A production line emulsion pump head assembly monitoring system

Abstract

The present application relates to the field of intelligent industrial monitoring, and more particularly to a production line emulsion pump head assembly monitoring system, comprising: a data analysis unit for determining whether to perform data enhancement according to the data missing degree of multi-source monitoring data text; an enhancement execution unit for replacing the original text with simulated data text having the maximum matching degree or the maximum working condition execution value according to the supplementary directional value; an entity analysis unit for determining whether to perform entity clustering according to the entity distribution density of the production line entity; an entity clustering unit for obtaining a to-be-selected entity combination by combining the directional value, constructing a combination set based on a repeated elimination condition, and screening a fault retrieval combination through time sequence effective coverage; and a fault monitoring unit for determining a fault handling strategy according to the matching between the fault retrieval combination and a pre-stored knowledge base. The present application realizes intelligent fault monitoring of the emulsion pump head assembly production line.

Description

Technical Field

[0001] This invention relates to the field of intelligent industrial monitoring, and in particular to a monitoring system for the assembly of emulsion pump heads on a production line. Background Technology

[0002] As a core component of cosmetic and daily chemical product packaging, the assembly quality of emulsion pump heads directly affects the user experience and sealing performance. With the increasing demands for product quality in the consumer market, emulsion pump head assembly lines are rapidly developing towards higher speeds, automation, and precision. Currently, various fully automated emulsion pump assembly equipment can automate the feeding, assembly, and testing of multiple components such as pump housings, pump balls, springs, pistons, upper pump rods, lower pump rods, and bottle caps. However, existing monitoring technologies in emulsion pump head assembly lines still suffer from limited data processing capabilities. The level of intelligence in fault monitoring is insufficient; existing monitoring systems mostly remain at the level of data acquisition and simple alarms, lacking the ability to intelligently match real-time monitoring data with structured knowledge bases, making it difficult to meet the actual needs of high-speed automated production lines for rapid fault response.

[0003] Chinese Patent Publication No. CN121525822A discloses a multi-source information fusion fault diagnosis method combining equipment atlas and AI models. In the offline stage, a historical dataset of industrial assets is prepared and an equipment topology knowledge graph is constructed. A dual machine learning diagnostic engine, including a "diagnostician" model and a "health assessment" model, is trained. An unstructured knowledge base is built based on technical documents. In the online diagnosis stage, the output of the "diagnostician" model is monitored, and a diagnostic process is triggered when the status is abnormal. The "diagnostic strategist" is activated to query the equipment atlas to obtain a set of related components. The "health assessment" model is called to obtain a quantitative health index. Health comparison is performed through knowledge retrieval and logical reasoning. When evidence conflicts, iterative reasoning is used until the root cause is confirmed. A structured evidence archive is generated. Finally, the "attribution analyzer" is activated to output a professional diagnostic report. However, the above technical solution has the following problems: it does not consider the dynamic characteristics of the scenario, such as the quality assessment of input data, handling of missing values, and the validity of actual entities, and lacks the ability to intelligently match real-time monitoring data with a structured knowledge base. Summary of the Invention

[0004] To address this issue, the present invention provides a production line emulsion pump head assembly monitoring system to overcome the problem that the existing technology lacks the ability to intelligently match real-time monitoring data with a structured knowledge base, making it difficult to meet the actual needs of high-speed automated production lines for rapid fault response.

[0005] To achieve the above objectives, the present invention provides a production line emulsion pump head assembly monitoring system, comprising: The data analysis unit is used to determine whether to perform data augmentation on the multi-source monitoring data text based on the data missingness corresponding to the multi-source monitoring data text. The enhanced execution unit is used to determine whether to adjust the selection criterion of the simulated data text from the selection of supplementary matching degree to the selection of the working condition execution value based on the supplementary orientation value, and to determine the simulated data text to replace the multi-source monitoring data text based on the selection criterion. The entity analysis unit is used to determine whether to perform entity clustering based on the entity distribution density of production line entities in multi-source monitoring data text. The entity clustering unit is used to perform entity clustering, obtain the entity combination to be selected by combining tendency value, construct several combination sets based on the duplication elimination condition, and select the entity combination to be selected in the combination set by the effective coverage of time series, which are respectively denoted as fault retrieval combination. The fault monitoring unit is used to match fault retrieval combinations with a pre-stored knowledge base to determine fault handling strategies.

