Flexible sheet product production quality traceability method, system, device and storage medium

Abstract

The application discloses a kind of soft board product production quality traceability method, system, device and storage medium, including obtaining the quality inspection data subset and process parameter data subset of soft board product under multiple batches, construct soft board production dataset;Soft board production dataset is preprocessed;At least one main defect type of soft board product is obtained from the soft board production dataset after preprocessing, and the Bayesian inference network of defect-process-process parameter is constructed;Using Bayesian inference network, according to the soft board production dataset after preprocessing, at least one key process under each main defect type of soft board product and at least one key process parameter under each key process are obtained.The application uses the mode of data analysis and data reasoning instead of the mode of artificial positioning quality problem to realize the traceability of soft board product production quality, and the labor cost is less, and the analysis efficiency is high, can realize efficient, fast and accurate production quality traceability.

Description

Technical Field

[0001] This invention relates to the field of flexible printed circuit board (FPCB) manufacturing, and more specifically to a method, system, device, and storage medium for tracing the production quality of FPCB products. Background Technology

[0002] As the number of flexible printed circuit boards (FPCs) entering the market continues to increase, the issues of product traceability efficiency and accuracy are becoming increasingly apparent, and the accuracy and efficiency of FPC product quality traceability are also becoming more and more prominent.

[0003] Currently, the production quality monitoring and traceability of flexible board products mainly rely on manual monitoring and location tracking. The main process is as follows:

[0004] (1) After a quality engineer discovers a quality problem, the problematic batch and the defect problem are reported back to the production process engineer;

[0005] (2) Production process engineers check the potentially related problematic processes based on the type of defect and their experience, and determine the cause through manual traceability experiments and other methods.

[0006] This method requires a lot of manpower, and the time to identify production quality problems may be delayed. It is also prone to failure of quality traceability due to human error, resulting in low accuracy and efficiency of quality traceability. Summary of the Invention

[0007] In view of this, the present invention provides a method, system, device and storage medium for tracing the production quality of flexible board products, in order to solve the problems of high labor costs, low accuracy and low efficiency in existing manual quality monitoring and traceability technologies.

[0008] This invention provides a method for tracing the production quality of flexible board products, comprising:

[0009] Obtain subsets of quality inspection data and process parameter data from multiple batches of flexible board products, and construct a flexible board production dataset.

[0010] The flexible board production dataset is preprocessed;

[0011] At least one major defect type of the flexible board product is obtained from the preprocessed flexible board production dataset, and a Bayesian inference network of defect-process-process parameters is constructed.

[0012] Using the Bayesian inference network, based on the preprocessed flexible board production dataset, at least one key process and at least one key process parameter for each of the major defect types of the flexible board product are obtained.

[0013] Optionally, the subset of process parameter data under each batch includes multiple processes of the flexible board product under the corresponding batch and the process parameter data of each process;

[0014] The acquisition of subsets of quality inspection data and process parameter data for flexible board products across multiple batches includes:

[0015] Obtain the part number of the flexible board product;

[0016] Based on the product part number, obtain the transit data of the flexible board product from the preset execution manufacturing system;

[0017] Based on the station data, multiple processes for the flexible board product are determined;

[0018] Using the data acquisition device configured on the flexible board product, the process parameter data of all processes in each batch of the flexible board product are collected;

[0019] Based on the process parameter data of all processes under each batch, a subset of process parameter data for each batch is obtained;

[0020] Using the quality inspection equipment configured on the flexible board product, a subset of quality inspection data for the flexible board product in multiple batches is obtained.

[0021] Optionally, each subset of quality inspection data under each batch includes the corresponding batch quality inspection results, total number of products, at least one product defect type, and the number of defective products for each product defect type.

[0022] Optionally, the preprocessing of the flexible board production dataset includes:

[0023] Based on the batch quality inspection results of all batches, the flexible board production dataset is divided into a normal batch data subset and an abnormal batch data subset;

[0024] Using all the process parameter data in the normal batch data subset, multiple process parameter ranges are identified;

[0025] The subset of abnormal batch data is discretized using all the process parameter ranges.

[0026] Optionally, the process parameter range includes a normal parameter range, a parameter range that is too small, a parameter range that is too small, a parameter range that is too large, and a parameter range that is too large.

[0027] Optionally, the step of obtaining at least one major defect type of the flexible board product from the preprocessed flexible board production dataset and constructing a Bayesian inference network of defect-process-process parameters includes:

[0028] At least one major defect type is determined from all product defect types in the discretized subset of the abnormal batch data;

[0029] Each of the major defect types is taken as a root node. The first child node corresponding to each root node is determined according to all the processes in the discretized abnormal batch data subset. The first conditional probability between each first child node and the corresponding root node is calculated.

[0030] Construct the first Bayesian network of defect-process based on all root nodes, all first child nodes, and all first conditional probabilities;

[0031] Each first child node is then used as a parent node. Based on all the process parameter data in the discretized subset of abnormal batch data, the second child node corresponding to each parent node is determined, and the second conditional probability between each second child node and its corresponding parent node is calculated.

[0032] Construct a second Bayesian network for process parameters based on all parent nodes, all second child nodes, and all second conditional probabilities;

[0033] Based on the first Bayesian network and the second Bayesian network, the Bayesian inference network for defect-process-process parameters is obtained.

[0034] Optionally, the step of using the Bayesian inference network to obtain at least one key process and at least one key process parameter for each major defect type of the flexible board product, based on the preprocessed flexible board production dataset, includes:

[0035] Using the Bayesian inference network, the posterior probability between each process parameter data and each major defect type in the discretized subset of abnormal batch data is calculated respectively.

[0036] Based on a preset probability threshold and all the posterior probabilities, at least one critical process and at least one critical process parameter for each of the major defect types of the flexible board product are determined.

[0037] Optionally, after obtaining at least one critical process step and at least one critical process parameter for each of the major defect types of the flexible board product, the method further includes:

[0038] Obtain the mining dataset of the flexible board product; wherein, the mining dataset includes at least one defect type to be mined from multiple batches of the flexible board product within a preset time period, at least one process to be mined under each defect type to be mined, and at least one process parameter to be mined corresponding to each process to be mined.

[0039] Based on the association rule mining method, association rule mining is performed on the mining dataset;

[0040] Based on the association rule mining results, all key process parameters under all the main defect types are verified to obtain the target key process parameter set for the flexible board product under each main defect type.

