Pearlescent material production data management system and method based on data integration analysis

Abstract

The invention relates to the technical field of data management, in particular to a pearlescent material production data management system and method based on data integration analysis. The method comprises the following steps: obtaining a process original data set of a pearlescent material production line, and carrying out standardization processing to obtain a standardized process data set; obtaining a quality detection data set of the pearlescent material; performing time interval division on the standardized process data set and the quality detection data set, and performing batch number marking to obtain a process-quality batch data pair; and based on the process-quality batch data pair, designing a modeling sample data set, and constructing a quality process coupling model. According to the invention, the intelligent, data-driven and high-quality stable control of the pearlescent material production process is comprehensively realized by constructing a closed-loop system integrating data integration, modeling prediction and regulation and control optimization.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to a data management system and method for pearlescent material production based on data integration and analysis. Background Technology

[0002] Traditional pearlescent material production often employs intermittent wet deposition processes. While the technology is mature, it suffers from low coating efficiency, uneven particle size control, and significant quality fluctuations. In recent years, with the widespread application of fluidized bed reactors and vapor deposition technology in materials engineering, pearlescent material production is gradually shifting towards continuous, large-scale, and intelligent processes. This is particularly evident in the preparation of gold-colored and silver-white / iridescent series products, where the fluidized bed-calcination coupling system has become the mainstream approach. Currently, existing technologies still have the following shortcomings: First, process and quality data sources are scattered and lack standardized formats, resulting in a lack of a standardized integration framework. Second, there is a time lag and sample misalignment between process and testing data, making it difficult to establish accurate correspondences. Third, most existing prediction models are static regressions, lacking real-time response capabilities and unable to support closed-loop optimization control. Fourth, there is a lack of anomaly identification mechanisms and dynamic adjustment path generation schemes for large-scale production scenarios, making it difficult to guarantee production efficiency and product consistency. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a pearlescent material production data management system and method based on data integration and analysis to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a data management method for pearlescent material production based on data integration and analysis includes the following steps:

[0005] Step S1: Obtain the original process dataset of the pearlescent material production line and perform standardization processing to obtain a standardized process dataset; obtain the quality inspection dataset of the pearlescent material; divide the standardized process dataset and the quality inspection dataset into time intervals and mark them with batch numbers to obtain process-quality batch data pairs;

[0006] Step S2: Based on the process-quality batch data, construct a quality-process coupling model for the design modeling sample dataset; extract the mapping relationship between process control and quality characterization in the quality-process coupling model to obtain the parameters of the quality-process coupling model, where the quality characterization indicators include surface gloss, particle size distribution uniformity and coating layer defect density;

[0007] Step S3: Collect the process parameters of the current production batch in real time, and perform quality index prediction analysis of the process parameters based on the parameters of the quality-process coupling model to obtain the prediction results;

[0008] Step S4: Perform deviation analysis based on the prediction results and preset quality control standards, identify abnormal parameters in the current process operation based on the deviation analysis results, formulate process optimization and adjustment plans, and obtain dynamic process adjustment data.

[0009] This invention enables the unified and standardized integration of distributed process parameters and discrete quality inspection information, establishing a structured data chain between process and quality, and improving data utilization efficiency and traceability accuracy. By introducing time interval division and batch marking mechanisms, it effectively solves the problem of data temporal misalignment in the production process, ensuring an accurate mapping relationship between the process and corresponding inspection results. By constructing the coupling relationship between process control parameters and quality characterization indicators, it realizes the extraction of quality evolution laws under the influence of complex multi-parameters, providing quantitative support for online quality prediction and process control. Based on real-time process data, it performs feedforward prediction of quality indicators, allowing for the prediction of deviation trends before product delivery. This approach proactively identifies potential quality issues, shortens feedback cycles, and reduces rework rates. Through deviation identification and anomaly tracking mechanisms, it quickly pinpoints key process nodes causing quality deviations, supporting cause diagnosis and responsibility parameter identification under multi-variable conditions. Combining sensitivity analysis and process boundary constraints, it generates process optimization and adjustment paths, ensuring that adjustment schemes are implemented within equipment capacity and process safety limits, thus improving the stability and accuracy of adjustments. The overall methodology supports dynamic closed-loop control logic throughout the entire process, addressing issues such as strong predictive staticity, control lag, and one-sided adjustment in existing technologies, effectively improving the consistency, stability, and process transparency of pearlescent material products under large-scale continuous production.

[0010] Preferably, the present invention also provides a pearlescent material production data management system based on data integration analysis, used to execute the above-described pearlescent material production data management method based on data integration analysis, wherein the pearlescent material production data management system based on data integration analysis includes:

[0011] The data integration module is used to acquire the original process dataset of the pearlescent material production line, and perform standardization processing to obtain a standardized process dataset; acquire the quality inspection dataset of the pearlescent material; divide the standardized process dataset and the quality inspection dataset into time intervals and mark them with batch numbers to obtain process-quality batch data pairs.

[0012] The model building module is used to construct a quality-process coupling model based on the design modeling sample dataset using process-quality batch data; it extracts the mapping relationship between process control and quality characterization in the quality-process coupling model to obtain the parameters of the quality-process coupling model, where the quality characterization indicators include surface gloss, particle size distribution uniformity and coating layer defect density.

[0013] The intelligent prediction module is used to collect the process parameters of the current production batch in real time, and perform quality index prediction analysis of the process parameters based on the parameters of the quality-process coupling model to obtain the prediction results.

[0014] The control and optimization module is used to perform deviation analysis based on the prediction results and the preset quality control standards, identify abnormal parameters in the current process operation based on the deviation analysis results, formulate process optimization and adjustment schemes, and obtain dynamic process adjustment data.

[0015] This invention comprehensively addresses the quality instability and efficiency bottlenecks caused by fragmented data, disjointed processes, and delayed response in traditional pearlescent material production. By achieving integrated acquisition, standardized formatting, and time alignment of multi-source process and quality data, it breaks down information silos, significantly improving data integrity and analyzability, and providing a high-quality sample foundation for subsequent modeling. Through the construction of a coupling mechanism between process parameters and quality indicators, it enables precise description of quality evolution trends in multivariate processes, helping to clarify key influencing factors and enhancing the interpretability and transparency of the production process. By real-time simulation of the target batch's quality performance during production, it allows for early detection of potential deviations, providing prior information for operational decisions and reducing the risk of batch defects. By correlating predicted deviations with process control variables, it promptly identifies unstable segments and outputs precise adjustment suggestions, transforming process control from experience-driven to data-driven. The system as a whole possesses dynamism, closed-loop capability, and adaptability, automatically switching characteristic parameter combinations according to different product systems. While maintaining quality consistency, it enhances adaptability to large-scale continuous production environments, providing key technical support for intelligent manufacturing and high-end product quality control in the pearlescent material industry. Attached Figure Description

[0016] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0017] Figure 1 This is a flowchart illustrating the steps of a pearlescent material production data management method based on data integration and analysis according to the present invention.

[0018] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0019] Figure 3 for Figure 1 A detailed flowchart of step S2;

[0020] Figure 4 This is a schematic diagram of the equipment used in the production process of pearlescent materials. Detailed Implementation

[0021] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0023] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a method for managing pearlescent material production data based on data integration and analysis, the method comprising the following steps:

[0025] Step S1: Obtain the original process dataset of the pearlescent material production line and perform standardization processing to obtain a standardized process dataset; obtain the quality inspection dataset of the pearlescent material; divide the standardized process dataset and the quality inspection dataset into time intervals and mark them with batch numbers to obtain process-quality batch data pairs;

[0026] Step S2: Based on the process-quality batch data, construct a quality-process coupling model for the design modeling sample dataset; extract the mapping relationship between process control and quality characterization in the quality-process coupling model to obtain the parameters of the quality-process coupling model, where the quality characterization indicators include surface gloss, particle size distribution uniformity and coating layer defect density;

[0027] Step S3: Collect the process parameters of the current production batch in real time, and perform quality index prediction analysis of the process parameters based on the parameters of the quality-process coupling model to obtain the prediction results;

[0028] Step S4: Perform deviation analysis based on the prediction results and preset quality control standards, identify abnormal parameters in the current process operation based on the deviation analysis results, formulate process optimization and adjustment plans, and obtain dynamic process adjustment data.

