Data annotation method and system for gold foil product testing

Abstract

The embodiment of the present invention discloses a data labeling method and system for gold foil product inspection. The method can provide comprehensive materials for subsequent analysis by obtaining a quality description data set covering quality description sentences and process parameter records; based on feature extraction processing, the semantic features of sentence context and parameter association features can be accurately captured; based on multi-dimensional strategy analysis processing of a dynamic matching algorithm, quality labeling results can be innovatively generated, thereby clearly presenting the mapping relationship between the defect type of gold foil product and the adjustment direction of process parameters; based on the quality labeling results, a process optimization strategy set is generated and fed back to the production control system to execute parameter regulation, which can quickly and accurately optimize the gold foil production process, improve product quality, reduce the defect rate, and enhance the intelligence and automation level of the production process.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data analysis technology, and specifically to a data labeling method and system for gold foil product detection. Background Art

[0002] In the area of gold foil production quality testing, quality requirements are increasing with its widespread application in numerous fields, such as art decoration and electronics. Traditional gold foil production quality testing relies primarily on manual experience and simple physical testing methods, resulting in a simplistic and one-sided assessment of product quality. Manual testing is not only inefficient but also prone to human error, making it difficult to comprehensively and accurately determine quality issues in gold foil products.

[0003] In addition, other existing technologies are unable to effectively integrate the complex relationship between quality description information and process parameters, making it difficult to deeply analyze the root causes of quality problems. On the other hand, when faced with complex and diverse types of gold foil product defects, there is a lack of intelligent analysis methods to determine the corresponding process parameter adjustment direction, making it difficult to achieve targeted process optimization. Summary of the Invention

[0004] The embodiments of the present invention provide a data annotation method and system for gold foil product inspection, which is used to solve the problem that the existing technology is difficult to comprehensively analyze quality problems and accurately optimize processes in gold foil product production quality inspection, thereby improving the quality and efficiency of gold foil production.

[0005] In the first aspect, an embodiment of the present invention provides a data labeling method for gold foil product inspection, which is applied to a data labeling system, and the method includes: obtaining a quality description data set of gold foil products, the quality description data set including multiple quality description statements and corresponding process parameter records; performing feature extraction processing on the quality description data set to obtain contextual semantic features of each quality description statement and parameter association features of the process parameter records; performing multi-dimensional strategy analysis processing on the contextual semantic features and the parameter association features based on a dynamic matching algorithm to generate quality labeling results of the quality description statements; the quality labeling results are used to indicate the mapping relationship between the defect type of the gold foil product and the adjustment direction of the process parameters; generating a process optimization strategy set based on the quality labeling results, and feeding the process optimization strategy set back to the gold foil production control system to execute parameter control operations.

[0006] In a second aspect, an embodiment of the present invention provides a data labeling system, including:

[0007] processor;

[0008] a storage device having a computer program stored thereon,

[0009] When the computer program is executed by the processor, the processor implements any of the data labeling methods for gold foil product detection.

[0010] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the data labeling method applied to gold foil product detection are implemented.

[0011] It can be seen that the embodiments of the present invention have the following beneficial effects: the embodiments of the present invention realize the in-depth mining and effective utilization of gold foil product quality data: by obtaining a quality description data set covering quality description statements and process parameter records, it can provide comprehensive materials for subsequent analysis; based on feature extraction processing, the context semantic features of sentences and parameter association features can be accurately captured; based on the multi-dimensional strategy analysis processing of the dynamic matching algorithm, quality labeling results can be innovatively generated, thereby clearly presenting the mapping relationship between the gold foil product defect type and the process parameter adjustment direction; based on the quality labeling results, a process optimization strategy set is generated and fed back to the production control system to execute parameter regulation, which can quickly and accurately optimize the gold foil production process, improve product quality, reduce the defect rate, and enhance the intelligence and automation level of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of a data labeling method for gold foil product detection provided by an embodiment of the present invention.

[0013] Figure 2 A schematic diagram of the basic structure of a data annotation system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0015] See also Figure 1 As shown in FIG, this figure is a flow chart of a data annotation method for gold foil product detection provided by an embodiment of the present invention, which can be applied to a data annotation system. Figure 1 As shown, the method may include steps 110 to 140.

[0016] Step 110: Acquire a quality description data set of the gold foil product, wherein the quality description data set includes a plurality of quality description statements and corresponding process parameter records.

[0017] In an embodiment of the present invention, the data annotation system of a gold foil production factory continuously collects various types of information during the production process. For example, during one production cycle, the data annotation system recorded multiple quality description statements, including but not limited to "fine scratches appear on the gold foil surface" and "uneven thickness of the gold foil." Simultaneously, a series of corresponding process parameters were recorded, such as temperature parameter A, whose value gradually changed from an initial value A1 to A2 during the cycle; pressure parameter B, which fluctuated from B1 to B2; and rolling speed parameter C, which varied between C1 and C3. These quality description statements and the corresponding process parameter records together constitute the quality description data set.

[0018] Step 120: performing feature extraction processing on the quality description data set to obtain contextual semantic features of each quality description statement and parameter association features of the process parameter record.

[0019] In the embodiment of the present invention, the system starts to extract features from the quality description data set, processing the quality description statements and process parameter records separately to obtain the required features.

[0020] Optionally, step 120 further includes:

[0021] Step 121: performing semantic unit segmentation processing on the quality description sentence to obtain multiple semantic unit sequences; each semantic unit sequence includes at least one semantic segment and a position identifier of the semantic segment in the sentence.

[0022] For example, for the quality description sentence "Minor scratches appear on the surface of the gold foil", the system uses semantic analysis tools commonly used in related technologies to perform semantic unit segmentation. It is segmented into semantic segments such as "gold foil", "surface", "appearance", "minor", and "scratches". Each semantic segment is assigned a position identifier in the sentence, such as "gold foil" in the first position, "surface" in the second position, etc., thus forming a sequence of multiple semantic units. The above-mentioned semantic unit sequence provides the basis for the subsequent in-depth analysis of the semantics of the sentence, making it easier for the system to further explore the relationship and potential information between the various parts of the sentence. Then, the segmented semantic unit sequence is passed to the next processing link, which is to call the pre-trained language feature encoder for context encoding processing.

[0023] Step 122: Calling a pre-trained language feature encoder to perform context encoding processing on the semantic unit sequence to generate context semantic features of the quality description sentence; the context semantic features include local semantic features of each semantic segment and context dependencies between adjacent semantic segments.

[0024] In a preferred embodiment, step 122 may include:

[0025] Step 1221: Perform part-of-speech tagging on the semantic segments in each semantic unit sequence to generate a tag sequence containing part-of-speech tags.

[0026] Taking the semantic unit sequence "Minor scratches appear on the surface of gold foil" as an example, the system performs part-of-speech tagging on each semantic segment. "Gold foil" is tagged as a noun, "surface" as a noun, "appear" as a verb, "minor" as an adjective, and "scratch" as a noun. This generates a tag sequence containing part-of-speech tags. This tag sequence helps the system more clearly understand the grammatical role of each semantic segment in the sentence, providing an important basis for the subsequent construction of a semantic dependency tree and hierarchical encoding. The tag sequence is then used as input to construct the semantic dependency tree.

