Coating abnormality analysis method and related device, storage medium

Abstract

The application discloses a coating abnormality analysis method and related equipment and a storage medium. The coating abnormality analysis method comprises the following steps: obtaining a coating abnormality feature input by a user, wherein the coating abnormality feature reflects a phenomenon of the current coating abnormality; obtaining event features of a plurality of coating abnormality basic events corresponding to the coating abnormality feature in a fault tree, and obtaining a target probability corresponding to the coating abnormality feature; calculating abnormal probabilities of the plurality of coating abnormality basic events based on the features of the plurality of coating abnormality basic events and the target probability; calculating decision game contribution degrees of the coating abnormality basic events based on the abnormal probabilities, and selecting at least one coating abnormality basic event as a key factor of the current coating abnormality based on the decision game contribution degrees of the coating abnormality basic events. The above scheme can alleviate the problem of missing key factors caused by relying on experience judgment or local data troubleshooting.

Description

Technical Field

[0001] This application relates to the field of fault analysis technology, and in particular to a coating anomaly analysis method and related equipment and storage medium. Background Technology

[0002] Currently, when coating abnormalities occur during industrial production, it is sometimes difficult to find the cause of the fault or the point at which the fault occurred. This is especially true for faults such as wear and aging of parts, excessively high operating temperature, and human error.

[0003] Traditional failure analysis methods often rely on experience-based judgment or partial data screening, making it difficult to systematically cover all possible influencing factors and leading to the omission of key factors. Summary of the Invention

[0004] This application provides at least one method, related equipment, and storage medium for coating anomaly analysis to alleviate the problem of missing key factors during coating anomaly analysis.

[0005] The first aspect of this application provides a coating anomaly analysis method, comprising: acquiring coating anomaly features input by a user, the coating anomaly features reflecting the phenomenon of the current coating anomaly; acquiring event features of several basic coating anomaly events corresponding to the coating anomaly features in a fault tree, and acquiring the target probability corresponding to the coating anomaly features; wherein, the fault tree includes several layers of events, the bottom layer of the fault tree is the basic event, and the target probability represents the probability of occurrence of the coating anomaly features; calculating the anomaly probability for the several basic coating anomaly events based on the features of the several basic coating anomaly events and the target probability; calculating the decision game contribution of each basic coating anomaly event based on the anomaly probability, and selecting at least one basic coating anomaly event as the key factor of the current coating anomaly based on the decision game contribution of each basic coating anomaly event; wherein, calculating the anomaly probability for the several basic coating anomaly events based on the event features of the several basic coating anomaly events and the target probability includes: extracting features from the event features and the target probability respectively, and fusion calculating the extracted features to obtain the anomaly probability for the several basic coating anomaly events.

[0006] The above solution, by obtaining the coating anomaly features input by the user, obtains the time features of the basic events of the coating anomaly corresponding to the coating anomaly features in the fault tree, thereby obtaining several basic events associated with the coating anomaly features, thus alleviating the problem of missing key factors due to relying on experience judgment or partial data investigation.

[0007] Among them, selecting at least one coating anomaly as the key factor of this coating anomaly based on the decision-making game contribution of each coating anomaly basic event includes: integrating the decision-making game contribution and target probability of each coating anomaly basic event to obtain the fault contribution of each coating anomaly basic event; and selecting at least one coating anomaly as the key factor of this coating anomaly based on the fault contribution of each coating anomaly basic event.

[0008] The above scheme obtains the fault contribution by integrating the target probability of coating anomaly features input by the user with the contribution of decision-making game, so that the coating anomaly analysis can dynamically adjust the weight according to the actual fault phenomenon, which alleviates the identification bias caused by the reliance on fixed rules in traditional methods, and thus helps to improve the accuracy of identifying key factors.

[0009] Among them, selecting at least one coating abnormality basic event as the key factor of this coating abnormality based on the fault contribution of each coating abnormality basic event includes: identifying several coating abnormality basic events with the highest fault contribution as key factors.

[0010] The above solution, by taking the most significant coating anomaly events as key factors, helps to further alleviate the problem of missing key factors.

[0011] The process involves extracting features from the event features and the target probability, and then fusing the extracted features to obtain the anomaly probability for several basic coating anomaly events. This includes using the event features and the target probability of the several basic coating anomaly events as input vectors for a conditional expert model. The conditional expert model then performs feature decomposition on the input vectors to extract a first feature and a second feature, and combines the first feature and the second feature to output the anomaly probability for the several basic coating anomaly events. The first feature is a general feature shared by the basic events corresponding to each intermediate event, and the second feature is a feature associated with the basic event corresponding to a specific intermediate event.

[0012] The above scheme calculates the anomaly probability of basic events of coating anomalies using a conditional expert model, and calculates the decision-making game contribution of basic events of coating anomalies based on the anomaly probability, which helps to improve the accuracy of key factor identification.