[0006] Furthermore, the data analysis unit responds to the condition that the data missing degree corresponding to the multi-source monitoring data text is greater than the preset data missing degree, and determines to perform data augmentation on the multi-source monitoring data text.

[0007] Furthermore, the execution unit responds to the condition that the supplementary orientation value is greater than the preset supplementary orientation value, and determines to select the simulated data text with the highest supplementary matching degree to replace the multi-source monitoring data text.

[0008] Furthermore, the condition that the supplementary orientation value of the execution unit response is less than or equal to the preset supplementary orientation value is enhanced, and the simulation data text with the largest operating condition execution value is selected to replace the multi-source monitoring data text.

[0009] Furthermore, the entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is greater than the preset entity distribution density, and determines all production line entities as a fault retrieval combination.

[0010] Furthermore, the entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is less than or equal to the preset entity distribution density, and determines to perform entity clustering for the production line entities.

[0011] Furthermore, the entity clustering unit performs entity clustering for production line entities, including: Get the combinations of entities to be selected whose combination tendency value is greater than the preset combination tendency value; Construct several sets of combinations, each set of combinations containing several combinations of entities to be selected; Obtain the effective temporal coverage of the combined set, select the combined set with the largest effective temporal coverage, and record each entity combination to be selected in the combined set as a fault retrieval combination. The duplication elimination condition is that the proportion of identical production line entities between any two candidate entity combinations within the combination set is less than a preset proportion.

[0012] Furthermore, for any set of combinations, the corresponding temporal effective coverage is determined based on the combination effectiveness of each candidate entity combination corresponding to the set of combinations; The effective coverage of the time series and the effectiveness of the combination are positively correlated.

[0013] Furthermore, for any combination of entities to be selected, the method for determining the validity of the corresponding combination is as follows: Obtain historical data of each production line entity in the candidate entity combination under different production scenarios, and count the number of times each production line entity appears in the monitoring data text under the condition of adjacent scenario change, which is recorded as the stable number; Determine the combinatorial effectiveness of the entity combination to be selected based on the number of stable pairs of each entity; The effectiveness of the combination is positively correlated with the number of stable times of each entity pair.

[0014] Furthermore, the data analysis unit determines that there is no need to perform data augmentation on the multi-source monitoring data text if the data missing degree corresponding to the multi-source monitoring data text is less than or equal to the preset data missing degree.

[0015] Compared with the prior art, the beneficial effects of the present invention are that the technical solution of the present invention detects the missing status of each preset key field in the multi-source monitoring data text through the data analysis unit, and performs data augmentation processing when the data missingness exceeds the standard. Compared with the prior art, which adopts a unified processing method for all monitoring data or directly performs knowledge base matching, the present invention can actively identify low-quality data and repair it at the source, effectively avoiding knowledge base matching failure caused by incomplete input data, and improving the robustness and reliability of the fault monitoring system.

[0016] Furthermore, in this invention, the variance of the supplementary matching degree of each simulated data text is calculated by the enhanced execution unit as the supplementary orientation value. When the supplementary orientation value is large, the simulated data text with the largest supplementary matching degree is selected; when the supplementary orientation value is small, the simulated data text with the largest working condition execution value is selected. This improves the effectiveness of matching the enhancement strategy with the data missing distribution characteristics, ensures the semantic coherence between the supplemented content and the original data, and ensures the adaptability of the replaced text to the current production scenario.

[0017] Furthermore, in this invention, the extracted production line entities are mapped to a fault knowledge graph through an entity analysis unit. The entity distribution density is calculated based on the average path length between entities. When the distribution density is greater than a preset threshold, the entities are directly combined into fault retrieval combinations; when the distribution density is less than or equal to the preset threshold, entity clustering is triggered. This achieves accurate matching between the entity combination method and the structural features of the knowledge graph, avoiding the waste of computational resources caused by over-clustering.