[0027] Step 1222: Construct a semantic dependency tree based on the annotation sequence to determine the subject-predicate relationship, modification relationship, and parallel relationship between the semantic segments.

[0028] Optionally, based on the tagging sequence of the above-mentioned part-of-speech tags, the system constructs a semantic dependency tree. In the sentence "Minor scratches appeared on the surface of the gold foil," "gold foil" is the subject of the entire sentence description, and "surface" describes the position of "gold foil," and the two have a modification relationship; "appearance" is an action, and its subject is "gold foil surface," which has a subject-predicate relationship; "minor" modifies "scratches," also a modification relationship. By constructing a semantic dependency tree, the system determines the complex relationships between various semantic fragments. These relationships will play a key role in the subsequent hierarchical encoding process, helping the system better capture the semantic information of the sentence. After the semantic dependency tree is constructed, hierarchical encoding is performed based on it.

[0029] Step 1223: Perform hierarchical encoding processing on the semantic unit sequence based on the semantic dependency tree: perform local encoding on each semantic segment to generate an initial semantic vector, perform relational aggregation on the initial semantic vector according to the subject-predicate relationship to generate an intermediate semantic vector, and perform contextual enhancement on the intermediate semantic vector based on the modification relationship and parallel relationship to generate an enhanced semantic vector.

[0030] Optionally, the system first performs local encoding on each semantic segment. For example, for the semantic segment of "gold foil", it is converted into an initial semantic vector V1 through the encoding algorithm commonly used in related technologies. Then, based on the subject-predicate relationship, the initial semantic vector of "gold foil surface" related to "appearance" is aggregated to obtain an intermediate semantic vector V2. Next, considering the modification relationship of "subtle" to "scratches" and the relationship between other related semantic segments, the intermediate semantic vector V2 is contextually enhanced to generate an enhanced semantic vector V3. The enhanced semantic vector V3 integrates the local information of the semantic segments and the contextual dependencies between them, and more comprehensively represents the semantic features of the semantic unit sequence. After the enhanced semantic vector is generated, the next step of processing is entered.

[0031] Step 1224: Arrange the enhanced semantic vectors in time sequence according to the position identifiers to generate contextual semantic features of the quality description sentence.

[0032] In this embodiment of the present invention, the system arranges the generated enhanced semantic vectors in chronological order according to the previously assigned position identifiers. For example, the enhanced semantic vector for "gold foil" is ranked first, that for "surface" is ranked second, and so on. This ultimately forms an ordered vector sequence, which serves as the contextual semantic feature of the quality description statement. This contextual semantic feature fully preserves the information of each semantic segment in the quality description statement and the contextual relationships between them, providing strong support for subsequent analysis based on the dynamic matching algorithm. Next, the system begins feature extraction from the process parameter records.

[0033] Step 123: performing parameter parsing processing on the process parameter records to extract parameter category identification and parameter time series variation pattern; the parameter time series variation pattern includes the fluctuation characteristics of parameter values over time and the correlation fluctuation trends between different parameters.

[0034] In a preferred embodiment, step 123 may include:

[0035] Step 1231: Divide the process parameter record into multiple parameter subsequences according to the time window, each parameter subsequence contains a set of parameter values within a preset time range.

[0036] For example, for the process parameter records of temperature parameter A, pressure parameter B, and rolling speed parameter C, the system divides them into time windows of every 10 minutes. In the first time window, the value set of temperature parameter A is [A11, A12, A13], the value set of pressure parameter B is [B11, B12, B13], and the value set of rolling speed parameter C is [C11, C12, C13]. Thus, multiple parameter subsequences are formed, and the above parameter subsequences provide basic data units for subsequent analysis of the variation characteristics of parameters in different time periods. After the division, trend analysis processing is performed on each parameter subsequence.

[0037] Step 1232: Perform trend analysis processing on each parameter subsequence, and extract the rising trend segment, falling trend segment, and stable trend segment.

[0038] Taking a parameter subsequence [A11, A12, A13] of temperature parameter A as an example, the system performs trend analysis by comparing the magnitude relationship of adjacent data points. If A11 < A12 < A13, then this subsequence shows a rising trend and is extracted as the rising trend segment; if A11 > A12 > A13, it is the falling trend segment; if the values of A11, A12, and A13 fluctuate little and are within a relatively stable range, it is the stable trend segment. By performing such trend analysis on each parameter subsequence, the system can clearly understand the variation trend of the parameter in different time periods and prepare for subsequent further feature extraction. After completing the trend analysis, corresponding feature calculations are performed on different trend segments.

[0039] Step 1233: Perform slope calculation processing on the rising trend segment to generate a first slope feature set; perform absolute value calculation processing on the falling trend segment to generate a second slope feature set; perform variance calculation processing on the stable trend segment to generate a fluctuation variance feature set.

[0040] For the rising trend segment of temperature parameter A, the system calculates the slope between adjacent data points. For example, if the rising trend segment is [A11, A12, A13], by calculating (A12 - A11) / (t2 - t1), (A13 - A12) / (t3 - t2), etc. (where t1, t2, t3 are the corresponding time points), a series of slope values are obtained, and these slope values form the first slope feature set. For the falling trend segment, the slope is also calculated, but the absolute value is taken to form the second slope feature set. For the stable trend segment, the variance of the parameter values is calculated to measure its fluctuation degree, and the fluctuation variance feature set is obtained. The above feature sets describe the variation characteristics of the parameter in different trend segments from different angles. Then, these feature sets are classified according to the parameter category identifier.

[0041] Step 1234: Classify the first slope feature set, the second slope feature set, and the volatility variance feature set according to the parameter category identifier to generate an intra-category trend feature matrix.

[0042] Among them, the system classifies the first slope feature set, the second slope feature set, and the fluctuation variance feature set of temperature parameter A into one category based on the parameter category identifier, the corresponding feature set of pressure parameter B into another category, and the rolling speed parameter C into another category. The feature sets in each category are then arranged in a certain order to form an intra-category trend feature matrix. For example, in the intra-category trend feature matrix of temperature parameter A, the first row contains the elements of the first slope feature set, the second row contains the elements of the second slope feature set, and the third row contains the elements of the fluctuation variance feature set. Such a matrix form is more convenient for the system to perform subsequent correlation calculations and feature fusion processing. After generating the intra-category trend feature matrix, correlation calculations are performed on the matrices between different parameter categories.

[0043] Step 1235: performing correlation calculation on the intra-category trend feature matrices between different parameter categories to generate a cross-category trend correlation matrix.

[0044] The system calculates the correlation between the intra-category trend feature matrix of temperature parameter A and the intra-category trend feature matrix of pressure parameter B, for example, by calculating the Pearson correlation coefficient between the matrix elements. For example, the intra-category trend feature matrix of temperature parameter A is M1, and that of pressure parameter B is M2. The correlation between each element in M1 and the corresponding position element in M2 is calculated to obtain a series of correlation values, which are then organized into a matrix, which is part of the cross-category trend correlation matrix between temperature parameter A and pressure parameter B. The same method is used to calculate the cross-category trend correlation matrix parts between temperature parameter A and rolling speed parameter C, and between pressure parameter B and rolling speed parameter C, and finally form a complete cross-category trend correlation matrix, which reflects the degree of trend correlation between different parameter categories. After the correlation calculation is completed, the matrix splicing processing is performed.