[0029] Preferably, step S1 includes the following steps:

[0030] Step S11: Configure edge acquisition terminals on the pearlescent material production line and collect raw process datasets in real time;

[0031] Step S12: Decode the communication protocol, unify the sampling frequency, and align the timestamps of the original process dataset to form a standardized process dataset;

[0032] Step S13: Obtain titanium dioxide-coated pearlescent material and perform quality inspection to obtain a quality inspection dataset;

[0033] Step S14: Divide the standardized process dataset and quality inspection dataset into time intervals, and establish sliding window and time mapping rules to obtain the material production path interval for each sample;

[0034] Step S15: Assign a unique batch number to the material production path range of each sample, identify and merge the process parameters and corresponding quality inspection results belonging to the same range, and generate process-quality batch data pairs.

[0035] In the embodiments of this invention, please refer to Figure 4First, WRM-200 edge acquisition terminals are installed and configured at key equipment nodes in the continuous pearlescent material production line. These terminals integrate sensor interfaces for temperature, pressure, flow, current, and voltage, and support the Modbus-TCP communication protocol. They are used to collect real-time data on the operating status of process equipment such as silos, cyclone preheaters, fluidized bed reactors, calcination systems, heat exchangers, and exhaust gas treatment devices, forming raw process datasets. The acquisition frequency is set to 1Hz. After acquisition, the local data synchronization module at the edge terminal performs communication protocol decoding, converting the Modbus format data into a unified key-value structure. Timestamp reconstruction and sampling frequency standardization are performed on each data channel, generating a standardized process dataset. The structure fields include acquisition time, equipment identifier, parameter type, parameter value, and unit. All data is standardized to ISO. The data was formatted in 8601 time format and accurate to milliseconds. Samples were then selected from the production batch of titanium dioxide-coated pearlescent material every 30 minutes. The D10, D50, and D90 particle size distributions were measured using a ZNS-820 laser particle size analyzer. The titanium dioxide coating rate was determined by titration. The gloss was measured using a BYK456 gloss meter, and the L, a, and b chromaticity values ​​were obtained using a tristimulus value measuring instrument. This formed a quality inspection dataset. All test data were recorded in JSON format, labeled with the test time and sample batch number. The test time corresponding to each data point in the quality inspection dataset was used as a reference benchmark, based on the average process residence time of the pearlescent material from feed to discharge (this process condition). (The time window interval is 21 minutes) In the standardized process dataset, a time window interval is constructed with a sliding step of 10 minutes. Linear interpolation is used to fit the boundary data to calculate the time interval between the raw material entry point and the discharge point of the corresponding sample, and the material production path interval is defined. After the path interval is defined, a unique batch number is assigned to each test sample, and all standardized process parameters within the time period are combined and integrated with the corresponding quality test value of the sample and written into a structured data table. The output is a process-quality batch data pair, and the fields include batch number, temperature sequence, gas flow rate sequence, residence time, coating rate, particle size distribution value, and gloss and color value.

[0036] This invention enables precise correlation between the process operation status and quality inspection results in pearlescent material production, constructing a data foundation system to support modeling and analysis. By deploying edge acquisition terminals and performing high-frequency, low-latency data acquisition of key process parameters across multiple device nodes, it ensures real-time and complete recording of production process information, significantly improving data coverage and continuity. Through unified communication protocol parsing and timestamp standardization, it resolves data structure differences caused by multiple devices and protocols, improving data consistency and system integration efficiency. By combining quality inspection results with time interval mapping and material path tracking, it accurately reconstructs the complete process history of each tested sample in the production line, eliminating the disconnect between traditional inspection data and process data. By introducing a sliding window mechanism and time mapping rules, it enhances the robustness of batch data matching and improves the accuracy of data pairing in situations with reaction lag and mixed-flow processes. Based on this, it establishes one-to-one data pairs between process and quality, providing accurate and high-quality input data for subsequent process rule extraction, process modeling, quality prediction, and control strategy formulation, achieving a closed-loop connection from data perception to information integration, effectively supporting the transformation of pearlescent material production from experience-driven to data-driven.

[0037] Preferably, step S13 includes the following steps:

[0038] Step S132: Locate the edge of the coating layer based on the reflected light intensity distribution data, extract the gradient of the transition zone of the coating layer, and generate the contour data of the coating layer.

[0039] Step S133: Perform sub-pixel-level segmentation based on the coating layer contour data, calculate the local thickness matrix, and obtain a heat map of the coating layer thickness distribution;

[0040] Step S134: Calculate the effective coverage area ratio based on the heat map of coating thickness distribution to obtain the coating rate data;

[0041] Step S135: Scan the surface of the coating layer of titanium dioxide-coated pearlescent material, construct the specular reflection intensity curve, and calculate the pearlescent effect consistency data;

[0042] Step S136: Identify low-reflection areas based on pearlescent effect consistency data, and locate defect boundaries by combining coating layer contour data, generating a defect distribution density map;

[0043] Step S137: Perform a weighted score based on the coverage rate data, defect distribution density, and pearlescent effect consistency, and output the quality inspection dataset.

[0044] In this embodiment of the invention, a titanium dioxide-coated pearlescent material sample is fixed on a three-dimensional precision displacement platform. The light source is a D65 standard LED ring array with an incident angle of 45°. A CCD industrial camera is positioned at 0° for reception, with a pixel resolution of 2448×2048, an exposure time of 2ms, and a gain of 0dB. After the acquired grayscale image is denoised by a two-dimensional Gaussian filter kernel pixel, a grayscale profile is taken every 0.5° radially. The gradient magnitude is calculated using the Sobel operator, and a gradient threshold T_g = 25 grayscale levels is set. Values ​​exceeding the threshold are... The locations are marked as edge points, and discrete edge points are connected into a continuous coating layer contour curve using quadratic curve fitting with sub-pixel precision. A grid is generated within the contour with a step size of 0.1 pixels. For each grid node, the transition range from light to dark grayscale is searched along the normal direction. Linear interpolation is used to obtain the sub-pixel position corresponding to the median grayscale value, denoted as the interface position. The distance between the interface and the contour is the local thickness. The thickness matrix size is consistent with the original image, with a thickness range of 0-500 nm. A heatmap of the coating layer thickness distribution is obtained by coloring at 10 nm equal intervals. Figure 2 The value threshold T_c = 30nm is used. Pixels above the threshold are considered effective coverage. The coverage rate is obtained by dividing the number of effective pixels by the total number of pixels. The sample is then placed in a coaxial illumination common-path reflectometer with a halogen lamp as the light source. The spectral range is 380-780nm, the step size is 2nm, and the detector rotation range is -80° to +80°. A specular reflection spectrum is sampled every 0.5°. The intensity at a single angle is obtained by integration, and a specular reflection intensity curve is plotted. The coefficient of variation (CV) of the intensity in the 20°-60° interval is calculated. CV < 5% is marked as a consistent region, and CV ≥ 5% is marked as a low-reflection region. Boolean intersection is performed between the low-reflection region and the coverage layer contour to extract the defect boundary. The defect pixel area is divided by the total area to obtain the defect distribution density map. Finally, the coverage rate weight is set to 0.4, the defect density weight to 0.4, and the consistency weight to 0.2. After weighted summation, the data is mapped to a quality range of 0-100 points, and the output includes a quality inspection dataset containing coverage rate, defect density, consistency score, and weighted total score.

