Automatic feeding control method and system for automotive parts electrophoresis line based on multiple sensors

Abstract

This invention relates to the field of automotive parts processing technology, and discloses an automatic feeding control method and system for an automotive parts electrophoresis line based on multiple sensors. The method includes: acquiring initial multi-source data through a sensor matrix; preprocessing the data to obtain a refined feature vector; acquiring feeding deviation values; if the feeding deviation exceeds a threshold, generating a feeding deviation signal; using a support vector machine classifier to classify and predict the set of deviation signal intensity indicators to obtain a classification result of the deviation factors; mapping the classification result to a control strategy library to match dynamic adaptation rules and obtain feeding parameter correction values; if environmental changes dominate the deviation, calculating a first adjustment coefficient using temperature compensation; performing a secondary calibration of the correction value based on the convergence result to generate a feeding command; updating system parameters and determining whether the deviation has been eliminated; and updating model parameters to obtain updated data acquisition rules. This method can achieve accurate identification and real-time correction of feeding deviations, improving control accuracy.

Description

Technical Field

[0001] This invention relates to the field of automotive parts processing technology, and in particular to an automatic feeding control method and system for automotive parts electrophoresis lines based on multiple sensors. Background Technology

[0002] Currently, automotive parts manufacturing is a crucial pillar of modern industry, and electrophoretic coating, as a key process in the surface treatment of automotive parts, is widely used for corrosion prevention and aesthetic enhancement. However, precise control of the feeding process in electrophoretic lines remains a core issue restricting production efficiency and product quality. Traditional feeding control relies on manual experience or simple automated equipment, which struggles to adapt to dynamic changes in complex production environments, leading to unstable quality. Therefore, researching a feeding control method based on IoT data acquisition technology that can dynamically adapt to varying production conditions has become an urgent need to improve the efficiency and reliability of electrophoretic lines.

[0003] Traditional methods struggle to accurately extract material feeding deviation signals from multi-source sensor data and identify the specific factors causing the deviation. For example, in actual production, sudden changes in ambient temperature can alter material flowability, leading to a deviation in feeding volume from expectations. However, existing systems may only detect flow anomalies without determining whether the deviation is caused by temperature changes or equipment malfunction. This incomplete information prevents the control system from adjusting the feeding strategy in a timely manner, ultimately resulting in uneven coating thickness and impacting product quality.

[0004] Existing electrophoresis line feeding control technology has the problem of insufficient control precision, as it is difficult to accurately identify and correct feeding deviations in real time. Summary of the Invention

[0005] This invention provides an automatic feeding control method and system for an automotive parts electrophoresis line based on multiple sensors, so as to achieve accurate identification and real-time correction of feeding deviation and improve control accuracy.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an automatic feeding control method for an automotive parts electrophoresis line based on multiple sensors, comprising:

[0007] Initial multi-source data is acquired through a sensor matrix, and the initial multi-source data is preprocessed and feature extracted to obtain a refined feature vector set.

[0008] Based on the refined feature vector set, the difference between the real-time feeding amount and the preset benchmark feeding amount is calculated to obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal strength indicators.

[0009] A preset support vector machine classifier is used to classify and predict the set of deviation signal intensity indicators to obtain the classification results of deviation factors.

[0010] The classification results of the deviation factors are mapped to a pre-established control strategy library, and the corresponding dynamic adaptation rules are matched to obtain a sequence of feed parameter correction values ​​for the deviation factors.

[0011] If the deviation factor classification result indicates that environmental change is the dominant deviation, then the first adjustment coefficient is obtained through the temperature compensation algorithm, it is determined whether the first adjustment coefficient is within the preset coefficient convergence range, the convergence judgment result is obtained, the feeding parameter correction value sequence is calibrated a second time, and the optimized feeding instruction is generated.

[0012] The control parameters of the electrophoresis line are updated in real time according to the optimized feeding instructions, the updated multi-source data is obtained, and it is determined whether the deviation has been eliminated and recorded as historical data.

[0013] Based on the historical data, the model parameters of the support vector machine classifier are iteratively updated to obtain the updated data collection rules.

[0014] Preferably, the step of acquiring initial multi-source data through a sensor matrix, performing data preprocessing and feature extraction on the initial multi-source data to obtain a refined feature vector set includes:

[0015] Real-time data is collected by a sensor matrix to obtain initial multi-source data. The initial multi-source data is then standardized to obtain an initial multi-source data set.

[0016] The initial multi-source data set is subjected to dimensionality reduction processing. The main feature components are extracted from the initial multi-source data set, and noise and redundant dimensions are removed to obtain a refined feature vector set.

[0017] Preferably, the step involves calculating the difference between the real-time feeding amount and the preset benchmark feeding amount based on the refined feature vector set to obtain a feeding deviation value. If the feeding deviation value exceeds a preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal strength indicators, including:

[0018] The real-time feeding amount is obtained from the refined feature vector set, and the real-time feeding amount is dynamically collected and standardized to obtain a standardized feeding amount dataset.

[0019] The difference between the standardized feeding amount dataset and the preset benchmark feeding amount is calculated to obtain the feeding deviation value;

[0020] If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated through logical judgment to obtain a set of deviation signals;

[0021] Based on the set of deviation signals, feature mapping is performed on the deviation signal intensity to obtain a set of deviation signal intensity indices.

[0022] Preferably, the step of using a preset support vector machine classifier to classify and predict the set of deviation signal intensity indicators to obtain the deviation factor classification result includes:

[0023] If the set of deviation signal strength indicators exceeds a preset deviation strength threshold, then the multi-source data at the time of deviation signal generation is standardized to obtain a standardized feature dataset.

[0024] The standardized feature dataset is input into a preset support vector machine classifier for classification prediction to obtain the classification result of the bias factor.

[0025] Preferably, the step of mapping the classification results of the deviation factors to a pre-established control strategy library, matching the corresponding dynamic adaptation rules, and obtaining a sequence of feed parameter correction values ​​for the deviation factors includes:

[0026] Based on the classification results of the deviation factors, a pre-established control strategy library is queried, and the classification results are used to map and match the corresponding dynamic adaptation rules to obtain a set of dynamic adaptation rules.

[0027] Based on the set of dynamic adaptation rules, the classification results of the deviation factors are analyzed. If the classification result of the deviation factors is environmental change, a first feeding parameter correction value is generated. If the classification result of the deviation factors is material characteristic fluctuation, a second feeding parameter correction value is generated, thus obtaining a set of feeding parameter correction values.

[0028] The set of feed parameter correction values ​​is sorted by priority to obtain a sorted sequence of feed parameter correction values.

[0029] Preferably, if the deviation factor classification result indicates that environmental change is the dominant deviation, then a first adjustment coefficient is obtained through a temperature compensation algorithm, it is determined whether the first adjustment coefficient is within the convergence range of a preset coefficient, a convergence judgment result is obtained, and the feeding parameter correction value sequence is calibrated a second time to generate an optimized feeding instruction, including:

[0030] Based on the initial multi-source data, a first set of environmental deviation values ​​is obtained through data filtering and extraction;

[0031] If the deviation factor classification result indicates that environmental change is the dominant deviation, an adjustment coefficient for the feeding flow rate is calculated using a temperature compensation algorithm based on the first set of environmental deviation values. If the temperature fluctuation exceeds a preset temperature fluctuation threshold, a first adjustment coefficient is generated.