[0045] Step 1236: Concatenate the intra-category trend feature matrix and the cross-category trend correlation matrix to generate the parameter time series variation pattern.

[0046] The system combines the previously generated intra-category trend feature matrices and cross-category trend association matrices according to specific rules. For example, the intra-category trend feature matrices for temperature parameter A, pressure parameter B, and rolling speed parameter C are arranged in sequence, and then the cross-category trend association matrix is added to the right. This creates a complete matrix structure, which represents the parameter temporal variation pattern. It integrates the parameter variation trend characteristics over different time periods and the correlation fluctuation trends between different parameters, comprehensively describing the dynamic variation characteristics of the process parameters. After the parameter temporal variation pattern is generated, feature fusion processing is performed based on the parameter category identification.

[0047] Step 124: Based on the parameter category identifier, feature fusion processing is performed on the parameter time series change pattern to generate parameter association features of the process parameter record; the parameter association features include multi-dimensional fluctuation correlation under the same parameter category and cross-category influence weights between different parameter categories.

[0048] Optionally, the system performs feature fusion on the parameter time series change pattern based on the parameter category identifier. For the same parameter category, such as temperature parameter A, the system analyzes the relationship between different feature sets in the trend feature matrix within its category to determine the multi-dimensional fluctuation correlation. For example, there may be a certain correlation between the first slope feature set and the fluctuation variance feature set, and the strength of this correlation is calculated by the algorithm commonly used in the relevant technology. For different parameter categories, the cross-category impact weight is determined based on the cross-category trend correlation matrix. For example, the impact weight of the change in temperature parameter A on pressure parameter B is quantified by analyzing the correlation value in the matrix. Finally, the multi-dimensional fluctuation correlation under the same parameter category and the cross-category impact weights between different parameter categories are integrated together to form the parameter association features of the process parameter records. The above parameter association features provide comprehensive and in-depth information about the process parameters for subsequent analysis based on the dynamic matching algorithm. After the feature extraction is completed, the multi-dimensional strategy analysis and processing stage is entered.

[0049] Step 130: Based on a dynamic matching algorithm, a multi-dimensional strategy analysis is performed on the context semantic features and the parameter association features to generate a quality labeling result of the quality description statement; the quality labeling result is used to indicate the mapping relationship between the gold foil product defect type and the process parameter adjustment direction.

[0050] In an exemplary embodiment, step 130 further includes:

[0051] Step 131: Map the context semantic features to the defect feature space to generate a defect feature vector containing a probability distribution of defect types.

[0052] Furthermore, the system uses a mapping function commonly used in related technologies to map the contextual semantic features of the previously generated quality description sentence, such as the contextual semantic feature vector corresponding to "fine scratches appear on the surface of the gold foil", to the defect feature space. In this space, different areas represent different defect types. Through mapping, a defect feature vector containing the probability distribution of defect types is obtained. For example, the probability value of this vector in the "scratch defect" area is 0.8, and the probability value in the "uneven thickness defect" area is 0.2, indicating that the situation described by this quality description sentence is more likely to be a scratch defect. After generating the defect feature vector, a similar mapping process is performed on the parameter association features.

[0053] Step 132: Map the parameter association feature to the control feature space to generate a control feature vector including parameter adjustment direction weights.

[0054] In this step, the system uses another commonly used feature mapping method to map the parameter association features of the process parameter records to the control feature space. In this space, different dimensions represent different parameter adjustment directions. For example, for the parameter association features of temperature parameter A, pressure parameter B, and rolling speed parameter C, a control feature vector is obtained after mapping. The weight of this vector in the direction of "increasing temperature parameter A" is 0.6, the weight in the direction of "reducing pressure parameter B" is 0.3, and the weight in the direction of "keeping rolling speed parameter C unchanged" is 0.1, indicating that according to the current parameter association features, increasing temperature parameter A is a more likely adjustment direction. After the mapping is completed, spatial alignment processing is performed.

[0055] Step 133: performing spatial alignment processing on the defect feature vector and the control feature vector to determine a mapping relationship between the defect type probability distribution and the parameter adjustment direction weight.

[0056] Step 133 may include:

[0057] Step 1330: Construct a dual-channel alignment network including a defect type dimension and a parameter adjustment dimension; input the defect feature vector into the defect type channel for feature enhancement processing to generate an enhanced defect feature vector; input the control feature vector into the parameter adjustment channel for feature enhancement processing to generate an enhanced control feature vector; perform cross-attention calculation processing on the enhanced defect feature vector and the enhanced control feature vector to generate a defect-parameter attention weight matrix; perform weighted summation processing on the enhanced defect feature vector and the enhanced control feature vector according to the defect-parameter attention weight matrix to generate a joint feature vector; perform full-connection mapping processing on the joint feature vector to generate a mapping relationship between the defect type probability distribution and the parameter adjustment direction weight.

[0058] Specifically, the system first constructs a dual-channel alignment network, which contains two channels: a defect type dimension and a parameter adjustment dimension. The defect feature vector is input into the defect type channel and enhanced through a series of operations such as convolution and pooling to obtain an enhanced defect feature vector. Similarly, the control feature vector is input into the parameter adjustment channel for enhancement processing to generate an enhanced control feature vector. Next, a cross-attention calculation is performed on these two enhanced vectors. For example, the degree of correlation between each element in the enhanced defect feature vector and each element in the enhanced control feature vector is calculated to obtain a defect-parameter attention weight matrix. Based on this matrix, a weighted summation is performed on the enhanced defect feature vector and the enhanced control feature vector to obtain a joint feature vector. Finally, the joint feature vector is processed using a fully connected mapping function to generate a mapping relationship between the defect type probability distribution and the parameter adjustment direction weight. For example, the resulting mapping relationship shows that "scratch defects" have a strong correlation with "increasing the rolling speed parameter C." After determining the mapping relationship, a dynamic matching matrix is constructed.

[0059] Step 134: construct a dynamic matching matrix according to the mapping relationship, perform singular value decomposition on the dynamic matching matrix, and obtain a defect type principal component vector and a parameter adjustment principal component vector.

[0060] Optionally, the system constructs a dynamic matching matrix based on the mapping relationship between the probability distribution of defect types and the weights of parameter adjustment directions. The rows of the matrix represent the defect types, the columns represent the parameter adjustment directions, and the matrix elements are the corresponding correlation values. This matrix is then subjected to singular value decomposition. Using algorithms commonly used in related technologies, the matrix is decomposed into the product of three matrices, from which the principal component vectors of the defect type and the principal component vectors of the parameter adjustment are extracted. These principal component vectors can more concisely represent the main characteristics of the defect type and the parameter adjustment direction. After the singular value decomposition is completed, the defect type and parameter adjustment direction are labeled.

[0061] Step 135: Determine the defect type label of the quality description statement based on the similarity calculation between the defect type principal component vector and the preset defect type standard vector, wherein the defect type label is used to characterize the defect type of the gold foil product.