[0045] This invention can simultaneously obtain a complete quality profile of thickness, coverage, defects, and pearlescent consistency in a single inspection process. It improves defect location accuracy from pixel level to sub-pixel level, controls thickness measurement error within ±3nm, coverage calculation deviation less than 0.5%, and low-reflection defect identification miss rate less than 1%. The inspection cycle time is shortened from 20min / sample for offline sampling to 3s / sample for online full inspection. The inspection data is directly used as input to the quality process coupling model, reducing the response time of subsequent process optimization closed loop from hours to minutes. The surface gloss CV value of the entire batch of pearlescent materials is reduced from 4.7% to 1.9%, the particle size distribution D50 fluctuation range is narrowed from ±8nm to ±2nm, and the coating layer defect density is reduced from 150ppm to 20ppm. Finally, the correlation between the weighted quality score and the first-pass acceptance rate of downstream customers reaches 0.97, significantly reducing rework and return costs.

[0046] Preferably, step S133 includes the following steps:

[0047] Step S1331: Identify the interface transition zone between the titanium dioxide coating layer and the mica substrate based on the coating layer contour data, and generate interface transition curve data;

[0048] Step S1332: Calculate the equivalent optical thickness of each pixel based on the interface transition curve data to obtain the optical thickness distribution matrix;

[0049] Step S1333: Solve for the local coating thickness based on the optical thickness distribution matrix to generate thickness gradient field data;

[0050] Step S1334: Compensate for the tilt projection error based on the thickness gradient field data, and correct to obtain the true thickness distribution data;

[0051] Step S1335: Construct a three-dimensional heat map containing the thickness-chromaticity mapping relationship based on the real thickness distribution data to obtain the heat map of the coating layer thickness distribution.

[0052] In this embodiment of the invention, the coating layer contour curve output in step S132 is taken, and Ti and Si elemental signals are collected point by point along the contour normal using scanning electron microscopy-energy dispersive spectroscopy (SEM-EDS) mode. The position where the Ti intensity drops to 5% and the Si intensity rises to 95% is marked as the transition start point between the titanium dioxide coating layer and the mica substrate, and the position where the Ti intensity rises to 95% and the Si intensity drops to 5% is marked as the transition end point. The interface transition curve data is obtained by connecting 50 equidistant sampling points between the start and end points. The coordinates of each point on the curve are mapped to the original CCD image coordinate system, and a 5-pixel × 5-pixel neighborhood is opened with the point as the center. The RGB mean value is read and converted into CIE XYZ tristimulus values. The refractive index of titanium dioxide (2.7), the refractive index of mica (1.58), the incident angle (45°), and the observation angle (0°) are substituted into the thin film interference equation. Where n is the direct emissivity of titanium dioxide or the refractive index of mica, R is obtained by normalizing the Y value to 0-1, λ is taken as 550nm, and the equivalent optical thickness d is calculated pixel by pixel. opt A 2448×2048 optical thickness distribution matrix is ​​generated; the gradients Gx and Gy of this matrix are obtained using a 3×1 Sobel kernel in the horizontal and vertical directions, respectively, and the amplitudes are synthesized. Regions with G < 1 nm / pixel are labeled as flat regions, and regions with G ≥ 1 nm / pixel are labeled as sloped regions. The thickness of the flat region is directly taken as d. opt The thickness of the inclined region is according to d local =d opt • cosθ correction, where θ is calculated from the gradient direction as θ = arctan(Gy / Gx), forming the thickness gradient field data; with the angle α = 20° between the sample surface normal and the microscope optical axis as a fixed tilt angle, the projection error Δd = d local ·(1―cosα), add Δd back to d local Obtain the true thickness distribution data; construct a 3D point cloud with the true thickness as the Z-axis and the corresponding pixel coordinates X and Y. The thickness range is 0-500nm, and the color is layered in 10nm increments. The chromaticity coordinates are calculated using the Lab conversion formula L=116f(Y / Yn)-16, a=500[f(X / Xn)-f(Y / Yn)], b=200[f(Y / Yn)-f(Z / Zn)], where Yn=100, Xn=95.047, Zn=108.883, and f(t)=t 1 / 3 When t>0.008856, otherwise f(t)=7.787t+16 / 116, the Lab color space is mapped to the RGB color space, and the final output is a heat map of the coating thickness distribution.

[0053] This invention advances the lateral resolution from the whole pixel level to the sub-pixel level, reducing the thickness measurement error to ±2nm and the interface positioning deviation to less than 0.05 pixels. This results in a 0.3% reduction in the subsequent coverage calculation error and a defect density statistical missed detection rate of less than 0.8%. By compensating for tilted projection errors with gradient field, the parallax distortion caused by the 20° observation angle is eliminated, and the linear correlation coefficient between the true thickness and the SEM cross-sectional measurement reaches 0.99. The three-dimensional heat map with thickness-color synchronous mapping directly converts thickness differences into visual color differences, shortening the manual re-inspection time from 15 minutes to 30 seconds. It can also be directly embedded into the downstream color matching system, achieving a brightness L fluctuation of ±0.5 and a and b color drift of <1.0 for the same batch of pearlescent powder, ultimately reducing the color difference return rate of the entire roll of pearlescent film by 90%.

[0054] Preferably, step S2 includes the following steps:

[0055] Step S21: Divide the process-quality batch data into input and output variables, extract key process parameters as input feature data, and extract titanium dioxide coating rate, D50 particle size, gloss and Lab color value as output quality indicators to obtain the modeling sample dataset;

[0056] Step S22: Normalize and impute missing values ​​in the modeling sample dataset, calculate the Pearson correlation coefficient between each input feature data and each output quality index, and calculate the information gain score between each input feature data and each output quality index to generate a feature sensitivity matrix.

[0057] Step S23: Based on the feature sensitivity matrix, select the main process control variables, and construct modeling subsets for each series of pearlescent materials, train the input-output mapping relationship, and obtain a preliminary quality-process coupling model;

[0058] Step S24: Perform K-fold cross-validation and residual analysis on the preliminary quality-process coupling model to evaluate the fitting accuracy of the preliminary quality-process coupling model in predicting coating rate, particle size and gloss, and optimize the model parameter combination to obtain the quality-process coupling model structure.

[0059] Step S25: Extract the prediction function expression, weight coefficient distribution and feature contribution score based on the quality-process coupling model structure to generate the quality-process coupling model parameter set.