[0018] Furthermore, this invention uses entity clustering units to filter candidate entity combinations based on combination tendency values, ensures entity differences between combinations within the same combination set through duplication elimination conditions, and selects the optimal set from multiple candidate combination sets through temporal effective coverage. This invention improves the coverage efficiency of fault retrieval. Attached Figure Description

[0019] Figure 1 This is a unit connection diagram of the production line emulsion pump head assembly monitoring system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating whether data augmentation is performed on multi-source monitoring data text in an embodiment of the present invention; Figure 3 This is a flowchart illustrating whether the selection criterion for simulated data text is adjusted from the selection of supplementary matching degree to the selection of operating condition execution value in an embodiment of the present invention; Figure 4 This is a flowchart illustrating whether entity clustering is performed in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0022] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0023] Please see Figure 1The diagram shown is a unit connection diagram of the production line emulsion pump head assembly monitoring system according to an embodiment of the present invention. The present invention provides a production line emulsion pump head assembly monitoring system, comprising: The data analysis unit is used to determine whether to perform data augmentation on the multi-source monitoring data text based on the data missingness corresponding to the multi-source monitoring data text. The enhanced execution unit is used to determine whether to adjust the selection criterion of the simulated data text from the selection of supplementary matching degree to the selection of the working condition execution value based on the supplementary orientation value, and to determine the simulated data text to replace the multi-source monitoring data text based on the selection criterion. The entity analysis unit is used to determine whether to perform entity clustering based on the entity distribution density of production line entities in multi-source monitoring data text. The entity clustering unit is used to perform entity clustering, obtain the entity combination to be selected by combining tendency value, construct several combination sets based on the duplication elimination condition, and select the entity combination to be selected in the combination set by the effective coverage of time series, which are respectively denoted as fault retrieval combination. The fault monitoring unit is used to match fault retrieval combinations with a pre-stored knowledge base to determine fault handling strategies.

[0024] Please see Figure 2 As shown, it is a flowchart of whether to perform data augmentation on multi-source monitoring data text in an embodiment of the present invention. The data analysis unit responds to the condition that the data missing degree corresponding to the multi-source monitoring data text is greater than the preset data missing degree, and determines to perform data augmentation on the multi-source monitoring data text.

[0025] In this embodiment of the invention, the acquisition of multi-source monitoring data text includes: textual descriptive records generated by on-site inspection personnel according to manually set inspection templates, or text describing the status of source data obtained by using a large language model to generate text for multiple source data from the production line. Specifically, by calling a pre-trained large language model, data from structured or numerical data sources such as equipment sensors, programmable logic controllers, and manufacturing execution systems are semantically converted and textualized to generate natural language text representing the current status of each source data. The above content is easily understood by those skilled in the art and will not be elaborated further. It is understood that the raw monitoring data generated by the emulsion pump head assembly line is mostly numerical parameters, such as pressing force, tightening torque, temperature, and position signals. These cannot be directly semantically matched with fault knowledge text in a pre-stored knowledge base. Through manual inspection entry or textual generation by a large language model, discrete numerical data can be transformed into semantically consistent textual descriptions, providing a feasible data foundation for subsequent entity analysis and knowledge base matching.

[0026] In this embodiment of the invention, the data analysis unit includes a missing information acquisition module for acquiring data missing information. The acquisition method for data missing information includes determining the data missing information based on the number of fields with missing content, wherein the data missing information is positively correlated with the number of fields with missing content. In this embodiment, a set of preset key fields in the multi-source monitoring data text is acquired. The preset key fields include, but are not limited to, workstation identifier fields, parameter type fields, parameter value fields, timestamp fields, and status description fields. The system detects whether each preset key field in the multi-source monitoring data text has missing content. Missing content includes field values ​​being empty, field values ​​being placeholders, and field values ​​being unrecognizable. The proportion of the number of fields with missing content to the total number of preset key fields is counted, and this proportion is recorded as the data missing information. It can be understood that the data missing information is used to quantify the degree of missing key information in the multi-source monitoring data text. The larger the data missing information, the less complete the effective information contained in the multi-source monitoring data text, the worse the support capability for subsequent entity analysis and knowledge base matching, and the higher the urgency for data augmentation.

