Three-proofing paint process optimization method and system based on big data analysis

Abstract

The application provides a three-proofing paint process optimization method and system based on big data analysis, first, a historical process data set of a three-proofing paint coating process is acquired, containing a process parameter sequence, coating quality detection results and environmental influence factor records, then feature extraction is performed on the historical process data set to obtain coating process features, quality index features and environmental features, then association rule mining is performed based on the coating process features, quality index features and environmental features to identify key process parameters and environmental influence conditions affecting coating quality, a process parameter adjustment scheme set is generated according to the identification results, the process parameter adjustment scheme set is simulated and verified to determine an optimal process parameter combination and environmental control requirement, and a process optimization instruction is generated, so that the quality and stability of the three-proofing paint coating process can be improved, the cost can be reduced, and the efficiency can be improved.

Description

Technical Field

[0001] This invention relates to the field of big data technology, and more specifically, to a method and system for optimizing conformal coating processes based on big data analysis. Background Technology

[0002] In the electronics manufacturing industry, conformal coating plays a crucial role in ensuring the reliability and stability of electronic products. Conformal coatings effectively prevent electronic products from being corroded by environmental factors such as humidity, salt spray, and mold, extending their lifespan. However, current optimization of conformal coating processes faces numerous challenges. Traditional process optimization methods rely heavily on engineers' experience and limited experimental data. Engineers typically adjust process parameters based on personal experience, but due to a lack of comprehensive and systematic data support, these adjustments are often arbitrary and fail to accurately identify the key factors affecting coating quality. Furthermore, the limited amount and high cost of experimental data cannot cover various combinations of process conditions and environmental factors, resulting in optimization results lacking universality and reliability. In addition, existing methods, when considering the impact of environmental factors on coating quality, often focus only on a single or a few environmental parameters, neglecting the complex interactions between environmental factors and their synergistic effects with process parameters. This makes it difficult to achieve ideal optimization results and meet the electronics manufacturing industry's demands for high-quality, high-efficiency production. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for optimizing conformal coating processes based on big data analysis, the method comprising:

[0004] Obtain a set of historical process data during the execution of the conformal coating process. The set of historical process data includes the process parameter sequence of different coating batches, coating quality test results and corresponding environmental influencing factor records.

[0005] Data feature extraction processing is performed on the historical process data set to obtain coating process features in the process parameter sequence, quality index features in the coating quality test results, and environmental features in the environmental impact factor records.

[0006] Based on the coating process characteristics, the quality index characteristics, and the environmental characteristics, association rule mining is performed to identify key process parameters and corresponding environmental impact conditions that affect coating quality.

[0007] A set of process parameter adjustment schemes is generated based on the key process parameters and the corresponding environmental impact conditions. The set of process parameter adjustment schemes includes multiple combinations of process parameters to be verified and corresponding environmental control requirements.

[0008] The set of process parameter adjustment schemes is simulated and verified to determine the optimal combination of process parameters and the corresponding environmental control requirements. Process optimization instructions are then generated based on the optimal combination of process parameters and the corresponding environmental control requirements.

[0009] In another aspect, embodiments of the present invention also provide a conformal coating process optimization system based on big data analysis, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code. The processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0010] Based on the above, this embodiment of the invention acquires a historical process data set during the conformal coating process, comprehensively covering process parameters, coating quality inspection results, and environmental influencing factors for different coating batches. Data feature extraction processing is performed on the historical process data set to accurately obtain coating process characteristics, quality index characteristics, and environmental characteristics. Based on these characteristics, association rule mining is performed to deeply identify key process parameters affecting coating quality and their corresponding environmental influencing conditions. A set of process parameter adjustment schemes is generated based on the identification results, containing multiple combinations of process parameters to be verified and corresponding environmental control requirements, providing various possibilities for process optimization. Simulation verification of the process parameter adjustment scheme set allows for the determination of the optimal combination of process parameters and corresponding environmental control requirements from numerous schemes, generating process optimization instructions. This comprehensively and accurately considers the interaction of various process parameters and environmental factors, significantly improving the quality and stability of the conformal coating process, reducing production costs, and increasing production efficiency. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the execution flow of the conformal coating process optimization method based on big data analysis provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of exemplary hardware and software components of the conformal coating process optimization system based on big data analysis provided in an embodiment of the present invention. Detailed Implementation

[0013] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a method for optimizing conformal coating processes based on big data analysis, as provided in one embodiment of the present invention. The following is a detailed description of this method for optimizing conformal coating processes based on big data analysis.

[0014] Step S110: Obtain the historical process data set during the execution of the conformal coating process. The historical process data set includes the process parameter sequence of different coating batches, coating quality test results and corresponding environmental influencing factor records.

[0015] Step S111: Collect the process parameter sequence of different coating batches during the coating process. The process parameter sequence includes coating equipment operating parameters, paint supply parameters, and coating operation parameters.

[0016] In this embodiment, the operating parameters of the coating equipment involve multiple operating indicators of the coating machine. For example, the rotational speed of the drive motor of the coating machine reflects the speed at which the coating machine operates, and its changes directly affect the uniformity of the coating. Another parameter is the tension of the transmission belt of the coating machine; different tensions lead to different vibrations of the transmission belt, which in turn affects the stability of the PCB board during the coating process.

[0017] Paint supply parameters include the pressure of the paint in the delivery pipeline, which determines the speed and flow rate of the paint sprayed from the nozzle. Paint temperature is also an important parameter; temperature variations affect paint viscosity, which is closely related to paint flowability and adhesion. Furthermore, paint atomization parameters, such as the size distribution of atomized particles, are also included in the paint supply parameters, directly impacting the surface quality of the coating.

[0018] The coating operation parameters cover the coating path planning, including the coordinates of the start and end points of the coating, as well as the path direction during the coating process. The number of coating repetitions is also a key parameter, as different numbers of repetitions will lead to differences in coating thickness. The distance between the nozzle and the PCB board during coating affects the coverage and uniformity of the coating, and is therefore also recorded in the coating operation parameters.

[0019] High-frequency acquisition was used to collect these parameters, ensuring that subtle changes were captured. The collected parameters were arranged in chronological order to form a sequence of process parameters for each coating batch, which was then associated with a batch identifier and stored in a database.

[0020] Step S112: Obtain coating quality test results for different coating batches. The coating quality test results include coating thickness distribution, coating adhesion performance test results, and coating protection performance test results.

[0021] The coating thickness distribution is detected using precision measuring instruments. Multiple detection points are set up in different areas of the PCB board, and the coating thickness at each point is measured. After the measurements are completed, the thickness values ​​of each detection point are organized into a matrix. This matrix provides a clear view of the coating thickness distribution on the PCB board, showing which areas are thicker and which are thinner.

[0022] Coating adhesion performance testing is conducted using specialized methods. Several test areas are selected on the PCB board, and an adhesion tester is used to perform a peel test on the coating, recording the force required to peel off. Simultaneously, the residue of the coating after peeling is observed, including the size and shape of the remaining area. This information collectively constitutes the coating adhesion performance test results.

[0023] Coating protection performance testing requires simulating the PCB board's operating environment. The coated PCB board is placed in a specific corrosive environment, and after a certain period, the appearance of the coating is observed, such as the appearance of bubbles, cracks, discoloration, etc. Problem areas are marked, and their severity is recorded. In addition, the coating's insulation performance is tested, detecting changes in insulation resistance after exposure to the corrosive environment. These data collectively constitute the coating protection performance test results.

[0024] The test results were categorized and organized according to the coating batches. Each batch of test results was accompanied by a detailed test report, including information such as the test method, test instrument, and test personnel, to facilitate subsequent traceability and analysis.

[0025] Step S113: Record the environmental influencing factors of different coating batches during the coating process. The environmental influencing factors record includes changes in ambient temperature, ambient humidity, and ambient cleanliness.

[0026] Ambient temperature changes are monitored using temperature sensors distributed throughout the workshop. The sensors are strategically placed to suit the workshop's spatial layout, installed around the coating equipment, in corners, and near ventilation openings to ensure a comprehensive view of temperature distribution and changes. The sensors collect temperature data at set intervals, generating a temperature change curve that records the dynamic temperature changes throughout the coating process.

[0027] Environmental humidity changes are monitored by humidity sensors, with monitoring points arranged in the same manner as temperature sensors. The humidity sensors record the humidity levels in the workshop in real time, generating a humidity change curve. This curve reveals humidity fluctuations during the coating process and differences in humidity across different areas.