[0041] Optionally, the association rule mining method performs association rule mining on the mining dataset, including:

[0042] Discretize the data set being mined;

[0043] Based on the Bayesian inference network, all candidate defect types are determined from all the defect types to be mined in the discretized mining dataset;

[0044] Select any of the candidate defect types, and based on the association rule mining method, concentrate the discretized mining data into all the process parameters to be mined under the selected candidate defect type, and determine it as the candidate set corresponding to the selected candidate defect type;

[0045] In the candidate set corresponding to the selected candidate defect type, the k process parameters related to the selected candidate defect type under each mining operation are determined as the candidate k-item set of the selected candidate defect type.

[0046] The set of all candidate k-itemsets of the selected candidate defect type is denoted as the first set C corresponding to the selected candidate defect type. k Where k is a positive integer with an initial value of 1;

[0047] Calculate the first set C respectively k The support of each candidate k-item set in the first set C is determined based on a preset minimum support and all the supports. k The process involves obtaining at least one frequent k-itemset corresponding to the selected candidate defect type, and denoting the set of all frequent k-itemsets corresponding to the selected candidate defect type as the second set L of the selected candidate defect type. k ;

[0048] Determine the second set L k Is it an empty set?

[0049] If so, then calculate the second set L respectively. k The confidence level of each of the frequent k-itemsets in the second set L is determined based on a preset minimum confidence level and all the aforementioned confidence levels. k The process involves obtaining the target frequent itemset corresponding to the selected candidate defect type from the discretized mining dataset, and completing the association rule mining for the selected candidate defect type in the discretized mining dataset; traversing each candidate defect type in the discretized mining dataset, and obtaining the target frequent itemset corresponding to each candidate defect type in the same way, thus completing the association rule mining for each candidate defect type;

[0050] Otherwise, let k = k + 1, for the second set L k Perform the join and pruning operations sequentially to obtain the second set L. k Using the same method, a second set L of the selected candidate defect types is obtained. k+1 The process is iterated until the second set L of the selected candidate defect types is reached. k+1 If the set is empty, end the iteration; and after the iteration ends, follow the second set L. k When the set is empty, the same method is used to obtain the target frequent itemset corresponding to each candidate defect type, and to complete the association rule mining for each candidate defect type.

[0051] Optionally, calculate the first set C corresponding to the i-th candidate defect type. k The first formula for the support of the j-th candidate k-itemset is:

[0052]

[0053] Calculate the second set L corresponding to the i-th candidate defect type. k The second formula for the confidence score of the t-th frequent k-itemset is:

[0054]

[0055] in, The first set C corresponding to the i-th candidate defect type k The support of the j-th candidate k-itemset, Let N be the number of batches in the j-th candidate k-item set that contain the i-th candidate defect type and have k process parameters to be mined, and let N be the total number of batches in the mining dataset. The second set L corresponding to the i-th candidate defect type kThe confidence level of the t-th frequent k-itemset in the given context. m is the number of batches in the t-th frequent k-item set where the i-th candidate defect type appears and there are k process parameters to be mined. i The number of batches in the mining dataset where the i-th candidate defect type appears.

[0056] Optionally, the step of verifying all the key process parameters under all the major defect types obtained based on the association rule mining results to obtain the target key process parameter set for the flexible board product under each major defect type includes:

[0057] Select any one of the main defect types and determine whether all the key process parameters under the selected main defect type are consistent with the target frequent itemset under the corresponding candidate defect type;

[0058] If so, the set of all the key process parameters under the selected main defect type is determined as the target key process parameter set corresponding to the selected main defect type;

[0059] If not, then the set of all process parameters to be mined in the target frequent itemset under the corresponding candidate defect type is determined as the set of target key process parameters corresponding to the selected main defect type.

[0060] By iterating through each of the major defect types, and following the same method, the set of target critical process parameters for the flexible board product under each of the major defect types is obtained.

[0061] Furthermore, the present invention also provides a production quality traceability system for flexible board products, comprising:

[0062] The data acquisition module is used to acquire subsets of quality inspection data and process parameter data of flexible board products in multiple batches, and to construct a flexible board production dataset.

[0063] The data processing module is used to preprocess the flexible board production dataset;

[0064] A network construction module is used to obtain at least one major defect type of the flexible board product from the preprocessed flexible board production dataset and construct a Bayesian inference network of defect-process-process parameters.

[0065] The inference module is used to utilize the Bayesian inference network to obtain at least one key process and at least one key process parameter for each key process of the flexible board product under each major defect type, based on the preprocessed flexible board production dataset.

[0066] Optionally, it also includes:

[0067] The association rule mining module is used to obtain the mining dataset of the flexible board product; wherein, the mining dataset includes at least one defect type to be mined from multiple batches of the flexible board product within a preset time period, at least one process to be mined under each defect type to be mined, and at least one process parameter to be mined corresponding to each process to be mined; it is also used to perform association rule mining on the mining dataset based on the association rule mining method.

[0068] The verification module is used to verify all the key process parameters under all the main defect types obtained based on the association rule mining results, so as to obtain the target key process parameter set of the flexible board product under each main defect type.

[0069] In addition, the present invention also provides a flexographic board product production quality traceability device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned flexographic board product production quality traceability method.

[0070] In addition, the present invention provides a computer-readable storage medium comprising: at least one instruction that, when executed, implements the method steps in the aforementioned method for tracing the production quality of flexible board products.

[0071] The beneficial effects of this invention are as follows: By constructing a flexible board production dataset based on subsets of quality inspection data and process parameter data from multiple batches of flexible board products, the quality inspection status of flexible board products in multiple batches and the process and technological parameters during production can be obtained. This facilitates the subsequent construction of a Bayesian inference network for defects, processes, and technological parameters using this big data. Typically, the quality inspection data of each batch can reflect the abnormalities of the corresponding batch. Through the constructed Bayesian inference network, the patterns of production anomalies on the flexible board production line can be summarized in the form of data inference learning. Furthermore, the relationship between defects in flexible board products and process and technological parameters can be analyzed, that is, at least one key process and at least one key technological parameter under each major defect type of flexible board product. This allows for tracing back to all key technological parameters on the flexible board production line that lead to each major defect type, thus achieving traceability of production quality.