[0060] In this embodiment of the invention, the process-quality batch data is first divided into fields. The input feature fields are defined as five key process parameters: reactor temperature (range set to 700℃ to 850℃), nitrogen flow rate (range set to 20 to 60 L / min), coating reaction residence time (set to 10 to 30 minutes), calcination section temperature (controlled at 750℃ to 900℃), and raw material feed rate (set to 10 to 40 kg / h). The output quality index fields are titanium dioxide coating rate (in percentage), D50 particle size (in micrometers), gloss (in GU), and L, a, and b values ​​in the Lab colorimetric values. (Units are CIE chromaticity coordinates). A modeling sample dataset was constructed by merging these values. The data format adopted a standard CSV structure, with at least 500 rows. After importing the modeling sample dataset into the data processing platform, normalization was performed on all input and output fields. The normalization rule was to scale the original values ​​to the [0,1] interval and use mean imputation to fill in any missing values. Fields with a missing value ratio exceeding 10% were removed from the modeling set. After normalization, based on bivariate correlation analysis, the Pearson correlation coefficient and information gain score (by analyzing the correlation between input features and output indicators) were calculated for each input feature and each output indicator. The output quality indicators are grouped and discretized, and the entropy decrease is measured to generate a 5×6 dimensional feature sensitivity matrix. Each cell in the matrix corresponds to the correlation strength between an input and an output. Based on the top two input features in the feature sensitivity matrix, reactor temperature and residence time are selected as the main process control variables from five process parameters. Independent modeling subsets are constructed for the gold and silver-white pearlescent materials, each containing no less than 200 records. A linear fit is performed between the input features and output indicators using the least squares method to obtain a preliminary quality-process coupling expression. The sample is then cross-validated. This episode is divided into 5 folds for sequential verification. The mean absolute error (MAE) and coefficient of determination (R²) of the predicted coating rate, particle size, and gloss are calculated in each fold. The best-performing combination of process parameters is selected as the structure of the quality-process coupling model with the best fitting quality. Based on the above structure, the final prediction function expression is extracted, which includes the weight coefficient value (retaining 4 decimal places) and constant term corresponding to each input feature. The output feature contribution scoring table is compiled. The scoring standard is the standardized weight ratio multiplied by the output variable change rate weight. A complete set of parameters for the quality-process coupling model is generated, including the function expression, coefficient distribution, and the ranking of the influence of each feature on quality changes.

[0061] This invention enables precise modeling of the correlation between process parameters and quality indicators in pearlescent material production, demonstrating significant practical application value and technological advantages. By systematically dividing input and output variables and constructing modeling samples, the rationality of the data structure and the relevance of modeling inputs are ensured, effectively improving modeling efficiency and accuracy. Normalization and missing value completion of the sample dataset help eliminate the influence of dimensions and enhance the model's tolerance to incomplete data, improving model robustness. Constructing a feature sensitivity matrix using both correlation and information gain indicators allows for objective evaluation of the impact of each process parameter on quality indicators, enabling scientific selection of key variables and avoiding interference from redundant inputs. Modeling different product series separately enhances the model's adaptability and generalization ability, improving the quality prediction accuracy of multi-category pearlescent materials. The use of K-fold cross-validation and residual analysis ensures the comprehensiveness and objectivity of model performance evaluation, avoiding overfitting or underfitting. Finally, by extracting the prediction function expression and weight distribution, the contribution of each input feature is clarified, providing quantitative basis and decision support for subsequent quality optimization and process adjustment, enhancing the data-driven intelligent control capability of the entire production line.

[0062] Of particular importance is that in step S21, the process-quality batch data pair is divided into input and output variables, and key process parameters are extracted as input feature data, including:

[0063] Based on process-quality batch data, the control parameters of pearlite interlayer spacing were extracted to obtain interlayer spacing process data;

[0064] The key parameters of the cooling rate are calculated based on the interlayer spacing process data to obtain the cooling process characteristic data;

[0065] The austenitizing temperature curve was extracted based on the cooling process characteristic data to obtain temperature history data;

[0066] The phase transition volume fraction of pearlite was calculated based on the temperature history data to obtain the phase transition degree data.

[0067] Pearl cluster size distribution is identified based on phase transition degree data, and pearl cluster morphological characteristic data are obtained;

[0068] Interlayer spacing process data, cooling process characteristic data, temperature history data, phase transition degree data, and pearlite morphological characteristic data are used as input feature data.

[0069] In this embodiment of the invention, firstly, based on process-quality batch data pairs, process parameters for controlling the interlayer spacing of the pearlite structure are extracted, including the final reaction temperature (set range of 700℃ to 850℃), the particle size of the sheet substrate (limited to 4μm to 15μm), the inlet air velocity (range of 3m / s to 6m / s), and the reaction residence time (range of 10 to 25 minutes). By performing mean normalization and variance standardization operations on the above parameters in different batches, the interlayer spacing process data is obtained and recorded in the form of a standardized matrix. After the interlayer spacing process data is extracted, based on the temperature decay curve of the calcination section and the ventilation rate data, the slope of the temperature change in the first 300 seconds in the cooling zone is selected, and the average value is calculated. Cooling rate (in °C / s) was used, and the airflow velocity, heat exchanger exhaust temperature, and ambient temperature difference were considered as influencing factors to construct cooling process characteristic data. This data includes three index fields: cooling rate value, cooling duration, and heat transfer amplitude. The cooling process characteristic data was spliced ​​with the temperature time series recorded by the measured thermocouple array to extract the temperature drop curve of the austenitization stage (750 °C to 500 °C). Sampling was performed at a time resolution of 1 second to obtain complete austenitization temperature history data, with a data length of no less than 300 sets. Based on the austenitization temperature history data, the Johnson-Mehl-Avrami type conversion formula V(t)=1―exp(―k·t) was used. n The process involves calculating the pearlite phase transition volume fraction at each time point in the transformation time series, where t is the reaction time, the conversion index n is set to 2, and the constant k is set to 0.003 according to experimental conditions. The results are retained to three decimal places and output as phase transition degree data. Based on the local maximum point of the volume fraction change rate in the phase transition degree data, combined with the crystal phase interface propagation rate parameter (experimentally measured value is 0.8 μm / s), the grain growth rate and termination point position in this batch of pearlite materials are determined. The pearlite particle size on each substrate is calculated using the regional statistical method to perform histogram distribution calculation, generating pearlite morphological characteristic data, including three items: average pearlite particle size, maximum particle size, and standard deviation of particle size distribution. Finally, the interlayer spacing process data, cooling process characteristic data, temperature history data, phase transition degree data, and pearlite morphological characteristic data are merged and processed to construct complete input characteristic data.

[0070] The variable extraction and construction process in this invention, starting from the physical essence, systematically captures the key process evolution paths affecting quality characteristics during the formation of pearlescent materials, significantly improving the physical relevance and engineering interpretability of the model input. Through layer-by-layer derivation and extraction of multi-dimensional data such as interlayer spacing, cooling rate, temperature history, phase transition behavior, and pearlite structure, the entire process characteristic expression from initial process operation to the evolution of the material's microstructure is realized, effectively reflecting the profound influence of process parameters on final quality indicators (such as coating uniformity, particle size distribution, and gloss effect). This process strengthens the integrated expression of thermodynamic, phase transition kinetic, and microstructural elements in the model input, enabling the subsequently established quality prediction model to not only possess high accuracy but also a sensitive response capability to complex process disturbances. This provides a highly interpretable and adaptable feature foundation for intelligent control systems, contributing to improved quality stability and controllability throughout the entire pearlescent material production process.

[0071] Preferably, step S3 includes the following steps:

[0072] Step S31: Collect the process parameters of the pearlescent material production line for the current production batch in real time to obtain the process parameters of the current production batch;

[0073] Step S32: Based on the parameters of the quality-process coupling model, perform model input mapping processing on the process parameters of the current production batch, execute the model inference process, and obtain the quality prediction result dataset of the current batch;

[0074] Step S33: Format and timestamp the quality prediction result dataset, and map it to the sheet substrate batch corresponding to the current production segment to obtain prediction results with batch identifiers.

