089] The spraying data processing rule setting module 202 is used to process and model the multi-dimensional spraying correlation data. Based on the validity requirements of film thickness detection data and the modeling accuracy specifications, it sets the data confidence screening threshold, multi-source data correlation matching rules and deviation sample generation standards.

[0090] The standardized coating deviation sample library construction module 203 is used to set data confidence screening thresholds, multi-source data association matching rules and deviation sample generation standards based on the validity requirements of film thickness detection data and modeling accuracy specifications. It performs integrated processing on multi-dimensional coating correlation data, completes low-confidence data removal, effective data association modeling, film thickness deviation sample extraction, and constructs a standardized film thickness deviation sample library. It reads the workpiece feature dimensions, film thickness deviation values, coating parameter types and sample quality assessment information of the sample library, and sorts them according to the contribution and correlation of the data to the film thickness deviation analysis.

[0091] The spraying parameter recommendation strategy formulation module 204 is used to determine the parameter recommendation generation strategy based on the sorted sample data, combined with the complexity of workpiece spraying features, the real-time requirements of production process and the upper limit of modeling computing power. At the same time, it sets the parameter recommendation weight, feature matching accuracy and model iteration optimization frequency according to the film thickness deviation sample features, workpiece feature dimensions and spraying process characteristics.

[0092] The spraying association pattern mining and training batch generation module 205 is used to recommend generation strategies and supporting parameters based on the determined parameters, adaptively mine the association patterns between film thickness deviation samples and spraying parameters according to the strategy, and form a training batch containing workpiece feature vectors, film thickness deviation labels and spraying parameter attributes to realize synchronous matching and association of workpiece association data of the same type and batch.

[0093] The spraying parameter recommendation and feedforward compensation module 206 is used to take the matched training batch as the algorithm input, learn the mapping relationship between film thickness deviation and workpiece characteristics and spraying parameters through multi-source data fusion modeling algorithm, and generate spraying parameter recommendation results by dynamically weighting and fusing workpiece identification, spatial position and spraying parameter characteristics. The recommended parameters and feedforward compensation control strategy are adapted by combining the degree of film thickness deviation, workpiece characteristic differences and spraying process requirements, and the parameter recommendation results are applied to the feedforward compensation control of the subsequent spraying process.

[0094] The computer-readable storage medium provided in the above embodiments of this application and the film thickness intelligent correction method based on multi-source data provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0095] The various embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for evaluating the film thickness intelligent correction method, electronic device, electronic device, and readable storage medium based on multi-source data are basically similar to the embodiments of the film thickness intelligent correction method based on multi-source data described above, and therefore are described relatively simply. Relevant parts can be referred to in the descriptions of the embodiments of the film thickness intelligent correction method based on multi-source data described above.