[0032] Determine whether the first adjustment coefficient is within the preset coefficient convergence range to obtain a convergence judgment result. If the convergence result shows that the first adjustment coefficient is not convergent, then perform a second calibration on the feeding parameter correction value sequence to obtain an optimized feeding command.

[0033] Preferably, the step of updating the electrophoresis line control parameters in real time according to the optimized feeding command, obtaining updated multi-source data, determining whether the deviation has been eliminated, and recording it as historical data includes:

[0034] The control parameters of the electrophoresis line are updated according to the optimized feeding instructions, the updated multi-source data is obtained, and the second set of environmental deviation values ​​is obtained through data filtering and extraction.

[0035] If the environmental deviation values ​​in the second set of environmental deviation values ​​do not exceed the preset environmental deviation range, then the deviation is determined to be eliminated, and the updated multi-source data is recorded as historical data.

[0036] Preferably, the step of iteratively updating the model parameters of the support vector machine classifier based on the historical data to obtain the updated data collection rules includes:

[0037] The historical data is preprocessed to obtain the first cleaned dataset;

[0038] Based on the first cleaned dataset, the parameters of the support vector machine classifier are iteratively updated, the classification boundary is adjusted, and an optimized set of classifier parameters is obtained.

[0039] If the optimized classifier parameter set is applied to a preset validation dataset and the recognition accuracy exceeds a preset accuracy threshold, then the validation dataset is processed through bias analysis to obtain a set of bias factors.

[0040] Based on the set of deviation factors, the subsequent data collection strategy is adjusted to obtain the updated data collection rules.

[0041] Secondly, the present invention provides an automatic feeding control system for an automotive parts electrophoresis line based on multiple sensors, comprising:

[0042] The feature vector acquisition module is used to acquire initial multi-source data through a sensor matrix, perform data preprocessing and feature extraction on the initial multi-source data, and obtain a refined feature vector set.

[0043] The feeding deviation detection module is used to calculate the difference between the real-time feeding amount and the preset benchmark feeding amount based on the refining feature vector set, and obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal intensity indicators.

[0044] The deviation factor classification module is used to classify and predict the set of deviation signal intensity indicators using a preset support vector machine classifier to obtain the deviation factor classification result.

[0045] The control strategy matching module is used to map the classification results of the deviation factors to a pre-established control strategy library, match the corresponding dynamic adaptation rules, and obtain a sequence of feed parameter correction values ​​for the deviation factors.

[0046] The environmental deviation adjustment module is used to obtain a first adjustment coefficient through a temperature compensation algorithm if the deviation factor classification result indicates that environmental change is the dominant deviation, determine whether the first adjustment coefficient is within the convergence range of a preset coefficient, obtain the convergence judgment result, perform secondary calibration on the feed parameter correction value sequence, and generate an optimized feed instruction.

[0047] The system update and verification module is used to update the control parameters of the electrophoresis line in real time according to the optimized feeding instructions, obtain updated multi-source data, determine whether the deviation has been eliminated and record it as historical data.

[0048] The classification model optimization module is used to iteratively update the model parameters of the support vector machine classifier based on the historical data, so as to obtain the updated data collection rules.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) This invention collects multi-source data through a sensor matrix, combines principal component analysis to reduce noise, and then uses a support vector machine classifier to iteratively optimize parameters. This can accurately identify deviation factors such as environmental changes and material property fluctuations, reduce misjudgments, and solve the problem of multi-source data interference.

[0051] (2) This invention generates feeding parameter correction values ​​by matching the classification results of deviation factors with the control strategy library, applies a temperature compensation algorithm for environmental changes, and verifies the adjustment effect in real time through feedback loop, thereby quickly correcting feeding deviations. This invention achieves accurate identification and real-time correction of feeding deviations in the feeding control of electrophoresis lines, thus improving control accuracy. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the automatic feeding control method for an automotive parts electrophoresis line based on multiple sensors provided in an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of the automatic feeding control system for an automotive parts electrophoresis line based on multiple sensors provided in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Reference Figure 1 The first embodiment of the present invention provides an automatic feeding control method for an automotive parts electrophoresis line based on multiple sensors, comprising the following steps:

[0056] S1, acquire initial multi-source data through sensor matrix, perform data preprocessing and feature extraction on the initial multi-source data to obtain a refined feature vector set;

[0057] S2, based on the set of refined feature vectors, calculate the difference between the real-time feeding amount and the preset benchmark feeding amount to obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, generate a deviation signal and perform feature mapping to obtain a set of deviation signal strength indicators.

[0058] S3, using a preset support vector machine classifier, classify and predict the set of deviation signal intensity indicators to obtain the classification result of deviation factors;

[0059] S4, map the classification results of the deviation factors to the pre-established control strategy library, match the corresponding dynamic adaptation rules, and obtain the feed parameter correction value sequence for the deviation factors;

[0060] S5, if the deviation factor classification result indicates that environmental change is the dominant deviation, then the first adjustment coefficient is obtained through the temperature compensation algorithm, it is determined whether the first adjustment coefficient is within the preset coefficient convergence range, the convergence judgment result is obtained, the feeding parameter correction value sequence is calibrated for a second time, and the optimized feeding instruction is generated.

[0061] S6. Update the control parameters of the electrophoresis line in real time according to the optimized feeding instruction, obtain the updated multi-source data, determine whether the deviation has been eliminated and record it as historical data.

[0062] S7. Based on the historical data, iteratively update the model parameters of the support vector machine classifier to obtain the updated data collection rules.

[0063] In step S1, initial multi-source data is acquired through a sensor matrix. Data preprocessing and feature extraction are then performed on the initial multi-source data to obtain a refined feature vector set, including:

[0064] S11, real-time data is collected through a sensor matrix to obtain initial multi-source data, and the initial multi-source data is standardized to obtain an initial multi-source data set;

[0065] S12, perform dimensionality reduction processing on the initial multi-source data set, extract the main feature components from the initial multi-source data set, remove noise and redundant dimensions, and obtain a refined feature vector set.

[0066] In step S11, real-time data is collected through a sensor matrix to obtain initial multi-source data. The initial multi-source data is then standardized to obtain an initial multi-source data set.

[0067] It should be noted that the sensor matrix refers to a combination of multiple types of sensors deployed at relevant monitoring locations on the electrophoresis line, such as pipelines. This matrix can collect initial multi-source data in real time during the material feeding process, including flow and pressure signals reflecting the material conveying status, as well as temperature and humidity signals affecting material properties. The flow signal monitors the amount of material conveyed per unit time, reflecting the feeding rate. In this embodiment, the flow data collection range is 0-1000 kg / h. The pressure signal monitors the fluid pressure within the pipeline, reflecting the stability of the conveying pressure. In this embodiment, the pressure data collection range is 0-100 kPa. The temperature and humidity signals refer to the data signals collected in real time by the temperature and humidity sensors in the sensor matrix during the electrophoresis line feeding process, reflecting the temperature and humidity status of the production environment. In this embodiment, the temperature data collection range is -20-80°C, and the humidity data collection range is 0-100%RH. The sampling frequency setting needs to consider the characteristics of the sensor types, taking into account data timeliness to capture dynamic changes in the data, while avoiding oversampling that increases the computational and storage burden. For example, the pressure sensor might be set to collect data 10 times per second, and the temperature and humidity sensor 5 times per second.