[0013] The fault tree includes several first intermediate events located in the first intermediate event layer; the conditional expert model includes a general expert network, a segmented expert network, and a feedforward network, and the segmented expert network includes several segmented sub-networks respectively associated with each first intermediate event; the conditional expert model performs feature decomposition on the input vector, extracts a first feature and a second feature, and combines the first feature and the second feature to output the anomalous probability for several basic coating anomalous events, including: inputting the input vector into the general expert network and outputting the first feature; and selecting the segmented sub-network associated with the first intermediate event corresponding to the coating anomalous feature from the segmented expert network for activation, and inputting the input vector into the activated segmented sub-network to output the second feature; inputting the first feature and the second feature into the feedforward network to obtain the anomalous probability for several basic coating anomalous events.

[0014] The above scheme calculates the anomaly probability through a general expert network, a segmented expert network, and a feedforward network. The segmented expert network includes segmented sub-networks associated with each first intermediate event. Thus, the coating anomaly features input by the user and the temporal features of the corresponding coating anomaly basic events are input into the segmented sub-networks associated with the first intermediate events corresponding to the coating anomaly features, which helps to further improve the accuracy of identifying key factors.

[0015] The fault tree includes, from top to bottom, a top event layer, several intermediate event layers, and a basic event layer. The highest layer of the several intermediate event layers is the first intermediate event layer. Each first intermediate event in the first intermediate event layer includes personnel, equipment, materials, methods, environment, and measurement.

[0016] The above solution uses the 5M1E systematic analysis method to identify the first intermediate event. The fault tree can cover multiple key impact dimensions, thereby alleviating the problem of missing key factors due to incomplete dimension coverage in traditional failure analysis.

[0017] The fault tree includes several second intermediate events located in the second intermediate event layer, each of which corresponds to at least one coating anomaly feature. The process of obtaining the event features of several coating anomaly basic events corresponding to the coating anomaly features in the fault tree includes: selecting the second intermediate event corresponding to the coating anomaly feature from the fault tree as the target intermediate event, and taking each basic event belonging to the target intermediate event in the fault tree as the coating anomaly basic event; and obtaining the event features of the coating anomaly basic events.

[0018] The above scheme can quickly locate relevant branches in the fault tree by matching coating anomaly features with the second intermediate event, which can alleviate the problems of large memory consumption and low processing efficiency caused by redundant calculations of scanning the entire fault tree.

[0019] In addition to selecting at least one basic event of coating abnormality as the key factor of this coating abnormality, the method also includes updating the probability corresponding to the coating abnormality feature based on the occurrence frequency of the basic event and user feedback.

[0020] The above solution, by combining the frequency of occurrence of basic events and user feedback to dynamically adjust the probability of coating abnormality features, helps to alleviate the problem of being unable to adapt to dynamic changes in fault modes, thereby improving the accuracy and timeliness of key factor identification.

[0021] Among them, the event characteristics of basic events are obtained by processing the parameters corresponding to the basic events. The parameter types corresponding to basic events include at least one of monitoring parameters and encoding parameters. The parameter types corresponding to each basic event are related to the attributes of the basic events. Monitoring parameters include numerical information, and encoding parameters are obtained by encoding or vectorizing non-numerical information.

[0022] The above scheme achieves analysis of non-numerical information by unifying the processing of monitoring parameters and encoding parameters, and by encoding or vectorizing non-numerical information to obtain encoding parameters, which helps to alleviate the problem of non-numerical information being unanalyzable.

[0023] The calculation of the decision-making game contribution of each coating abnormality basic event based on the abnormality probability includes: calculating the Shapley value of each coating abnormality basic event by combining the abnormality probability and the event characteristics of each coating abnormality basic event, and using the Shapley value as the decision-making game contribution of each coating abnormality basic event.

[0024] The above scheme, by dynamically weighting and fusing the contribution of decision-making games with the probability of abnormal coating features, can quantitatively assess the impact of failure probability on each branch of the fault tree, thereby improving the accuracy of identifying key factors.

[0025] The second aspect of this application provides an electronic device including a memory and a processor coupled to each other, the processor being used to execute program instructions stored in the memory to implement the coating anomaly analysis method in the first aspect described above.

[0026] A third aspect of this application provides a computer-readable storage medium having program instructions stored thereon, which, when executed by a processor, implement the coating anomaly analysis method described in the first aspect above.

[0027] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. Attached Figure Description

[0028] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0029] Figure 1 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0030] Figure 2 This is a schematic diagram of a fault tree in one embodiment of this application;

[0031] Figure 3 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0032] Figure 4 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0033] Figure 5 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0034] Figure 6 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0035] Figure 7 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0036] Figure 8 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0037] Figure 9 This is a flowchart illustrating an embodiment of the coating anomaly analysis method of this application;

[0038] Figure 10 This is a schematic diagram of the framework of an embodiment of the electronic device of this application;

[0039] Figure 11 This is a schematic diagram of a framework of an embodiment of the computer-readable storage medium of this application. Detailed Implementation

[0040] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0041] In the following description, specific details such as particular system architectures, interfaces, and technologies are presented for illustrative purposes rather than for limiting purposes, in order to provide a thorough understanding of this application.

[0042] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "many" in this document means two or more. Moreover, the term "at least one" in this document means any combination of at least two of any one or more of a plurality of objects. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0043] Please see Figure 1 This application provides a method for analyzing coating anomalies, including:

[0044] Step S110: Obtain the coating anomaly features input by the user.