[0028] Changes in environmental cleanliness are detected using an air particle counter. This counter counts the number of dust particles of different sizes in the air, with detection points positioned above and around the coating area. The data is recorded chronologically to form a cleanliness change sequence, reflecting variations in workshop air quality during the coating process.

[0029] These environmental impact factor records are associated with the corresponding coating batches, and the records indicate the model of the testing instrument, calibration status, and testing personnel to ensure the reliability of the data.

[0030] Step S114: The process parameter sequence, the coating quality test results, and the environmental influencing factor records are associated according to the coating batch to establish a historical process data set containing batch identification information. The batch identification information is used to uniquely identify the execution time and corresponding production line number of each coating batch.

[0031] A unique batch identifier is generated for each coating batch, consisting of the execution time and the production line number. The execution time is accurate to the minute, and the production line number is determined according to the workshop's production line naming rules. For example, "Line-A-202406151030" indicates that the coating batch on production line A started at 10:30 AM on June 15, 2024.

[0032] During the association process, the process parameter sequence, coating quality test results, and environmental impact factor records are bound together using batch identification information. Specifically, an association table is created in the database using the batch identifier as the key. This table contains the storage addresses of the corresponding process parameter sequence, quality test results, and environmental impact factor records for that batch.

[0033] Using the above association method, when querying data related to a specific batch, simply entering the batch identifier will quickly retrieve all relevant information for that batch. Furthermore, to facilitate data statistics and analysis, indexes have been established by time, production line, and other dimensions to improve data query efficiency.

[0034] Step S120: Perform data feature extraction processing on the historical process data set to obtain coating process features in the process parameter sequence, quality index features in the coating quality test results, and environmental features in the environmental impact factor records.

[0035] Data feature extraction involves extracting information from raw data that reflects its essential characteristics. For process parameter sequences, it is necessary to analyze their variation patterns over different time periods and the relationships between parameters in order to extract the characteristics of the coating process.

[0036] Coating quality inspection results contain a large amount of test data. By statistically analyzing this data, key indicators that characterize coating quality can be extracted, forming quality indicator features. Environmental impact factor records need to be analyzed from both temporal and spatial dimensions to extract the characteristics of environmental changes, i.e., environmental features.

[0037] In the process of feature extraction, appropriate algorithms and tools are needed to ensure that the extracted features are representative and discriminative. Simultaneously, the extracted features should be standardized so that different types of features can be compared and analyzed on the same dimension.

[0038] Step S121: Perform time series analysis on the process parameter sequence in the historical process data set, extract the trend characteristics of the process parameter sequence in different time periods and the correlation characteristics between parameters, and combine the trend characteristics and the correlation characteristics to form coating process characteristics.

[0039] Time series analysis first requires determining the time granularity of the analysis. Based on the characteristics of the coating process, the entire coating process is divided into several consecutive time periods, the length of which is determined according to the stability of the process and the frequency of parameter changes.

[0040] Within each time period, trend analysis is performed on the process parameter sequence to determine whether the parameters show an upward, downward, or stable trend, as well as the rate of change of the trend. Simultaneously, the correlation between different process parameters is analyzed, such as which parameters change synchronously and which parameters have interdependent relationships.

[0041] The changing trend characteristics of each time period and the correlation characteristics between parameters are combined in chronological order to form the characteristics of the entire coating process. These characteristics can comprehensively reflect the dynamic changes of the coating process and the interaction between parameters.

[0042] Step S1211: Divide the process parameter sequence into multiple consecutive time windows according to the stage division criteria of the coating process, with each time window corresponding to a process stage of the coating process.

[0043] The stages of the coating process are determined based on the flow of the coating process. For example, the coating process can be divided into a pretreatment stage, a primer coating stage, a curing stage, a topcoat coating stage, and a secondary curing stage. Each stage has its specific process requirements and operating steps, and therefore the corresponding time window lengths are also different.

[0044] The pretreatment stage mainly involves cleaning and drying the PCB board; the time window length is determined by the pretreatment process time. The primer coating stage involves applying the first coat; the time window length depends on the coating area and speed. The curing stage's time window is set based on the curing temperature and the coating's characteristics to ensure the coating fully cures.

[0045] By using the above method, the process parameter sequence is divided into time windows corresponding to each process stage, so that the parameter changes in each stage can be analyzed separately, thereby extracting the characteristics of each stage more accurately.

[0046] Step S1212: Perform trend analysis on the process parameter sequence within each time window, calculate the slope change characteristics, extreme point distribution characteristics and stationarity characteristics of the parameter sequence, and combine them to form the change trend characteristics.

[0047] The slope change characteristics are calculated by selecting multiple time points of the parameter sequence within a time window, calculating the ratio of the parameter change between two adjacent time points to the time interval, and obtaining multiple slope values. The distribution range and trend of the above slope values ​​reflect the rate of change characteristics of the parameter within the time window.

[0048] The distribution characteristics of extreme points involve identifying the maximum and minimum points of the parameter sequence within a time window and recording their locations within that window. By analyzing the distribution of extreme points, we can understand the fluctuations of the parameters during that period and identify potential outliers.

[0049] Stationarity characteristics are determined by calculating the fluctuation range of the parameter sequence within a time window. Specifically, the standard deviation of the parameter sequence is calculated; the smaller the standard deviation, the more stable the parameter changes within that time window. Combining slope variation characteristics, extreme point distribution characteristics, and stationarity characteristics forms the trend characteristics of the parameter sequence within that time window.

[0050] Step S1213: Perform correlation analysis on the process parameter sequences within different time windows, calculate the correlation coefficient matrix and cross-covariance characteristics between different process parameters, and combine them to form the correlation characteristics between parameters.

[0051] The correlation coefficient matrix is ​​calculated by taking any two process parameters within different time windows and calculating their correlation coefficients. The magnitude of the correlation coefficient reflects the degree of linear correlation between the two parameters; positive values ​​indicate a positive correlation, negative values ​​indicate a negative correlation, and the larger the absolute value, the stronger the correlation. The correlation coefficient matrix is ​​formed by arranging the correlation coefficients between all parameters in matrix form.

[0052] Cross-covariance characteristics are used to analyze the correlation between two process parameters under different time lags. For example, the covariance between parameter A at time t and parameter B at time t+1, and the covariance between parameter A at time t+1 and parameter B at time t, can be calculated. By analyzing the magnitude and sign of these covariance values, it can be determined whether a lag correlation exists between the two parameters.

[0053] By combining the correlation coefficient matrix and the cross-covariance feature, a correlation feature between parameters is formed, which can reflect the mutual influence relationship between different process parameters at different stages.

[0054] Step S1214: The variation trend characteristics and correlation characteristics between parameters of the same coating batch in each time window are spliced ​​together in chronological order to generate the coating process characteristics of the coating batch.

[0055] The coating process of the same batch includes multiple stages, each with its own trend characteristics and correlations between parameters. During splicing, the characteristics of the pretreatment stage are placed first, followed by the characteristics of the primer coating stage, curing stage, topcoat coating stage, and secondary curing stage, in that order.

[0056] The spliced ​​coating process characteristics include the parameter change trends and the correlation between parameters at each stage of the entire coating process, which can comprehensively reflect the overall situation of the coating process of this batch.

[0057] Step S122: Perform statistical feature analysis on the coating quality test results in the historical process data set, calculate the distribution characteristics and variation characteristics of each test index in the coating quality test results, and combine the distribution characteristics and variation characteristics to form quality index characteristics.

[0058] Coating quality inspection results include multiple inspection indicators, each with a large amount of data. Statistical feature analysis involves in-depth analysis of historical process data to extract features that reflect the data distribution patterns and degree of variation.

[0059] Distribution characteristics mainly include the central tendency and dispersion of the data, such as indicators reflecting central tendency like the mean, median, and mode, and indicators reflecting dispersion like variance and standard deviation. Variation characteristics include the range of data variation and extreme values.

[0060] By calculating and organizing the distribution and variation characteristics of various detection indicators, they are combined to form quality indicator characteristics. These quality indicator characteristics can comprehensively reflect the quality status of the coating and facilitate correlation analysis with other characteristics.

[0061] Step S1221: Perform statistical analysis on the coating thickness distribution in the coating quality inspection results, calculate the mean, median, mode and quantile characteristics of the coating thickness, and combine them to form the coating thickness distribution characteristics.