[0068] Standardization refers to linearly normalizing the collected initial multi-source data to map it to a unified interval of 0-1, thereby eliminating dimensional differences and obtaining an initial multi-source data set. For example, if the pressure data collection range is 0-100 kPa, and the original pressure value is 50 kPa, the normalized data is (50-0) ÷ (100-0) = 0.5; if the temperature data collection range is -20-80°C, and the original temperature is 30°C, the normalized data is (30-(-20)) ÷ (80-(-20)) = 0.5; if the humidity data collection range is 0-100%RH, and the original humidity is 60%RH, the normalized data is (60-0) ÷ (100-0) = 0.6.

[0069] In step S12, the initial multi-source data set is subjected to dimensionality reduction processing. The main feature components are extracted from the initial multi-source data set, and noise and redundant dimensions are removed to obtain a refined feature vector set.

[0070] It should be noted that in this embodiment, principal component analysis is used to reduce the dimensionality of the initial multi-source dataset. Specifically, the covariance matrix of the standardized initial multi-source data is calculated first. For example, by quantifying the linear correlation between different dimensions such as pressure and temperature, and temperature and humidity, if the covariance between pressure and temperature is 0.7, indicating a strong positive correlation, and the covariance between temperature and humidity is 0.2, indicating a weak correlation, then it is clear that data redundancy mainly exists in the pressure and temperature dimensions. Next, eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors. For example, after decomposition, the eigenvalues ​​are λ1=5.2, λ2=1.8, and λ3=0.5, and the corresponding eigenvectors represent the pressure, temperature, and humidity dimensions, respectively. The larger the eigenvalue, the more original data information the corresponding eigenvector carries. Subsequently, through orthogonal transformation, the standardized data is projected onto a low-dimensional coordinate system composed of the eigenvectors corresponding to the first k largest eigenvalues, such as λ1 and λ2, making the new dimensions independent to eliminate redundancy. Simultaneously, the contribution rate of each principal component is calculated. For example, λ1÷(λ1+λ2+λ3)=5.2÷7.5≈69.3% and λ2÷(λ1+λ2+λ3)=1.8÷7.5=24%, with a cumulative contribution rate of 93.3%, thus identified as the principal feature components. The humidity dimension corresponding to λ3 has a contribution rate of only 6.7%, and is determined to be a redundant dimension. During the process, outliers caused by transient sensor malfunctions, such as sudden changes in 50°C high-temperature data, are also filtered out. This type of invalid data is considered noise. Finally, a refined feature vector set that retains the key information of the original data is obtained, achieving data dimensionality reduction.

[0071] In step S2, based on the refined feature vector set, the difference between the real-time feeding amount and the preset benchmark feeding amount is calculated to obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal strength indicators, including:

[0072] S21, obtain the real-time feeding amount from the refined feature vector set, dynamically collect and standardize the real-time feeding amount to obtain a standardized feeding amount dataset;

[0073] S22, calculate the difference between the standardized feeding amount dataset and the preset benchmark feeding amount to obtain the feeding deviation value;

[0074] S23, if the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated through logical judgment to obtain a set of deviation signals;

[0075] S24. Based on the set of deviation signals, perform feature mapping on the deviation signal intensity to obtain a set of deviation signal intensity indices.

[0076] In step S21, the real-time feeding amount is obtained from the set of refined feature vectors, and the real-time feeding amount is dynamically collected and standardized to obtain a standardized feeding amount dataset.

[0077] It should be noted that the standardization of real-time feeding rate refers to mapping the real-time feeding rate data to a unified range of 0-1 through linear normalization, thereby eliminating the difference in dimensions. For example, if the feeding rate data collection range is 0-1000 kg / h, when the real-time feeding rate is 500 kg / h, the normalized data is (500-0) ÷ (1000-0) = 0.5.

[0078] In step S22, the difference between the standardized feeding amount dataset and the preset benchmark feeding amount is calculated to obtain the feeding deviation value.

[0079] It should be noted that the preset baseline feeding rate is set based on the average feeding rate under historical normal operating conditions. For example, if the average feeding rate under historical normal operating conditions is 600 kg / h, and it is normalized according to the linear normalization standard mentioned above, the preset baseline feeding rate is 0.6. In this embodiment, the feeding deviation is calculated using the absolute difference. For example, if the standardized real-time feeding rate is 0.5 and the preset baseline feeding rate is 0.6, the feeding deviation value is |0.5-0.6|=0.1.

[0080] In step S23, if the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated through logical judgment to obtain a set of deviation signals.

[0081] It should be noted that the preset feeding deviation threshold is set based on the average value of historical feeding deviation data during normal production of the electrophoresis line. For example, if the average deviation of the real-time feeding amount from the baseline value during three months of normal production is 0.08, then the preset feeding deviation threshold is set to 0.08 to ensure coverage of normal fluctuations. A deviation signal is generated when the feeding deviation value exceeds the preset feeding deviation threshold, indicating an abnormal feeding amount. For example, if the real-time feeding deviation value is 0.12, exceeding the preset feeding deviation threshold of 0.08, a deviation signal will be generated. The deviation signal set summarizes relevant information for all deviation signals, such as the corresponding timestamp and the standardized real-time feeding amount at the time of the deviation.

[0082] In step S24, based on the set of deviation signals, feature mapping is performed on the deviation signal intensity to obtain a set of deviation signal intensity indices.

[0083] It should be noted that the deviation signal strength is an indicator that quantifies the severity and correlation characteristics of the deviation signal. When performing feature mapping on the deviation signal strength, the correlation matrix must be constructed first, followed by the mapping operation. During the construction of the correlation matrix, based on historical production data, the linear correlation between the deviation signal and various sensor data, such as temperature, pressure, and humidity, is calculated using the Pearson correlation coefficient. The Pearson correlation coefficient ranges from -1 to 1; the closer the absolute value is to 1, the stronger the correlation. For example, a correlation coefficient of 0.8 between the deviation signal and temperature indicates a strong positive correlation, 0.75 between the deviation signal and pressure indicates a strong positive correlation, and 0.2 between the deviation signal and humidity indicates a weak correlation.

[0084] Next, the absolute values ​​of each correlation coefficient were normalized to obtain the proportion of influence of each sensor data on the feeding deviation, i.e., the correlation weight. The sum of the absolute values ​​of the total correlation coefficients is 0.8 + 0.75 + 0.2 = 1.75. The influence of temperature is 0.8 ÷ 1.75 ≈ 45.7%, rounded to 46%; the influence of pressure is 0.75 ÷ 1.75 ≈ 42.9%, rounded to 43%; and the influence of humidity is 0.2 ÷ 1.75 ≈ 11.4%, rounded to 11%. Based on this, the correlation weights of temperature, pressure, and humidity in the correlation matrix are set to 0.46, 0.43, and 0.11, respectively. The matrix dimension is consistent with the number of sensor types. For example, a 3×3 matrix is ​​constructed for 3 types of sensors, where rows represent sensor types and columns represent deviation signal characteristics.

[0085] Upon entering the feature mapping stage, a specific deviation signal is first extracted from the deviation signal set. For example, at a certain deviation moment, the feeding rate is 400 kg / h, normalized to 0.4. The preset baseline feeding rate is 0.6, the preset feeding deviation threshold is 0.1, and the feeding deviation value of 0.2 exceeds the threshold. Real-time sensor data for that moment are then extracted, such as temperature 35°C, normalized to 0.55; pressure 60 kPa, normalized to 0.6; and humidity 70% RH, normalized to 0.7. Next, the normalized values ​​of each sensor data for this deviation signal are substituted into the correlation matrix for calculation. That is, the deviation value is multiplied by the correlation weight corresponding to each sensor and then summed to obtain 0.55 × 0.46 + 0.6 × 0.43 + 0.7 × 0.11 = 0.588. This value is the intensity index of the deviation signal. All deviation signals in the deviation signal set are processed in this way, converting each deviation signal into a corresponding intensity index. The summation of these indices yields the set of deviation signal intensity indices.