[0027] In this embodiment of the invention, the preset data missing degree setting method includes: collecting historical monitoring data text samples from the emulsion pump head assembly production line, using different values ​​as data missing degree thresholds for knowledge base matching tests, and using the accuracy of the matching results as the evaluation index. The data missing degree threshold corresponding to the simulation result with the best accuracy is recorded as the preset data missing degree. The smaller the data missing degree, the higher the completeness of the multi-source monitoring data text, which can be directly used for subsequent entity analysis and knowledge base matching; the larger the data missing degree, the more key information is missing in the text, and if it is directly used for knowledge base matching, it will lead to a significant decrease in the accuracy of the retrieval results. The greater the user's sensitivity to the accuracy of fault monitoring, the smaller the preset data missing degree. The accuracy of the matching results refers to the proportion of the number of times the correct fault handling strategy is successfully matched after the text is matched with the pre-stored knowledge base during the knowledge base matching test, out of the total number of matching tests.

[0028] Please see Figure 3 As shown, it is a flowchart of whether to adjust the selection criterion of the simulated data text from the selection of supplementary matching degree to the selection of the working condition execution value in an embodiment of the present invention. The enhancement execution unit is used to perform data enhancement to replace the multi-source monitoring data text with simulated data text. In response to the condition that the supplementary orientation value is greater than the preset supplementary orientation value, it determines to select the simulated data text with the largest supplementary matching degree to replace the multi-source monitoring data text.

[0029] In the embodiment of the present invention, the manner for acquiring the simulated data text includes: invoking a pre-trained large language model, taking the keywords identical to the preset keyword library extracted from the multi-source monitoring data text as conditional input, generating a plurality of knowledge texts with complete semantics, which are recorded as simulated data texts. Each simulated data text is a semantic completion result of the same multi-source monitoring data text in different completion directions. In this embodiment, there exists a multi-source monitoring data text "Torque is low, the torque sensor needs to be calibrated", and the corresponding simulated data texts are "The tightening torque at the locking cover tightening station is low, which is caused by the zero drift of the torque sensor, and the torque sensor needs to be recalibrated", and "The tightening torque at the locking cover tightening station is low, and after inspection, it is found that the thread size of the incoming locking cover material is too small, and the incoming material batch needs to be replaced". In this embodiment, the preset keyword library is a pre-constructed set covering common fault keywords of each station in the emulsion pump head assembly production line. The construction manner of the preset keyword library is: collecting historical fault text data of the emulsion pump head assembly production line, including but not limited to maintenance work orders, fault reports, and equipment logs, performing word segmentation processing and word frequency statistics on the historical fault text data, eliminating general words such as stop words like "de", "shi", "jin xing", etc., and retaining professional words that appear frequently in fault descriptions, such as "torque", "press-fitting", "spring", "locking cover", "calibration", "sensor", "drift", "thread", "stuck", etc.; classifying and storing the high-frequency professional words according to station categories to construct the preset keyword library. It can be understood that the preset keyword library is a matching benchmark for the large language model to identify core semantic elements, and the larger the coverage of the keyword library, the more accurate the model's understanding ability is.

[0030] In the embodiment of the present invention, the enhancement execution unit comprises a matching degree acquisition module, configured to acquire a supplementary matching degree. The manner for acquiring the supplementary matching degree includes: extracting each production line entity appearing in the simulated data text, where the production line entities include station entities, equipment entities, component entities, fault phenomenon entities, and fault cause entities, acquiring the co-occurrence frequency of each production line entity in the historical multi-source monitoring data text, that is, the number of times any two production line entities appear simultaneously in the same historical multi-source monitoring data text, calculating the average value of the pairwise co-occurrence frequencies of all production line entities in the simulated data text, which is taken as the supplementary matching degree of the simulated data text. It can be understood that when each production line entity in the simulated data text frequently co-occurs in the historical multi-source monitoring data text, it indicates that the combination relationship between these entities has strong domain rationality, and the completion content of the simulated data text is more in line with the actual fault propagation law of emulsion pump head assembly production, so the supplementary matching degree is relatively high.

[0031] In this embodiment of the invention, the enhanced execution unit includes a direction value acquisition module for acquiring supplementary direction values. The acquisition method for the supplementary direction values ​​includes: determining the supplementary direction value based on the supplementary matching degree corresponding to each simulated data text, wherein the difference between the supplementary direction value and the supplementary matching degree corresponding to each simulated data text is positively correlated. In this embodiment, the supplementary matching degree corresponding to each simulated data text is acquired, and the variance of the supplementary matching degree is recorded as the supplementary direction value. It can be understood that when the variance of the supplementary matching degree of each simulated data text is large, it indicates that there is a significant difference in the supplementary matching degree among the simulated data texts. At this time, a clear distinction should be made based on the supplementary matching degree, and therefore the simulated data text with the largest supplementary matching degree should be selected. When the variance of the supplementary matching degree of each simulated data text is small, it indicates that the difference in the supplementary matching degree is not significant. At this time, the supplementary matching degree is insufficient to distinguish the selection. The selection is based on the working condition execution value, and the simulated data text with the largest working condition execution value is selected to avoid introducing information that is inconsistent with the current production environment.