[0062] The mean characteristic of coating thickness is the average value obtained by summing the thickness values ​​of all detection points and dividing by the number of detection points. It reflects the average thickness level of the coating. The median characteristic is the value of the middle position after arranging the thickness values ​​of all detection points in ascending order. If the number of detection points is even, the average of the two middle values ​​is taken. It is not affected by extreme values ​​and better reflects the central position of the data.

[0063] The mode feature refers to the thickness value that appears most frequently among all the thickness values ​​detected, reflecting the most common value for coating thickness. The quantile feature divides the sorted thickness values ​​according to a set proportion, such as the 1 / 4 quantile, 1 / 2 quantile (i.e., the median), and 1 / 3 quantile. These quantiles reflect the numerical distribution of coating thickness at different proportional positions. Therefore, combining these features forms the coating thickness distribution feature, which comprehensively reflects the distribution of coating thickness on the PCB board.

[0064] Step S1222: Perform statistical analysis on the coating adhesion performance test results in the coating quality test results, calculate the fluctuation range characteristics, dispersion characteristics and central tendency characteristics of the adhesion performance test values, and combine them to form coating adhesion performance characteristics.

[0065] The fluctuation range of adhesion performance test values ​​is obtained by subtracting the minimum value from the maximum value, reflecting the degree of difference in adhesion performance between different test points. The dispersion characteristic is determined by calculating the standard deviation of the test values; the larger the standard deviation, the more dispersed the distribution of adhesion performance test values ​​and the worse the stability.

[0066] The central tendency features, including the mean, median, and mode, are calculated using the same methods as those used in the coating thickness distribution features. They reflect the overall level of the adhesion performance test values. Combining the fluctuation range features, dispersion features, and central tendency features forms the coating adhesion performance features, which comprehensively reflect the coating adhesion performance status.

[0067] Step S1223: Perform statistical analysis on the coating protection performance test results in the coating quality test results, calculate the trend characteristics, stability characteristics and extreme value characteristics of the protection performance test values, and combine them to form the coating protection performance characteristics.

[0068] The trend characteristics of protective performance test values ​​are obtained by analyzing the changes in test values ​​at different test time points or under different test conditions. For example, in corrosion testing, does the trend of protective performance test values ​​gradually decrease or remain stable over time?

[0069] Stability characteristics are determined by calculating the coefficient of variation (COP) of the measured values. The COP is the ratio of the standard deviation to the mean, reflecting the relative dispersion of the measured values. The smaller the COP, the more stable the protective performance. Extreme value characteristics refer to abnormally large or small values ​​in the measured values, as well as the frequency and location of these values, reflecting potential weaknesses in the coating's protective performance. Combining these characteristics forms the coating's protective performance characteristics, which comprehensively reflect the coating's protective performance status.

[0070] Step S1224: Perform feature fusion processing on the coating thickness distribution characteristics, coating adhesion performance characteristics and coating protective performance characteristics to generate quality index characteristics of coating quality test results.

[0071] Feature fusion processing is performed using feature stitching. First, the sub-features in the coating thickness distribution feature are arranged in a predetermined order to form a feature vector. Then, the coating adhesion performance feature and the coating protection performance feature are also formed into feature vectors.

[0072] Finally, these three feature vectors are concatenated in sequence to form the quality index features. This concatenation method preserves the original information of each feature while integrating them together, facilitating subsequent correlation analysis and model training.

[0073] Step S123: Perform spatiotemporal feature analysis on the environmental impact factor records in the historical process data set, extract the variation amplitude features and spatiotemporal distribution features of the environmental impact factors during the coating process, and combine the variation amplitude features and the spatiotemporal distribution features to form environmental features.

[0074] Environmental influencing factors change over time and space during the coating process. Spatiotemporal feature analysis aims to capture the characteristics of these changes. The magnitude of change reflects the degree of change of environmental factors over time, while the spatiotemporal distribution characteristics reflect the spatial distribution of environmental factors and their temporal patterns of change.

[0075] By analyzing records of environmental influencing factors, the variation range characteristics and spatiotemporal distribution characteristics are calculated and combined to form environmental characteristics, which can comprehensively reflect the environmental conditions during the coating process.

[0076] Step S1231: Calculate the variation characteristics of environmental influencing factors during the coating process. For changes in ambient temperature, calculate the range of variation of the ratio of temperature difference to time difference at different time points. For changes in ambient humidity, use the same method to calculate. For changes in ambient cleanliness, calculate the range of variation of the ratio of the difference in the number of dust particles to time difference at different time points.

[0077] For ambient temperature, multiple different combinations of time points are selected from the temperature records of the coating process. For example, time points t1 and t2 are selected, the temperature difference between these two time points is calculated, and then divided by the time difference between t2 and t1 to obtain a ratio. By performing the above calculation on all possible combinations of time points during the coating process, multiple ratios are obtained. The range between the maximum and minimum values ​​of these ratios represents the characteristic of the ambient temperature variation.

[0078] The calculation method for the variation range of ambient humidity is the same as that for temperature, except that the temperature value is replaced by a humidity value. The variation range of ambient cleanliness is determined by dividing the difference in the number of dust particles at different time points by the corresponding time difference, and then determining the range of variation of these ratios.

[0079] Step S1232: Extract the spatiotemporal distribution characteristics of environmental influencing factors. For ambient temperature, record the temperature values ​​of temperature sensors at different locations at different time points to form a two-dimensional temperature distribution matrix. For ambient humidity and ambient cleanliness, use the same method to form humidity distribution matrices and cleanliness distribution matrices respectively.

[0080] In the coating workshop, temperature sensors installed at different locations record temperature values ​​at multiple time points during the coating process. These temperature values ​​are arranged in order of sensor location and time point, forming a two-dimensional matrix. The rows of the matrix represent different time points, the columns represent different sensor locations, and each element in the matrix is ​​the temperature value at the corresponding time point and location; this is the temperature distribution matrix.

[0081] The humidity distribution matrix is ​​formed in the same way as the temperature distribution matrix, except that the matrix elements are humidity values. The cleanliness distribution matrix, on the other hand, is a two-dimensional matrix formed by recording the number of dust particles at different times using air particle counters at different locations.

[0082] Step S1233: Combine the variation amplitude features and spatiotemporal distribution features to form environmental features.

[0083] First, the variation range characteristics of ambient temperature, ambient humidity, and ambient cleanliness are arranged in order to form a variation range feature vector.

[0084] Then, the temperature distribution matrix, humidity distribution matrix, and cleanliness distribution matrix are each expanded into a one-dimensional vector by rows. For example, the temperature distribution matrix has m rows and n columns, which, after expansion, becomes a one-dimensional vector containing m×n elements.

[0085] Finally, the variation amplitude feature vector is concatenated with these three expanded matrix vectors in sequence to form the environmental features. This concatenation method can simultaneously include the variation amplitude and spatiotemporal distribution information of environmental factors.

[0086] Step S124: Perform feature standardization processing on the coating process features, the quality index features, and the environmental features, and align the standardized features according to the coating batch to generate a structured feature set.

[0087] Feature standardization is performed to eliminate the influence of differences in units and numerical ranges between different features. For each sub-feature in the coating process, the mean and standard deviation of the sub-feature are calculated across all coating batches. Then, the mean is subtracted from the sub-feature value for each batch, and the result is divided by the standard deviation to obtain the standardized sub-feature value.

[0088] The same method was used to standardize the quality indicator characteristics and environmental characteristics. After standardization, the standardized coating process characteristics, quality indicator characteristics, and environmental characteristics of each coating batch were linked to ensure that the three characteristics of each batch corresponded one-to-one.

[0089] The associated features are sorted according to batch identifiers to form a structured feature set. In this structured feature set, each entry contains complete feature information for a coated batch, which facilitates subsequent association rule mining and model training.

[0090] Step S130: Based on the coating process characteristics, the quality index characteristics, and the environmental characteristics, perform association rule mining to identify key process parameters and corresponding environmental influence conditions that affect coating quality.

[0091] The purpose of association rule mining is to identify the intrinsic relationships between coating process characteristics, environmental characteristics, and quality indicator characteristics. By analyzing these relationships, it is determined which combinations of process parameters and environmental conditions will have a significant impact on coating quality.

[0092] First, the structured feature set is used as input data and processed using an association rule mining algorithm. Specifically, it can find frequent patterns and associations between features to generate a series of association rules.