[0075] In this embodiment of the invention, WRM-200 edge acquisition terminals are first deployed in key equipment sections of the pearlescent material production line, with the acquisition frequency set to 1Hz. Process parameters for the current production batch are collected from operating equipment including the screw feeder, cyclone preheater, coated fluidized bed reactor, calcination section, heat exchanger, and cooling section. These parameters mainly include the coating reaction temperature (set range 720℃ to 850℃), gas phase transport velocity (range 25 to 60 L / min), calcination temperature (set range 750℃ to 900℃), reaction residence time (set range 10 to 25 minutes), and raw material feed rate (set range 12 to 38 kg / h). The real-time acquired parameters are uniformly converted into a time series format to form the process parameters for the current production batch. The process parameters for this batch are then mapped according to the field order of the quality-process coupling model parameters output in step S25, and the input data is normalized (the processing method is to convert the original data into a normalized data). The normalized value (after subtracting the training mean and dividing by the training standard deviation) is input into the constructed quality prediction function expression for term-by-term function calculation. The weighted sum of each parameter and the corresponding input value is added, and a constant term is added. The predicted values ​​of titanium dioxide coating rate (in %), D50 particle size (in μm), gloss (in GU), and colorimetry Lab for the current batch are output respectively, forming a quality prediction result dataset. All prediction results retain three significant figures. Subsequently, the quality prediction result dataset is formatted, with a unified field structure of "timestamp + batch number + prediction index name + prediction value", where the timestamp is accurate to milliseconds. Each data point is marked with a unique sheet substrate batch identifier according to the material number tracking mechanism. The batch number is generated in the format of "QG + year month day + production period number". The corresponding batch process parameters are associated with the prediction results one by one, generating prediction results with batch identifiers.

[0076] This invention enables real-time perception and batch-level tracking management of quality status during the pearlescent material production process, significantly improving the intelligence and responsiveness of the production line. By real-time acquisition and model inference of current batch process parameters, quality trends can be predicted in advance during production, potential deviation risks can be identified in a timely manner, and delayed exposure of quality problems can be avoided, thereby reducing the defect rate and rework costs. The structured and time-stamped processing of the model inference results gives the quality prediction data high temporal consistency and traceability, facilitating integration with the production execution system and supporting the formulation of refined process control strategies. Simultaneously, accurately mapping the prediction results to specific substrate batches achieves product-level quality prediction and tracking, providing a data foundation for establishing functions such as quality early warning, trend analysis, and precise scheduling, comprehensively improving the level of visual management and quality control efficiency in pearlescent material production.

[0077] Preferably, step S32 involves performing model input mapping processing on the process parameters of the current production batch based on the quality-process coupling model parameters, and executing the model inference process, including:

[0078] Extract pearlite interlayer spacing process parameters for the current production batch based on the parameters of the quality-process coupling model;

[0079] The model input is normalized based on the pearlite interlayer spacing process parameters to obtain standardized interlayer spacing data.

[0080] Pearlite phase transition calculations were performed based on standardized interlayer spacing data to obtain pearlite phase transition degree data;

[0081] The pearlite size distribution characteristics were calculated based on the pearlite phase transition degree data to obtain pearlite morphological data;

[0082] Material strength inference is performed based on pearloid morphology data to obtain strength performance prediction data;

[0083] The mapping relationship between toughness indexes is calculated based on the strength performance prediction data to obtain the ductile-brittle transition temperature data;

[0084] A comprehensive quality evaluation is performed based on strength performance prediction data and ductile-brittle transition temperature data to generate a dataset of quality prediction results for the current batch.

[0085] In this embodiment of the invention, firstly, based on the variable structure defined in the quality-process coupling model parameters output in step S25, process parameters related to pearlite interlayer spacing in the current production batch are extracted, including fluidized bed reactor temperature (set range of 720℃ to 850℃), airflow velocity (controlled between 3.0 and 5.5 m / s), reaction time (set between 10 and 30 minutes), and raw material particle size (set between 4 and 12 μm). These values ​​are used as pearlite interlayer spacing process parameters. The extracted pearlite interlayer spacing process parameters are normalized by using a linear scaling method to normalize each parameter to the [0,1] interval, resulting in standardized interlayer spacing data. Based on this standardized interlayer spacing data, the degree of pearlite phase transition is calculated. The degree of phase transition is numerically approximated according to the isothermal phase transition theory, using an exponential function form V(t) = 1 - exp(-k·t). nThe process involves converting the phase transition volume fraction under the current process conditions, where t is the reaction time (in minutes), k is a fixed value of 0.005, and n is the nucleation index set to 2.0. The resulting pearlite phase transition degree data is output, retained to three decimal places. Using this volume fraction data, combined with the system's preset pearlite growth rate constant (set to 0.7 μm / min), the particle size within the pearlite growth range is sequentially calculated to obtain the pearlite size distribution characteristics in the current production batch. Pearlite morphological data, including average particle size, particle size distribution variance, and maximum particle size, are output, in micrometers. This pearlite morphological data is then input into the strength inference module, based on the Griffith fracture criterion... The structural form is substituted one by one, where E is the elastic modulus (taken as a fixed value of 120 GPa) and γ is the interfacial energy (taken as a value of 1.2 J / m). 2 ), where 'a' represents the maximum particle size (in meters), and the material's fracture strength σ is calculated after unifying the units, in MPa, outputting predicted strength performance data; based on this strength data and the system's built-in critical value for the fracture toughness of pearlescent substrate, KIC = 1.5 MPa·m 0.5 Based on the Ashby ductile-brittle transition standard, the ductile-brittle transition temperature of the current batch is calculated using the form TDB = A―B·σ, with A set to 350K and B set to 0.35K / MPa, yielding ductile-brittle transition temperature data in K. Finally, the fracture strength data and ductile-brittle transition temperature data of the batch are weighted and summarized. The scoring structure is: comprehensive quality value = 0.6 × strength normalized value + 0.4 × toughness stability index. The strength normalized value is normalized according to the maximum predicted value, and the toughness stability index is scored according to the ideal state below 300K. The output is a standard format quality prediction result dataset, with fields including batch number, timestamp, predicted coating strength, predicted transition temperature, and comprehensive quality score.

[0086] This invention introduces a multi-stage structure-performance mapping mechanism to decouple and reconstruct key process and material characteristics, such as pearlite interlayer spacing, phase transformation behavior, microstructure, strength properties, and toughness indices, step by step. This achieves a high-precision inference path from process parameters to comprehensive quality assessment results. Its advantages include transforming complex production data into physically meaningful standardized features, improving the consistency of model inputs and the interpretability of prediction outputs; establishing a logical chain between material structure and macroscopic properties through the derivation of phase transformation degree and pearlite size, enhancing the model's sensitivity to microstructural changes; further expanding the dimensions of quality assessment by introducing strength prediction and brittle-ductility temperature indices, enabling prediction results to cover not only appearance and particle size indices but also key mechanical properties, effectively supporting a comprehensive assessment of material functionality; and finally, the resulting quality prediction dataset exhibits high accuracy, physical consistency, and process relevance, providing a solid data foundation and decision support for subsequent quality control and dynamic process optimization.

[0087] Preferably, step S4 includes the following steps:

[0088] Step S41: Compare and analyze the prediction results with the preset quality control standards, calculate the difference between the prediction results of the current batch and the preset target standards, and obtain the predicted quality deviation data;

[0089] Step S42: Based on the predicted quality deviation data and the current production batch process parameters, use the quality-process coupling model to trace back the main process control variables that affect the deviation of the prediction results, identify the abnormal parameter combinations in the current process operation, and obtain the abnormal parameter identification results;

[0090] Step S43: Perform sensitivity analysis on the abnormal parameter identification results, calculate the rate of change of quality prediction response within different parameter adjustment ranges, construct the optimization objective function, and obtain process optimization adjustment path data;

[0091] Step S44: Based on the process optimization adjustment path data and the preset standard process segment sample, formulate a process adjustment plan and form a dynamic process adjustment suggestion table;

[0092] Step S45: Verify the process safety and equipment boundary constraints of the dynamic process adjustment suggestion table, and output executable dynamic process adjustment data.