[0072] The present invention provides a method, system, device, and storage medium for tracing the production quality of flexible board products. By using data analysis and data reasoning to replace manual identification of quality problems, it achieves traceability of the production quality of flexible board products, reduces labor costs, improves analysis efficiency, and avoids quality traceability failures due to human error. It enables efficient, fast, and accurate production quality traceability, helping enterprises optimize the production of flexible board products based on the results of quality traceability and improve enterprise profits. Attached Figure Description

[0073] The features and advantages of the invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the drawings:

[0074] Figure 1 A flowchart of a production quality traceability method for flexible board products according to Embodiment 1 of the present invention is shown;

[0075] Figure 2 This illustrates a flowchart of obtaining subsets of quality inspection data and process parameter data for flexible board products across multiple batches in Embodiment 1 of the present invention.

[0076] Figure 3 The flowchart illustrating the preprocessing of the flexible board production dataset in Embodiment 1 of the present invention is shown.

[0077] Figure 4 The flowchart of constructing a Bayesian inference network for defect-process-process parameters in Embodiment 1 of the present invention is shown;

[0078] Figure 5 A model diagram of the Bayesian inference network for defect-process-process parameters in Embodiment 1 of the present invention is shown;

[0079] Figure 6 This document illustrates a flowchart of obtaining at least one key process step and at least one key process parameter for each major defect type in Embodiment 1 of the present invention.

[0080] Figure 7 A flowchart of another method for tracing the production quality of flexible board products according to Embodiment 1 of the present invention is shown;

[0081] Figure 8 This diagram illustrates the structure of a production quality traceability system for flexible board products according to Embodiment 2 of the present invention.

[0082] Figure 9 The diagram shows a structural diagram of another flexible board product production quality traceability system according to Embodiment 2 of the present invention. Detailed Implementation

[0083] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0084] Example 1

[0085] like Figure 1 As shown, a method for tracing the production quality of flexible board products includes:

[0086] S1: Obtain subsets of quality inspection data and process parameter data for flexible board products in multiple batches, and construct a flexible board production dataset;

[0087] S2: Preprocess the flexible board production dataset;

[0088] S3: Obtain at least one major defect type of the flexible board product from the preprocessed flexible board production dataset, and construct a Bayesian inference network of defect-process-process parameters;

[0089] S4: Using the Bayesian inference network, based on the preprocessed flexible board production dataset, obtain at least one key process for the flexible board product under each of the major defect types and at least one key process parameter for each key process.

[0090] A flexible board production dataset is constructed based on subsets of quality inspection data and process parameter data from multiple batches of flexible board products. This dataset provides information on the quality inspection status of flexible board products across multiple batches, as well as the process and parameters involved in the production process. This facilitates the subsequent construction of a Bayesian inference network connecting defects, processes, and process parameters using this big data. Typically, the quality inspection data for each batch reflects the abnormalities of that batch. Through the constructed Bayesian inference network, patterns of production anomalies on the flexible board production line can be summarized using data inference learning. This allows for the analysis of the relationships between defects in flexible board products and process and parameter conditions. Specifically, it identifies at least one critical process and at least one critical process parameter for each major defect type in the flexible board product. This enables the tracing of all critical process parameters on the flexible board production line that led to each major defect type, thus achieving traceability of production quality.

[0091] The flexible board product production quality traceability method in this embodiment uses data analysis and data reasoning to replace manual identification of quality problems, realizing the traceability of flexible board product production quality. This reduces labor costs and improves analysis efficiency, preventing quality traceability failures due to human error. It achieves efficient, fast, and accurate production quality traceability, helping companies optimize flexible board product production based on the results of quality traceability and improve corporate profits.

[0092] Specifically, the subset of process parameter data for each batch includes multiple processes of the flexible board product in the corresponding batch, as well as the process parameter data for each process.

[0093] The aforementioned subset of process parameter data facilitates subsequent deduction of the relationship between key processes and their corresponding key process parameter data.

[0094] Preferably, such as Figure 2 As shown, S1 includes:

[0095] S11: Obtain the part number of the flexible board product;

[0096] S12: Based on the product part number, obtain the transit data of the flexible board product from the preset execution manufacturing system;

[0097] S13: Determine multiple processes for the flexible board product based on the station data;

[0098] S14: Using the data acquisition device configured on the flexible board product, collect the process parameter data of all processes of the flexible board product in each batch;

[0099] S15: Based on the process parameter data of all processes under each batch, obtain the process parameter data subset of each batch;

[0100] S16: Using the quality inspection equipment configured on the flexible board product, a subset of quality inspection data for the flexible board product in multiple batches is obtained.

[0101] Flexible board production lines typically include several common pieces of equipment, such as Manufacturing Execution Systems (MES), data acquisition devices, and quality inspection equipment. The MES system stores transit data for the flexible board products, including the start and end times of each process within each batch. Therefore, by identifying the part number on the flexible board product, the corresponding transit data can be exported from the MES system. This transit data identifies all processes corresponding to that part number. Combined with the process parameter data for the entire batch collected by the data acquisition device, the process parameter data for each process can be segmented, resulting in one-to-one process parameter data for each process. This allows for the creation of a subset of process parameter data for each batch. The quality inspection equipment performs quality checks on each batch of flexible board products, and the resulting data provides a subset of quality inspection data for each batch.

[0102] Specifically, each subset of quality inspection data under each batch includes the corresponding batch quality inspection results, total number of products, at least one product defect type, and the number of defective products for each product defect type.

[0103] When quality inspection equipment inspects each batch of flexible board products, it provides various quality inspection data, including batch inspection results, total number of products, product defect type, and number of defective products. The batch inspection result indicates whether the batch is normal or abnormal. A batch is considered normal if its pass rate meets the standard, and abnormal if it does not. Regardless of whether it's a normal or abnormal batch, each batch will contain defective products. The total number of products refers to the total number of flexible board products produced in each batch. The product defect type refers to the defect type of the defective products in each batch. The total number of defective products is the total number of defective products in each batch, and the number of defective products is the number of products with each defect type in each batch. This embodiment uses the subset of quality inspection data for each batch to facilitate accurate subsequent analysis of the relationship between defects and processes, as well as process parameter data.

[0104] Preferably, such as Figure 3 As shown, S2 includes:

[0105] S21: Based on the batch quality inspection results of all batches, the flexible board production dataset is divided into a normal batch data subset and an abnormal batch data subset;

[0106] S22: Using all the process parameter data in the normal batch data subset, multiple process parameter ranges are identified;

[0107] S23: Discretize the subset of abnormal batch data using all the process parameter ranges.