[0086] In step S3, a preset support vector machine classifier is used to classify and predict the set of deviation signal strength indicators, obtaining the classification results of deviation factors, including:

[0087] S31, if the set of deviation signal strength indicators exceeds the preset deviation strength threshold, then the multi-source data at the time of deviation signal generation is standardized to obtain a standardized feature dataset.

[0088] S32, the standardized feature dataset is input into a preset support vector machine classifier for classification prediction to obtain the classification result of the deviation factor.

[0089] In step S31, if the set of deviation signal strength indicators exceeds a preset deviation strength threshold, the multi-source data at the time of deviation signal generation is standardized to obtain a standardized feature dataset.

[0090] It should be noted that the deviation intensity threshold is set based on the average value of historical deviation signal intensity indicators. For example, by statistically analyzing the intensity indicators of all deviation signals during normal production of the electrophoresis line over the past three months, the average value, such as 0.6, is set as the preset deviation intensity threshold. Standardization processing is performed on the multi-source data at the time of deviation signal generation. First, based on the previously constructed correlation matrix, indicators with strong correlation to deviation factors need to be selected. For example, in the correlation matrix, the correlation weight of temperature is 0.46, the correlation weight of pressure is 0.43, and the correlation weight of humidity is 0.11. Indicators with correlation weights below 0.2 contribute very little to the classification and are therefore eliminated; thus, the correlation weight of humidity is removed. Next, standardization processing is performed. Since the selected core feature indicators have different dimensions or numerical ranges due to different sensor types, a linear normalization method is used to map them to a unified interval of 0-1 to eliminate differences. For example, two types of core signal strength indicators are extracted from historical datasets. One type is temperature-related indicators, with historical values ​​ranging from 0.2 to 1.0. Assuming the temperature value at the time the deviation signal occurred was 0.6, standardizing these indicators yields a temperature-related standardized indicator of (0.6-0.2) ÷ (1.0-0.2) = 0.5. The other type is pressure-related indicators, with historical values ​​ranging from 0.4 to 1.0. Assuming the pressure value at the time the deviation signal occurred was 0.8, standardizing these indicators yields a pressure-related standardized indicator of (0.8-0.4) ÷ (1.0-0.4) ≈ 0.67. Through this process of combining feature extraction and standardization, a standardized feature dataset containing a temperature-related standardized indicator of 0.5 and a pressure-related standardized indicator of 0.67 is finally obtained.

[0091] In step S32, the standardized feature dataset is input into a preset support vector machine classifier for classification prediction to obtain the classification result of the bias factor.

[0092] It should be noted that the preset support vector machine (SVM) classifier is trained using a historically standardized feature dataset. Specifically, the historically standardized feature dataset is first divided into a training set and a validation set in an 8:2 ratio. Then, the core parameters of the SVM classifier are initialized. A radial basis function (RBF) kernel suitable for small sample sizes and high-dimensional data is selected, and a penalty coefficient C=1.0 is set to balance the model's fit to the training samples and its generalization ability. The kernel function parameter γ=0.1 is set, with the initial value based on the historical data feature distribution. The training set is then input into the initial SVM classifier. The classifier learns the correspondence between features and deviation categories through the following logic: by calculating the distance between the feature vector of each data point in the training set and two types of deviation factors—environmental change and material property fluctuation—it finds the optimal classification hyperplane. For example, feature vectors with temperature correlation indicators above 0.6 and pressure correlation indicators below 0.2 are closer to the environmental change category, while feature vectors with pressure correlation indicators above 0.7 and temperature correlation indicators below 0.3 are closer to the material property fluctuation category. During training, the model's classification accuracy is tested in real-time using a validation set. For example, a dataset of 1000 data points includes 800 standardized feature data points as the training set and 200 as the validation set. If 180 data points in the validation set are correctly classified, the model accuracy is 90%. If the accuracy does not reach the preset threshold (e.g., 90%), parameters are adjusted. For instance, the penalty coefficient C is adjusted to 1.5 to enhance the model's attention to outliers; or γ is adjusted to 0.2 to optimize the kernel function's mapping effect on features. Training is then repeated until the validation set accuracy reaches the target. In real-time production, the standardized feature dataset is input into a preset support vector machine classifier. The classifier identifies the dominant bias factor for each standardized feature data point as environmental change or material property fluctuation. The classification results of the bias factors for each standardized feature data point are then summarized to obtain a set of bias factor categories.

[0093] In step S4, the classification results of the deviation factors are mapped to a pre-established control strategy library, and corresponding dynamic adaptation rules are matched to obtain a sequence of feed parameter correction values ​​for the deviation factors, including:

[0094] S41, Based on the classification results of the deviation factors, query the pre-established control strategy library, and use the classification results to map and match the corresponding dynamic adaptation rules to obtain a set of dynamic adaptation rules;

[0095] S42, Based on the set of dynamic adaptation rules, analyze the classification results of the deviation factors. If the classification result of the deviation factors is environmental change, generate a first feeding parameter correction value. If the classification result of the deviation factors is material characteristic fluctuation, generate a second feeding parameter correction value, and obtain a set of feeding parameter correction values.

[0096] S43, sort the set of feeding parameter correction values ​​by priority to obtain the sorted sequence of feeding parameter correction values.

[0097] In step S41, based on the classification results of the deviation factors, a pre-established control strategy library is queried, and the classification results are used to map and match the corresponding dynamic adaptation rules to obtain a set of dynamic adaptation rules.

[0098] It should be noted that the pre-established control strategy library is a set of rules built based on historical production data, process standards, and experience in handling deviations from the electrophoresis line. The library stores the correspondence between various deviation factor categories and dynamic adaptation rules, including temperature compensation parameter adjustment rules corresponding to environmental changes and viscosity adjustment strategies corresponding to material property fluctuations. For example, under the environmental change category, based on historical processing experience, it is further subdivided into compensation rules for a ±2℃ temperature fluctuation corresponding to a ±3% feed flow rate and adjustment rules for a ±5% humidity fluctuation corresponding to a ±0.2MPa feed pressure. Under the material property fluctuation category, it is further subdivided into adjustment rules for a ±8cP material viscosity corresponding to a ±50r / min stirring speed and correction rules for a ±2% solid content corresponding to a ±1.5% feed concentration. When using classification results to map and match dynamic adaptation rules, for example, if the set of deviation factors is environmental change and temperature fluctuation of +3℃, the rule corresponding to temperature fluctuation of +3℃ under the environmental change category is accurately located in the control strategy library, and the dynamic adaptation rule of feeding flow rate +4.5% is extracted. If the set also contains material characteristic fluctuation - viscosity +12cP, then the rule of stirring speed +75r / min is matched synchronously, and finally the dynamic adaptation rule set is formed.

[0099] In step S42, the deviation factor classification result is analyzed according to the dynamic adaptation rule set. If the deviation factor classification result is environmental change, a first feeding parameter correction value is generated. If the deviation factor classification result is material characteristic fluctuation, a second feeding parameter correction value is generated, thus obtaining a feeding parameter correction value set.