[0093] Then, the generated association rules are screened and evaluated, retaining those with statistical significance and practical meaning. By analyzing these rules, key process parameters and corresponding environmental impact conditions that have a significant impact on coating quality are extracted.

[0094] Step S131: Construct an association analysis dataset containing the coating process features, the quality index features, and the environmental features, wherein each record in the association analysis dataset corresponds to a feature combination of a coating batch.

[0095] From a structured set of features, the coating process characteristics, quality index characteristics, and environmental characteristics of each coating batch are integrated into a single record. Each record is uniquely identified by a batch identifier and contains all characteristic information of that batch during the coating process.

[0096] When constructing the association analysis dataset, it is necessary to check the completeness of each record to ensure that no feature values ​​are missing. For records with missing values, the missing information is supplemented by comparing them with the original data of that batch. Simultaneously, the data format is standardized to ensure that all feature values ​​are represented consistently, facilitating subsequent algorithmic processing.

[0097] Step S132: Perform feature selection processing on the correlation analysis dataset to retain relevant features that affect coating quality.

[0098] The purpose of feature selection is to reduce the number of features, remove those features that are irrelevant to coating quality or have a weak correlation, and improve the efficiency and accuracy of association rule mining.

[0099] A feature importance assessment method is used to calculate the correlation between each feature and the quality indicator features. For example, for each sub-feature in the coating process features and environmental features, its correlation coefficient with each sub-feature in the quality indicator features is calculated.

[0100] A correlation threshold is set, and features with absolute correlation coefficients greater than this threshold are retained, while features with absolute correlation coefficients less than or equal to this threshold are removed. This method yields a simplified correlation analysis dataset containing only relevant features that affect coating quality.

[0101] Step S133: Use an association rule mining algorithm to perform rule mining on the processed association analysis dataset to generate an association rule set containing antecedent features and consequent features. The antecedent features include process parameter features and environmental features in the coating process features, and the consequent features include quality index features.

[0102] Choose a suitable association rule mining algorithm, such as the Apriori algorithm. Input the processed association analysis dataset into the algorithm, and the algorithm will mine frequent itemsets according to the set parameters.

[0103] Frequent itemsets are feature combinations that appear frequently in a dataset. Association rules are generated from frequent itemsets, each consisting of antecedent and consequent features. Antecedent features are combinations of process parameters and environmental features during the coating process, while consequent features are certain sub-features from the quality indicator features.

[0104] For example, an association rule might be "process parameter A is in a certain state and environmental characteristic B is in a certain state → quality indicator C reaches a certain level". Through this method, a large number of association rules are generated, forming an association rule set.

[0105] Step S1331: Discretize the continuous features in the processed association analysis dataset, dividing the continuous features into multiple discrete intervals, with each discrete interval corresponding to a feature state.

[0106] Continuous features refer to those features that can take any value, such as temperature and pressure. Discretization is the process of dividing the above continuous numerical range into several non-overlapping intervals.

[0107] The method for dividing the feature into intervals can be determined based on the distribution of the feature. For example, for a continuous feature, it can be divided into three intervals: "low," "medium," and "high," according to its value, with each interval corresponding to a feature state. The number of intervals is determined based on the actual distribution of the feature and process requirements, ensuring that the feature values ​​within each interval have similar meaning and impact.

[0108] After discretization, the value of each continuous feature is converted into the corresponding feature state, which is convenient for association rule mining algorithms to process.

[0109] Step S1332: Perform transactional processing on the discretized association analysis dataset, converting the feature combination of each coating batch into a transaction item set, wherein the transaction item set contains multiple feature items.

[0110] Transactional processing represents each coating batch's feature combination as a transaction item set. Each feature item is a discretized feature state. For example, in the coating process features of a certain coating batch, the state of process parameter A is "low" and the state of process parameter B is "medium"; in the environmental features, the state of ambient temperature is "suitable" and the state of ambient humidity is "normal".

[0111] Therefore, the transaction item set of this batch includes feature items such as "process parameter A - low", "process parameter B - medium", "ambient temperature - suitable", and "ambient humidity - normal", as well as discrete state items of the quality index features corresponding to this batch.

[0112] Step S1333: Set the minimum support threshold and minimum confidence threshold for the association rule mining algorithm. The minimum support threshold is used to control the frequency of rule occurrence in the dataset, and the minimum confidence threshold is used to control the reliability of the rule.

[0113] The minimum support threshold refers to the minimum frequency at which the antecedent and consequent features must appear simultaneously in an association rule. For example, if the minimum support threshold is set to a certain percentage, only association rules that appear at a frequency of at least that percentage in the dataset will be retained.

[0114] The minimum confidence threshold refers to the minimum probability required for the consequent feature to occur given the presence of the antecedent feature. For example, if the minimum confidence threshold is a certain percentage, then only association rules where the probability of the consequent feature occurring given the presence of the antecedent feature is not less than that percentage will be considered.

[0115] The setting of these two thresholds needs to be determined through multiple trials and adjustments, taking into account the actual dataset and process requirements, to ensure that the discovered association rules have both a certain degree of universality and high reliability.

[0116] Step S1334: Run the association rule mining algorithm to perform frequent itemset mining on the transaction itemset and generate a set of frequent itemsets that meet the minimum support threshold.

[0117] Association rule mining algorithms perform combination analysis on the feature items in the transaction item set and calculate the frequency of each feature item combination in all transaction item sets, i.e., the support.

[0118] The combination of features whose support is not lower than the minimum support threshold is selected; these combinations are called frequent itemsets. For example, if the combination of the features "process parameter A - low" and "quality index C - good" occurs frequently enough to reach the minimum support threshold, then they constitute a frequent itemset.

[0119] The frequent itemset set contains all combinations of feature items that meet the minimum support requirement, and these combinations form the basis for generating association rules.

[0120] Step S1335: Generate a set of association rules based on the set of frequent itemsets. Each association rule includes antecedent features and consequent features. The antecedent features are selected from the process parameter features and environmental features in the coating process features, and the consequent features are selected from the quality index features.

[0121] For each frequent itemset, select some features as antecedent features and the remaining quality-related features as consequent features to form an association rule. For example, if the frequent itemset contains "process parameter A - low", "ambient temperature - suitable", and "quality index C - good", an association rule can be generated: "process parameter A - low and ambient temperature - suitable → quality index C - good".

[0122] When generating association rules, it is necessary to ensure that the antecedent features only include process parameter features and environmental features from the coating process features, and the consequent features only include quality index features. Through this method, a large number of association rules are generated, forming an association rule set.

[0123] Step S134: Perform rule evaluation processing on the set of association rules, calculate the support feature, confidence feature and lift feature of each association rule, and filter out strong association rules based on the support feature, the confidence feature and the lift feature.

[0124] Rule evaluation is a process of selecting rules with practical significance and high reliability from a set of association rules. Support, confidence, and lift are commonly used evaluation metrics.

[0125] Support features reflect the frequency of association rules in the dataset, confidence features reflect the reliability of the rules, and lift features reflect the influence of antecedent features on consequent features. By calculating these three features, each association rule is quantitatively evaluated, and then strongly associated rules are selected based on a set threshold.

[0126] Step S1341: For each association rule in the association rule set, calculate the frequency of the rule in the association analysis dataset as a support feature, where the support feature represents the probability that the antecedent feature and the consequent feature in the rule appear simultaneously.

[0127] The support feature of an association rule is obtained by counting the number of coating batches in the association analysis dataset where both the antecedent and consequent features appear simultaneously, and then dividing the result by the total number of batches in the association analysis dataset.

[0128] For example, if the antecedent and consequent features of a certain association rule appear 20 times in 100 coating batches, and the total number of batches is 200, then the support feature of this rule is 20 / 200.

[0129] Step S1342: Calculate the confidence feature of each association rule, where the confidence feature represents the conditional probability of the consequent feature occurring given that the antecedent feature is present.

[0130] The statistical association analysis dataset contains the number of batches in which the antecedent feature appears, and the number of batches in which the consequent feature also appears. Dividing the number of batches in which the consequent feature also appears by the number of batches in which the antecedent feature appears gives the confidence score of the association rule.

[0131] For example, if there are 50 batches with the antecedent feature, and 40 of them also have the consequent feature, then the confidence score of the rule is 40 / 50.

[0132] Step S1343: Calculate the lift feature of each association rule. The lift feature represents the degree of influence of the antecedent feature on the consequent feature and is used to measure the association strength of the rule.