[0108] By dividing the flexible board production dataset into normal batch data subsets and abnormal batch data subsets, it is easier to identify the qualified process parameter ranges for the flexible board product in each process based on the normal batch data subsets. Then, the qualified process parameter ranges in each process are used as discretization standards to discretize the abnormal batch data subsets, which facilitates the subsequent construction of a Bayesian inference network to achieve production quality traceability through data inference.

[0109] Specifically, the process parameter range includes the normal parameter range, the parameter too small range, the parameter too small range, the parameter too large range, and the parameter too large range.

[0110] By using the above five process parameter ranges, the subset of abnormal batch data can be better discretized, making the data inference results of the subsequently constructed Bayesian inference network more accurate.

[0111] In one specific embodiment of this example, process parameter data for each process step in a certain number of batches are taken from a subset of normal batch data. The mean and standard deviation of each process parameter data for each process step in all selected batches are then calculated. For example, the mean and standard deviation of the first process parameter data P1X1 for all batches in process P1 are calculated as mean... P1X1 and sigma P1X1 Then [mean] P1X1 -3sigma P1X1 ,mean P1X1 +3sigma P1X1 The data range is used as the normal parameter range corresponding to the process parameter data P1X1 for all batches. P1X1 -6sigma P1X1 ,mean P1X1 -3sigma P1X1 The data range is used as the smaller interval of the process parameter data P1X1 for all batches, and values ​​less than the mean are considered. P1X1 -6sigma P1X1 The data range is used as the interval where the parameter P1X1 corresponding to the process parameter data is too small for all batches. P1X1 +3sigma P1X1 ,mean P1X1 +6sigma P1X1The data range is used as the larger range of the process parameter data P1X1 for all batches, and will be greater than the mean. P1X1 +6sigma P1X1 The data range is used as the excessively large range for the process parameter data P1X1 across all batches. The five process parameter ranges for other process parameter data follow the same principle, and will not be elaborated further in this embodiment. Finally, the five process parameter ranges for all the above-mentioned process parameter data are used to discretize all process parameter data in the subset of abnormal batch data.

[0112] Preferably, such as Figure 4 As shown, S3 includes:

[0113] S31: Determine at least one of the major defect types from all the product defect types in the discretized subset of the abnormal batch data;

[0114] S32: Take each of the major defect types as the root node, determine the first child node corresponding to each root node according to all the processes in the discretized abnormal batch data subset, and calculate the first conditional probability between each first child node and the corresponding root node.

[0115] S33: Construct the first Bayesian network of defect-process based on all root nodes, all first child nodes and all first conditional probabilities;

[0116] S34: Each first child node is then used as a parent node. Based on all the process parameter data in the discretized subset of abnormal batch data, the second child node corresponding to each parent node is determined, and the second conditional probability between each second child node and its corresponding parent node is calculated.

[0117] S35: Construct a second Bayesian network for process parameters based on all parent nodes, all second child nodes, and all second conditional probabilities;

[0118] S36: Based on the first Bayesian network and the second Bayesian network, obtain the Bayesian inference network of defect-process-process parameters.

[0119] By identifying the main defect types, the amount of data can be reduced, thus lowering the complexity of subsequent data inference. Using each identified main defect type as the root node and all corresponding processes as their first child nodes, a directed model graph between the first and second layers can be constructed. This is a Bayesian network describing the relationship between main defect types and processes—the defect-process first Bayesian network. Similarly, using each process as the parent node and all corresponding process parameter data as their second child nodes, a directed model graph between the second and third layers can be constructed. This is a Bayesian network describing the relationship between processes and process parameter data—the process-process parameter second Bayesian network. Finally, a three-layer Bayesian network is obtained from the first and second Bayesian networks—the defect-process-process parameter Bayesian inference network. By constructing a Bayesian inference network as described above, an inference model can be built that accurately reflects the relationship between major defect types, processes, and process parameter data. This facilitates subsequent mathematical reasoning to trace the main influencing factors (i.e., key process parameters) affecting the quality of flexible board products when a certain major defect type (i.e., a quality problem) occurs. This provides guidance for flexible board production lines in future production, improving the quality of flexible board products and the company's production efficiency.

[0120] The model diagram of the Bayesian inference network for defect-process-process parameters constructed in this embodiment is shown below. Figure 5 As shown, in Figure 5 In this diagram, the main defect types are represented by Y1, Y2, ..., Yn, the processes are represented by P1, P2, ..., Pm, and the process parameter data under process P1 are represented by P1X1, P1X2, ..., P1Xk1. The process parameter data under other processes are represented in the same way, and will not be repeated here.

[0121] In this embodiment S31, the main defect types can be obtained in two ways. One is to pre-determine all main defect types based on manual experience; the other is to calculate the percentage of each product defect type in each batch and determine the product defect types with the highest percentages as the main defect types. In this embodiment, the main defect types include black holes, hole misalignment, short circuits, and open circuits.

[0122] In S32 and S34 of this embodiment, the conditional probabilities between the root node and the first child node, and between the parent node and the second child node, are calculated using conventional calculation methods in existing Bayesian network model techniques. Specific details will not be elaborated here.

[0123] Preferably, such as Figure 6 As shown, S4 includes:

[0124] S41: Using the Bayesian inference network, calculate the posterior probability between each process parameter data and each major defect type in the discretized subset of the abnormal batch data;

[0125] S42: Based on a preset probability threshold and all the posterior probabilities, determine at least one critical process and at least one critical process parameter for each of the major defect types of the flexible board product.

[0126] By utilizing the constructed Bayesian inference network, respectively The posterior probability between process parameter data and major defect types is calculated, and combined with a preset probability threshold, the key process parameters that affect the major defect types are identified, thereby enabling the localization of quality problems in flexible board products.

[0127] In this embodiment S41, for Figure 5 The formula for calculating the posterior probability between the main defect type and process parameter data in the Bayesian inference network shown is as follows:

[0128]

[0129] Wherein, P(P s X ks |Y l ) is the Pth s Process parameter data P for each process step s X ks For the posterior probability of the occurrence of the l-th major defect type, P(P s X ks ,Y l ) indicates the occurrence of the l-th major defect type and the existence of the P-th defect type. s Process parameter data P for each process step s X ks The probability, P(Y) l ) represents the probability of the occurrence of the l-th major defect type.