[0100] It should be noted that when the classification result is environmental change, the process of generating the first feeding parameter correction value is as follows: the rules for environmental change in the dynamic adaptation rule set are called. For example, for every 3°C increase in temperature above the standard value, the feeding amount increases by 2%; for every 5% increase in humidity above the standard value, the feeding rate decreases by 1%. Combined with real-time collected environmental data, for example, if the current temperature is 6°C higher than the standard value and the humidity is 10% higher, the first feeding parameter correction value is calculated according to the rules. The feeding amount needs to be increased by (6÷3)×2%=4%, and the feeding rate needs to be decreased by (10÷5)×1%=2%. When the classification result indicates fluctuations in material properties, the process of generating the second feeding parameter correction value involves calling the corresponding rules. For example, for every 8 cP increase in material viscosity above the standard value, the feeding pressure increases by 4 kPa; for every 2% decrease in solid content below the standard value, the feeding concentration increases by 1.5%. Combining real-time material detection data, for example, if the viscosity is 16 cP higher and the solid content is 4% lower, the calculated feeding pressure increase is (16 ÷ 8) × 4 kPa = 8 kPa, and the feeding concentration increase is (4 ÷ 2) × 1.5% = 3%. These values ​​are the second feeding parameter correction values. Finally, the generated first and second feeding parameter correction values ​​are summed to obtain the feeding parameter correction value set.

[0101] In step S43, the set of feeding parameter correction values ​​is sorted by priority to obtain a sorted sequence of feeding parameter correction values.

[0102] It should be noted that the specific process of prioritization is as follows: First, the parameter type and associated attributes corresponding to each correction value are extracted from the set of feed parameter correction values. Then, the quantitative scores of each dimension in the evaluation rules are matched. For example, the quality impact is 0-5 points, based on the magnitude of the impact of the correction value on the core quality indicators of electrophoretic coating, such as coating thickness and uniformity. An impact exceeding the process threshold by 100% (e.g., coating thickness fluctuation exceeding 5μm) earns 5 points; 50%-100% (e.g., fluctuation of 3-5μm) earns 3-4 points; 20%-50% (e.g., fluctuation of 1-3μm) earns 1-2 points; and <20% (e.g., fluctuation of <1μm) earns 0 points. Here, the process threshold refers to the allowable fluctuation range set on the standard value of the core quality indicators of electrophoretic coating. For example, the coating thickness fluctuation range is 0-5 points. The standard layer thickness is 100μm, with an allowable fluctuation range of ±5μm. The impact is calculated by dividing the actual fluctuation value by the allowable fluctuation value. Safety risk is scored from 0 to 3 points, determined by the ratio of the actual load to the rated pressure of the equipment associated with the correction value. A ratio ≥80% (e.g., pump pressure above 80kPa) earns 3 points; 50%-80% (e.g., 60-80kPa) earns 2 points; 20%-50% (e.g., 30-60kPa) earns 1 point; and <20% (e.g., <30kPa) earns 0 points. Response efficiency is scored from 0 to 2 points, based on the stable production system duration after correction in historical data. Stability of ≤10 minutes earns 2 points; stability of 11-30 minutes earns 1 point; and stability of >30 minutes earns 0 points. The total priority score for each correction value is calculated. For example, a +8kPa increase in feeding pressure corresponds to fluctuations in material properties and quality. An 8kPa pressure fluctuation corresponds to a 3μm fluctuation in coating thickness, an impact of 50%, earning 3 points. Regarding safety risks, the adjusted pressure reaches 88kPa, while the pump's rated pressure is 100kPa, representing 88% of the risk, earning 3 points. Historical data shows stability 8 minutes after adjustment, earning 2 points. The total score is 3+3+2=8 points. A +4% increase in feeding amount corresponds to environmental changes and quality. A 4% fluctuation in feeding amount corresponds to a 6μm fluctuation in coating thickness, earning 5 points. Regarding safety risks, only the feeding amount is adjusted, without affecting the equipment's rated threshold, earning 0 points. Historical data shows stable material supply within 10 minutes after adjustment, earning 2 points. The total score is 5+0+2=7 points. A 2% reduction in the feeding rate is an auxiliary adjustment that affects quality; a 2% rate fluctuation corresponds to a 0.8μm fluctuation in coating thickness, earning 1 point. Safety risk does not involve critical equipment load values, earning 0 points. Response efficiency stabilizes after 25 minutes of adjustment, earning 1 point. The total score is 1+0+1=2 points. Finally, the parameters are sorted from highest to lowest total score to form a sequence of corrected feeding parameters. For example, first increase the feeding pressure by 8kPa, then increase the feeding amount by 4%, and finally reduce the feeding rate by 2%, prioritizing the resolution of high-risk, high-impact parameter deviations.

[0103] In step S5, if the deviation factor classification result indicates that environmental change is the dominant deviation, a first adjustment coefficient is obtained through a temperature compensation algorithm. It is then determined whether the first adjustment coefficient falls within a preset coefficient convergence range, and a convergence judgment result is obtained. The feeding parameter correction value sequence is then calibrated a second time to generate an optimized feeding instruction, including:

[0104] S51, Based on the initial multi-source data, a first set of environmental deviation values ​​is obtained through data filtering and extraction;

[0105] S52, if the deviation factor classification result indicates that environmental change is the dominant deviation, the adjustment coefficient of the feeding flow rate is calculated using a temperature compensation algorithm based on the first set of environmental deviation values. If the temperature fluctuation exceeds the preset temperature fluctuation threshold, a first adjustment coefficient is generated.

[0106] S53, determine whether the first adjustment coefficient is within the preset coefficient convergence range, and obtain the convergence judgment result. If the convergence result shows that the first adjustment coefficient is not converged, then perform a second calibration on the feeding parameter correction value sequence to obtain the optimized feeding command.

[0107] In step S51, based on the initial multi-source data, a first set of environmental deviation values ​​is obtained through data filtering and extraction.

[0108] It should be noted that the steps for data filtering and extracting the first set of environmental deviation values ​​are as follows: First, based on the production requirements of the electrophoresis line, the normal fluctuation ranges of temperature and humidity are preset to be 25-30°C and 50-60%, respectively, and the average value of the normal fluctuation range in the historical normal operation data is set. Then, the temperature and humidity data collected by the real-time sensor are smoothed and filtered. The window size is set to 5, and the arithmetic mean of the data within the window is calculated. If the deviation between the original data and the average value is greater than 3°C for temperature or greater than 5%RH for humidity, it is determined to be a jump value and replaced with the average value. For example, if the temperature data is 28°C, 29°C, 50°C, 27°C, and 28°C, after removing the jump value of 50°C, the average value is 28°C, and the current temperature is replaced with 28°C. Finally, the temperature and humidity data that exceed the normal range are selected and the absolute value of the deviation is calculated. This is the first environmental deviation value. For example, if the temperature is 32℃, which exceeds the upper limit by 30℃, the deviation value is |32℃-30℃|=2℃; if the humidity is 48%RH, which is below the lower limit of 50%RH, the deviation value is |48%RH-50%RH|=2%RH. These are then summarized to form the first environmental deviation value set.

[0109] In step S52, if the deviation factor classification result indicates that environmental change is the dominant deviation, the adjustment coefficient of the feeding flow rate is calculated using a temperature compensation algorithm based on the first set of environmental deviation values. If the temperature fluctuation exceeds the preset temperature fluctuation threshold, a first adjustment coefficient is generated.