[0045] Coating anomaly characteristics reflect the phenomena of this coating anomaly. For example, coating anomaly characteristics may include "frequent production interruptions" or "equipment damage caused by misoperation." When "frequent production interruptions" occur, the user can input the coating anomaly characteristic on the interactive interface.

[0046] Step S120: Obtain the event features of several basic events of coating abnormality corresponding to the coating abnormality features in the fault tree, and obtain the target probability corresponding to the coating abnormality features.

[0047] The fault tree consists of several layers of events. The bottom layer of the fault tree consists of basic events, and the target probability represents the probability of the occurrence of the coating anomalous features.

[0048] Please refer to Figure 2 , Figure 2 An exemplary structure of a fault tree in this application is illustrated. The fault tree, from top to bottom, may include a top event layer, several intermediate event layers, and a basic event layer. Furthermore... Figure 2 In the fault tree, the intermediate event layer can also include a first intermediate event layer. When the top event is "coating film width anomaly," the coating film width anomaly can be decomposed into six first intermediate events according to the 5M1E theory: personnel, equipment, materials, methods, environment, and measurement. The basic event layer is located at the bottom of the fault tree, such as... Figure 2 The issues mentioned include "unfamiliarity with equipment operation" and "insufficient understanding of the coating process." In a fault tree, the top event layer can be understood as an abnormal phenomenon, while the intermediate event layers and the basic event layer can be understood as the causes of that abnormal phenomenon. Combined with... Figure 2For the anomaly of "abnormal coating film width," the cause can be decomposed into six intermediate events at the intermediate event layer: personnel, equipment, materials, methods, environment, and measurement. Each intermediate event layer can be further decomposed into corresponding basic events; in other words, there is a correspondence between the basic events and the intermediate events. For example, the intermediate event "personnel" can be further decomposed into the intermediate events "insufficient operator skills" and "operational errors," both of which describe "personnel." "Insufficient operator skills" can then be further decomposed into the basic events "unfamiliarity with equipment operation" and "insufficient understanding of the coating process," and so on.

[0049] In some embodiments, the target probability represents the probability of occurrence of coating anomaly features, and the probability of occurrence of coating anomaly features can be assigned by combining the historical frequency of occurrence of the phenomenon with expert experience. The coating anomaly analysis system can automatically associate the probability values ​​corresponding to coating anomaly features based on historical cases and expert experience, such as the probability of "discontinuous production coating" being 0.35. For example, if the historical frequency of "frequent production interruptions" is high, a higher probability can be assigned to it. The probability corresponding to each coating anomaly feature can be assigned according to the actual scenario, and this application does not impose any restrictions.

[0050] In some embodiments, coating anomaly features can be associated with an intermediate event layer. For example... Figure 2 As shown, the intermediate event layer "Personnel" can be associated with multiple coating anomaly features, such as "frequent production interruptions," "continuous deviation of coating film width from standard," and "coating film roll not meeting established standards," etc. Intermediate events can be further subdivided; for example, the intermediate event "Personnel" can be subdivided into intermediate events such as "insufficient operator skills" and "operational errors." Each intermediate event can be associated with a coating anomaly feature; for example, "insufficient operator skills" can be associated with the coating anomaly feature "frequent production interruptions." The lowest-level intermediate events can be further subdivided into basic events; for example, "insufficient operator skills" can be subdivided into "unfamiliarity with equipment operation" and "insufficient understanding of coating process," etc. "Unfamiliarity with equipment operation" and "insufficient understanding of coating process" can also be associated with the coating anomaly feature "frequent production interruptions." Therefore, when a user inputs a coating anomaly feature, the system can retrieve the corresponding basic coating anomaly event based on the fault tree and obtain the target probability corresponding to the coating anomaly feature. In some embodiments, coating anomaly features can also be associated with various basic events, meaning that each basic event has a corresponding coating anomaly feature.

[0051] For example, the top event in the top event layer of the fault tree can be "coating film width anomaly". In other words, the coating anomaly analysis method provided in this application can be used to analyze the key factors causing the coating film width anomaly when it occurs in the coating process. The coating anomaly analysis method of this application can also be applied to analyze key factors when other anomalies occur in other processes, and this application does not impose any limitations.

[0052] In some embodiments, the event characteristics of a basic event can be obtained by feature encoding or normalization of the parameters of the basic event. For example... Figure 2 As shown, basic events can include "abnormal motor speed," "ambient temperature," "ambient humidity," and so on. Based on the coating anomaly characteristics input by the user, after determining the corresponding coating anomaly basic event in the fault tree, parameters corresponding to the coating anomaly basic event, such as motor speed, ambient temperature, and ambient humidity, can be obtained from the cloud database platform and subjected to feature encoding or normalization processing to obtain event features.

[0053] Step S130: Based on the event characteristics and target probabilities of several basic coating abnormalities, calculate the abnormal probability for several basic coating abnormalities.