[0130] In this embodiment S42, the posterior probability can reflect the influence of process parameter data on the main defect types. Therefore, by using a preset probability threshold, when the posterior probability of a certain process parameter data is greater than the preset probability threshold, it indicates that the process parameter data has a greater influence on the main defect types, and it is the key process parameter to be found. Following the same comparison method, all key process parameters corresponding to all main defect types can be found.

[0131] Preferably, such as Figure 7 As shown, after S4, it also includes:

[0132] S5: Obtain the mining dataset of the flexible board product; wherein, the mining dataset includes at least one defect type to be mined from multiple batches of the flexible board product within a preset time period, at least one process to be mined under each defect type to be mined, and at least one process parameter to be mined corresponding to each process to be mined.

[0133] S6: Based on the association rule mining method, perform association rule mining on the mining dataset;

[0134] S7: Based on the association rule mining results, verify all the key process parameters under all the main defect types to obtain the target key process parameter set for the flexible board product under each main defect type.

[0135] After inferring at least one key process parameter corresponding to each major defect type through a Bayesian inference network, a mining dataset is provided, and association rules are formed on the mining dataset based on the association rule mining method. The obtained association rule mining results are verified with the at least one key process parameter corresponding to each major defect type obtained through inference. This ensures that all key process parameters corresponding to each major defect type can be consistent across all batches. This further identifies these key process parameters as the key factors leading to the corresponding quality problems, enabling precise location of quality problems in flexible board product manufacturing and more accurate quality traceability.

[0136] Specifically, in embodiment S5, the mining dataset includes at least one defect type to be mined from multiple batches of flexible board products within a preset time period, at least one process to be mined under each defect type, and at least one process parameter to be mined corresponding to each process. That is, the mining dataset includes at least three types of data: defect type, process, and process parameter. It may also include the total number of products and the number of defective products per batch, similar to the flexible board production dataset. In other words, the data structure of the mining dataset is the same as that of the flexible board production dataset. The flexible board production dataset can be large datasets collected from the flexible board production line at any point in the past, while the mining dataset consists of consecutive batches of data collected from the flexible board production line in a recent period. The specific time period within the preset time period in the mining dataset can be determined according to actual circumstances.

[0137] Preferably, S6 includes:

[0138] S61: Discretize the mining dataset;

[0139] S62: Based on the Bayesian inference network, determine all candidate defect types from all the defect types to be mined in the discretized mining dataset;

[0140] S63: Select any one of the candidate defect types, and based on the association rule mining method, concentrate the discretized mining data into all the process parameters to be mined under the selected candidate defect type, and determine it as the candidate set corresponding to the selected candidate defect type;

[0141] S64: In the candidate item set corresponding to the selected candidate defect type, the k process parameters related to the selected candidate defect type under each excavation process are determined as the candidate k-item set of the selected candidate defect type.

[0142] S65: The set of all candidate k-itemsets of the selected candidate defect type is denoted as the first set C corresponding to the selected candidate defect type. k Where k is a positive integer with an initial value of 1;

[0143] S66: Calculate the first set C respectively. k The support of each candidate k-item set in the first set C is determined based on a preset minimum support and all the supports. k The process involves obtaining at least one frequent k-itemset corresponding to the selected candidate defect type, and denoting the set of all frequent k-itemsets corresponding to the selected candidate defect type as the second set L of the selected candidate defect type. k ;

[0144] S67: Determine the second set L k Is it an empty set?

[0145] If so, then calculate the second set L respectively. k The confidence level of each of the frequent k-itemsets in the second set L is determined based on a preset minimum confidence level and all the aforementioned confidence levels. k The process involves obtaining the target frequent itemset corresponding to the selected candidate defect type from the discretized mining dataset, and completing the association rule mining for the selected candidate defect type in the discretized mining dataset. Then, iterating through each candidate defect type in the discretized mining dataset, the process continues to obtain the target frequent itemset corresponding to each candidate defect type, and completing the association rule mining for each candidate defect type.

[0146] Otherwise, let k = k + 1, for the second set L k Perform the join and pruning operations sequentially to obtain the second set L. k Using the same method, a second set L of the selected candidate defect types is obtained. k+1 The process is iterated until the second set L of the selected candidate defect types is reached.k+1 If the set is empty, end the iteration; and after the iteration ends, follow the second set L. k When the set is empty, the same method is used to obtain the target frequent itemset corresponding to each candidate defect type, and to complete the association rule mining for each candidate defect type.

[0147] In this embodiment, the association rule mining method is specifically the Apriori algorithm, which is a frequent itemset mining algorithm for association rules. In S61, the mining dataset is discretized. This facilitates the subsequent use of a Bayesian inference network to determine candidate defect types corresponding to the main defect type, and also facilitates the mining of target frequent itemsets based on the discretized process parameters, thus realizing association rule mining. In S62, the candidate defect types determined by the Bayesian inference network can be easily verified against all key process parameters under the main defect type using the association rule mining results.

[0148] In steps S63-S67, the mining of association rules for any selected candidate defect type first involves identifying all process parameters to be mined under each process to be mined that are related to the candidate defect type as corresponding candidate item sets. Then, within these candidate item sets, we search for item sets related to the candidate defect type that contain k process parameters to be mined. Since k is an initial positive integer of 1, it can be incremented sequentially from 1. Therefore, we can sequentially determine the set of the 1st process parameter, the set of the 2nd process parameter, and so on, up to the set of k process parameters most relevant to the candidate defect type. This process aggregates all candidate defect types and their related abnormal process parameters to be mined, generating different candidate k-itemsets. For each determined k value, the set of all candidate k-itemsets for the selected candidate defect type is the first set C corresponding to that selected candidate defect type. k .

[0149] For example, when k is initially set to 1, the processes to be excavated associated with the selected candidate defect type are P1, P2, P3, and P4. The candidate item set corresponding to process P1 contains only one most relevant process parameter to be excavated, forming a candidate 1-itemset in process P1 corresponding to the selected candidate defect type. Similarly, the candidate item set corresponding to process P2 contains only one most relevant process parameter to be excavated, forming a candidate 1-itemset in process P2 corresponding to the selected candidate defect type. The candidate item set corresponding to process P3 contains only one most relevant process parameter to be excavated, forming a candidate 1-itemset in process P3 corresponding to the selected candidate defect type. The candidate item set corresponding to process P4 contains only one most relevant process parameter to be excavated, forming a candidate 1-itemset in process P4 corresponding to the selected candidate defect type. These four candidate 1-itemsets constitute the first set C1 corresponding to the selected candidate defect type.