Claims

1. A method for intelligent film thickness correction based on multi-source data, characterized in that, include: Acquire multi-dimensional spraying-related data, including workpiece identification information, workpiece spatial location data, spraying process parameters, and film thickness detection results; Multi-dimensional spraying correlation data is processed and modeled for analysis. Based on the validity requirements of film thickness detection data and the modeling accuracy specifications, data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards are set. Based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, we set data confidence screening thresholds, multi-source data association matching rules, and deviation sample generation standards. We then integrated the processing of multi-dimensional spraying correlation data, completed the removal of low-confidence data, effective data association modeling, and film thickness deviation sample extraction, and constructed a standardized film thickness deviation sample library. We read the workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information from the sample library and sorted them according to the contribution and correlation of the data to the film thickness deviation analysis. Based on the sorted sample data, the parameter recommendation generation strategy is determined by combining the complexity of workpiece spraying features, the real-time requirements of production process and the upper limit of modeling computing power. At the same time, the parameter recommendation weight, feature matching accuracy and model iteration optimization frequency are set according to the characteristics of film thickness deviation samples, workpiece feature dimensions and spraying process characteristics. Based on the established parameter recommendation generation strategy and supporting parameters, the correlation between film thickness deviation samples and spraying parameters is adaptively mined to form a training batch containing workpiece feature vectors, film thickness deviation labels, and spraying parameter attributes. Correlation data of workpieces of the same type and batch are synchronously matched through an association mechanism. This includes pattern mining and feature extraction of the correlation characteristics between the parameter recommendation generation strategy and film thickness deviation samples and spraying parameters, generating sample deviation features and spraying parameter correlation features; dimensional integration and correlation analysis of the sample deviation features and spraying parameter correlation features are performed to generate a multi-dimensional film thickness parameter correlation feature set; based on the multi-dimensional film thickness parameter correlation feature set, a strategy adaptation model is used to specifically construct sample mining rules, parameter correlation logic, and feature matching mechanisms, generating a standardized training batch containing workpiece feature vectors, film thickness deviation labels, and spraying parameter attributes; the standardized training batch is synchronously matched with the correlation data of workpieces of the same type and batch through an association mechanism, completing the entire process of data integration and matching for mining the correlation patterns between film thickness deviation samples and spraying parameters. Using the matched training batches as algorithm input, a multi-source data fusion modeling algorithm learns the mapping relationship between film thickness deviation and workpiece features and spraying parameters. It dynamically weights and fuses these parameters according to workpiece identification, spatial location, and spraying parameter features to generate recommended spraying parameters. The recommended parameters and feedforward compensation control strategies are then adapted based on the degree of film thickness deviation, workpiece feature differences, and spraying process requirements. The recommended parameters are applied to the feedforward compensation control of subsequent spraying processes. This includes model training and pattern extraction of the learning logic, feature mapping dimensions, and weight allocation rules of the multi-source data fusion modeling algorithm, generating film thickness deviation feature vectors, workpiece feature association dimensions, and spraying parameter mapping rules to form the basic information for parameter recommendation modeling. Furthermore, the algorithm is designed and quantified to define the weighted fusion logic, dynamic adjustment range, and recommendation accuracy target for workpiece identification, spatial location, and spraying parameter features, generating feature weighting coefficients, parameter adjustment boundaries, and recommendation result achievement thresholds to form parameter recommendation generation constraint information. The system decomposes and formulates requirements for film thickness deviation, workpiece feature differences, and coating process requirements, including adaptation analysis dimensions, strategy matching rules, and compensation control logic. This generates deviation adaptation judgment conditions, process matching benchmarks, and feedforward compensation execution specifications, forming compensation strategy adaptation execution information. It integrates parameter recommendation modeling basic information, parameter recommendation generation constraint information, and compensation strategy adaptation execution information, and performs full-process collaborative processing on film thickness deviation feature mapping, weighted recommendation of coating parameters, and feedforward compensation strategy adaptation. This generates coating parameter recommendation results that are both adaptable and accurate, and applies them to feedforward compensation control in subsequent coating processes.

2. The method as described in claim 1, characterized in that, Multi-dimensional spraying correlation data is processed and modeled for analysis. Based on the validity requirements of film thickness detection data and the accuracy specifications of modeling, data confidence screening thresholds, multi-source data correlation matching rules, and deviation sample generation standards are set, including: Based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, the dimensions and correlation logic of the data confidence screening threshold, multi-source data association matching rules, and deviation sample generation standards are integrated to clarify the core constraints and adaptation requirements of each indicator. Based on the quality control requirements constructed from film thickness deviation samples, the standard for setting indicators was designed, and the definition rules for core indicators and auxiliary constraints were clarified. Among them, the core indicators include the data confidence screening threshold and the multi-source data association matching rules, and the auxiliary constraints include the allowable range of modeling data validity deviation. In accordance with the stability requirements of multi-source data modeling for spraying, we set up confidence thresholds for dynamic identification and correlation matching rules to optimize the indicators based on data correlation, ensuring that deviation samples meet the requirements for modeling and analysis. The constraints, quality control requirements, and optimization rules of the indicators are integrated and processed to generate basic data for membrane thickness modeling configuration, which includes indicator types, setting specifications, correlation logic, and optimization strategies.