[0032] In this embodiment of the invention, the method for setting the preset supplementary orientation value includes: collecting historical data enhancement case samples from the emulsion pump head assembly production line, using different values ​​as supplementary orientation value thresholds to test the enhancement effect, and using the accuracy rate of the enhanced knowledge base matching as the evaluation index. The average value of the supplementary orientation value thresholds corresponding to each simulation result whose accuracy meets user requirements is recorded as the preset supplementary orientation value. The accuracy rate of the enhanced knowledge base matching is the proportion of the number of times the correct fault handling strategy is successfully matched out of the total number of matches after replacing the original multi-source monitoring data text with simulated data text, inputting the enhanced multi-source monitoring data text into the fault monitoring unit, and matching it with the pre-stored knowledge base.

[0033] Specifically, the condition that the supplementary orientation value of the enhanced execution unit response is less than or equal to the preset supplementary orientation value is used to determine and select the simulation data text with the largest working condition execution value to replace the multi-source monitoring data text.

[0034] In this embodiment of the invention, the enhanced execution unit further includes an execution value acquisition module for acquiring operating condition execution values. The acquisition method for these operating condition execution values ​​includes: extracting each production line entity appearing in the simulated data text; acquiring the total number of times each production line entity appears in the production management system, equipment operation logs, maintenance records, and historical multi-source monitoring data text within a preset historical time window before the current moment, recording this as the actual usage frequency of each entity; and calculating the average of the actual usage frequencies of all production line entities in the simulated data text as the operating condition execution value of the simulated data text. It is understood that the more times a production line entity is used in actual production, the more common the corresponding workstation, equipment, or fault mode is, and the higher the hit rate and accuracy of fault retrieval based on that entity. The larger the preset historical time window, the larger the amount of data used to acquire the operating condition execution values, and the more accurate the calculation. The higher the user's requirement for the accuracy of acquiring the operating condition execution values, the larger the preset historical time window.

[0035] Please see Figure 4 As shown, it is a flowchart of whether entity clustering is performed in an embodiment of the present invention. The entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is greater than the preset entity distribution density, and determines that all production line entities are recorded as a fault retrieval group.

[0036] In this embodiment of the invention, the entity analysis unit includes a distribution density acquisition module for obtaining entity distribution density. The acquisition method includes: mapping each production line entity extracted from multi-source monitoring data text to corresponding nodes in a fault knowledge graph; obtaining the shortest path length between any two entities in the entity set, where a direct connection is counted as 1, a connection through an intermediate node as 2, and so on; calculating the average of all pairwise shortest path lengths between entities, denoted as the average path length; the entity distribution density is negatively correlated with the average path length, i.e., the smaller the average path length, the more clustered the entities are in the knowledge graph, and the greater the entity distribution density. In this embodiment, the entity distribution density is the reciprocal of the average path length. It is understood that the entity distribution density is used to determine the structural tightness of production line entities in the fault knowledge graph. The higher the entity distribution density, the closer and more interconnected the production line entities are in the fault knowledge graph, and there are clear fault propagation paths or causal relationships between them. This entity set has high structural cohesion and is suitable for direct use as a fault retrieval combination. The lower the entity distribution density, the more dispersed the production line entities are in the knowledge graph, and there is a lack of direct structural relationships between them. There are undiscovered or unconnected entity paths, and entity clustering is required.

[0037] In this embodiment of the invention, the method for setting the preset entity distribution density includes: collecting historical multi-source monitoring data text samples from the emulsion pump head assembly production line; mapping the entities in each sample to a fault knowledge graph; calculating entity distribution density thresholds using different values ​​and performing simulations; and using the accuracy of the final knowledge base matching as the evaluation index. The entity distribution density threshold corresponding to the simulation result whose accuracy meets user requirements is selected and denoted as the preset entity distribution density. Furthermore, as an alternative method, if the number of historical multi-source monitoring data text samples is too small to support the acquisition of preset values, samples can be manually constructed to simulate the preset values. The acquisition of other preset values ​​in this invention follows the same principle, which is already known to those skilled in the art and will not be elaborated upon here.