[0133] First, calculate the probability that the consequent feature appears alone in the association analysis dataset, which is the number of batches in which the consequent feature appears divided by the total number of batches.

[0134] Then, the confidence feature of the association rule is divided by the probability of the consequent feature appearing alone, and the result is the lift feature. If the lift feature is greater than 1, it means that the occurrence of the antecedent feature helps the occurrence of the consequent feature; if it is equal to 1, it means that there is no obvious association between the antecedent feature and the consequent feature; if it is less than 1, it means that the occurrence of the antecedent feature has an inhibitory effect on the occurrence of the consequent feature.

[0135] Step S1344: Set the support threshold, confidence threshold, and lift threshold, and filter out the association rules that satisfy the following conditions: support feature is greater than the support threshold, confidence feature is greater than the confidence threshold, and lift feature is greater than the lift threshold as strong association rules.

[0136] The setting of support threshold, confidence threshold, and lift threshold needs to be combined with actual business needs and data conditions. For example, the support threshold is set according to the requirement of rule universality, the confidence threshold is set according to the requirement of rule reliability, and the lift threshold is usually set to 1 to filter out rules with positive impact.

[0137] By comparing the support, confidence, and lift features of association rules with their corresponding thresholds, association rules that satisfy all conditions are retained; these rules are called strong association rules.

[0138] Step S135: Analyze the antecedent features in the strong association rules to identify the key process parameters and corresponding environmental influence conditions that have a significant impact on the coating quality.

[0139] For each strongly associated rule, the antecedent features are analyzed to extract the process parameter features and environmental features contained therein. For example, if the antecedent features of a strongly associated rule are "process parameter A - low", "process parameter B - medium", and "ambient temperature - suitable", then the corresponding process parameters are A and B, and the corresponding environmental influence condition is "ambient temperature is suitable".

[0140] By summarizing and statistically analyzing the antecedent features of all strongly correlated rules, we can identify the process parameters and environmental influence conditions that appear frequently in multiple strongly correlated rules. These are the key process parameters and corresponding environmental influence conditions that have a significant impact on coating quality.

[0141] Step S140: Generate a set of process parameter adjustment schemes based on the key process parameters and the corresponding environmental impact conditions. The set of process parameter adjustment schemes includes multiple combinations of process parameters to be verified and corresponding environmental control requirements.

[0142] Based on the identified key process parameters and their corresponding environmental impact conditions, several possible process parameter adjustment schemes are designed. These schemes need to cover different combinations of key process parameters and their corresponding environmental control requirements.

[0143] Step S141: Determine the adjustable process parameter range based on the key process parameters, wherein the adjustable process parameter range includes the value range of each key process parameter.

[0144] Based on the technical specifications of the coating equipment, the performance parameters of the coatings, and past production experience, determine the reasonable range of values ​​for each key process parameter. For example, for the key process parameter of the drive motor speed of the coating machine, determine its adjustable speed range according to the maximum and minimum allowable speeds of the equipment and the requirements of the coating quality.

[0145] The range of values ​​for each key process parameter needs to ensure that the equipment can operate normally and will not have a negative impact on the coating quality when the parameter is adjusted within that range.

[0146] Step S142: Determine the range of environmental control requirements based on the corresponding environmental impact conditions. The range of environmental control requirements includes the environmental temperature control range, the environmental humidity control range, and the environmental cleanliness control range.

[0147] Based on the environmental requirements of the conformal coating process and the control capabilities of the workshop's environmental control system, the control ranges for ambient temperature, humidity, and cleanliness are determined. For example, according to the conformal coating's instructions, its optimal coating ambient temperature has a suitable range. Taking into account the workshop's air conditioning system's regulation capabilities, the control range for ambient temperature is determined.

[0148] The determination of the environmental humidity control range and the environmental cleanliness control range are similar, requiring comprehensive consideration of process requirements and actual control capabilities.

[0149] Step S143: Using experimental design methods, generate multiple sets of process parameter combinations to be verified and corresponding environmental control requirements within the adjustable process parameter range and the environmental control requirements range.

[0150] Choose an appropriate experimental design method, such as orthogonal experimental design or uniform experimental design. Determine the factors and levels of the experiment based on the number of key process parameters and the adjustable range of each parameter.

[0151] For example, given three key process parameters, each with three different levels, orthogonal experimental design can cover all possible combinations of different levels of each parameter by using a small number of experimental combinations. The multiple combinations of process parameters and corresponding environmental control requirements generated through experimental design methods constitute the solutions to be validated.

[0152] Step S144: Perform validity verification on the generated multiple sets of process parameter combinations to be verified and their corresponding environmental control requirements. After removing process parameter combinations and corresponding environmental control requirements that do not meet the process feasibility, organize the process parameter combinations and corresponding environmental control requirements that have passed the validity verification into a set of process parameter adjustment schemes.

[0153] Organize professional technicians to evaluate each set of solutions to be verified, check whether the combination of process parameters is within the safe operating range of the equipment, whether the environmental control requirements can be achieved through the environmental control equipment in the workshop, and whether the solution complies with the coating usage specifications, etc.

[0154] Solutions that do not meet process feasibility requirements, such as parameter combinations exceeding equipment capacity or environmental control requirements that cannot be met, will be eliminated. The approved solutions will be compiled into a set of process parameter adjustment solutions.

[0155] Step S150: Perform simulation verification on the set of process parameter adjustment schemes to determine the optimal combination of process parameters and the corresponding environmental control requirements, and generate process optimization instructions based on the optimal combination of process parameters and the corresponding environmental control requirements.

[0156] When conducting simulation verification, the first step is to build a virtual environment that can simulate the entire process of conformal coating. This virtual environment will reproduce various conditions in actual production as closely as possible, including the operating status of the coating equipment, the physicochemical properties of the coating, and the influence of environmental factors.

[0157] Each set of process parameter combinations to be verified and the corresponding environmental control requirements from the set of process parameter adjustment schemes are input into the virtual environment. The virtual environment will simulate a series of physicochemical changes in the coating process, such as the flow, adhesion, and curing of the coating, based on the input parameters, and record various key data during the simulation process, such as coating thickness changes, temperature field distribution, and humidity field distribution.

[0158] During the simulation, the status at each time point can be monitored and recorded in real time. For example, in the simulated paint spraying stage, the atomization state of the paint after it is sprayed from the nozzle, its trajectory in the air, and its spread on the PCB board surface can be recorded. In the simulated curing stage, the degree of curing in different areas can be recorded over time.

[0159] After the simulation is completed, the simulated coating quality data can be analyzed. This data is compared with preset quality standards to evaluate the coating quality under the given combination of process parameters and environmental control requirements. Evaluation indicators include coating thickness uniformity, adhesion strength, and protective performance, among other aspects.

[0160] For each set of solutions to be validated, the above simulation and evaluation process is repeated. Then, all solutions are ranked according to the evaluation results. The ranking is based on a comprehensive consideration of the achievement of various quality indicators, prioritizing those solutions that perform well in multiple indicators.

[0161] After sorting, the optimal solution is selected based on the combination of process parameters and environmental control requirements corresponding to the best overall performance. Subsequently, a process optimization instruction is generated based on this optimal solution. The process optimization instruction details the specific numerical range of each process parameter that needs adjustment, as well as the specific environmental control requirements, such as temperature being controlled within a certain range and humidity being maintained at a certain level.

[0162] For example, step S151: Construct a simulation model of the conformal coating process, wherein the simulation model of the conformal coating process includes a coating process simulation module, a quality prediction module and an environmental impact simulation module.

[0163] When constructing a coating process simulation module, it is necessary to refer to the mechanical structure and working principle of the coating equipment. For example, to simulate the motion trajectory of a coating robot, it is necessary to input information such as the robot's joint parameters and motion speed limits, and obtain the motion path of the robot's end effector through kinematic calculations.

[0164] The paint spraying simulation needs to consider physical parameters such as paint viscosity, surface tension, and density, as well as parameters such as nozzle orifice diameter and spray pressure. By calculating the flow state of the paint in the nozzle, the atomization process after spraying, and its movement in the air, the deposition process of paint on the PCB board surface is simulated.

[0165] The construction of the quality prediction module requires the collection of a large amount of historical coating quality data as training samples. These samples include data on coating thickness, adhesion strength, and protective performance under different combinations of process parameters. Machine learning algorithms are used to train on these samples to establish a mapping relationship between process parameters and coating quality.