[0110] It should be noted that the specific calculation process of the temperature compensation algorithm is as follows: First, the difference between the actual temperature and the standard temperature is extracted from the first set of environmental deviation values. Then, a preset temperature-flow correlation model is called. This model is based on the historical production data of the electrophoresis line for the past two years, containing 1500 sets of effective data on temperature deviation and corresponding flow rate adjustments. It is established through linear regression analysis. For example, with the temperature deviation value as the independent variable x and the feed flow rate adjustment ratio as the dependent variable y, the historical data is fitted to obtain the linear equation y=kx+b, where k is the flow rate adjustment coefficient corresponding to a unit temperature deviation, and b is the basic correction value. After fitting calculation, k=0.8% / ℃ and b=0. The final model rule is that for every 1℃ increase in temperature above the standard temperature, the feed flow rate needs to be increased by 0.8%; for every 1℃ decrease in temperature below the standard temperature, the flow rate needs to be decreased by 0.8%. The system determines whether the temperature fluctuation in the first set of environmental deviation values ​​exceeds a preset temperature fluctuation threshold, such as an allowable normal temperature fluctuation upper limit of ±1℃. If the temperature fluctuation does not exceed the threshold, such as a fluctuation of only 0.5℃, then no adjustment coefficient needs to be generated. If the temperature fluctuation exceeds the threshold, such as a fluctuation of 3℃, then a specific adjustment value is calculated using a temperature compensation algorithm, such as 3℃ × 0.8% = 2.4%, which is the first adjustment coefficient. Summarizing the first adjustment coefficients yields the set of adjustment coefficients.

[0111] In step S53, it is determined whether the first adjustment coefficient is within the preset coefficient convergence range to obtain a convergence judgment result. If the convergence result shows that the first adjustment coefficient is not converged, the feeding parameter correction value sequence is calibrated a second time to obtain an optimized feeding command.

[0112] It should be noted that the preset coefficient convergence range is a safe and effective adjustment range set based on equipment performance and process accuracy requirements. For example, the upper limit of the flow rate adjustment accuracy of the conveying pump is ±4%, and the response error of the feeding valve is ≤±1%. This determines the adjustment range within which the equipment can operate stably. Then, referring to process accuracy standards such as the feeding flow rate fluctuation needing to be ≤5% to ensure coating uniformity, the intersection of the equipment's stable range and process requirements is taken, and the final convergence range is set to ±5%. This means that when the adjustment coefficient is within this range, the feeding parameters can quickly stabilize at the standard value after adjustment, without oscillation or increased deviation. The control unit compares the first adjustment coefficient with this range to obtain a convergence judgment result. If the coefficient is within the range, it is considered convergent; otherwise, it is considered non-convergent. If the convergence result shows that the first adjustment coefficient is non-convergent, the feeding parameter correction value is calibrated a second time. For example, if the calculated first adjustment coefficient is +5.5%, exceeding the ±5% range, it indicates that direct adjustment may cause excessive fluctuations in the feeding flow rate. For instance, if the original flow rate is 500 kg / h, after adjustment by 5.5%, it becomes 527.5 kg / h, exceeding the allowable fluctuation range of the process. At this point, a secondary calibration algorithm is invoked, combining historical adjustment data and real-time deviation trends for correction. Specifically, the extent by which the non-convergent first adjustment coefficient exceeds the preset convergence range is first calculated. Then, this excess is multiplied by a pre-set calibration weight, and the result is the calibration amount used for secondary calibration. In this embodiment, the calibration weight is determined based on nearly one year of historical non-convergent adjustment data, fitted using linear regression analysis. With the excess as the independent variable and the optimal calibration ratio as the dependent variable, a weight of 2.7 is ultimately determined. This weight has been verified to ensure that for common excesses of 1%-3%, the calibrated coefficient falls within the convergence range without over-correction. Subtracting the calibration amount from the first adjustment coefficient, (5.5%-5%)×2.7=1.35%, the correction value after the second calibration is 5.5%-1.35%=4.15%, corresponding to an adjustment coefficient of 4.15%. Within the convergence range, the final optimized feeding command is generated with a feeding flow rate adjusted to 500×(1+4.15%)=520.75kg / h, ensuring stable feeding operation after adjustment.

[0113] In step S6, the control parameters of the electrophoresis line are updated in real time according to the optimized feeding command, the updated multi-source data is obtained, and it is determined whether the deviation has been eliminated and recorded as historical data, including:

[0114] S61, update the control parameters of the electrophoresis line according to the optimized feeding instruction, obtain the updated multi-source data, and extract the second set of environmental deviation values ​​through data filtering;

[0115] S62, if the environmental deviation value in the second set of environmental deviation values ​​does not exceed the preset environmental deviation range, then the deviation is determined to be eliminated, and the updated multi-source data is recorded as historical data.

[0116] In step S61, the control parameters of the electrophoresis line are updated according to the optimized feeding command, the updated multi-source data is obtained, and the second set of environmental deviation values ​​is obtained through data filtering and extraction.

[0117] It should be noted that the acquisition steps for the second set of environmental deviation values ​​are the same as those for the first set of environmental deviation values. Temperature and humidity data that exceed the normal range or reach the fluctuation threshold in the updated multi-source data are filtered out and summarized to form the second set of environmental deviation values.

[0118] In step S62, if the environmental deviation value in the second set of environmental deviation values ​​does not exceed the preset environmental deviation range, then the deviation is determined to be eliminated, and the updated multi-source data is recorded as historical data.

[0119] It should be noted that the preset environmental deviation range is a standard threshold for determining whether the environment has no impact on material feeding. This range is set based on the electrophoresis line's production process requirements, equipment tolerance, and historical normal production data. For example, the allowable temperature fluctuation range is ±1℃, and the allowable humidity fluctuation range is ±5%. As long as the environmental deviation value is within this range, it means that environmental factors will not cause the material feeding to deviate from the standard value. Each data point in the second set of environmental deviation values ​​is compared with the preset environmental deviation range one by one. If all deviation values ​​are within the range, it means that the material feeding deviation previously caused by environmental changes has been resolved. Therefore, the deviation is determined to be eliminated, and the updated multi-source data is recorded as historical data.

[0120] In step S7, based on the historical data, the model parameters of the support vector machine classifier are iteratively updated to obtain the updated data collection rules, including:

[0121] S71, perform data preprocessing on the historical data to obtain the first cleaned dataset;

[0122] S72, based on the first cleaned dataset, iteratively update the parameters of the support vector machine classifier, adjust the classification boundary, and obtain the optimized classifier parameter set;

[0123] S73, if the recognition accuracy exceeds the preset accuracy threshold when the optimized classifier parameter set is applied to the preset verification dataset, then the verification dataset is processed by bias analysis to obtain a set of bias factors.

[0124] S74, Based on the set of deviation factors, adjust the subsequent data acquisition strategy to obtain the updated data acquisition rules.

[0125] In step S71, the historical data is preprocessed to obtain the first cleaned dataset.

[0126] It should be noted that the specific method for preprocessing historical data is as follows: first, perform quality diagnosis on the historical data to identify outliers such as temperature values ​​exceeding the 20-25℃ process range, missing values ​​such as humidity data gaps caused by sensor offline, and redundant values ​​such as continuously repeated normal parameters; then, use the 3σ principle to remove outliers and use the mean to fill in missing values; finally, delete redundant data to obtain the first cleaned dataset.

[0127] In step S72, the parameters of the support vector machine classifier are iteratively updated based on the first cleaned dataset, the classification boundary is adjusted, and an optimized set of classifier parameters is obtained.