[0038] Specifically, the entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is less than or equal to the preset entity distribution density, and then determines to perform entity clustering for the production line entities.

[0039] Specifically, the entity clustering unit performs entity clustering for production line entities, including: Get the combinations of entities to be selected whose combination tendency value is greater than the preset combination tendency value; Construct as many combination sets as possible, with each combination set containing several combinations of entities to be selected; Obtain the effective temporal coverage of the combined set, select the combined set with the largest effective temporal coverage, and record each entity combination to be selected in the combined set as a fault retrieval combination. The duplication elimination condition is that the proportion of identical production line entities between any two candidate entity combinations within the combination set is less than a preset proportion.

[0040] In this embodiment of the invention, the method for obtaining the candidate entity combination includes: constructing several production line entity subsets from all production line entities extracted from multi-source monitoring data text. Each production line entity subset includes some production line entities from the multi-source monitoring data text but does not include all production line entities. All possible subset combinations in the complete entity set are traversed and denoted as candidate entity combinations. As many candidate entity combinations as possible are generated, and each candidate entity combination must satisfy the condition that its combination tendency value is greater than a preset combination tendency value. It is understood that generating as many candidate entity combinations as possible can fully cover the potential different combination methods between production line entities, ensuring that no entity combinations with high fault retrieval value are missed, and providing a sufficient candidate basis for subsequent selection of the optimal combination set through effective temporal coverage.

[0041] In this embodiment of the invention, the method for setting the combination tendency value includes: obtaining the pairwise co-occurrence frequency of each production line entity in the candidate entity combination in historical multi-source monitoring data text, that is, the number of times any two production line entities appear simultaneously in the same historical text; calculating the average of the pairwise co-occurrence frequencies of all entity pairs in the candidate entity combination as the co-occurrence support of the combination; in addition, obtaining the frequency of each entity pair in the candidate entity combination being verified as a valid combination in historical fault retrieval, that is, the number of times the entity pair was used for knowledge base matching and successfully matched the correct fault handling strategy; calculating the average of the valid verification frequencies of all entity pairs as the verification confidence of the combination; and performing a comprehensive analysis based on the co-occurrence support and verification confidence to obtain the combination tendency value. In this embodiment, the comprehensive analysis of co-occurrence support and validation confidence includes multiplying the co-occurrence support and validation confidence by their respective processing coefficients and then summing the results to obtain the combination tendency value. The processing coefficients for co-occurrence support and validation confidence are different. It can be understood that co-occurrence support reflects the frequency with which production line entities co-occur in historical fault records, measuring the strength of the potential association between entities. Validation confidence reflects the frequency with which the entity combination is validated as valid in actual knowledge base matching. Users can determine the corresponding processing coefficients based on the influence weights of co-occurrence support and validation confidence on the combination tendency value. The greater the influence weight, the larger the corresponding processing coefficient. This is easily understood by those skilled in the art and will not be elaborated upon here.

[0042] Specifically, for any set of combinations, the corresponding temporal effective coverage is determined based on the combination effectiveness of each candidate entity combination corresponding to the set of combinations; The temporal effective coverage and the combinatorial effectiveness are positively correlated. In this embodiment, for a combinatorial set, the average combinatorial effectiveness of each candidate entity combination in the combinatorial set is calculated and denoted as the temporal effective coverage.

[0043] In this embodiment of the invention, the method for setting the combination validity includes: acquiring historical data of each production line entity in the candidate entity combination under different production scenarios, counting the number of times each production line entity appears in both production scenarios under adjacent scenario changes, and recording this as the stable count; determining the combination validity of the candidate entity combination based on the average of the stable counts of each entity pair. Specifically, if there are changes in product model, process parameter configuration, or production batch in the production line scenario, which constitute different production scenarios that are adjacent in time sequence, then these are considered adjacent scenario changes. It can be understood that the combination validity is used to determine the effective stability of the candidate entity combination in the production time dimension. The higher the combination validity, the more consistently and stably each production line entity in the candidate entity combination appears in different production scenarios, the higher the temporal stability of the entity combination relationship, and the stronger the long-term reliability of the combination as a fault retrieval combination.