[0062] Optionally, the system calculates the similarity between the defect type principal component vector and various preset standard vectors for various defect types. For example, a "scratch defect standard vector" and a "thickness unevenness defect standard vector" are preset. The system compares the similarity between the defect type principal component vector and these standard vectors by calculating cosine similarity or other methods. If the similarity with the "scratch defect standard vector" is the highest, at 0.9, the defect type for the quality description statement is determined to be "scratch defect," which clarifies the possible defect type for the gold foil product. After determining the defect type label, parameter adjustment direction labeling is performed.

[0063] Step 136: Determine the parameter adjustment direction label based on the direction matching process between the parameter adjustment principal component vector and the historical optimization strategy vector, where the parameter adjustment direction label is used to represent the process parameter adjustment direction.

[0064] Optionally, the system directionally matches the parameter adjustment principal component vector with the historical optimization strategy vector. The historical optimization strategy vector records the parameter adjustment directions that were successful in similar situations in the past. By calculating the angle between the vectors and other methods, the system determines the directional consistency between the parameter adjustment principal component vector and the historical optimization strategy vector. If the historical optimization strategy vector related to "increasing rolling speed parameter C" is found to be consistent in direction with a similarity of 0.8, the parameter adjustment direction is marked as "increasing rolling speed parameter C." This marking provides directional guidance for process parameter adjustments. After the marking is completed, the results are combined.

[0065] Step 137: Combine the defect type annotation and the parameter adjustment direction annotation to generate a quality annotation result of the quality description statement.

[0066] The system combines the identified defect type annotation ("scratch defect") and the parameter adjustment direction annotation ("increase rolling speed parameter C") to form a quality annotation result, a quality description statement. This result clearly indicates the mapping between the gold foil product defect type and the process parameter adjustment direction, providing a key basis for the subsequent generation of process optimization strategies. After generating the quality annotation results, the process optimization strategy set is generated and parameter adjustment operations are executed.

[0067] Step 140: Generate a process optimization strategy set based on the quality labeling result, and feed the process optimization strategy set back to the gold foil production control system to perform parameter control operations.

[0068] In an alternative embodiment, generating a set of process optimization strategies according to the quality annotation results in step 140 includes:

[0069] Step 141: extracting a defect type priority list and a parameter adjustment direction priority list from the quality marking result.

[0070] Optionally, the system conducts an in-depth analysis of the generated quality labeling results. For example, for the quality labeling results of a series of quality description statements, the frequency of occurrence of different defect types and the severity assessment values are counted, and they are arranged in order from high to low to form a defect type priority list. For example, among multiple quality labeling results, "scratch defects" appear more frequently and have a higher assessed severity, so they are ranked higher in the defect type priority list; "uneven thickness defects" appear relatively less frequently and have a slightly lower severity, so they are ranked behind them. At the same time, based on the importance assessment of the parameter adjustment direction in the quality labeling results and the degree of correlation with the defect type, a parameter adjustment direction priority list is generated. For example, "increasing the rolling speed parameter C", which is closely associated with the "scratch defect" and appears many times, is at the top of the parameter adjustment direction priority list. The above priority list provides a basis for the subsequent determination of the optimization weight. After the extraction is completed, the optimization weight is determined based on the list.

[0071] Step 142: Determine a first optimization weight according to the defect type priority list, and determine a second optimization weight according to the parameter adjustment direction priority list.

[0072] Among them, for the defect type priority list, the system adopts the weight allocation algorithm commonly used in related technologies. For example, the first optimization weight is determined based on the comprehensive consideration of the proportion of the frequency of the defect type to the total frequency and the severity assessment value. For example, "scratch defect" accounts for 40% of the total frequency, and the severity assessment value is 8 points (out of 10 points). Through the existing weight normalization algorithm, its first optimization weight is 0.6. For the parameter adjustment direction priority list, the second optimization weight is also determined based on factors such as the closeness of its correlation with the defect type and the stability in multiple quality labeling results. For example, "increasing the rolling speed parameter C" is closely related to "scratch defect" and appears stable, so its second optimization weight is determined to be 0.5. After determining the optimization weight, weighted fusion processing is performed.

[0073] Step 143: Perform weighted fusion processing on the first optimization weight and the second optimization weight to generate a comprehensive optimization weight.

[0074] The system uses a weighted fusion algorithm to combine the first and second optimization weights. For example, using a simple weighted average method, setting the weight coefficient of the first optimization weight to 0.6 and the weight coefficient of the second optimization weight to 0.4, the overall optimization weight = 0.6 × first optimization weight + 0.4 × second optimization weight = 0.6 × 0.6 + 0.4 × 0.5 = 0.56. This overall optimization weight comprehensively considers the importance of defect type and parameter adjustment direction, providing a unified metric for screening historical process parameter adjustment records. After the overall optimization weight is generated, screening is performed.

[0075] Step 144: Filter and process historical process parameter adjustment records according to the comprehensive optimization weight to obtain a set of candidate adjustment strategies.

[0076] Optionally, the system traverses historical process parameter adjustment records and evaluates each record based on the defect type and parameter adjustment direction involved, combined with comprehensive optimization weights. For example, a historical record that adopted the adjustment strategy of "increasing rolling speed parameter C" to address "scratch defects" is evaluated based on the comprehensive optimization weights and the record's historical effectiveness in resolving defects. Records with high scores are screened to form a set of candidate adjustment strategies. The strategies in this set are initially selected based on historical experience and current quality labeling results and have certain feasibility and optimization potential. After the screening is completed, the candidate adjustment strategy set is evaluated for feasibility.

[0077] Step 145: Perform feasibility evaluation on the candidate adjustment strategy set and generate a result list including evaluation weights.

[0078] In detail, step 145 includes:

[0079] Step 1450: Split each candidate adjustment strategy into a sequence of parameter adjustment steps, each parameter adjustment step including an adjustment target value and an adjustment time range; perform conflict detection processing on the parameter adjustment step sequence to identify conflicting step pairs with time overlap or conflicting target values; generate a conflict assessment score based on the number and severity of the conflicting step pairs; perform simulation execution processing on the parameter adjustment step sequence to predict the parameter change curve and defect type probability change curve after execution; generate a parameter matching score based on the matching degree between the parameter change curve and the target parameter range; generate a defect improvement score based on the difference between the defect type probability change curve and the target defect threshold; perform weighted sum processing on the conflict assessment score, the parameter matching score and the defect improvement score to generate a result list of evaluation weights of the candidate adjustment strategy.