[0166] During training, the algorithm's parameters can be continuously adjusted to improve prediction accuracy. For example, for neural network models, the number of hidden layers, the number of neurons, and the type of activation function can be adjusted. After training, the quality prediction module can quickly predict various coating quality indicators based on the input process and environmental parameters.

[0167] The construction of the environmental impact simulation module needs to consider the influence mechanisms of environmental factors such as temperature, humidity, and cleanliness on the coating process. For example, temperature affects the curing speed and viscosity of the coating, humidity affects the drying time and adhesion performance of the coating, and cleanliness affects the surface quality of the coating.

[0168] The module incorporates mathematical models of these influencing mechanisms, calculating their impact on various stages of the coating process by inputting environmental parameter variation curves. For example, when simulating the effect of temperature on curing, the curing reaction rate at different times can be calculated based on the temperature-time variation curve, thus obtaining the distribution of the degree of curing.

[0169] Step S1511: Construct a coating process simulation module, which includes a robot motion submodule and a paint spraying submodule. The robot motion submodule simulates the movement path and posture changes of the robotic arm based on input parameters such as running speed and movement trajectory; the paint spraying submodule simulates the atomization effect and adhesion of the paint after it is sprayed from the nozzle based on parameters such as paint supply flow rate and nozzle working pressure.

[0170] The core of the robot motion submodule is the kinematics solver. When the running speed and trajectory parameters are input, the kinematics solver calculates the rotation angle and motion speed of each joint based on the robot arm's structural parameters. Using these calculations, the robot arm's motion path and posture changes in three-dimensional space can be simulated in real time.

[0171] During the simulation, collision detection can be performed on the robotic arm's movement to ensure that its trajectory does not collide with surrounding equipment or PCB boards. If a potential collision is detected, the motion parameters can be adjusted in time, and the path can be replanned.

[0172] In the paint spraying submodule, the first step is to establish a paint flow model. Based on the paint supply flow rate and nozzle operating pressure, the pressure and velocity distribution of the paint inside the nozzle are calculated. Then, the atomization process of the paint after it is sprayed from the nozzle is simulated using an atomization model to obtain information such as the size distribution and velocity distribution of the atomized particles.

[0173] The movement of atomized coating particles in the air is affected by forces such as gravity and air resistance. By calculating the magnitude and direction of these forces, the trajectory of the coating particles can be simulated. When the coating particles reach the PCB board surface, their adhesion and spreading process can be simulated based on the surface morphology and the viscosity of the coating, thereby obtaining the initial morphology of the coating.

[0174] Step S1512: Construct a quality prediction module. This quality prediction module adopts a deep learning model. The input layer receives feature parameters from the simulation results of the coating process and the simulation results of the environmental impact. The hidden layer contains multiple convolutional layers and fully connected layers. The convolutional layers are used to extract local features, and the fully connected layers are used to synthesize global features. The output layer outputs the prediction results of coating quality, such as coating thickness distribution prediction and adhesion performance prediction.

[0175] When building a deep learning model, the number of neurons in the input layer needs to match the dimensionality of the feature parameters in the simulation results of the coating process and the environmental impact. For example, if the input feature parameters include data in multiple dimensions such as the thickness distribution, temperature distribution, and humidity distribution of the coating, then the number of neurons in the input layer needs to cover these dimensions.

[0176] The settings for convolutional layers need to be determined based on the spatial correlation of the features. For example, for data with spatial distribution characteristics, such as coating thickness distribution, by setting appropriate convolutional kernel size and number, local features at different scales can be extracted, such as regions with abrupt changes in thickness and uniform regions.

[0177] Fully connected layers integrate the local features extracted by convolutional layers and obtain global features through nonlinear transformations. Activation functions are added between fully connected layers to increase the model's nonlinear expressive power; commonly used activation functions include ReLU and sigmoid.

[0178] The number of neurons in the output layer is determined by the number of quality indicators that need to be predicted. For example, if it is necessary to predict three indicators—coating thickness uniformity, adhesion strength, and protective performance—then the output layer will have three neurons, each corresponding to a predicted value for one of these three indicators.

[0179] During model training, a large amount of historical data can be used as training samples. Process and environmental parameters from the historical data are input into the model to obtain predicted quality results, which are then compared with actual quality inspection results to calculate the error. The weights and biases in the model are adjusted using the backpropagation algorithm to continuously reduce the error until the model's prediction accuracy reaches the preset requirements.

[0180] Step S1513: Construct an environmental impact simulation module, which includes a temperature impact submodule, a humidity impact submodule, and a cleanliness impact submodule. The temperature impact submodule simulates the effect of ambient temperature on the coating curing speed; the humidity impact submodule simulates the effect of ambient humidity on the coating viscosity; and the cleanliness impact submodule simulates the effect of airborne dust particles on the coating surface quality.

[0181] In the temperature effect submodule, a relationship model between temperature and curing reaction rate needs to be established. This model references the curing kinetic parameters of the coating and calculates the curing reaction rate at different times based on the ambient temperature variation curve. By integrating the change in curing reaction rate over time, the curing degree distribution of the coating is obtained.

[0182] In the humidity-affected submodule, the first step is to determine the relationship between humidity and coating viscosity. By fitting experimental data to establish a functional relationship between humidity and viscosity, and inputting a curve showing changes in ambient humidity, the coating viscosity at different times can be calculated. Changes in coating viscosity affect its flow properties and adhesion properties, thus impacting the quality of the coating.

[0183] The cleanliness impact submodule can simulate the trajectory of airborne dust particles. Based on the size, density, and airflow velocity of the dust particles, their distribution within the coated area is calculated. When dust particles come into contact with the uncured coating surface, their impact on the surface smoothness and potential defects such as pinholes and impurities can be simulated.

[0184] Step S152: Input each set of process parameter combinations to be verified and the corresponding environmental control requirements from the set of process parameter adjustment schemes into the conformal coating process simulation model.

[0185] When inputting parameters, it is essential to ensure that the format and units of the parameters are consistent with the requirements of the simulation model. For example, the temperature unit in process parameters must be uniformly set to degrees Celsius, and the pressure unit to Pascals. For each scheme to be validated, the parameters will be input into the model in the same format.

[0186] During the input process, parameters can be checked for validity to ensure they are within the equipment's allowable range. For example, the coating robot's movement speed cannot exceed its maximum rated speed, and the nozzle's working pressure cannot exceed its pressure resistance limit. If invalid parameters are found, the operator can be prompted to make corrections.

[0187] Step S153: Simulate the coating process using the coating process simulation module to obtain the coating process simulation results, which include robotic arm motion trajectory data, paint spray trajectory data, and preliminary coating formation data.

[0188] The robotic arm's motion trajectory data will be recorded in the form of three-dimensional coordinates, including the position and orientation of the robotic arm's end effector at each moment. The above data can be displayed through visualization software to intuitively observe whether the robotic arm's motion path is reasonable.

[0189] Paint spray trajectory data includes the movement trajectory information of each paint particle. By analyzing this data, we can understand the distribution of paint during the spraying process and determine whether there are areas of uneven spraying. Preliminary coating formation data records information such as the initial thickness distribution and surface morphology of the coating on the PCB board surface.

[0190] Step S154: Simulate the impact of environmental factors on the coating process through the environmental impact simulation module to obtain environmental impact simulation results. These results include data on the impact of temperature on curing speed, data on the impact of humidity on coating viscosity, and data on the impact of cleanliness on the coating surface.

[0191] The effect of temperature on curing speed will be presented as a curve showing the degree of curing over time and space. This curve allows us to understand the curing progress in different areas and determine whether there is incomplete or over-curing.

[0192] Data on the effect of humidity on coating viscosity will record the coating viscosity values ​​at different times, as well as the impact of viscosity changes on the coating's flow properties. Data on the effect of cleanliness on the coating surface will record the attachment location and quantity of dust particles on the coating surface, as well as the degree of influence of these particles on the surface roughness and smoothness of the coating.

[0193] Step S155: Input the simulation results of the coating process and the simulation results of the environmental impact into the quality prediction module to predict the coating quality result, which includes the coating thickness distribution prediction result, adhesion performance prediction result, and protective performance prediction result.

[0194] When inputting simulation results, the data needs to be preprocessed, such as normalized and standardized, to meet the input requirements of the quality prediction module. The preprocessed data stream enters the model through the input layer, undergoes multiple layers of computation within the model, and finally yields the prediction results from the output layer.