[0044] Specifically, the data analysis unit determines that no data augmentation is needed for the multi-source monitoring data text if the data missing degree corresponding to the multi-source monitoring data text is less than or equal to the preset data missing degree.

[0045] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0046] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A monitoring system for emulsion pump head assembly on a production line, characterized in that, include: The data analysis unit is used to determine whether to perform data augmentation on the multi-source monitoring data text based on the data missingness corresponding to the multi-source monitoring data text. The enhanced execution unit is used to determine whether to adjust the selection criterion of the simulated data text from the selection of supplementary matching degree to the selection of the operating condition execution value based on the supplementary orientation value, and to determine the simulated data text to replace the multi-source monitoring data text based on the selection criterion. The entity analysis unit is used to determine whether to perform entity clustering based on the entity distribution density of production line entities in multi-source monitoring data text. The entity clustering unit is used to perform entity clustering, obtain the entity combination to be selected by combining tendency value, construct several combination sets based on the duplication elimination condition, and select the entity combination to be selected in the combination set by the effective coverage of time series, which are respectively denoted as fault retrieval combination. The fault monitoring unit is used to match fault retrieval combinations with a pre-stored knowledge base to determine fault handling strategies.

2. The production line emulsion pump head assembly monitoring system according to claim 1, characterized in that, The data analysis unit responds to the condition that the data missing degree corresponding to the multi-source monitoring data text is greater than the preset data missing degree, and determines to perform data augmentation on the multi-source monitoring data text.

3. The production line emulsion pump head assembly monitoring system according to claim 2, characterized in that, The execution unit responds to the condition that the supplementary orientation value is greater than the preset supplementary orientation value, and then selects the simulated data text with the highest supplementary matching degree to replace the multi-source monitoring data text.

4. The production line emulsion pump head assembly monitoring system according to claim 3, characterized in that, Enhance the condition that the supplementary orientation value of the execution unit response is less than or equal to the preset supplementary orientation value, and determine to replace the multi-source monitoring data text with the simulation data text that has the largest operating condition execution value.

5. The production line emulsion pump head assembly monitoring system according to claim 4, characterized in that, The entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is greater than the preset entity distribution density, and determines all production line entities as a fault retrieval combination.

6. The production line emulsion pump head assembly monitoring system according to claim 5, characterized in that, The entity analysis unit responds to the condition that the entity distribution density of the production line entities in the multi-source monitoring data text is less than or equal to the preset entity distribution density, and then determines to perform entity clustering for the production line entities.

7. The production line emulsion pump head assembly monitoring system according to claim 6, characterized in that, The entity clustering unit performs entity clustering for production line entities, including: Get the combinations of entities to be selected whose combination tendency value is greater than the preset combination tendency value; Construct several sets of combinations, each set of combinations containing several combinations of entities to be selected; Obtain the effective temporal coverage of the combined set, select the combined set with the largest effective temporal coverage, and record each entity combination to be selected in the combined set as a fault retrieval combination. The duplication elimination condition is that the proportion of identical production line entities between any two candidate entity combinations within the combination set is less than a preset proportion.

8. The production line emulsion pump head assembly monitoring system according to claim 7, characterized in that, For any set of combinations, the corresponding effective temporal coverage is determined based on the combinatorial effectiveness of each combination of entities to be selected for the set. The effective coverage of the time series and the effectiveness of the combination are positively correlated.

9. The production line emulsion pump head assembly monitoring system according to claim 8, characterized in that, The method for determining the validity of any combination of entities to be selected is as follows: Obtain historical data of each production line entity in the candidate entity combination under different production scenarios, and count the number of times each production line entity appears in the monitoring data text under the condition of adjacent scenario change, which is recorded as the stable number; Determine the combinatorial effectiveness of the entity combination to be selected based on the number of stable pairs of each entity; The effectiveness of the combination is positively correlated with the number of stable times of each entity pair.

10. The production line emulsion pump head assembly monitoring system according to claim 1, characterized in that, The data analysis unit determines that no data augmentation is needed for the multi-source monitoring data text if the data missing degree corresponding to the multi-source monitoring data text is less than or equal to the preset data missing degree.

Images