[0080] Take, for example, a candidate adjustment strategy that includes adjustments to temperature parameter A, pressure parameter B, and rolling speed parameter C. This strategy is broken down into a sequence of parameter adjustment steps, such as "adjust temperature parameter A from its current value to target value A3 during the time period t1-t2," and "adjust pressure parameter B from its current value to target value B4 during the time period t2-t3." The system first performs conflict detection on this sequence, checking for time overlap or conflicting target values. If the operations of "adjusting temperature parameter A to A3" and "adjusting pressure parameter B to B4" partially overlap in time and may affect each other, they are identified as a conflicting step pair. A conflict assessment score is generated based on the number of conflicting step pairs and the potential severity of their impact on the production process. For example, if there are three conflicting step pairs with a severe impact, the conflict assessment score is 0.3 (out of 1). The parameter adjustment step sequence is then simulated and executed. Using a simulation model of the production process, the parameter change curves and defect type probability curves after execution are predicted. For example, after the simulation, the parameter variation curve shows that temperature parameter A approaches the target parameter range well. Based on its degree of match with the target range, a parameter matching score (e.g., 0.8) is generated. The defect type probability variation curve shows that the probability of a "scratch defect" decreases from 0.8 to 0.5. Based on the difference from the target defect threshold (set at 0.3), a defect improvement score (e.g., 0.6) is generated. Finally, according to the preset weightings (e.g., 0.2 for the conflict assessment score, 0.3 for the parameter matching score, and 0.5 for the defect improvement score), a weighted sum is calculated: 0.2 × 0.3 + 0.3 × 0.8 + 0.5 × 0.6 = 0.54, resulting in the evaluation weight for the candidate adjustment strategy. This evaluation is performed for each strategy in the candidate adjustment strategy set, generating a result list containing the evaluation weights. After the evaluation is complete, a strategy is selected.

[0081] Step 146: Sort the result list according to the evaluation weights, and select candidate adjustment strategies that exceed a preset score threshold to generate the process optimization strategy set.

[0082] Optionally, the system sorts the result list in descending order based on the evaluation weight. For example, the candidate adjustment strategy with the highest evaluation weight is ranked first, and so on. A score threshold is preset, such as 0.5. The system selects candidate adjustment strategies with evaluation weights exceeding this threshold, combines these strategies together, and forms a process optimization strategy set. The strategies in this set are determined after multiple rounds of screening and evaluation, have high feasibility and optimization effects, and can provide effective parameter control guidance for the gold foil production control system. After the process optimization strategy set is generated, it is fed back to the gold foil production control system to perform parameter control operations.

[0083] Based on the above, step 140 of feeding back the process optimization strategy set to the gold foil production control system to perform parameter control operations includes:

[0084] Step 1471: Convert the process optimization strategy set into a control instruction sequence, each control instruction including a parameter identifier, a target value, and an execution timestamp.

[0085] The system parses and converts each strategy in the process optimization strategy set. For example, a strategy such as "adjust rolling speed parameter C from current value C1 to target value C2 during the time period t1-t2" is converted into a control instruction with the parameter identifier "rolling speed parameter C," the target value "C2," and the execution timestamp "t1" (indicating the start time of execution). This conversion is performed on all strategies in the process optimization strategy set, forming a control instruction sequence that clearly defines the specific requirements and execution schedule for each parameter adjustment, making it easier for the gold foil production control system to receive and execute it. After the conversion is completed, a timing verification process is performed.

[0086] Step 1472: Perform timing verification on the control instruction sequence to ensure that the execution timestamp complies with the operation interval limit of the production equipment.

[0087] Optionally, the gold foil production equipment sets corresponding operation interval restrictions. For example, a certain time interval is required between two adjustment operations on the same parameter to ensure stable operation of the equipment. The system checks the execution timestamp of each control instruction in the control instruction sequence to determine whether it meets the operation interval restrictions of the equipment. For example, for two consecutive control instructions for rolling speed parameter C, the execution timestamp of the first instruction is t1 and the second is t2. The operation interval time required by the equipment is Δt. If t2-t1<Δt, it does not meet the requirements. The timestamps that do not meet the requirements are marked to prepare for subsequent adjustments. After the timing check is completed, the timestamp adjustment process is performed.

[0088] Step 1473: Perform timestamp adjustment processing on the control instruction sequence according to the verification result to generate an optimized control instruction sequence.

[0089] During the verification process, the system adjusts timestamps found to be inconsistent with the operation interval limits. For example, the execution timestamp t2 of the second control instruction for rolling speed parameter C is adjusted to t1 + Δt. This adjustment ensures that all execution timestamps in the control instruction sequence comply with the production equipment's operation interval limits, generating an optimized control instruction sequence that ensures orderly execution of parameter control operations within the acceptable limits of the production equipment, avoiding equipment failures or production anomalies caused by improper timing. Once the adjustment is complete, the optimized control instruction sequence is sent to the gold foil production control system.

[0090] Step 1474: Send the optimized control instruction sequence to the parameter execution module of the gold foil production control system, and instruct the parameter execution module to perform parameter control operations according to the optimized control instruction sequence.

[0091] The system sends an optimization control instruction sequence to the parameter execution module of the gold foil production control system via a communication interface. Upon receiving the instruction sequence, the parameter execution module sequentially adjusts the corresponding process parameters according to the parameter identifier, target value, and execution timestamp in the instruction. For example, upon receiving an instruction with the parameter identifier "rolling speed parameter C," the target value "C2," and the execution timestamp "t1," the module adjusts rolling speed parameter C to C2 at time t1. This enables precise control of process parameters during gold foil production, improving product quality. After the instruction is sent, the execution status begins to be monitored in real time.

[0092] Step 1475: Monitor the execution status of the parameter control operation in real time. When an execution exception is detected, generate a rollback instruction or an adaptive adjustment instruction according to the exception type to update the optimization control instruction sequence.

[0093] Furthermore, the system incorporates a monitoring module into the gold foil production control system to track the execution of parameter control operations in real time. For example, it monitors whether the rolling speed parameter C has been successfully adjusted to the target value C2, and whether the equipment's operating status is normal during the adjustment process. If an abnormality occurs during the adjustment process, such as a failure in equipment feedback adjustment or a parameter value not falling within the expected range, the monitoring module detects the type of abnormality. If the abnormality is caused by a parameter setting error, the system generates a rollback instruction to restore the parameters to their pre-adjustment state. If the abnormality is caused by external environmental factors, the system generates adaptive adjustment instructions based on a preset adaptive adjustment strategy, such as appropriately adjusting the target value or execution time. The generated instructions are added to the optimization control instruction sequence, and the sequence is updated to ensure that subsequent parameter control operations can proceed smoothly, safeguarding the stability of the gold foil production process and product quality. After executing the parameter control operation, the subsequent monitoring and optimization phase begins.

[0094] In a non-limiting embodiment, after feeding back the process optimization strategy set to the gold foil production control system to perform parameter control operations, the method further includes:

[0095] Step 210: Real-time collection of the actual value set of process parameters after executing the parameter control operation and the corresponding gold foil quality inspection data set; comparing the actual value set of process parameters with the target parameter values in the process optimization strategy set item by item, and calculating the real-time control deviation of each parameter; generating a dynamic compensation strategy set based on the real-time control deviation, the dynamic compensation strategy set including the compensation adjustment amount and priority for each parameter; superimposing and performing conflict detection processing on the dynamic compensation strategy set, identifying parameter adjustment conflict pairs and generating conflict resolution weights; correcting the dynamic compensation strategy set based on the conflict resolution weights, and generating an optimized compensation instruction sequence; feeding back the optimized compensation instruction sequence to the gold foil production control system to execute the compensation control operation, and synchronously updating the process optimization strategy set.