[0195] The coating thickness distribution prediction results will be displayed as a two-dimensional image, with different colors representing different thickness values, intuitively reflecting the coating thickness uniformity. The adhesion performance prediction results and protection performance prediction results will be presented in numerical form, with these values ​​corresponding to the respective quality levels.

[0196] Step S156: Evaluate the coating quality prediction results of each set of process parameter combinations to be verified and the corresponding environmental control requirements, and determine the evaluation index values.

[0197] The evaluation metrics are set with reference to relevant industry standards and the company's quality requirements. For example, the evaluation metric for coating thickness uniformity can be the standard deviation of the thickness; the smaller the standard deviation, the better the uniformity. The evaluation metric for adhesion performance can be the minimum adhesion strength; the larger the minimum value, the more reliable the adhesion performance.

[0198] For each set of prediction results, specific values ​​for each evaluation indicator are calculated. Then, based on the importance of each indicator, different weights are assigned, and a weighted sum is obtained to obtain a comprehensive evaluation indicator value. This comprehensive evaluation indicator value can be used to measure the overall quality level of the set of solutions.

[0199] Step S157: Sort the process parameter combinations and corresponding environmental control requirements in the process parameter adjustment scheme set according to the evaluation index values, and select the process parameter combination and corresponding environmental control requirements with the optimal evaluation index values ​​as the optimal process parameter combination and corresponding environmental control requirements.

[0200] During the ranking process, all schemes are first initially ranked from highest to lowest based on their comprehensive evaluation index values. For schemes with the same comprehensive evaluation index value, a secondary ranking is then conducted based on the merits of individual indicators. For example, in the case of identical comprehensive scores, the scheme with better coating thickness uniformity is given priority.

[0201] The sorting method described above allows for the rapid selection of the best-performing solution. After determining the optimal solution, it is necessary to further validate it to ensure that it maintains good performance under various possible interference factors.

[0202] Step S158: Generate a process optimization instruction based on the optimal combination of process parameters and the corresponding environmental control requirements. The process optimization instruction is used to guide the parameter adjustment of the conformal coating process.

[0203] Process optimization instructions are written in a standardized format, clearly specifying the exact value or range of each process parameter that needs to be adjusted. For example, the operating speed of the coating robot should be adjusted to a certain range, and the working pressure of the nozzle should be set to a certain value.

[0204] The instructions also specify the detailed environmental control requirements, such as maintaining the workshop temperature within a certain range, humidity at a certain level, and cleanliness at a certain grade. To facilitate understanding and execution by operators, the instructions also include corresponding adjustment procedures and precautions.

[0205] Step S210: Train the simulation model of the conformal coating process to improve its prediction accuracy.

[0206] Training the model requires a large amount of historical data as a training set. This data includes process parameters, environmental parameters, and corresponding actual coating quality inspection results from past production. The data is then divided into a training set and a validation set according to a predetermined ratio. The training set is used to learn the model parameters, while the validation set is used to evaluate the model's training effectiveness.

[0207] During training, process and environmental parameters from the training set can be input into the model to obtain predicted quality results. Then, the error between the predicted results and the actual detection results is calculated, and the weights and bias parameters in the model are adjusted using the backpropagation algorithm to reduce the error.

[0208] Training is iterated multiple times, and the model is evaluated using a validation set after each iteration. If the error on the validation set continues to decrease, it indicates that the model is continuously optimizing; when the error on the validation set stabilizes or begins to rise, training is stopped to avoid overfitting.

[0209] Step S211: Collect the process parameter sequence, environmental influencing factor records, and corresponding actual coating quality test results of the new coating batch to form a training dataset.

[0210] When collecting new data, it is essential to ensure the completeness and accuracy of the data. For process parameter sequences, the changes in every key parameter during the coating process must be recorded; for environmental influencing factors, the measurement data of parameters such as temperature, humidity, and cleanliness must be accurate; for actual coating quality test results, standardized testing methods must be used to ensure the reliability of the test data.

[0211] The collected data will be organized and stored in a standardized format. To facilitate model training, the data will also be preprocessed, such as removing outliers and imputing missing values. The preprocessed dataset will be merged with historical datasets to continuously enrich the scale and diversity of the training data.

[0212] Step S212: Input the process parameter sequence and environmental influencing factors recorded in the training dataset into the conformal coating process simulation model to obtain the model's prediction results.

[0213] The process of inputting data is similar to step S152, and it is necessary to ensure that the data format and units meet the requirements of the model. The model will simulate the coating process based on the input parameters and output the predicted quality results. The above prediction results will correspond one-to-one with the actual coating quality detection results in the training dataset, and will be used for subsequent error calculation and model parameter adjustment.

[0214] Step S213: Calculate the error between the predicted result and the actual coating quality test result, and use the mean absolute error and root mean square error as error evaluation indicators.

[0215] Mean absolute error (MAE) is the average of the absolute errors between the predicted and actual results, reflecting the average magnitude of the error. Root mean square error (RMSE) is the square root of the average of the squares of the errors; it is more sensitive to larger errors and reflects the degree of dispersion of the error.

[0216] By calculating these two error metrics, we can gain a comprehensive understanding of the model's prediction accuracy. A large error indicates that the model's performance needs improvement, requiring further adjustment of model parameters or additional training data.

[0217] Step S214: Adjust the parameters of the model based on the error results. For the coating process simulation module, adjust the parameters in the coating flow model and atomization model; for the quality prediction module, adjust the weights and biases in the deep learning model; for the environmental impact simulation module, adjust the coefficients in the temperature, humidity, and cleanliness impact models.

[0218] When adjusting the parameters of the coating process simulation module, you can refer to the phenomena observed during the actual coating process. For example, if the simulated coating atomization effect differs significantly from the actual situation, you can adjust parameters such as the surface tension coefficient and viscosity coefficient in the atomization model.

[0219] For the quality prediction module, the connection weights between layers and the bias values ​​of neurons in the deep learning model can be adjusted based on the backpropagation results of the error. The adjustment range is determined by the magnitude of the error and the model's learning rate. The learning rate should be set appropriately; too high a rate may lead to model instability, while too low a rate will slow down the training process.

[0220] The parameters of the environmental impact simulation module are adjusted to reflect the actual impact of the environment on coating quality. For example, if the effect of temperature on the curing rate is more significant in reality than in the model simulation, the reaction rate coefficient in the temperature impact model can be increased.

[0221] Step S215: Repeat steps S212-S214 until the model's prediction error is less than the set error threshold, and complete the model training.

[0222] Each repetition of the process optimizes the model; by continuously adjusting the parameters, the model's prediction accuracy gradually improves. When the model's prediction error falls below a set threshold, it indicates that the model meets the needs of actual production and can be used to guide process optimization.

[0223] After training is complete, the model can be saved for direct use in subsequent process optimization. The training process and parameter settings can also be recorded.

[0224] Figure 2 The illustration shows exemplary hardware and software components of a conformal coating process optimization system 100 based on big data analysis, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the conformal coating process optimization system 100 based on big data analysis and to perform the functions in this application.

[0225] For example, the conformal coating process optimization system 100 based on big data analytics may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the conformal coating process optimization system 100 based on big data analytics may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The conformal coating process optimization system 100 based on big data analytics also includes an I / O interface 150 between the computer and other input / output devices.

[0226] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned method for optimizing conformal coating process based on big data analysis is implemented.