[0054] "The probability of an anomaly for a number of basic coating anomalies" indicates the probability of an anomaly occurring under the parameters corresponding to these basic coating anomalies. In some embodiments, an anomaly probability threshold can be set, such as 80% or 75%. When the anomaly probability exceeds the anomaly probability threshold, it indicates that an anomaly exists in the basic coating anomaly event; otherwise, the basic coating anomaly event does not have an anomaly.

[0055] In some embodiments, feature extraction can be performed on event features and target probabilities respectively, and the extracted features can be fused and calculated to obtain the abnormal probability for several basic events of coating abnormality.

[0056] Step S140: Calculate the decision-making game contribution of each coating anomaly basic event based on the anomaly probability, and select at least one coating anomaly basic event as the key factor of this coating anomaly based on the decision-making game contribution of each coating anomaly basic event.

[0057] The above solution, by obtaining the coating anomaly features input by the user, obtains the time features of the basic events of the coating anomaly corresponding to the coating anomaly features in the fault tree, thereby obtaining several basic events associated with the coating anomaly features, thus alleviating the problem of missing key factors due to relying on experience judgment or partial data investigation.

[0058] In some implementation methods, please refer to Figure 3 Step S140 may include:

[0059] Step S241: Calculate the decision-making game contribution of each coating anomaly basic event based on the anomaly probability, and integrate the decision-making game contribution of each coating anomaly basic event with the target probability to obtain the fault contribution of each coating anomaly basic event.

[0060] Step S242: Based on the fault contribution of each coating abnormality basic event, select at least one coating abnormality basic event as the key factor of this coating abnormality.

[0061] The fault contribution is obtained by fusing the decision-making game contribution and the target probability (i.e., the probability of coating anomaly features). For example, the product of the decision-making game contribution and the probability of coating anomaly features can be used as the fault contribution. For instance, based on the fault contribution from high to low, K events (K is a positive integer) with the highest fault contribution exceeding a preset threshold can be selected as key factors. For example, in coating anomaly analysis, the top five basic events with the highest fault contribution are given priority. The value of K can be selected according to the actual scenario, and this application does not impose any restrictions.

[0062] The above scheme, by weighting and integrating the contribution of decision-making games with the probability of coating anomaly features, helps to improve the accuracy of coating anomaly analysis and alleviate the problem of misjudgment of key factors caused by relying on experience-based judgment.

[0063] In some implementation methods, please refer to Figure 4 In step S242, based on the fault contribution of each coating anomaly basic event, at least one coating anomaly basic event is selected as the key factor of this coating anomaly, including:

[0064] Step S342: Identify the several coating anomaly basic events with the highest failure contribution as key factors.

[0065] For example, please combine Figure 2 Assuming the user-input coating anomaly characteristics include "uniform coating film width," "continuous expansion of the anomaly," and "inability to detect minute film width changes," the basic events of coating anomalies can include basic events associated with the three intermediate events of "equipment," "method," and "environment," such as "coating roller wear," "abnormal motor speed," and "decreased sensor accuracy." The failure contribution of each basic event of coating anomalies, from largest to smallest, is as follows: "coating roller wear," "abnormal motor speed," "decreased sensor accuracy," "failure to establish a fluctuation warning," and "large temperature and humidity fluctuations." The K basic events with the highest failure contribution can be selected as key factors.

[0066] For example, in practical applications, the sorting and filtering process can be completed automatically, and the K key factors with the highest contribution to the failure, such as coating roller wear and abnormal motor speed, can be output and presented to on-site production managers through an interactive interface for quick confirmation and troubleshooting.

[0067] The above solution, by taking the most significant coating anomaly events as key factors, helps to further mitigate the problem of overlooking key factors, reduces the time required for coating anomaly analysis, accelerates the problem localization and resolution process, and improves the stability of the production line and product quality.

[0068] In some implementation methods, please refer to Figure 5 Step S130 involves feature extraction for both event features and target probabilities. The extracted features are then fused and calculated to obtain the anomaly probabilities for several basic coating anomaly events, including:

[0069] Step S431: Use the event features and target probabilities of several basic coating abnormalities as the input vector of the conditional expert model, so that the conditional expert model can perform feature decomposition on the input vector, extract the first feature and the second feature, and combine the first feature and the second feature to output the abnormal probability for several basic coating abnormalities.

[0070] The first feature is the general feature of the basic event corresponding to each intermediate event, and the second feature is the feature associated with the basic event corresponding to a specific intermediate event.

[0071] By using the event features and target probabilities of several basic coating abnormalities, an input vector can be generated. The event features of the basic coating abnormalities (such as process parameter settings, coating speed, etc.) can be converted into numerical vectors through normalization and combined with the coating abnormality feature probability (e.g., 0.35) to form the input vector. For example, the input vector can be [process parameter settings, 0.35].

[0072] The above scheme calculates the anomaly probability of basic events of coating anomalies using a conditional expert model, and calculates the decision-making game contribution of basic events of coating anomalies based on the anomaly probability, which helps to improve the accuracy of key factor identification.

[0073] In some implementations, the fault tree includes a plurality of first intermediate events located at a first intermediate event layer. For example, please refer to... Figure 2 The first intermediate event in the first intermediate event layer can include personnel, equipment, materials, methods, environment, and measurements.