[0093] In this embodiment of the invention, the coating rate (in %), D50 particle size (in μm), gloss (in GU), and Lab chromaticity value contained in the quality prediction result dataset are first compared item by item with the preset quality control standards built into the production system. The standard coating rate target is set at 40%, the D50 target particle size is 10.0 μm, the gloss is controlled within the range of 85±5 GU, the chromaticity value L is required to be greater than 80, and a and b are controlled within ±5. The arithmetic difference between the predicted value and the target value is calculated, and the deviation value of each quality indicator is recorded to obtain the predicted quality deviation data of the current batch. All deviation values ​​are uniformly retained to two decimal places. The quality process coupling model parameter structure is called to perform item-by-item weighted reverse lookup processing on the actual process parameters of the current production batch, and the deviation index of each indicator is calculated. The contribution rate under the previous process settings was used to extract the two process variables with the greatest impact on coating rate and particle size, which were identified as key abnormal parameters. These included calcination temperature (exceeding the set upper limit of 860℃) and reactor gas flow rate (below the set lower limit of 3.0m / s). The abnormal parameter identification results were combined, and the data format included parameter name, deviation direction, deviation magnitude, and historical threshold comparison label. Numerical perturbation operations were performed on the identified abnormal parameters. The predicted coating rate and particle size change trends were calculated at adjustment increments of ±5%, ±10%, and ±15%. The predicted response change rate was obtained by dividing the incremental change rate by the parameter adjustment magnitude. The process variable with the largest response change rate was selected as the first adjustment object. Combining the minimization of coating rate deviation and the convergence of particle size error as objectives, the optimization objective function was constructed as ΔY=∑(w i ·|X i ′―X i |), where X i ′ represents the value of the parameter to be adjusted, w i To predict sensitivity weights, process optimization adjustment path data is output. The optimized adjustment path is compared with the standard process segment samples archived in the system database. Historical sample records with similar parameter combinations under the same equipment and raw material conditions are selected. The corresponding temperature setpoint, wind speed parameter, reaction residence time, and raw material particle size combination are extracted from the historical samples. Combined with the current deviation direction, a dynamic process adjustment suggestion table is generated. The fields include adjustment parameter items, recommended target values, recommended variation range, and adjustment sequence number. Boundary checks are performed on all recommended parameters in the dynamic process adjustment suggestion table. Among them, the reactor temperature should not exceed 880℃, the wind speed should not be less than 2.5m / s, the raw material feed rate should not exceed 40kg / h, and the calcination time should not be less than 8 minutes. By sequentially checking the upper and lower limits of each suggested parameter value and comparing it with the equipment adjustment capability, the instruction items that do not meet the conditions are eliminated, and the final executable dynamic process adjustment data is output.

[0094] This invention enables intelligent identification, causal analysis, and optimized response to predicted quality deviations, comprehensively enhancing the closed-loop control capability and process adaptability of pearlescent material production. By performing difference analysis between predicted results and standard quality requirements, the degree of deviation of the actual process from the quality target can be quantitatively identified, providing a clear direction for subsequent adjustments. Using a coupled model to reverse-analyze the sources of deviations, the key process variable combinations leading to quality anomalies can be accurately identified, avoiding the subjectivity and blindness of traditional experience-based judgments. Furthermore, sensitivity analysis is used to construct an optimization objective function, enabling the process adjustment scheme to possess quantitative optimization and response prediction capabilities, ensuring the scientific and targeted nature of the adjustment measures. Dynamic adjustment suggestions developed based on historical high-performing process samples enhance the practicality and feasibility of the scheme, while verification of process safety and equipment boundary conditions ensures that the final output data is operationally executable and equipment compatible, fundamentally improving the reliability, stability, and real-time response efficiency of process adjustments, effectively supporting the achievement of high-quality, low-fluctuation intelligent manufacturing goals.

[0095] Of particular importance, in step S42, based on the predicted quality deviation data and the current production batch process parameters, the main process control variables affecting the deviation of the prediction results are traced back using the quality-process coupling model to identify abnormal parameter combinations in the current process operation, including:

[0096] Based on the predicted quality deviation data, the deviation threshold of pearlite strength performance is extracted to obtain key quality deviation indicators;

[0097] By reverse matching the cooling rate influence coefficient in the quality process coupling model based on the key quality deviation index, the cooling rate sensitivity data is obtained.

[0098] By scanning the interlayer spacing control curve of the current production batch based on cooling rate sensitivity data, abnormal fluctuation sections are identified, and abnormal interlayer spacing fluctuation data are obtained.

[0099] By correlating the austenitizing temperature time series records with the abnormal fluctuation data of interlayer spacing, the time periods of temperature deviation were extracted to obtain abnormal temperature process data;

[0100] The phase transition kinetic parameter offset was calculated based on temperature process anomaly data to obtain pearlite phase transition anomaly characteristics.

[0101] Based on the abnormal characteristics of pearlite phase transformation, the rolling process parameter database is traced back to locate the abnormal parameter combination and obtain the abnormal parameter identification results.

[0102] In this embodiment of the invention, the predicted quality deviation data is first filtered by field selection, and the difference between the predicted pearlite strength performance value and the target strength control standard is extracted. If the predicted strength is lower than 125 MPa and the deviation exceeds 5%, it is determined to be a key quality deviation indicator, and the deviation direction is marked as negative fluctuation. Based on this key quality deviation indicator, the response coefficients of cooling section process parameters affecting strength performance are extracted from the parameter set of the quality-process coupling model. Specifically, this includes the coefficient value of the change in average cooling rate on the predicted strength, with the unit of value being MPa·s / ℃. The cooling rate sensitivity data is split and statistically analyzed by process section, and the cooling rate is recorded. The impact threshold is defined as follows: below 2.0 MPa·s / ℃ is the low-sensitivity response zone, 2.0 to 3.5 is the medium-sensitivity zone, and above 3.5 is the high-sensitivity zone. Combining the real-time cooling section data of the current production batch, the cooling curve of the entire process from the completion of coating to the recovery of ambient temperature is extracted. Under the condition of a sampling period of 1 second, the first derivative analysis of cooling rate is performed to screen out continuous fluctuation sections with a slope change rate greater than a set threshold. The continuous fluctuation duration is set as the anomaly judgment standard, and the set of abnormal points in the section is identified and the interlayer spacing abnormal fluctuation data is generated. The data structure is the interval start and end time, the maximum rate change value and the corresponding process section number. The abnormal fluctuation data is compared with the austenitizing temperature time series records of the current batch to identify the heating temperature range corresponding to the occurrence of the abnormal fluctuation. The time period in which the temperature exceeds the set temperature control range by more than ±20℃ is extracted to generate temperature process anomaly data. The fields include the start and end time of the deviation, the maximum deviation, and the duration. Based on the temperature process anomaly data and combined with the system's built-in phase transformation kinetic parameter table, the heating rate and cooling rate of the batch during the abnormal period are jointly analyzed. The phase transformation response deviation is evaluated by the two-interval phase transformation rate ratio calculation method (V1 / V2). If the ratio is greater than 1, the deviation is considered positive. If 4 is selected, it is determined to be a phase transition delayed anomaly. Pearlite phase transition anomaly feature data is generated, and the structure fields include offset type, occurrence stage, and associated parameter values. Finally, the phase transition anomaly features of this batch are used as query conditions to enter the archived rolling process parameter library in the pearlite material production database. Batch instances with historical anomaly records in the corresponding batch parameters such as reactor airflow rate, calcination temperature, and holding time are selected. The top two process parameter combinations are selected according to the anomaly frequency and defined as the anomaly parameter combination. The output is the anomaly parameter identification result, and the fields include parameter name, offset direction, over-limit amplitude, and number of hit anomaly cases.