[0096] After parameter control is executed, the system begins real-time data collection. For example, at regular intervals (e.g., 5 minutes), the system collects actual process parameter values, including the actual value A4 of temperature parameter A, the actual value B5 of pressure parameter B, and the actual value C3 of rolling speed parameter C. Simultaneously, the produced gold foil undergoes quality inspection, acquiring corresponding gold foil quality inspection data, such as the number of scratches on the foil surface and thickness deviation. Next, the actual process parameter values are compared with the target parameter values in the process optimization strategy set. For example, if the target value of temperature parameter A in the process optimization strategy set is A3 and the actual value is A4, the real-time control deviation is calculated as |A4 - A3| / A3. Based on the real-time control deviation of each parameter, a dynamic compensation strategy set is generated. For example, for temperature parameter A, if the deviation is large, a compensation adjustment of ΔA is determined, and its priority is given based on its impact on gold foil quality. For rolling speed parameter C, if the deviation is small, a compensation adjustment of ΔC is assigned, giving it a lower priority. The dynamic compensation strategy set is then overlaid to perform conflict detection. For example, consider two conflicting compensation strategies: one for increasing temperature parameter A and one for decreasing it. These strategies are identified as a parameter adjustment conflict pair, and conflict resolution weights are generated based on the severity of the conflict. For example, the strategy of increasing temperature parameter A is more conducive to improving gold foil quality, with a conflict resolution weight of 0.8, while the weight for decreasing temperature parameter A is 0.2. Based on these weights, the dynamic compensation strategy set is modified to generate an optimized compensation instruction sequence. Finally, the optimized compensation instruction sequence is fed back to the gold foil production control system to execute compensation control operations, while simultaneously updating the process optimization strategy set, allowing the strategy set to be continuously optimized based on actual conditions.

[0097] In a non-limiting embodiment, after feeding back the process optimization strategy set to the gold foil production control system to perform parameter control operations, the method further includes:

[0098] Step 220: Collect the defect type distribution data and process parameter adjustment records of the gold foil products after parameter control to construct an incremental training data set; perform incremental feature extraction processing on the quality description statements in the incremental training data set to generate incremental context semantic features; perform incremental parameter parsing processing on the adjusted process parameter records to generate incremental parameter association features; perform feature fusion processing on the incremental context semantic features and historical context semantic features to generate a global semantic feature matrix; perform time series splicing processing on the incremental parameter association features and historical parameter association features to generate a global parameter feature matrix; perform incremental training on the dynamic matching algorithm based on the global semantic feature matrix and the global parameter feature matrix, and update the mapping relationship between the defect feature space and the control feature space; perform real-time annotation processing on the newly added quality description data set according to the updated dynamic matching algorithm to generate incremental quality annotation results and update the process optimization strategy set.

[0099] In this embodiment, the system collects data after parameter adjustment, such as defect type distribution data for gold foil products showing a decrease in the proportion of "scratch defects" and a change in the proportion of "uneven thickness defects," as well as process parameter adjustment records, such as temperature parameter A adjusted from A1 to A2 and pressure parameter B adjusted from B1 to B2. This data is used to construct an incremental training dataset. Incremental feature extraction is performed on the quality description statements in the dataset. For example, for the newly appearing quality description statement "the gold foil edge has a slight curl," incremental contextual semantic features are generated using the previous feature extraction method. Incremental parameter parsing is performed on the adjusted process parameter records to obtain incremental parameter association features. The incremental contextual semantic features are fused with historical contextual semantic features, for example, by concatenating or weighted summing the newly generated feature vectors with the previously accumulated feature vectors to generate a global semantic feature matrix. Similarly, the incremental parameter association features are concatenated with the historical parameter association features in a time-series manner to form a global parameter feature matrix. Based on these two matrices, the dynamic matching algorithm is incrementally trained, and the mapping relationship between the defect feature space and the control feature space is updated by adjusting the algorithm's parameters, for example. Finally, the updated algorithm is used to perform real-time annotation processing on the newly added quality description data set, generating incremental quality annotation results, and updating the process optimization strategy set based on these results, so that the system can continuously adapt to new data and situations, and improve the ability to detect gold foil product quality and optimize the process.

[0100] In a non-limiting embodiment, after feeding back the process optimization strategy set to the gold foil production control system to perform parameter control operations, the method further includes:

[0101] Step 230: Acquire quality evaluation data of multiple batches of gold foil products after the parameter control operation is executed, the evaluation data including defect type statistics, process parameter fluctuation range and production energy consumption index; normalize the defect type statistics to generate a defect improvement rate, standardize the process parameter fluctuation range to generate a parameter stability coefficient, and perform dimensionless conversion on the production energy consumption index to generate an energy efficiency factor; perform weighted summation of the defect improvement rate, parameter stability coefficient and energy efficiency factor according to a preset weight configuration to generate a comprehensive optimization evaluation value; if the comprehensive optimization evaluation value is lower than a preset optimization threshold, perform multi-objective optimization processing on the process optimization strategy set, adjust the parameter adjustment direction weight to balance the conflict between the defect improvement rate and the energy efficiency factor; resend the process optimization strategy set after multi-objective optimization to the gold foil production control system, and iteratively execute the parameter control operation and evaluation processing.

[0102] Optionally, the system acquires quality assessment data from multiple batches of gold foil products. For example, the system counts the number of defects, such as "scratch defects" and "uneven thickness defects," across multiple batches of gold foil products and calculates defect type statistics. The fluctuation range of process parameters after parameter manipulation is recorded, such as the fluctuation of temperature parameter A between A2 and A3. Energy consumption indicators during the production process, such as total power consumption, are also calculated. Defect type statistics are normalized, for example, by comparing the number of defects in the current batch with the number before manipulation, and calculating the defect improvement rate (= (number of defects before manipulation - current number of defects) / number of defects before manipulation). The fluctuation range of process parameters is standardized and converted into a parameter stability coefficient using a conversion formula commonly used in related technologies. Production energy consumption indicators are dimensionlessly converted, such as converting total power consumption into an energy efficiency factor. Based on a preset weighting configuration, the defect improvement rate is weighted to 0.4, the parameter stability coefficient to 0.3, and the energy efficiency factor to 0.3. A weighted summation is performed to obtain a comprehensive optimization evaluation value. If the comprehensive optimization evaluation value is lower than the preset optimization threshold, it means that the current process optimization strategy set is not ideal and multi-objective optimization processing is required. For example, when it is found that the defect improvement rate has increased but the energy efficiency factor has decreased, the parameter adjustment direction weight is adjusted, such as appropriately reducing the adjustment intensity of parameters that have a greater impact on energy consumption, and increasing the parameter adjustment weight that is more effective in improving defects. The process optimization strategy set after multi-objective optimization is resent to the gold foil production control system, and then the parameter control operation and evaluation processing are performed again, and the iterative cycle is repeated. After the new round of parameter control operation is executed, the system obtains the quality evaluation data of multiple batches of gold foil products again, repeats the above-mentioned normalization, standardization, dimensionless conversion and weighted summation steps, and calculates a new comprehensive optimization evaluation value. If the new evaluation value still does not reach the preset optimization threshold, the process optimization strategy set will continue to be optimized for multiple objectives, and the parameter adjustment direction weight will be further adjusted. For example, if analysis finds that "increasing the rolling speed parameter C" can effectively reduce the probability of "scratch defects", but leads to a significant increase in production energy consumption, during multi-objective optimization, the weight of "increasing the rolling speed parameter C" should be appropriately reduced, while increasing the weight of other parameter adjustment directions that can both improve defects and reduce energy consumption, such as the weight of "fine-tuning the pressure parameter B to optimize the rolling process".