[0150] When candidate k-itemsets are generated and the first set C formed by all candidate k-itemsets is obtained... k Next, by calculating the support of each candidate k-itemset and combining it with the preset minimum support, several candidate k-itemsets that best meet the support conditions can be obtained, called frequent k-itemsets (the remaining candidate k-itemsets that do not meet the support conditions are eliminated, which can reduce the complexity of association rule mining). The k process parameters to be mined in each frequent k-itemset are the process parameters most relevant to the selected candidate defect type; these frequent k-itemsets form the second set L corresponding to the selected candidate defect type. k .

[0151] The above forms the second set L k The process is iterative and loop-based for different k values. For each different k value, when the corresponding second set L is formed... k Then, determine the second set L. k Is it an empty set? If so, then it means that the second set L is empty. k Once the most frequent k-itemsets in the selected candidate defect types have been completely determined and the most relevant process parameters are identified, the iteration can be terminated; otherwise, iteration needs to continue, and the second set L can be obtained in the same way. k until the final second set L is obtained. k Given an empty set, identify the most relevant process parameters that are already fully determined.

[0152] After the iteration terminates, the second set L is calculated. kThe confidence level of each frequent k-itemset is combined with the preset minimum confidence level to obtain the target frequent itemset, which is the set of process parameters that are most likely to be abnormal and thus cause quality problems when the selected candidate defect type appears in the batch. In other words, the target frequent itemset corresponding to the selected candidate defect type is obtained, and the association rule mining of the candidate defect type is completed.

[0153] For each candidate defect type, association rule mining is performed in the same way. This allows us to infer from the candidate defect type that the defect problem in this batch is caused by a single abnormal process parameter or a combination of several abnormal process parameters. From this, we can further infer that the abnormal production parameters of these processes have a high probability of causing the candidate defect problem in this batch, thus further realizing the quality traceability of flexible board products.

[0154] Specifically, in S66, the first set C corresponding to the i-th candidate defect type is calculated. k The first formula for the support of the j-th candidate k-itemset is:

[0155]

[0156] In S67, the second set L corresponding to the i-th candidate defect type is calculated. k The second formula for the confidence score of the t-th frequent k-itemset is:

[0157]

[0158] in, The first set C corresponding to the i-th candidate defect type k The support of the j-th candidate k-itemset, Let N be the number of batches in the j-th candidate k-item set that contain the i-th candidate defect type and have k process parameters to be mined, and let N be the total number of batches in the mining dataset. The second set L corresponding to the i-th candidate defect type k The confidence level of the t-th frequent k-itemset in the given context. m is the number of batches in the t-th frequent k-item set where the i-th candidate defect type appears and there are k process parameters to be mined. i The number of batches in the mining dataset where the i-th candidate defect type appears.

[0159] Preferably, S7 includes:

[0160] S71: Select any one of the main defect types, and determine whether all the key process parameters under the selected main defect type are consistent with the target frequent itemset under the corresponding candidate defect type;

[0161] If so, the set of all the key process parameters under the selected main defect type is determined as the target key process parameter set corresponding to the selected main defect type;

[0162] If not, then the set of all process parameters to be mined in the target frequent itemset under the corresponding candidate defect type is determined as the set of target key process parameters corresponding to the selected main defect type.

[0163] S72: Traverse each of the major defect types and obtain the set of target key process parameters for the flexible board product under each of the major defect types using the same method.

[0164] Since the candidate defect types are derived from the Bayesian inference network, they correspond to the main defect types. All key process parameters under each main defect type are compared with all process parameters to be mined in the target frequent itemset of the corresponding candidate defect type. If they are consistent, it indicates that both the results obtained from Bayesian inference network mathematical reasoning and those obtained from association rule mining are accurate, and the set of all key process parameters under the main defect type can be directly determined as the corresponding target key process parameter set. If they are inconsistent, it indicates that the results obtained from Bayesian inference network mathematical reasoning are not accurate enough, and the set of all process parameters to be mined in the target frequent itemset of the candidate defect type can be determined as the corresponding target key process parameter set. This corrects all key process parameters of the corresponding main defect type obtained from mathematical reasoning, effectively improving the accuracy of quality traceability.

[0165] Example 2

[0166] like Figure 8 As shown, a flexible board product manufacturing quality traceability system includes:

[0167] The data acquisition module is used to acquire subsets of quality inspection data and process parameter data of flexible board products in multiple batches, and to construct a flexible board production dataset.

[0168] The data processing module is used to preprocess the flexible board production dataset;

[0169] A network construction module is used to obtain at least one major defect type of the flexible board product from the preprocessed flexible board production dataset and construct a Bayesian inference network of defect-process-process parameters.

[0170] The inference module is used to utilize the Bayesian inference network to obtain at least one key process and at least one key process parameter for each key process of the flexible board product under each major defect type, based on the preprocessed flexible board production dataset.

[0171] The flexible board product production quality traceability system in this embodiment uses data analysis and data reasoning to replace manual identification of quality problems, realizing traceability of flexible board product production quality. This reduces labor costs and improves analysis efficiency, preventing quality traceability failures due to human error. It achieves efficient, fast, and accurate production quality traceability, helping companies optimize flexible board product production based on the results of quality traceability and improve corporate profits.

[0172] Preferably, such as Figure 9 As shown, it also includes:

[0173] The association rule mining module is used to obtain the mining dataset of the flexible board product; wherein, the mining dataset includes at least one defect type to be mined from multiple batches of the flexible board product within a preset time period, at least one process to be mined under each defect type to be mined, and at least one process parameter to be mined corresponding to each process to be mined; it is also used to perform association rule mining on the mining dataset based on the association rule mining method.

[0174] The verification module is used to verify all the key process parameters under all the main defect types obtained based on the association rule mining results, so as to obtain the target key process parameter set of the flexible board product under each main defect type.

[0175] By mining and validating association rules, we can ensure that all key process parameters corresponding to each major defect type remain consistent across all batches. This further identifies these key process parameters as the critical factors leading to the corresponding quality problems, enabling precise location of quality issues in flexible board production and more accurate quality traceability.