[0128] It should be noted that the specific steps for iteratively updating the parameters of the support vector machine classifier are as follows: First, the first cleaned dataset is divided into an iterative training set and an iterative test set according to a preset ratio, such as 7:3. The former is used for parameter adjustment training, and the latter is used to verify the optimization effect. Then, based on the initial parameters of the support vector machine classifier, such as C=1.0 and γ=0.1, the adapted radial basis kernel function is used to carry out multiple rounds of iterative adjustment on the penalty coefficient C and the kernel function parameter γ. The adjustment of C balances the model's tolerance for classification errors of new samples, and the adjustment of γ optimizes the mapping precision of the kernel function to the features in the first cleaned dataset. After each round of adjustments, the iterative training set is input into the classifier. The classifier dynamically fine-tunes the classification hyperplane that distinguishes between environmental changes and material property fluctuations by calculating the distance between the data feature vectors and the two types of bias factors. For example, it makes the newly added low temperature + low viscosity feature vector more accurately classified into the environmental change category. The classification accuracy is then verified using the iterative test set. If the accuracy does not reach the preset accuracy threshold, such as ≥95%, the preset accuracy threshold is set based on historical classification accuracy and production stability targets. The parameter combination is then adjusted, such as C being adjusted to 1.5 and γ to 0.12. The training and verification process is repeated until the accuracy under a certain parameter combination consistently meets the target and the classification boundary adapts to the sample distribution in the first cleaned dataset. Finally, the parameter combination and the optimized classification boundary are integrated to obtain the optimized classifier parameter set.

[0129] In step S73, if the recognition accuracy exceeds the preset accuracy threshold when the optimized classifier parameter set is applied to the preset verification dataset, then the verification dataset is processed through bias analysis to obtain a set of bias factors.

[0130] It should be noted that the validation dataset refers to historical data with clear deviation category labels, such as deviation sample data that has not been updated in the classifier parameters within the past month. Its purpose is to objectively test the generalization ability of the optimized classifier. The specific method of processing the validation dataset using deviation analysis technology is as follows: First, the optimized classifier labels each sample in the validation dataset with either environmental change-dominated deviation or material characteristic fluctuation-dominated deviation. The validation dataset is then split into an environmental deviation group and a material deviation group according to the labels. Each group is analyzed separately, and the percentage of samples within each group whose features exceed the normal range of the process is counted. Features strongly correlated with deviations are selected. In this embodiment, a sample percentage exceeding 70% is considered strongly correlated. The determination of strong correlation is based on the historical deviation data patterns of the electrophoresis line. Weakly correlated features with a percentage below 30% within each group are removed, and a set of deviation factors is formed. For example, the validation dataset contains 200 samples, and the optimized classifier labels 100 environmental deviations and 100 material deviations. After processing, it was found that 82% of the samples in the environmental deviation group had temperature fluctuations exceeding 2℃ and 75% had humidity fluctuations exceeding 8%, while 88% of the samples in the material deviation group had material viscosity exceeding 10%. The final set of deviation factors was {temperature fluctuations exceeding 2℃, humidity fluctuations exceeding 8%, and material viscosity exceeding 10%}.

[0131] In step S74, the subsequent data acquisition strategy is adjusted according to the set of deviation factors to obtain the updated data acquisition rules.

[0132] It should be noted that the process of adjusting the subsequent data acquisition strategy is based on the set of deviation factors. The acquisition frequency and accuracy of the original data acquisition scheme are optimized in a targeted manner to ensure that subsequent acquisition can accurately cover the causes of deviation and reduce redundant data. For example, the original temperature data was collected once every 5 minutes. Since temperature fluctuations exceeding 2°C are a strongly correlated factor, the frequency was adjusted to once every 2 minutes to ensure timely detection of anomalies exceeding the threshold. For indicators in the set that are sensitive to accuracy, the accuracy is improved. For example, if the material viscosity exceeds 10%, it is necessary to accurately identify subtle changes. The original acquisition accuracy was 0.1 Pa / s, and it is now adjusted to 0.01 Pa / s. For indicators with low accuracy requirements, such as humidity fluctuations exceeding 8%, it is only necessary to identify larger changes, so the original accuracy, such as 1%, is maintained. This balances accuracy and data processing costs, and the updated data acquisition rules are obtained after summarizing.

[0133] In summary, this invention employs principal component analysis to reduce noise from multi-source data, extracts and refines feature vectors for deviation detection, calculates the difference between the feeding amount and the preset benchmark based on the feature vectors, identifies deviation signals, and combines a support vector machine classifier to analyze historical data to identify deviation factors. The classification results are mapped to a control strategy library, matched with dynamic adaptation rules, and a feeding parameter correction value is generated. For deviations dominated by environmental changes, a temperature compensation algorithm is applied to adjust the feeding flow rate, ensuring that the correction value converges to a stable range. The optimized feeding command updates the parameters of this invention in real time. The adjustment effect is verified through feedback loops, historical data is recorded, and the classifier model is iteratively optimized. This achieves accurate identification and real-time correction of feeding deviations in electrophoresis line feeding control, improving control accuracy.

[0134] Reference Figure 2 The second embodiment of the present invention provides an automatic feeding control system for an automotive parts electrophoresis line based on multiple sensors, including:

[0135] The feature vector acquisition module is used to acquire initial multi-source data through a sensor matrix, perform data preprocessing and feature extraction on the initial multi-source data, and obtain a refined feature vector set.

[0136] The feeding deviation detection module is used to calculate the difference between the real-time feeding amount and the preset benchmark feeding amount based on the refining feature vector set, and obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal intensity indicators.

[0137] The deviation factor classification module is used to classify and predict the set of deviation signal intensity indicators using a preset support vector machine classifier to obtain the deviation factor classification result.

[0138] The control strategy matching module is used to map the classification results of the deviation factors to a pre-established control strategy library, match the corresponding dynamic adaptation rules, and obtain a sequence of feed parameter correction values ​​for the deviation factors.

[0139] The environmental deviation adjustment module is used to obtain a first adjustment coefficient through a temperature compensation algorithm if the deviation factor classification result indicates that environmental change is the dominant deviation, determine whether the first adjustment coefficient is within the convergence range of a preset coefficient, obtain the convergence judgment result, perform secondary calibration on the feed parameter correction value sequence, and generate an optimized feed instruction.

[0140] The system update and verification module is used to update the control parameters of the electrophoresis line in real time according to the optimized feeding instructions, obtain updated multi-source data, determine whether the deviation has been eliminated and record it as historical data.

[0141] The classification model optimization module is used to iteratively update the model parameters of the support vector machine classifier based on the historical data, so as to obtain the updated data collection rules.

[0142] It should be noted that the multi-sensor-based automatic feeding control system for automotive parts electrophoresis lines provided in this embodiment of the invention is used to execute all process steps of the multi-sensor-based automatic feeding control method for automotive parts electrophoresis lines described above. The working principles and beneficial effects of the two are one-to-one, and therefore will not be repeated.

[0143] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the above-described embodiments of the automatic feeding control method for automotive parts electrophoresis lines based on multi-sensor technology. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments.