[0074] Conditional expert models can include general expert networks, fragmented expert networks, and feedforward networks, with each fragmented expert network comprising several fragmented sub-networks associated with a specific first intermediate event. Please refer to [reference needed]. Figure 6 In step S431, the conditional expert model performs feature decomposition on the input vector, extracts the first feature and the second feature, and combines the first feature and the second feature to output the anomaly probability for several basic events of coating anomalies, including:

[0075] Step S5321: Input the input vector into the general expert network and output the first feature; and, from the segmented expert network, select the segmented sub-network associated with the first intermediate event corresponding to the coating anomaly feature, activate it, and input the input vector into the activated segmented sub-network to output the second feature.

[0076] Step S5322: Input the first feature and the second feature into the feedforward network to obtain the anomaly probability for several basic events of coating anomalies.

[0077] The sliced ​​expert network comprises multiple sliced ​​sub-networks, each associated with a specific first intermediate event, such as the sliced ​​sub-network associated with the "method" event. In practical applications, when a user selects a coating anomaly feature such as "discontinuous coating in production," the network can automatically identify the association of this feature with the first intermediate event of "method," thereby activating the corresponding sliced ​​sub-network—in other words, loading the model parameters of the corresponding sliced ​​sub-network. The input vector is simultaneously fed into the general expert network to generate the first feature and into the activated sliced ​​sub-network to generate the second feature. The first and second features can be concatenated and then fed into the feedforward network to calculate the anomaly probability of the coating anomaly basic event. For example, for the basic event of "improper process parameter settings," the input vector can include process parameter values ​​(such as coating temperature) and coating anomaly feature probabilities. The general expert network extracts the first feature across scenarios, the sliced ​​sub-networks extract the method branch features, and the feedforward network fuses the features to output the anomaly probability.

[0078] For example, when a user selects "compliant but poor cell consistency" as the coating anomaly feature, the slab sub-network related to "measurement" can be activated; when a user selects "compliant but poor cell consistency" or "frequent production interruptions" as the coating anomaly features, the slab sub-network related to "measurement" and "personnel" can be activated, and the outputs of the slab sub-networks related to "measurement" and "personnel" can be spliced ​​together and the anomaly probability can be calculated through a feedforward network.

[0079] For example, the feedforward network can be a multi-layer feedforward neural network, such as a multi-layer perceptron (MLP). This application does not limit the structure of the feedforward network.

[0080] In some embodiments, the anomalous probability of the feedforward network output in step S5322 can be: ;in, For activation function, This is the weight matrix. For general expert networks, For a sharding expert network, For weight compensation, This represents the probability of an anomaly.

[0081] The above solution dynamically selects the sub-networks that match the coating anomaly features, loading only the model parameters relevant to the current analysis. For example, when the user selects a coating anomaly feature such as "discontinuous coating in production," the solution can automatically identify the feature's association with the first intermediate event in the "method" section, thereby loading the corresponding sub-network's model parameters. This helps reduce memory usage and improve model loading speed. Furthermore, focusing on feature extraction from specific intermediate events improves the accuracy of identifying key factors.

[0082] In some implementations, the fault tree includes, from top to bottom, a top event layer, several intermediate event layers, and a basic event layer. The highest layer of the several intermediate event layers is the first intermediate event layer, and each first intermediate event in the first intermediate event layer includes personnel, equipment, materials, methods, environment, and measurement.

[0083] Each of the first intermediate events in the first intermediate event layer was determined using the 5M1E systematic analysis method. The 5M1E systematic analysis method refers to a framework that systematically decomposes the root causes of coating anomalies from six dimensions: personnel, equipment, materials, methods, environment, and measurement. Personnel encompasses operator skills and training; equipment refers to coating equipment and its operating status; materials involve the characteristics of the coating substrate and auxiliary materials; methods include process parameter settings and operating procedures; environment refers to production conditions such as temperature and humidity; and measurement involves the accuracy of testing equipment. Combined with... Figure 2 The fault tree structure can be categorized into two types: personnel as the primary intermediate event, further subdivided into secondary intermediate events such as insufficient operator skills and operational errors; similarly, equipment can be categorized into primary intermediate events such as coating roller wear and abnormal motor speed. The fault tree structure utilizes the 5M1E framework to achieve hierarchical decomposition from the top event (e.g., abnormal coating film width) to the basic events, ensuring that each intermediate event corresponds to a clearly defined scope of influence from the coating anomaly. This facilitates the system's ability to locate relevant basic events based on user-selected coating anomaly characteristics (e.g., "coating discontinuity"), and subsequently analyze key factors.

[0084] The above scheme calculates the anomaly probability through a general expert network, a segmented expert network, and a feedforward network. The segmented expert network includes segmented sub-networks associated with each first intermediate event. Thus, the coating anomaly features input by the user and the temporal features of the corresponding coating anomaly basic events are input into the segmented sub-networks associated with the first intermediate events corresponding to the coating anomaly features, which helps to further improve the accuracy of identifying key factors.