[0176] The functions of each module in the flexible board product production quality traceability system described in this embodiment are the same as the steps in the flexible board product production quality traceability method in Embodiment 1. Therefore, for details not covered in this embodiment, please refer to Embodiment 1 and... Figures 1 to 7 The specific details will not be elaborated here.

[0177] Example 3

[0178] A flexographic board product manufacturing quality traceability device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the method steps in the flexographic board product manufacturing quality traceability method of Embodiment 1.

[0179] By using computer programs stored in memory and running on a processor, data analysis and reasoning are employed to replace manual methods of locating quality problems, thus enabling traceability of the production quality of flexible board products. This reduces labor costs, improves analytical efficiency, and eliminates traceability failures caused by human error. It achieves efficient, fast, and accurate production quality traceability, helping companies optimize the production of flexible board products based on the traceability results and increase their profits.

[0180] 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 computer device, connecting all parts of the computer device through various interfaces and lines.

[0181] Memory can be used to store computer programs and / or models. The processor performs various functions of the computer device by running or executing the computer programs and / or models stored in the memory, and by accessing data stored in the memory. Memory can primarily include a program storage area and a data storage area. The program storage area can store the operating system and at least one application program required for a function (e.g., sound playback, image playback, etc.); the data storage area can store data created based on the use of the mobile phone (e.g., audio data, video data, etc.). Furthermore, memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital cards (SD cards), flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0182] It should be understood that each block of a flowchart and / or block diagram, and combinations of blocks in a flowchart and / or block diagram, can be implemented by a computer program. These computer programs can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that instructions executable by the processor of the computer or other programmable data processing device generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0183] These computer programs may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0184] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0185] This embodiment also provides a computer-readable storage medium, which includes at least one instruction that, when executed, implements the method steps in the flexible board product production quality traceability method of Embodiment 1.

[0186] By executing a computer storage medium containing at least one instruction, and using data analysis and reasoning to replace manual identification of quality problems, traceability of flexible board product production quality is achieved. This reduces labor costs, improves analysis efficiency, and prevents quality traceability failures due to human error. It enables efficient, fast, and accurate production quality traceability, helping companies optimize flexible board product production based on the results of quality traceability and improve corporate profits.

[0187] Similarly, for details not covered in Embodiment 3, please refer to Embodiments 1 and 2. Figures 1 to 9 The specific details will not be elaborated here.

[0188] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method of flexible sheet product production quality traceability, characterized by, The method comprises the following steps: acquiring quality inspection data subsets and process parameter data subsets of flexible board products in multiple batches, and constructing a flexible board production dataset; preprocessing the flexible board production dataset; acquiring at least one main defect type of the flexible board products from the preprocessed flexible board production dataset, and constructing a Bayesian inference network of defects-process-technological parameters; acquiring at least one key process and at least one key technological parameter of each key process under each main defect type of the flexible board products according to the preprocessed flexible board production dataset by using the Bayesian inference network; acquiring a mining dataset of the flexible board products; wherein the mining dataset comprises at least one to-be-mined defect type, at least one to-be-mined process and at least one to-be-mined technological parameter corresponding to each to-be-mined process of multiple batches of the flexible board products in a preset time period; and performing association rule mining on the mining dataset based on an association rule mining method; verifying all the key technological parameters under all the main defect types according to the association rule mining result, and obtaining a target key technological parameter set of the flexible board products under each main defect type; the process parameter data subset under each batch comprises multiple processes and technological parameter data of each process of the flexible board products in the corresponding batch; and the quality inspection data subset under each batch comprises a batch quality inspection result, a total number of products, at least one product defect type and a number of defective products of each product defect type; the preprocessed flexible board production dataset specifically comprises a discretized abnormal batch data subset; and the step of acquiring at least one main defect type of the flexible board products from the preprocessed flexible board production dataset and constructing a Bayesian inference network of defects-process-technological parameters comprises the following steps: determining at least one main defect type from all the product defect types in the discretized abnormal batch data subset; taking each main defect type as a root node respectively, determining a first child node corresponding to each root node according to all the processes in the discretized abnormal batch data subset, and calculating a first conditional probability between each first child node and the corresponding root node; constructing a first Bayesian network of defects-process according to all the root nodes, all the first child nodes and all the first conditional probabilities; taking each first child node as a parent node respectively, determining a second child node corresponding to each parent node according to all the technological parameter data in the discretized abnormal batch data subset, and calculating a second conditional probability between each second child node and the corresponding parent node; constructing a second Bayesian network of process-technological parameters according to all the parent nodes, all the second child nodes and all the second conditional probabilities; obtaining the Bayesian inference network of defects-process-technological parameters according to the first Bayesian network and the second Bayesian network.

2. The flexible sheet product production quality traceability method according to claim 1, characterized by, The method for acquiring quality inspection data subsets and process parameter data subsets of flexible board products in multiple batches comprises the following steps: Obtaining a product number of the flexible board product; Obtaining, according to the product number, pass data of the flexible board product from a preset execution manufacturing system; Determining a plurality of processes of the flexible board product according to the pass data; Collecting, by using a data collection device arranged on the flexible board product, the process parameter data of all processes of the flexible board product under each batch; Obtaining, respectively according to the process parameter data of all processes under each batch, the process parameter data subset under each batch; Obtaining, by using a quality inspection device arranged on the flexible board product, the quality inspection data subset of the flexible board product under a plurality of batches.

3. The flexible sheet product production quality traceability method according to claim 1, characterized by, The preprocessing of the flexible board production data set comprises: According to the batch quality inspection results under all batches, the flexible board production data set is divided into a normal batch data subset and an abnormal batch data subset; Calibrating a plurality of process parameter intervals by using all the process parameter data in the normal batch data subset; Discretizing the abnormal batch data subset by using all the process parameter intervals.

4. The flexible sheet product production quality traceability method according to claim 3, characterized by, The process parameter intervals include a normal parameter interval, a parameter slightly small interval, a parameter too small interval, a parameter slightly large interval and a parameter too large interval.

5. The flexible sheet product production quality traceability method according to claim 1, characterized by, The obtaining, by using the Bayesian inference network, of at least one key process and at least one key process parameter under each main defect type of the flexible board product according to the preprocessed flexible board production data set comprises: Calculating, by using the Bayesian inference network, the posterior probability between each process parameter data in the discretized abnormal batch data subset and each main defect type; According to a preset probability threshold and all the posterior probabilities, at least one key process and at least one key process parameter under each main defect type of the flexible board product are determined.