[0103] The adjusted process optimization strategy set is resent to the gold foil production control system to execute parameter control operations. During this process, the system continuously monitors the execution status of parameter control operations in real time to ensure that each control instruction is executed accurately and without error. If any abnormal situation occurs during execution, such as equipment feedback that a parameter cannot be adjusted to the target value, the system will immediately generate corresponding processing instructions based on the type of abnormality. For example, a rollback instruction will restore the parameter to its pre-adjustment state, or an adaptive adjustment instruction will be generated to attempt to adjust the target value or execution time, etc., to ensure the continuity and stability of the production process.

[0104] In addition, during each iteration, the system also collects the actual values of the process parameters after the parameter control operation and the corresponding gold foil quality inspection data set in real time according to the method of steps 210 and 220, generates and processes the dynamic compensation strategy, and constructs an incremental training data set to perform incremental training on the dynamic matching algorithm. By continuously collecting new data, updating algorithms, and adjusting strategies, the system can more accurately grasp the relationship between defect types and process parameters in the gold foil production process, gradually optimize process parameters, improve the quality of gold foil products, reduce production energy consumption, and enhance the stability and efficiency of the entire production system.

[0105] Throughout this continuous iterative process, each parameter adjustment and evaluation is based on actual production data, fully considering factors such as gold foil product quality, process parameter stability, and production energy consumption. Through continuous adjustment and optimization, the set of process optimization strategies becomes increasingly refined, better adapting to the various changes and demands of the gold foil production process. For example, as factors such as aging production equipment and fluctuations in raw material quality affect the system, the system can promptly adjust process parameters based on collected data to ensure that gold foil products maintain a consistently high level of quality.

[0106] Furthermore, with the continuous accumulation of data and continuous optimization of the algorithm, the dynamic matching algorithm will achieve even more accurate quality labeling results for quality description statements, more precisely indicating the mapping relationship between gold foil product defect types and process parameter adjustment directions. This will further improve the effectiveness and pertinence of the process optimization strategy set, enabling the gold foil production control system to more scientifically and rationally execute parameter control operations, thereby achieving intelligent, efficient, and high-quality gold foil production. The entire process forms a virtuous cycle, continuously driving the gold foil production process to a higher level of development and meeting market demand for high-quality gold foil products.

[0107] It is worth mentioning that when implementing the above technical solutions, technical personnel in the relevant field can normalize the process parameter characteristics based on the feature normalization method in the existing technology (such as Z-score or Min-Max normalization). By converting the slope characteristics and variance characteristics of different dimensional parameters such as temperature, pressure, and rolling speed into dimensionless standardized values, the influence of magnitude differences on the calculation of cross-category trend correlation matrices can be eliminated, thereby ensuring the calculation validity of statistical quantities such as the Pearson correlation coefficient.

[0108] For the dynamic matching algorithm, the cross-attention mechanism in the Transformer architecture can be used to implement a dual-channel alignment network. Specifically, the BERT pre-trained model is used as the language feature encoder and fine-tuned based on the domain corpus. The Stanford CoreNLP tool is introduced for grammatical analysis when constructing the semantic dependency tree. At the same time, the LSTM timing network is used to simulate the evolution of the production process after the process parameters are adjusted. By defining physical models such as the rolling force-temperature coupling equation, accurate prediction of the parameter change curve can be achieved.

[0109] For further optimization of the logical closed loop, a multi-task joint optimization framework can be constructed, and the quality labeling result generation and energy efficiency evaluation can be integrated into a unified loss function. The NSGA-II multi-objective optimization algorithm is used to balance the contradiction between defect improvement rate and energy consumption indicators. A synchronous update mechanism for parameter adjustment weights and feature space mapping is designed, and the TensorFlow model hot update interface is used to realize dynamic adaptation of strategy weights during incremental training.

[0110] In the process simulation and verification phase, a finite element model of the gold foil rolling process can be established based on the ANSYS simulation platform. By setting the material constitutive equation and contact boundary conditions, the stress distribution and surface deformation after parameter adjustment can be accurately simulated. The simulated microstructure can be scratch detected and the thickness measured using OpenCV image processing technology to verify the quality improvement effect.

[0111] In addition, by integrating the OPC UA industrial communication protocol, real-time interaction between optimized control instructions and production equipment PLCs can be achieved. The Modbus TCP protocol can be used to collect sensor data and feed it back to the compensation strategy generation module, ultimately forming a complete technical closed loop from quality description analysis to dynamic parameter control.

[0112] In summary, the above technical solution, through innovative technical means, can deeply explore the potential connection between quality description and process parameters, generate accurate quality labeling results and process optimization strategies, and effectively solve the problem that existing technologies are difficult to comprehensively analyze quality problems and accurately optimize processes in gold foil product production quality inspection, thereby improving the quality and efficiency of gold foil production.

[0113] In detail, the embodiments of the present invention realize the in-depth mining and effective utilization of gold foil product quality data: by obtaining a quality description data set covering quality description statements and process parameter records, it can provide comprehensive materials for subsequent analysis; based on feature extraction processing, it can accurately capture the contextual semantic features of sentences and parameter association features; based on the multi-dimensional strategy analysis processing of the dynamic matching algorithm, it can innovatively generate quality labeling results, thereby clearly presenting the mapping relationship between the gold foil product defect type and the process parameter adjustment direction; based on the quality labeling results, a process optimization strategy set is generated and fed back to the production control system to execute parameter regulation, which can quickly and accurately optimize the gold foil production process, improve product quality, reduce the defect rate, and enhance the intelligence and automation level of the production process.

[0114] See also Figure 2 As shown in FIG. 2 , this figure is a schematic diagram of the basic structure of a data annotation system 200 provided by an embodiment of the present invention. The data annotation system 200 includes:

[0115] Processor 201;

[0116] a storage device 202 having a computer program 2020 stored thereon;

[0117] When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the data labeling methods for gold foil product detection.

[0118] Based on the above, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0119] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems or devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the methods.