[0085] In some implementations, the fault tree includes a plurality of second intermediate events located in the second intermediate event layer, each second intermediate event corresponding to at least one coating anomaly feature. Please refer to... Figure 7 Step S120 may include:

[0086] Step S1201: Select the second intermediate event corresponding to the coating anomaly feature from the fault tree as the target intermediate event, and take each basic event belonging to the target intermediate event in the fault tree as the coating anomaly basic event, obtain the event features of the coating anomaly basic event, and obtain the target probability corresponding to the coating anomaly feature.

[0087] Select the second intermediate event corresponding to the coating anomaly feature from the fault tree. In other words, each second intermediate event can correspond to one coating anomaly feature. For example, please refer to... Figure 2 Assuming the second intermediate event "insufficient operator skills" corresponds to the coating anomaly feature "frequent production interruptions," meaning "insufficient operator skills" may cause "frequent production interruptions," the corresponding second intermediate event can be determined based on the correspondence between the second intermediate event and the coating anomaly feature.

[0088] For example, in the fault tree, a mapping relationship between the second intermediate event and the corresponding coating anomaly feature can be established, so that the corresponding second intermediate event can be obtained after the user inputs the coating anomaly feature.

[0089] It is worth noting that the intermediate event layers of a fault tree can include all other levels except the top event layer and the basic event layer. Therefore, intermediate event layers can include level 1, level 2, level 3, etc., specifically determined according to the structure of the fault tree, and this application does not impose any restrictions. Please refer to... Figure 2 , Figure 2The intermediate event layer of the fault tree includes two layers: a "first intermediate event layer" comprising "personnel," "equipment," "materials," "methods," "environment," and "measurement." Each first intermediate event layer can be further subdivided into a "second intermediate event layer." For example, "personnel" can be subdivided into "insufficient operator skills" and "operational errors." In other words, the "first intermediate event layer" and the "second intermediate event layer" are different layers in this case. In some embodiments, the intermediate event layer of the fault tree may also include one layer. For example, "personnel" can be directly subdivided into basic events such as "unfamiliarity with equipment operation" and "insufficient understanding of coating processes." In this case, the "first intermediate event layer" and the "second intermediate event layer" are the same layer.

[0090] Combination Figure 2 The second intermediate event layer of the fault tree can include intermediate events such as "insufficient operator skills," "operational errors," and "coating equipment malfunction." Each intermediate event can correspond to at least one coating anomaly feature; for example, "insufficient operator skills" can correspond to "frequent production interruptions." When acquiring basic coating anomaly events, the intermediate events corresponding to the coating anomaly features in the fault tree can be identified first as target intermediate events. For example, when the user selects "frequent production interruptions," "insufficient operator skills" can be identified as the target intermediate event. Then, all basic events under this target intermediate event can be extracted as basic coating anomaly events, such as unfamiliarity with equipment operation or insufficient understanding of the coating process.

[0091] The above scheme can quickly locate relevant branches in the fault tree by matching coating anomaly features with the second intermediate event, which can alleviate the problems of large memory consumption and low processing efficiency caused by redundant calculations of scanning the entire fault tree.

[0092] In some implementation methods, please refer to Figure 8 After selecting at least one basic event of coating anomaly as the key factor of this coating anomaly, the method may further include:

[0093] Step S750: Update the probability corresponding to the coating anomaly feature based on the frequency of occurrence of basic events and user feedback.

[0094] In coating anomaly analysis, once a basic coating anomaly event is identified as a key factor in the current anomaly, the frequency of this event in historical coating anomaly data can be automatically collected. This represents the cumulative number of times the event has been recorded as a cause of the coating anomaly. Feedback from on-site operators regarding the analysis results can also be collected, such as confirmations or corrections. Based on this data, a sliding time window analysis strategy can be employed, for example, using the most recent 30 days as a time window to dynamically adjust the probability values ​​corresponding to coating anomaly characteristics. For instance, if the "coating roller wear" event is identified as a key factor in three consecutive coating anomalies, its corresponding coating anomaly characteristic probability can be automatically updated from the initial value of 0.35 to 0.42; or, based on real-time corrections from user feedback, the probability value can be manually adjusted to match the latest production experience.

[0095] The above solution, through a dynamic update mechanism, adjusts the probability of coating anomaly features in real time by combining the frequency of basic events and user feedback. This alleviates the shortcomings of relying on static probabilities, which cannot adapt to dynamic changes in fault modes, and improves the accuracy and timeliness of key factor identification. Furthermore, it reduces the frequency of manual intervention and lowers maintenance costs.

[0096] In some implementations, the event characteristics of a basic event are obtained by processing the parameters corresponding to the basic event. The parameter types corresponding to the basic event include at least one of monitoring parameters and encoding parameters. The parameter types corresponding to each basic event are related to the attributes of the basic event. The monitoring parameters include numerical information, and the encoding parameters are obtained by encoding or vectorizing non-numerical information.