[0103] This invention can start from specific quality performance deviations and gradually trace back to key process links such as cooling rate, interlayer spacing fluctuations, temperature anomalies, and phase change behavior, constructing a complete quality-process causal path. By introducing quantitative indicators such as sensitivity and kinetic offset, the anomaly identification process is data-supported and reproducible, avoiding the problems of strong human subjectivity and many adjustment blind spots in traditional experience analysis. At the same time, through cross-level variable correlation analysis and parameter library backtracking, the judgment of abnormal operating conditions is not limited to a single variable, but identifies synergistic anomalies between variable combinations, improving the completeness and accuracy of anomaly identification results. The final output of abnormal parameter combinations provides a clear direction for process control, which helps to quickly close the loop for optimization and adjustment, and supports the production system to achieve a higher-order leap from "problem discovery" to "root cause location" and "proactive intervention".

[0104] Preferably, in step S43, sensitivity analysis is performed on the abnormal parameter identification results, the rate of change of quality prediction response is calculated within different parameter adjustment ranges, and the optimization objective function is constructed, including:

[0105] Extract the set of key process variables for the current batch based on the results of abnormal parameter identification;

[0106] Design multi-level orthogonal experiments based on the current batch of key process variables, and generate a parameter perturbation simulation dataset;

[0107] The quality response surface is calculated based on the parameter perturbation simulation dataset to obtain variable-sensitive gradient data;

[0108] By coupling variable sensitivity gradient data with pearl optical interference effect, a dynamic optimization objective function is constructed.

[0109] Solve for the optimal adjustment path of the dynamic optimization objective function and generate a process optimization instruction set;

[0110] The robustness of the optimization path is verified based on the process optimization instruction set, and process optimization adjustment path data is obtained.

[0111] In this embodiment of the invention, firstly, based on the abnormal parameter identification results, the key process variable set for the current batch is extracted. The screening condition is that the process parameter items with deviations greater than the set threshold values ​​are selected. The set thresholds are: coating reaction temperature deviation ±15℃, gas flow rate deviation ±10L / min, and calcination holding time deviation ±3 minutes. Finally, reactor temperature, calcination time, and gas flow rate are selected as the key process variable set. Based on this variable set, a five-level three-factor orthogonal experimental design is carried out. The levels are set as five adjustment levels for each variable, namely 5%, 10%, 15%, 20%, and 25% above and below the current set value. L25(5 3A standard orthogonal array design was used to generate a parameter perturbation simulation dataset containing 25 sets of variable combinations. All combined parameters were strictly limited within the safety upper and lower limits set by the process control system. The quality-process coupling function was input into this parameter perturbation simulation dataset, and the predicted coating rate, particle size D50, and gloss value for each combination were obtained through direct substitution calculation. The rate of change of the predicted results for each process parameter under different levels of variation was calculated. A three-dimensional quality response surface was constructed using the difference quotient form ΔY / ΔX. The maximum gradient of change of each parameter on the predicted index was extracted and organized into variable-sensitive gradient data, recorded in units of prediction units / process units (e.g., % / ℃). Based on this, the variable-sensitive gradient data was coupled with the phase gain law of film gloss under the optical interference effect of pearlescent materials. The reference coefficient for the rate of change of the influence of film thickness and interference peak on gloss was set to 0.35 GU / μm. The indirect response term of coating thickness was introduced into each sensitive gradient to construct an optical enhancement weight function. The gain weight of each parameter on optical performance was adjusted to form a dynamic optimization objective function Z=∑(Wopt·|Y i —Y target |), where Y i Y represents the predicted index value for the simulation group. target The objective value is set as the quality control target value. An exhaustive calculation is performed on this objective function to determine the parameter combination corresponding to the minimum Z value. The output is a process optimization instruction set, including recommended temperature, recommended wind speed, and recommended insulation time, with the adjustment priority order marked. Subsequently, based on the real-time process parameters of the current batch, the parameter settings in the process optimization instruction set are sequentially embedded into the simulated process path. Three rounds of disturbance tests are performed, and the fluctuation range of the predicted results is observed to remain within ±5% of the target threshold. If stable, its robustness is confirmed to be qualified, ultimately forming validated process optimization adjustment path data.

[0112] This invention achieves a comprehensive response exploration of multi-factor, multi-level combinations under controllable cost conditions by extracting key process variables and designing orthogonal experiments. By generating quality response surfaces using simulated data, the nonlinear impact of variable changes on quality indicators can be quantitatively analyzed, enhancing the targeting and predictability of control strategies. The combination of variable-sensitive gradient data and pearlescent optical interferometry allows optimization to focus not only on numerical fitting results but also on the synergistic optimization needs of material properties and appearance. The constructed dynamic optimization objective function possesses solvability and multi-dimensional adaptability, laying the foundation for generating efficient and executable process optimization instructions. Robustness verification of the optimization path ensures that the recommended adjustment scheme has good stability and anti-disturbance capability under actual production conditions, significantly improving the reliability of the process adjustment process and the success rate of final quality control.

[0113] Preferably, the present invention also provides a pearlescent material production data management system based on data integration analysis, used to execute the above-described pearlescent material production data management method based on data integration analysis, wherein the pearlescent material production data management system based on data integration analysis includes:

[0114] The data integration module is used to acquire the original process dataset of the pearlescent material production line, and perform standardization processing to obtain a standardized process dataset; acquire the quality inspection dataset of the pearlescent material; divide the standardized process dataset and the quality inspection dataset into time intervals and mark them with batch numbers to obtain process-quality batch data pairs.

[0115] The model building module is used to construct a quality-process coupling model based on process-quality batch data, and extract the mapping relationship between process control and quality characterization to obtain the parameters of the quality-process coupling model.

[0116] The intelligent prediction module is used to collect the process parameters of the current production batch in real time, and perform quality index prediction analysis based on the parameters of the quality-process coupling model to obtain the prediction results;

[0117] The control and optimization module is used to perform deviation analysis based on the prediction results and preset quality control standards, identify abnormal parameters in the current process operation, formulate process optimization and adjustment plans, and obtain dynamic process adjustment data.

[0118] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is not limited by the foregoing description. Thus, all changes falling within the meaning and scope of the equivalents of the application are intended to be included within the scope of the invention.

[0119] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for managing pearlescent material production data based on data integration and analysis, characterized in that, Includes the following steps: Step S1: Obtain the original process dataset of the pearlescent material production line and perform standardization processing to obtain a standardized process dataset; Obtain a quality inspection dataset for pearlescent materials; The standardized process dataset and quality inspection dataset are divided into time intervals and labeled with batch numbers to obtain process-quality batch data pairs. Step S2: Based on the process-quality batch data, construct a quality-process coupling model for the design modeling sample dataset; extract the mapping relationship between process control and quality characterization in the quality-process coupling model to obtain the parameters of the quality-process coupling model, where the quality characterization indicators include surface gloss, particle size distribution uniformity and coating layer defect density; Step S3: Collect the process parameters of the current production batch in real time, and perform quality index prediction analysis of the process parameters based on the parameters of the quality-process coupling model to obtain the prediction results; Step S4: Perform deviation analysis based on the prediction results and preset quality control standards, identify abnormal parameters in the current process operation based on the deviation analysis results, formulate process optimization and adjustment plans, and obtain dynamic process adjustment data.

2. The pearlescent material production data management method based on data integration analysis according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Configure edge acquisition terminals on the pearlescent material production line and collect raw process datasets in real time; Step S12: Decode the communication protocol, unify the sampling frequency, and align the timestamps of the original process dataset to form a standardized process dataset; Step S13: Obtain titanium dioxide-coated pearlescent material and perform quality inspection to obtain a quality inspection dataset; Step S14: Divide the standardized process dataset and quality inspection dataset into time intervals, and establish sliding window and time mapping rules to obtain the material production path interval for each sample; Step S15: Assign a unique batch number to the material production path range of each sample, identify and merge the process parameters and corresponding quality inspection results belonging to the same range, and generate process-quality batch data pairs.