[0227] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for optimizing conformal coating processes based on big data analysis, characterized in that, include: Obtain a set of historical process data during the execution of the conformal coating process, including the process parameter sequence of different coating batches, coating quality test results and corresponding environmental influencing factor records; Data feature extraction processing is performed on the historical process data set to obtain the corresponding coating process features, quality index features and environmental features; A correlation analysis dataset containing coating process features, quality index features, and environmental features is constructed, where each record in the correlation analysis dataset corresponds to a feature combination of a coating batch; feature selection processing is performed on the correlation analysis dataset to retain relevant features that affect coating quality; An association rule mining algorithm is used to process the association analysis dataset, generating a set of association rules containing antecedent and consequent features. Antecedent features include process parameters and environmental features from the coating process, while consequent features include quality indicators. The association rule set is then evaluated by calculating the support, confidence, and lift features of each rule. Strong association rules are selected based on these features. The antecedent features of the strong association rules are analyzed to identify key process parameters and corresponding environmental conditions that significantly affect coating quality. A set of process parameter adjustment schemes is generated based on key process parameters and corresponding environmental impact conditions, including multiple combinations of process parameters to be verified and corresponding environmental control requirements. A simulation model for conformal coating process is constructed, including a coating process simulation module, a quality prediction module, and an environmental impact simulation module. Each set of process parameters to be verified and the corresponding environmental control requirements from the set of process parameter adjustment schemes are input into the conformal coating process simulation model. The coating process is simulated through the coating process simulation module to obtain the coating process simulation results. The environmental impact simulation module simulates the impact of environmental factors on the coating process to obtain the environmental impact simulation results. The coating process simulation results and the environmental impact simulation results are input into the quality prediction module to predict the coating quality results. The coating quality prediction results for each set of process parameter combinations to be verified and the corresponding environmental control requirements are evaluated to determine the evaluation index values. Based on the evaluation index values, the process parameter combinations and corresponding environmental control requirements in the process parameter adjustment scheme set are sorted, and the process parameter combination and corresponding environmental control requirements with the optimal evaluation index values ​​are selected as the optimal process parameter combination and corresponding environmental control requirements. Based on the optimal process parameter combination and corresponding environmental control requirements, a process optimization instruction is generated, which is used to guide the parameter adjustment of the conformal coating process.

2. The method for optimizing conformal coating process based on big data analysis according to claim 1, characterized in that, The acquisition of the historical process data set during the execution of the conformal coating process includes: Collect process parameter sequences for different coating batches during the coating process, the process parameter sequences including coating equipment operating parameters, paint supply parameters and coating operation parameters; Obtain coating quality test results for different coating batches, including coating thickness distribution, coating adhesion performance test results, and coating protective performance test results; Record the environmental impact factors of different coating batches during the coating process. The environmental impact factor records include changes in ambient temperature, changes in ambient humidity, and changes in ambient cleanliness. The process parameter sequence, coating quality test results, and environmental influencing factor records are associated according to coating batches to establish a historical process data set containing batch identification information. The batch identification information is used to uniquely identify the execution time and corresponding production line number of each coating batch.

3. The method for optimizing conformal coating process based on big data analysis according to claim 1, characterized in that, The process of extracting data features from the historical process data set yields coating process features in the process parameter sequence, quality index features in the coating quality detection results, and environmental features in the environmental impact factor records, including: Time series analysis is performed on the process parameter sequences in the historical process data set to extract the trend characteristics of the process parameter sequences in different time periods and the correlation characteristics between parameters. The trend characteristics and the correlation characteristics are combined to form coating process characteristics. Statistical feature analysis is performed on the coating quality test results in the historical process data set to calculate the distribution characteristics and variation characteristics of each test index in the coating quality test results, and the distribution characteristics and variation characteristics are combined to form quality index characteristics. Spatiotemporal feature analysis is performed on the environmental impact factor records in the historical process data set to extract the variation amplitude and spatiotemporal distribution characteristics of the environmental impact factors during the coating process. The variation amplitude characteristics and the spatiotemporal distribution characteristics are combined to form environmental features. The coating process features, quality index features, and environmental features are standardized, and the standardized coating process features, quality index features, and environmental features are aligned according to coating batches to generate a structured feature set.

4. The method for optimizing conformal coating process based on big data analysis according to claim 3, characterized in that, The process parameter sequence in the historical process data set is subjected to time series analysis to extract the changing trend characteristics and correlation characteristics between the process parameter sequences in different time periods. The changing trend characteristics and correlation characteristics are combined to form coating process characteristics, including: The process parameter sequence is divided into multiple consecutive time windows according to the stage division criteria of the coating process, and each time window corresponds to a process stage of the coating process. Trend analysis is performed on the process parameter sequence within each time window to calculate the slope change characteristics, extreme point distribution characteristics, and stationarity characteristics of the parameter sequence, and these are combined to form the change trend characteristics. Correlation analysis was performed on the process parameter sequences within different time windows to calculate the correlation coefficient matrix and cross-covariance characteristics between different process parameters, and these were combined to form the correlation characteristics between parameters. The coating process characteristics of the same coating batch are generated by splicing the trend characteristics and correlation characteristics between parameters in different time windows in chronological order.

5. The method for optimizing conformal coating process based on big data analysis according to claim 3, characterized in that, The statistical feature analysis of the coating quality inspection results in the historical process data set is performed to calculate the distribution characteristics and variation characteristics of each inspection index in the coating quality inspection results. The distribution characteristics and variation characteristics are then combined to form quality index characteristics, including: Statistical analysis was performed on the coating thickness distribution in the coating quality inspection results. The mean, median, mode and quantile characteristics of the coating thickness were calculated and combined to form the coating thickness distribution characteristics. Statistical analysis was performed on the coating adhesion performance test results in the coating quality test results. The fluctuation range characteristics, dispersion characteristics and central tendency characteristics of the adhesion performance test values ​​were calculated and combined to form the coating adhesion performance characteristics. Statistical analysis was performed on the coating protection performance test results in the coating quality test results, and the trend characteristics, stability characteristics and extreme value characteristics of the protection performance test values ​​were calculated and combined to form the coating protection performance characteristics. The coating thickness distribution characteristics, coating adhesion performance characteristics, and coating protective performance characteristics are fused together to generate quality index characteristics of the coating quality test results.

6. The method for optimizing conformal coating process based on big data analysis according to claim 1, characterized in that, The association rule set is subjected to rule evaluation processing, calculating the support feature, confidence feature, and lift feature of each association rule, and strong association rules are selected based on the support feature, the confidence feature, and the lift feature, including: For each association rule in the association rule set, the frequency of the rule's occurrence in the association analysis dataset is calculated as a support feature, where the support feature represents the probability that the antecedent feature and consequent feature of the rule occur simultaneously. Calculate the confidence feature for each association rule, where the confidence feature represents the conditional probability that the consequent feature will occur given that the antecedent feature will occur. Calculate the lift feature of each association rule, whereby the lift feature represents the degree of influence of the antecedent feature on the consequent feature, and is used to measure the association strength of the rule; Set support threshold, confidence threshold, and lift threshold, and filter out association rules that satisfy the following conditions: support feature is greater than support threshold, confidence feature is greater than confidence threshold, and lift feature is greater than lift threshold as strong association rules.

7. The method for optimizing conformal coating process based on big data analysis according to claim 1, characterized in that, The process parameter adjustment scheme set generated based on the key process parameters and corresponding environmental impact conditions includes: The adjustable process parameter range is determined based on the key process parameters, and the adjustable process parameter range includes the value range of each key process parameter. The environmental control requirements range is determined based on the corresponding environmental impact conditions. The environmental control requirements range includes the environmental temperature control range, the environmental humidity control range, and the environmental cleanliness control range. Multiple sets of process parameter combinations and corresponding environmental control requirements to be verified are generated within the range of adjustable process parameters and the range of environmental control requirements. The generated multiple sets of process parameter combinations to be verified and their corresponding environmental control requirements are subjected to validity testing. After removing process parameter combinations and corresponding environmental control requirements that do not meet the process feasibility, the process parameter combinations and corresponding environmental control requirements that have passed the validity test are compiled into a set of process parameter adjustment schemes.

8. The method for optimizing conformal coating process based on big data analysis according to claim 1, characterized in that, The process employs an association rule mining algorithm to perform rule mining on the processed association analysis dataset, generating a set of association rules containing antecedent and consequent features, including: The continuous features in the processed association analysis dataset are discretized by dividing the continuous features into multiple discrete intervals, each of which corresponds to a feature state. The discretized association analysis dataset is processed in a transactional manner, and the feature combination of each coating batch is converted into a transaction item set, which contains multiple feature items. Set a minimum support threshold and a minimum confidence threshold for the association rule mining algorithm. The minimum support threshold is used to control the frequency of rule occurrence in the dataset, and the minimum confidence threshold is used to control the reliability of the rule. Run the association rule mining algorithm to perform frequent itemset mining on the transaction itemset and generate a set of frequent itemsets that meet the minimum support threshold; A set of association rules is generated based on the set of frequent itemsets. Each association rule includes antecedent features and consequent features. The antecedent features are selected from process parameter features and environmental features in the coating process features, and the consequent features are selected from quality index features.

9. A three-proof coating process optimization system based on big data analysis, characterized in that, The device includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the three-conformal coating process optimization method based on big data analysis as described in any one of claims 1-8.

Images