[0144] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0145] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0146] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0147] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0148] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0149] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0150] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An automatic feeding control method for an automotive parts electrophoresis line based on multiple sensors, characterized in that, include: Initial multi-source data is acquired through a sensor matrix, and the initial multi-source data is preprocessed and feature extracted to obtain a refined feature vector set. Based on the refined feature vector set, the difference between the real-time feeding amount and the preset benchmark feeding amount is calculated to obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal strength indicators. A preset support vector machine classifier is used to classify and predict the set of deviation signal intensity indicators to obtain the classification results of deviation factors. The classification results of the deviation factors are mapped to a pre-established control strategy library, and the corresponding dynamic adaptation rules are matched to obtain a sequence of feed parameter correction values ​​for the deviation factors. If the deviation factor classification result indicates that environmental change is the dominant deviation, then the first adjustment coefficient is obtained through the temperature compensation algorithm, it is determined whether the first adjustment coefficient is within the preset coefficient convergence range, the convergence judgment result is obtained, the feeding parameter correction value sequence is calibrated a second time, and the optimized feeding instruction is generated. The control parameters of the electrophoresis line are updated in real time according to the optimized feeding instructions, the updated multi-source data is obtained, and it is determined whether the deviation has been eliminated and recorded as historical data. Based on the historical data, the model parameters of the support vector machine classifier are iteratively updated to obtain the updated data collection rules.

2. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 1, characterized in that, The process involves acquiring initial multi-source data through a sensor matrix, performing data preprocessing and feature extraction on the initial multi-source data to obtain a refined feature vector set, including: Real-time data is collected by a sensor matrix to obtain initial multi-source data. The initial multi-source data is then standardized to obtain an initial multi-source data set. The initial multi-source data set is subjected to dimensionality reduction processing. The main feature components are extracted from the initial multi-source data set, and noise and redundant dimensions are removed to obtain a refined feature vector set.

3. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 1, characterized in that, The step involves calculating the difference between the real-time feeding amount and the preset benchmark feeding amount based on the refined feature vector set to obtain a feeding deviation value. If the feeding deviation value exceeds a preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal strength indicators, including: The real-time feeding amount is obtained from the refined feature vector set, and the real-time feeding amount is dynamically collected and standardized to obtain a standardized feeding amount dataset. The difference between the standardized feeding amount dataset and the preset benchmark feeding amount is calculated to obtain the feeding deviation value; If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated through logical judgment to obtain a set of deviation signals; Based on the set of deviation signals, feature mapping is performed on the deviation signal intensity to obtain a set of deviation signal intensity indices.

4. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 1, characterized in that, The step of using a preset support vector machine classifier to classify and predict the set of deviation signal intensity indicators to obtain the classification results of deviation factors includes: If the set of deviation signal strength indicators exceeds a preset deviation strength threshold, then the multi-source data at the time of deviation signal generation is standardized to obtain a standardized feature dataset. The standardized feature dataset is input into a preset support vector machine classifier for classification prediction to obtain the classification result of the bias factor.

5. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 4, characterized in that, The step of mapping the classification results of the deviation factors to a pre-established control strategy library, matching the corresponding dynamic adaptation rules, and obtaining a sequence of feed parameter correction values ​​for the deviation factors includes: Based on the classification results of the deviation factors, a pre-established control strategy library is queried, and the classification results are used to map and match the corresponding dynamic adaptation rules to obtain a set of dynamic adaptation rules. Based on the set of dynamic adaptation rules, the classification results of the deviation factors are analyzed. If the classification result of the deviation factors is environmental change, a first feeding parameter correction value is generated. If the classification result of the deviation factors is material characteristic fluctuation, a second feeding parameter correction value is generated, thus obtaining a set of feeding parameter correction values. The set of feed parameter correction values ​​is sorted by priority to obtain a sorted sequence of feed parameter correction values.

6. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 1, characterized in that, If the deviation factor classification result indicates that environmental change is the dominant deviation, then a first adjustment coefficient is obtained through a temperature compensation algorithm, it is determined whether the first adjustment coefficient is within the preset coefficient convergence range, a convergence judgment result is obtained, and the feeding parameter correction value sequence is calibrated a second time to generate an optimized feeding instruction, including: Based on the initial multi-source data, a first set of environmental deviation values ​​is obtained through data filtering and extraction; If the deviation factor classification result indicates that environmental change is the dominant deviation, an adjustment coefficient for the feeding flow rate is calculated using a temperature compensation algorithm based on the first set of environmental deviation values. If the temperature fluctuation exceeds a preset temperature fluctuation threshold, a first adjustment coefficient is generated. Determine whether the first adjustment coefficient is within the preset coefficient convergence range to obtain a convergence judgment result. If the convergence result shows that the first adjustment coefficient is not convergent, then perform a second calibration on the feeding parameter correction value sequence to obtain an optimized feeding command.

7. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 6, characterized in that, The process of updating the electrophoresis line control parameters in real time according to the optimized feeding command, obtaining updated multi-source data, determining whether the deviation has been eliminated, and recording it as historical data includes: The control parameters of the electrophoresis line are updated according to the optimized feeding instructions, the updated multi-source data is obtained, and the second set of environmental deviation values ​​is obtained through data filtering and extraction. If the environmental deviation values ​​in the second set of environmental deviation values ​​do not exceed the preset environmental deviation range, then the deviation is determined to be eliminated, and the updated multi-source data is recorded as historical data.

8. The automatic feeding control method for automotive parts electrophoresis line based on multiple sensors according to claim 1, characterized in that, The step of iteratively updating the model parameters of the support vector machine classifier based on the historical data to obtain the updated data collection rules includes: The historical data is preprocessed to obtain the first cleaned dataset; Based on the first cleaned dataset, the parameters of the support vector machine classifier are iteratively updated, the classification boundary is adjusted, and an optimized set of classifier parameters is obtained. If the optimized classifier parameter set is applied to a preset validation dataset and the recognition accuracy exceeds a preset accuracy threshold, then the validation dataset is processed through bias analysis to obtain a set of bias factors. Based on the set of deviation factors, the subsequent data collection strategy is adjusted to obtain the updated data collection rules.

9. An automatic feeding control system for an automotive parts electrophoresis line based on multiple sensors, characterized in that, include: The feature vector acquisition module is used to acquire initial multi-source data through a sensor matrix, perform data preprocessing and feature extraction on the initial multi-source data, and obtain a refined feature vector set. The feeding deviation detection module is used to calculate the difference between the real-time feeding amount and the preset benchmark feeding amount based on the refining feature vector set, and obtain the feeding deviation value. If the feeding deviation value exceeds the preset feeding deviation threshold, a deviation signal is generated and feature mapping is performed to obtain a set of deviation signal intensity indicators. The deviation factor classification module is used to classify and predict the set of deviation signal intensity indicators using a preset support vector machine classifier to obtain the deviation factor classification result. The control strategy matching module is used to map the classification results of the deviation factors to a pre-established control strategy library, match the corresponding dynamic adaptation rules, and obtain a sequence of feed parameter correction values ​​for the deviation factors. The environmental deviation adjustment module is used to obtain a first adjustment coefficient through a temperature compensation algorithm if the deviation factor classification result indicates that environmental change is the dominant deviation, determine whether the first adjustment coefficient is within the convergence range of a preset coefficient, obtain the convergence judgment result, perform secondary calibration on the feed parameter correction value sequence, and generate an optimized feed instruction. The system update and verification module is used to update the control parameters of the electrophoresis line in real time according to the optimized feeding instructions, obtain updated multi-source data, determine whether the deviation has been eliminated and record it as historical data. The classification model optimization module is used to iteratively update the model parameters of the support vector machine classifier based on the historical data, so as to obtain the updated data collection rules.

Images