[0097] The event characteristics of basic events are obtained by processing the parameters corresponding to the basic events, such as through normalization. Monitoring parameters can include numerical information, such as coating speed, equipment rotation speed, or temperature—real-time data that can be directly measured. Encoded parameters are obtained by encoding or vectorizing non-numerical information, which refers to information that cannot be represented numerically, such as personnel information. Encoded parameters can be obtained by desensitizing equipment fault codes, operator skill levels, or environmental conditions before feature encoding or vectorization. For example, unique thermal encoding can be performed for different personnel or equipment.

[0098] The above scheme achieves analysis of non-numerical information by unifying the processing of monitoring parameters and encoding parameters, and by encoding or vectorizing non-numerical information to obtain encoding parameters, which helps to alleviate the problem of non-numerical information being unanalyzable.

[0099] In some implementation methods, please refer to Figure 9 Step S140 may include:

[0100] Step S1402: Calculate the Shapley value of each coating anomaly basic event by combining the anomaly probability and the event characteristics of each coating anomaly basic event, and use the Shapley value as the decision game contribution of each coating anomaly basic event.

[0101] Based on the Shapley value principle in game theory, the decision-making game contribution of each basic event of coating anomalies can be calculated by combining anomaly probabilities and event characteristics. Specifically, the decision-making game contribution of basic events of coating anomalies... It can be calculated using the following formula: Where F is the feature set, S is the feature subset, and i is the event feature of the basic event of the coating anomaly. This represents the probability of an anomaly.

[0102] In some embodiments, the decision-making game contribution of each of the coating anomaly basic events is calculated based on the anomaly probability, and the decision-making game contribution of each of the coating anomaly basic events and the target probability are fused to obtain the fault contribution of each of the coating anomaly basic events. This can be obtained by multiplying the decision-making game contribution and the target probability.

[0103] By using the event characteristics and anomaly probabilities of basic events in coating anomalies as input, the SHAP framework is used to analyze the marginal contribution of each basic event and generate quantitative indicators. The decision game contribution reflects the causal weight of basic events in coating anomalies, providing an objective basis for subsequent failure contribution calculations and reducing biases caused by reliance on human experience. Therefore, the interpretability mechanism of SHAP can improve the accuracy of key factor identification. Simultaneously, the dynamic weighted fusion of decision game contribution and coating anomaly feature probabilities shifts failure analysis from passive detection to proactive analysis, improving the efficiency and stability of coating anomaly analysis.

[0104] In some embodiments, after identifying the key factors contributing to the coating anomaly, this information can be proactively communicated to the on-site team. Notification methods may include email, SMS, internal enterprise communication software, or direct display on the production management system interface. On-site production managers and technicians will receive the notification and immediately conduct initial verification to improve the accuracy and timeliness of the information. Subsequently, the on-site team can conduct a step-by-step review. For example, detailed parameter settings in the current production process, such as temperature, pressure, and humidity, can be recorded and compared with the key factors. Through on-site observation and data comparison, specific failure points and potential risks can be identified. After confirming the problem, specific improvement measures can be developed, such as adjusting process parameters, optimizing equipment operation, and strengthening personnel training, and these measures can be implemented quickly. Throughout the process, the coating anomaly analysis system can continuously monitor the improvement effects, ensuring the stability of the production process and product quality.

[0105] In some embodiments, a probability correction for coating anomalies can be triggered each time new data arrives, thereby capturing the evolution trend of nonlinear relationships between variables. Update requests can also be submitted automatically and confirmed by engineers.

[0106] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0107] Please see Figure 10 , Figure 10 This is a schematic diagram of a framework of an embodiment of the electronic device of this application. The electronic device 80 includes a memory 81 and a processor 82 coupled to each other. The processor 82 is used to execute program instructions stored in the memory 81 to implement the steps in any of the above-described embodiments of the coating anomaly analysis method. In a specific implementation scenario, the electronic device 80 may include, but is not limited to, a microcomputer or a server. In addition, the electronic device 80 may also include mobile devices such as laptops and tablets, which are not limited here.

[0108] Specifically, processor 82 controls itself and memory 81 to implement the steps in any of the above-described coating anomaly analysis method embodiments. Processor 82 can also be referred to as a CPU (Central Processing Unit). Processor 82 may be an integrated circuit chip with signal processing capabilities. Processor 82 can also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor. Furthermore, processor 82 can be implemented using integrated circuit chips.

[0109] Please see Figure 11 , Figure 11 This is a schematic diagram of a framework of an embodiment of the computer-readable storage medium of this application. The computer-readable storage medium 90 stores program instructions 901 that can be executed by a processor. The program instructions 901 are used to implement the steps in any of the above embodiments of the coating anomaly analysis method.

[0110] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0111] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0113] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for analyzing coating anomalies, characterized in that, include: Obtain coating anomaly features input by the user, which reflect the phenomenon of coating anomaly in this instance; Obtain event features of several basic events of coating abnormality corresponding to the coating abnormality feature in the fault tree, and obtain the target probability corresponding to the coating abnormality feature; wherein, the fault tree includes several layers of events, the bottom layer of the fault tree is the basic event, and the target probability represents the probability of occurrence of the coating abnormality feature. Based on the event characteristics of the several basic coating abnormalities and the target probability, the abnormal probability for the several basic coating abnormalities is calculated. The abnormal probability represents the probability of an abnormality occurring under the parameters corresponding to the basic coating abnormality. The decision game contribution of each coating anomaly basic event is calculated based on the anomaly probability, and at least one coating anomaly basic event is selected as the key factor of this coating anomaly based on the decision game contribution of each coating anomaly basic event. The decision game contribution reflects the causal weight of the basic event in the coating anomaly. The step of calculating the anomalous probability for the several basic coating abnormalities based on the event features and the target probability includes: extracting features from the event features and the target probability respectively, and fusion calculation of the extracted features to obtain the anomalous probability for the several basic coating abnormalities.