3. The pearlescent material production data management method based on data integration and analysis according to claim 2, characterized in that, Step S13 includes the following steps: Step S131: Obtain titanium dioxide-coated pearlescent material, acquire surface reflection images under a standard light source, and obtain reflected light intensity distribution data; Step S132: Locate the edge of the coating layer based on the reflected light intensity distribution data, extract the gradient of the transition zone of the coating layer, and generate the contour data of the coating layer. Step S133: Perform sub-pixel-level segmentation based on the coating layer contour data, calculate the local thickness matrix, and obtain a heat map of the coating layer thickness distribution; Step S134: Calculate the effective coverage area ratio based on the heat map of coating thickness distribution to obtain the coating rate data; Step S135: Scan the surface of the coating layer of titanium dioxide-coated pearlescent material, construct the specular reflection intensity curve, and calculate the pearlescent effect consistency data; Step S136: Identify low-reflection areas based on pearlescent effect consistency data, and locate defect boundaries by combining coating layer contour data, generating a defect distribution density map; Step S137: Perform a weighted score based on the coverage rate data, defect distribution density, and pearlescent effect consistency, and output the quality inspection dataset.

4. The pearlescent material production data management method based on data integration and analysis according to claim 3, characterized in that, Step S133 includes the following steps: Step S1331: Identify the interface transition zone between the titanium dioxide coating layer and the mica substrate based on the coating layer contour data, and generate interface transition curve data; Step S1332: Calculate the equivalent optical thickness of each pixel based on the interface transition curve data to obtain the optical thickness distribution matrix; Step S1333: Solve for the local coating thickness based on the optical thickness distribution matrix to generate thickness gradient field data; Step S1334: Compensate for the tilt projection error based on the thickness gradient field data, and correct to obtain the true thickness distribution data; Step S1335: Construct a three-dimensional heat map containing the thickness-chromaticity mapping relationship based on the real thickness distribution data to obtain the heat map of the coating layer thickness distribution.

5. The pearlescent material production data management method based on data integration analysis according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Divide the process-quality batch data into input and output variables, extract key process parameters as input feature data, and extract titanium dioxide coating rate, D50 particle size, gloss and Lab color value as output quality indicators to obtain the modeling sample dataset; Step S22: Normalize and impute missing values ​​in the modeling sample dataset, calculate the Pearson correlation coefficient between each input feature data and each output quality index, and calculate the information gain score between each input feature data and each output quality index to generate a feature sensitivity matrix. Step S23: Based on the feature sensitivity matrix, select the main process control variables, and construct modeling subsets for each series of pearlescent materials, train the input-output mapping relationship, and obtain a preliminary quality-process coupling model; Step S24: Perform K-fold cross-validation and residual analysis on the preliminary quality-process coupling model to evaluate the fitting accuracy of the preliminary quality-process coupling model in predicting coating rate, particle size and gloss, and optimize the model parameter combination to obtain the quality-process coupling model structure. Step S25: Extract the prediction function expression, weight coefficient distribution and feature contribution score based on the quality-process coupling model structure to generate the quality-process coupling model parameter set.

6. The pearlescent material production data management method based on data integration analysis according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Collect the process parameters of the pearlescent material production line for the current production batch in real time to obtain the process parameters of the current production batch; Step S32: Based on the parameters of the quality-process coupling model, perform model input mapping processing on the process parameters of the current production batch, execute the model inference process, and obtain the quality prediction result dataset of the current batch; Step S33: Format and timestamp the quality prediction result dataset, and map it to the sheet substrate batch corresponding to the current production segment to obtain prediction results with batch identifiers.

7. The pearlescent material production data management method based on data integration analysis according to claim 6, characterized in that, Step S32 involves mapping the process parameters of the current production batch to the model input based on the parameters of the quality-process coupling model, and performing the model inference process, including: Extract pearlite interlayer spacing process parameters for the current production batch based on the parameters of the quality-process coupling model; The model input is normalized based on the pearlite interlayer spacing process parameters to obtain standardized interlayer spacing data. Pearlite phase transition calculations were performed based on standardized interlayer spacing data to obtain pearlite phase transition degree data; The pearlite size distribution characteristics were calculated based on the pearlite phase transition degree data to obtain pearlite morphological data; Material strength inference is performed based on pearloid morphology data to obtain strength performance prediction data; The mapping relationship between toughness indexes is calculated based on the strength performance prediction data to obtain the ductile-brittle transition temperature data; A comprehensive quality evaluation is performed based on strength performance prediction data and ductile-brittle transition temperature data to generate a dataset of quality prediction results for the current batch.

8. The pearlescent material production data management method based on data integration analysis according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Compare and analyze the prediction results with the preset quality control standards, calculate the difference between the prediction results of the current batch and the preset target standards, and obtain the predicted quality deviation data; Step S42: Based on the predicted quality deviation data and the current production batch process parameters, use the quality-process coupling model to trace back the main process control variables that affect the deviation of the prediction results, identify the abnormal parameter combinations in the current process operation, and obtain the abnormal parameter identification results; Step S43: Perform sensitivity analysis on the abnormal parameter identification results, calculate the rate of change of quality prediction response within different parameter adjustment ranges, construct the optimization objective function, and obtain process optimization adjustment path data; Step S44: Based on the process optimization adjustment path data and the preset standard process segment sample, formulate a process adjustment plan and form a dynamic process adjustment suggestion table; Step S45: Verify the process safety and equipment boundary constraints of the dynamic process adjustment suggestion table, and output executable dynamic process adjustment data.

9. The pearlescent material production data management method based on data integration analysis according to claim 8, characterized in that, Step S43 involves sensitivity analysis of the abnormal parameter identification results, calculation of the rate of change in the quality prediction response within different parameter adjustment ranges, and construction of the optimization objective function, including: Extract the set of key process variables for the current batch based on the results of abnormal parameter identification; Design multi-level orthogonal experiments based on the current batch of key process variables, and generate a parameter perturbation simulation dataset; The quality response surface is calculated based on the parameter perturbation simulation dataset to obtain variable-sensitive gradient data; By coupling variable sensitivity gradient data with pearl optical interference effect, a dynamic optimization objective function is constructed. Solve for the optimal adjustment path of the dynamic optimization objective function and generate a process optimization instruction set; The robustness of the optimization path is verified based on the process optimization instruction set, and process optimization adjustment path data is obtained.

10. A pearlescent material production data management system based on data integration and analysis, characterized in that, For executing the pearlescent material production data management method based on data integration analysis as described in claim 1, the pearlescent material production data management system based on data integration analysis includes: The data integration module is used to acquire the original process dataset of the pearlescent material production line, and perform standardization processing to obtain a standardized process dataset; acquire the quality inspection dataset of the pearlescent material; divide the standardized process dataset and the quality inspection dataset into time intervals and mark them with batch numbers to obtain process-quality batch data pairs. The model building module is used to construct a quality-process coupling model based on the design modeling sample dataset using process-quality batch data; it extracts the mapping relationship between process control and quality characterization in the quality-process coupling model to obtain the parameters of the quality-process coupling model, where the quality characterization indicators include surface gloss, particle size distribution uniformity and coating layer defect density. The intelligent prediction module is used to collect the process parameters of the current production batch in real time, and perform quality index prediction analysis of the process parameters based on the parameters of the quality-process coupling model to obtain the prediction results. The control and optimization module is used to perform deviation analysis based on the prediction results and the preset quality control standards, identify abnormal parameters in the current process operation based on the deviation analysis results, formulate process optimization and adjustment schemes, and obtain dynamic process adjustment data.

Images