Claims

1. A data annotation method for gold foil product detection, characterized in that: include: Acquire a quality description data set of a gold foil product, wherein the quality description data set includes a plurality of quality description statements and corresponding process parameter records; Performing feature extraction processing on the quality description data set to obtain contextual semantic features of each quality description statement and parameter association features of the process parameter record; Mapping the context semantic features to the defect feature space to generate a defect feature vector containing a probability distribution of defect types; mapping the parameter association features to the control feature space to generate a control feature vector containing a weight of a parameter adjustment direction; performing spatial alignment processing on the defect feature vector and the control feature vector to determine a mapping relationship between the probability distribution of defect types and the weight of the parameter adjustment direction; constructing a dynamic matching matrix based on the mapping relationship, performing singular value decomposition processing on the dynamic matching matrix to obtain a defect type principal component vector and a parameter adjustment principal component vector; determining the defect type label of the quality description statement based on a similarity calculation processing between the defect type principal component vector and a preset defect type standard vector, the defect type label is used to characterize the defect type of the gold foil product; determining the parameter adjustment direction label based on a directional matching processing between the parameter adjustment principal component vector and the historical optimization strategy vector, the parameter adjustment direction label is used to characterize the process parameter adjustment direction; combining the defect type label and the parameter adjustment direction label to generate a quality labeling result of the quality description statement; the quality labeling result is used to indicate the mapping relationship between the defect type of the gold foil product and the process parameter adjustment direction; Extracting a defect type priority list and a parameter adjustment direction priority list from the quality labeling result; determining a first optimization weight according to the defect type priority list, and determining a second optimization weight according to the parameter adjustment direction priority list; Performing weighted fusion processing on the first optimization weight and the second optimization weight to generate a comprehensive optimization weight; The historical process parameter adjustment records are screened according to the comprehensive optimization weights to obtain a set of candidate adjustment strategies; a feasibility assessment is performed on the candidate adjustment strategy set to generate a result list containing assessment weights; the result list is sorted according to the assessment weights, and candidate adjustment strategies that exceed a preset score threshold are selected to generate a process optimization strategy set, and the process optimization strategy set is fed back to the gold foil production control system to execute parameter control operations.

2. The method according to claim 1, characterized in that The feature extraction process is performed on the quality description data set to obtain contextual semantic features of each quality description statement and parameter association features of the process parameter record, including: Performing semantic unit segmentation processing on the quality description sentence to obtain multiple semantic unit sequences; each semantic unit sequence includes at least one semantic segment and a position identifier of the semantic segment in the sentence; Calling a pre-trained language feature encoder to perform context encoding processing on the semantic unit sequence to generate context semantic features of the quality description sentence; the context semantic features include local semantic features of each semantic segment and context dependency relationships between adjacent semantic segments; Performing parameter parsing on the process parameter records to extract parameter category identifiers and parameter time series variation patterns; the parameter time series variation patterns include fluctuation characteristics of parameter values over time and correlation fluctuation trends between different parameters; Based on the parameter category identifier, feature fusion processing is performed on the parameter time series change pattern to generate parameter association features of the process parameter record; the parameter association features include multi-dimensional fluctuation correlation under the same parameter category and cross-category influence weights between different parameter categories.

3. The method according to claim 2, characterized in that The calling of a pre-trained language feature encoder to perform context encoding processing on the semantic unit sequence to generate contextual semantic features of the quality description sentence includes: Perform part-of-speech tagging on the semantic segments in each semantic unit sequence to generate a tag sequence containing part-of-speech tags; Constructing a semantic dependency tree based on the annotation sequence to determine the subject-verb relationship, modification relationship, and parallel relationship between the semantic segments; Performing hierarchical encoding processing on the semantic unit sequence based on the semantic dependency tree: performing local encoding on each semantic segment to generate an initial semantic vector, performing relational aggregation on the initial semantic vector according to the subject-predicate relationship to generate an intermediate semantic vector, and performing contextual enhancement on the intermediate semantic vector based on the modification relationship and the parallel relationship to generate an enhanced semantic vector; The enhanced semantic vectors are arranged in time sequence according to the position identifiers to generate contextual semantic features of the quality description sentence.

4. The method according to claim 2, characterized in that The performing parameter parsing on the process parameter records to extract parameter category identifiers and parameter time series change patterns includes: Dividing the process parameter records into multiple parameter subsequences according to time windows, each parameter subsequence containing a set of parameter values within a preset time range; Perform trend analysis on each parameter subsequence to extract rising trend segments, falling trend segments and stable trend segments; Performing slope calculation processing on the rising trend segment to generate a first slope feature set; performing slope absolute value calculation processing on the falling trend segment to generate a second slope feature set; performing variance calculation processing on the stationary trend segment to generate a fluctuation variance feature set; Classify the first slope feature set, the second slope feature set, and the volatility variance feature set according to the parameter category identifier to generate an intra-category trend feature matrix; The correlation calculation is performed on the intra-category trend feature matrix between different parameter categories to generate the cross-category trend correlation matrix; The intra-category trend feature matrix and the cross-category trend association matrix are spliced to generate the parameter time series change pattern.

5. The method according to claim 1, wherein The spatial alignment processing of the defect feature vector and the control feature vector to determine the mapping relationship between the defect type probability distribution and the parameter adjustment direction weight includes: Construct a dual-channel alignment network that includes defect type dimension and parameter adjustment dimension; Inputting the defect feature vector into a defect type channel for feature enhancement processing to generate an enhanced defect feature vector; Inputting the control feature vector into the parameter adjustment channel for feature enhancement processing to generate an enhanced control feature vector; Performing cross-attention calculation processing on the enhanced defect feature vector and the enhanced control feature vector to generate a defect-parameter attention weight matrix; Performing weighted summation processing on the enhanced defect feature vector and the enhanced control feature vector according to the defect-parameter attention weight matrix to generate a joint feature vector; Fully connected mapping processing is performed on the joint feature vector to generate a mapping relationship between the defect type probability distribution and the parameter adjustment direction weight.

6. The method according to claim 1, characterized in that The feasibility evaluation process is performed on the candidate adjustment strategy set to generate a result list containing evaluation weights, including: Split each candidate adjustment strategy into a sequence of parameter adjustment steps, where each parameter adjustment step includes an adjustment target value and an adjustment time range; Performing conflict detection on the parameter adjustment step sequence to identify conflicting step pairs with overlapping time or conflicting target values; generating a conflict assessment score based on the number and severity of the conflicting step pairs; Performing simulation execution processing on the parameter adjustment step sequence, and predicting the parameter change curve and defect type probability change curve after execution; Generate a parameter matching score based on the matching degree between the parameter change curve and the target parameter range; generate a defect improvement score based on the difference between the defect type probability change curve and the target defect threshold; A weighted summation process is performed on the conflict evaluation score, the parameter matching score, and the defect improvement score to generate a result list of evaluation weights of the candidate adjustment strategies.

7. The method according to claim 1, characterized in that Feeding back the process optimization strategy set to the gold foil production control system to perform parameter control operations includes: Converting the process optimization strategy set into a control instruction sequence, each control instruction including a parameter identifier, a target value, and an execution timestamp; Performing timing verification on the control instruction sequence to ensure that the execution timestamp complies with the operation interval limit of the production equipment; Performing timestamp adjustment processing on the control instruction sequence according to the verification result to generate an optimized control instruction sequence; Sending the optimized control instruction sequence to a parameter execution module of a gold foil production control system, instructing the parameter execution module to execute parameter control operations according to the optimized control instruction sequence; The execution status of the parameter control operation is monitored in real time. When an execution abnormality is detected, a rollback instruction or an adaptive adjustment instruction is generated according to the abnormality type to update the optimized control instruction sequence.

8. A data annotation system, characterized in that: include: processor; A storage device having a computer program stored thereon, wherein when the computer program is executed by the processor, the processor implements the data labeling method for gold foil product detection as described in any one of claims 1 to 7.

Images