2. The method according to claim 1, characterized in that, The selection of at least one of the basic events of coating anomaly as the key factor of this coating anomaly, based on the decision-making game contribution of each of the basic events of coating anomaly, includes: By integrating the decision-making game contribution of each of the coating anomaly basic events and the target probability, the fault contribution of each of the coating anomaly basic events is obtained. Based on the fault contribution of each of the basic events of the coating abnormality, at least one of the basic events of the coating abnormality is selected as the key factor of this coating abnormality.

3. The method according to claim 2, characterized in that, The selection of at least one of the coating anomaly basic events as the key factor of this coating anomaly based on the fault contribution of each of the coating anomaly basic events includes: The key factors are identified as the several coating anomaly basic events with the highest failure contribution among the coating anomaly basic events.

4. The method according to claim 1, characterized in that, The step of extracting features from the event features and the target probability, and then fusing and calculating the extracted features to obtain the anomaly probability for the several basic coating anomaly events, includes: The event features of the plurality of coating anomaly basic events and the target probability are used as the input vector of the conditional expert model, so that the conditional expert model performs feature decomposition on the input vector, extracts a first feature and a second feature, and outputs the anomaly probability for the plurality of coating anomaly basic events by combining the first feature and the second feature; the first feature is a general feature of the basic event corresponding to each intermediate event, the second feature is a feature associated with the basic event corresponding to a specific intermediate event, and the specific intermediate event is the highest-level intermediate event corresponding to the coating anomaly feature.

5. The method according to claim 4, characterized in that, The fault tree includes a number of first intermediate events located in the first intermediate event layer; the conditional expert model includes a general expert network, a fragmented expert network, and a feedforward network, and the fragmented expert network includes a number of fragmented sub-networks that are respectively associated with each of the first intermediate events. The conditional expert model performs feature decomposition on the input vector, extracts a first feature and a second feature, and combines the first feature and the second feature to output the anomaly probability for the plurality of basic coating anomaly events, including: The input vector is fed into the general expert network, and the first feature is output; and, From the segmented expert network, select the segmented sub-network associated with the first intermediate event corresponding to the coating anomaly feature, activate it, and input the input vector into the activated segmented sub-network to output the second feature; The first feature and the second feature are input into the feedforward network to obtain the anomaly probability for the plurality of coating anomaly basic events.

6. The method according to any one of claims 1 to 5, characterized in that, The fault tree includes, from top to bottom, a top event layer, several intermediate event layers, and a basic event layer. The highest layer of the several intermediate event layers is the first intermediate event layer. Each first intermediate event in the first intermediate event layer includes personnel, equipment, materials, methods, environment, and measurement.

7. The method according to any one of claims 1 to 5, characterized in that, The fault tree includes a number of second intermediate events located in the second intermediate event layer, and each second intermediate event corresponds to at least one coating anomaly feature. The acquisition of event features of several basic events of coating anomalies corresponding to the coating anomaly features in the fault tree includes: Select the second intermediate event corresponding to the coating anomaly feature from the fault tree as the target intermediate event, and take each basic event belonging to the target intermediate event in the fault tree as the coating anomaly basic event. Obtain the event characteristics of the basic events of the coating abnormality.

8. The method according to claim 1, characterized in that, After selecting at least one of the basic events of the coating anomaly as the key factor of this coating anomaly, the method further includes: Based on the frequency of occurrence of the basic events and user feedback, the probability corresponding to the coating anomaly feature is updated.

9. The method according to claim 1, characterized in that, The event characteristics of the basic event are obtained by processing the parameters corresponding to the basic event. The parameter types corresponding to the basic event include at least one of monitoring parameters and encoding parameters. The parameter types corresponding to each basic event are related to the attributes of the basic event. The monitoring parameters include numerical information, and the encoding parameters are obtained by encoding or vectorizing non-numerical information.

10. The method according to claim 1, characterized in that, The decision-making game contribution of each of the basic events of the coating anomaly calculated based on the anomaly probability includes: By combining the anomaly probability and the event characteristics of each coating anomaly basic event, the Shapley value of each coating anomaly basic event is calculated, and the Shapley value is used as the decision-making game contribution of each coating anomaly basic event.

11. An electronic device, characterized in that, The method includes a memory and a processor coupled to each other, the processor being configured to execute program instructions stored in the memory to implement the coating anomaly analysis method according to any one of claims 1 to 10.

12. A computer-readable storage medium having program instructions stored thereon, characterized in that, When the program instructions are executed by the processor, they implement the coating anomaly analysis method according to any one of claims 1 to 10.

Images