3. The method as described in claim 1, characterized in that, Based on the requirements for the validity of film thickness detection data and the specifications for modeling accuracy, data confidence screening thresholds, multi-source data association matching rules, and deviation sample generation standards are set. Multi-dimensional spraying correlation data is processed in an integrated manner, completing the removal of low-confidence data, effective data association modeling, and extraction of film thickness deviation samples. A standardized film thickness deviation sample library is constructed. The workpiece feature dimensions, film thickness deviation values, spraying parameter types, and sample quality assessment information of the sample library are read. The data are ranked according to their contribution and correlation to the film thickness deviation analysis, including: Combining the multi-source data fusion processing logic with the film thickness modeling data screening execution requirements, a data validity identification algorithm and a multi-dimensional data association verification mechanism are introduced to modularly decompose and functionally locate the spraying associated data processing link, generating parameter-marked spraying data processing nodes; By combining deterministic processing rules with modeling data coupling implementation strategies, the parameter-marked spraying data processing nodes are processed, the processing steps are matched with the algorithm functions, and preliminary spraying data processing setup data is generated. Based on the requirements of film thickness modeling data processing process, algorithm function information and data processing construction specifications generation rules, the preliminary spraying data processing construction data is systematically organized to complete the construction of the basic framework for integrated processing of multi-dimensional related data of spraying. Based on the correlation of spraying data processing links and the logical relationship of data processing structure, core data processing nodes are automatically generated, and abnormal matches are marked and structural adjustments are made based on the node algorithm adaptability detection mechanism. According to the preset data processing construction mode generation rules, the multi-source data processing and parsing algorithm is associated and integrated with the data processing basic framework and core processing nodes to realize the algorithm coupling of low confidence data removal, effective data association modeling and film thickness deviation sample extraction. A standardized film thickness deviation sample library is constructed, and the workpiece feature dimensions, film thickness deviation values, spraying parameter types and sample quality assessment information in the sample library are read. The data are sorted according to their contribution and correlation to the film thickness deviation analysis.

4. The method as described in claim 1, characterized in that, Based on the sorted sample data, a parameter recommendation generation strategy is determined by considering the complexity of workpiece coating features, the real-time requirements of the production process, and the upper limit of modeling computing power. Simultaneously, based on the characteristics of film thickness deviation samples, the dimensionality of workpiece features, and the characteristics of the coating process, parameter recommendation weights, feature matching accuracy, and model iteration optimization frequency are set, including: Multi-dimensional constraint analysis technology is used to integrate and process process adaptation requirements, and generate workpiece spraying feature complexity assessment results, production process real-time requirement indicators, and modeling computing power upper limit threshold data. The parameter recommendation generation requirements are integrated with process constraint standards, sample feature association rules and process characteristic adaptation requirements to establish a precise matching relationship between strategy formulation and multi-dimensional constraints, and generate an adaptive parameter recommendation generation strategy. Based on the parameter recommendation generation strategy, the core configuration requirements are extracted. The film thickness deviation sample features are used as the input dimension, the workpiece feature dimension is used as the core parameter, and the spraying process characteristics are used as the judgment basis. This enables the scientific setting of parameter recommendation weight, feature matching accuracy and model iteration optimization frequency. Based on the recommended adaptation strength and process execution priority requirements, the set core configuration parameters are quantitatively calibrated and optimized, the adaptation weight, feature matching range and iteration update cycle of various parameters are clarified, and parameter configuration standards that meet the requirements of the spraying process are generated.

5. A film thickness intelligent correction system based on multi-source data, characterized in that, The system is configured to execute the film thickness intelligent correction method based on multi-source data as described in any one of claims 1 to 4 by executing executable instructions.

6. An electronic device, characterized in that, include: First processor; and memory for storing executable instructions of the first processor; The first processor is configured to execute the film thickness intelligent correction method based on multi-source data according to any one of claims 1 to 4 by executing the executable instructions.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the second processor, it implements the intelligent film thickness correction method based on multi-source data as described in any one of claims 1 to 4.

Images