6. The flexible sheet product production quality traceability method according to claim 1, characterized by, The association rule mining method for the mining data set comprises: Discretizing the mining data set; According to the Bayesian inference network, all candidate defect types are determined from all the to-be-mined defect types in the discretized mining data set; Selecting any one of the candidate defect types, based on the association rule mining method, all the to-be-mined process parameters in the discretized mining data set under the selected candidate defect type are determined as a candidate item set corresponding to the selected candidate defect type; In the candidate item set corresponding to the selected candidate defect type, k to-be-mined process parameters related to the selected candidate defect type under each to-be-mined process are determined as a candidate k-item set of the selected candidate defect type. A set of all the candidate k-item sets of the selected candidate defect type is denoted as a first set C corresponding to the selected candidate defect type k ; wherein k is a positive integer with an initial value of 1. calculating support of each of the candidate k-item sets in the first set C k k obtaining at least one frequent k-item set corresponding to the selected candidate defect type from the first set C k ;​ determining whether the second set L k is empty; If yes, the confidence of each of the second set L k is calculated, and according to a preset minimum confidence and all the confidences, a target frequent item set corresponding to the selected candidate defect type is obtained from the second set L k , and the association rule mining of the selected candidate defect type in the discretized mining data set is completed; each of the candidate defect types in the discretized mining data set is traversed, and the target frequent item set corresponding to each of the candidate defect types is obtained in the same way, and the association rule mining of each of the candidate defect types is completed. Otherwise, let k = k + 1, for the second set L k Perform the join and pruning operations sequentially to obtain the second set L. k Using the same method, a second set L of the selected candidate defect types is obtained. k+1 The process is iterated until the second set L of the selected candidate defect types is reached. k+1 If the set is empty, end the iteration; and after the iteration ends, follow the second set L. k When the set is empty, the same method is used to obtain the target frequent itemset corresponding to each candidate defect type, and to complete the association rule mining for each candidate defect type.

7. The flexible sheet product production quality traceability method according to claim 6, characterized by, calculating a first set C corresponding to the ith candidate defect type k The first formula for the support of the jth candidate k-item set in the middle is: ； calculating a second set L corresponding to the ith candidate defect type k The second formula of the confidence of the tth frequent k-item set in the middle is: ； wherein, is a first set C corresponding to the ith candidate defect type k is a support of the jth candidate k-item set in the ith candidate defect type, is a number of batches in which the ith candidate defect type appears in the jth candidate k-item set and there are k process parameters to be mined, and N is a total number of batches in the mining dataset, is a second set L corresponding to the ith candidate defect type k is a confidence of the tth frequent k-item set in the ith candidate defect type, is a number of batches in which the ith candidate defect type appears in the tth frequent k-item set and there are k process parameters to be mined, and m i is a number of batches in which the ith candidate defect type appears in the mining dataset.

8. The flexible sheet product production quality traceability method according to claim 6, characterized by, The verification of all the key process parameters under all the main defect types obtained according to the association rule mining result to obtain a target key process parameter set of the flexible board product under each main defect type comprises: If yes, the set of all the key process parameters under the selected main defect type is determined as the target key process parameter set corresponding to the selected main defect type. If no, the set of all the process parameters to be mined in the target frequent item set under the corresponding candidate defect type is determined as the target key process parameter set corresponding to the selected main defect type. The target key process parameter set of the flexible plate product under each main defect type is obtained by traversing each main defect type and using the same method. It comprises:

9. A flexible sheet product production quality traceability system characterized by, A data acquisition module is configured to acquire quality inspection data subsets and process parameter data subsets of a flexible plate product in multiple batches and construct a flexible plate production dataset; A data processing module is configured to preprocess the flexible plate production dataset; A network construction module is configured to acquire at least one main defect type of the flexible plate product from the preprocessed flexible plate production dataset and construct a defect-process-process parameter Bayesian inference network; An inference module is configured to acquire at least one key process and at least one key process parameter under each key process of the flexible plate product under each main defect type by using the Bayesian inference network and based on the preprocessed flexible plate production dataset; An association rule mining module is configured to acquire a mining dataset of the flexible plate product; wherein the mining dataset comprises at least one defect type to be mined, at least one process to be mined under each defect type to be mined, and at least one process parameter to be mined corresponding to each process to be mined of the flexible plate product in multiple batches within a preset time period; and is further configured to perform association rule mining on the mining dataset based on an association rule mining method; A verification module is configured to verify all the key process parameters under all the main defect types based on the association rule mining result to obtain a target key process parameter set of the flexible plate product under each main defect type; The process parameter data subset under each batch comprises multiple processes and process parameter data of each process of the flexible plate product under the corresponding batch; and the quality inspection data subset under each batch comprises a corresponding batch quality inspection result, a total number of products, at least one product defect type, and a number of defective products of each product defect type; The preprocessed flexible plate production dataset specifically comprises a discretized abnormal batch data subset; the network construction module acquires at least one main defect type of the flexible plate product from the preprocessed flexible plate production dataset and constructs a defect-process-process parameter Bayesian inference network, specifically comprising: At least one main defect type is determined from all the product defect types in the discretized abnormal batch data subset; ​ each of the main defect types is taken as a root node respectively, a first child node corresponding to each root node is determined according to all processes in the discretized abnormal batch data subset, and a first conditional probability between each first child node and the corresponding root node is calculated; a first Bayesian network of defects-processes is constructed according to all root nodes, all first child nodes and all first conditional probabilities; each of the first child nodes is taken as a parent node respectively, a second child node corresponding to each parent node is determined according to all process parameter data in the discretized abnormal batch data subset, and a second conditional probability between each second child node and the corresponding parent node is calculated; a second Bayesian network of processes-process parameters is constructed according to all parent nodes, all second child nodes and all second conditional probabilities; the Bayesian inference network of defects-processes-process parameters is obtained according to the first Bayesian network and the second Bayesian network.

10. A flexible sheet product production quality traceability apparatus characterized by comprising: The computer readable storage medium comprises at least one instruction which, when executed, implements the flexible sheet product production quality traceability method according to any one of claims 1 to 8.

11. A computer readable storage medium, characterized in that, The computer readable storage medium comprises at least one instruction which, when executed, implements the flexible sheet product production quality traceability method according to any one of claims 1 to 8.

Images