A production abnormality root cause positioning method and system based on multi-round commonality analysis

Abstract

The application provides a production abnormality root cause positioning method and system based on multi-round commonality analysis, relates to the technical field of industrial data processing, and comprises the following steps: S1, locating an attribute field corresponding to a target quality index from a data record set, and constructing a remaining attribute field into an analysis set; S2, traversing each attribute field in the analysis set, and performing first commonality analysis on each attribute field: performing single-dimension correlation test on the attribute field and the target quality index to determine a to-be-verified field; S3, performing second commonality analysis on each to-be-verified field: taking the to-be-verified field as a fixed variable, traversing the remaining attribute fields in the analysis set, respectively forming field combinations with the remaining attribute fields, performing multi-dimension correlation test on each field combination and the target quality index, and determining a credible field; and S4, positioning a root cause conclusion from the credible field based on the first test result and the second test result, which can effectively eliminate the interference of confusing factors.

Description

Technical Field

[0001] This invention relates to the field of industrial data processing technology, and specifically to a method and system for locating the root causes of production anomalies based on multi-round commonality analysis. Background Technology

[0002] In industrial production, controlling product quality indicators is a crucial aspect of production management. When a quality indicator exhibits abnormal fluctuations, such as a sudden increase in the percentage of open-circuit failures, it is necessary to quickly identify the root cause of the anomaly in order to implement targeted containment and improvement measures.

[0003] In actual production environments, production data associated with product quality indicators are usually stored in the quality management system in tabular form. These data typically include multiple attribute fields, such as batch numbers at different levels, factory names, production line numbers, equipment numbers, material suppliers, incoming material quantities, various processing parameters, transit times, etc. Each row of records contains the specific values ​​of the corresponding observed entity under each attribute field.

[0004] When quality indicators show anomalies, engineers typically need to manually investigate data and pinpoint root causes in tables containing numerous attribute fields. This manual approach has significant shortcomings. Firstly, when tables contain a large number of attribute fields, the workload of grouping, calculating, and comparing each field is enormous, making the entire process time-consuming, labor-intensive, and inefficient. Secondly, the common practice of manually testing single variables one by one is susceptible to confounding factors. For example, a batch of materials might show a significant association with anomalies in quality indicators in a single-variable comparison, but this association might actually be due to the entire batch being processed on a faulty piece of equipment, rather than a problem with the material itself. Engineers cannot comprehensively verify each suspect factor independently after controlling for other variables during manual investigation. If such false positive associations are not eliminated, they can lead to incorrect attribution, causing unnecessary material freezes or equipment adjustments. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for locating the root causes of production anomalies based on multi-round commonality analysis. This method automatically filters and verifies multiple candidate analysis dimensions through multi-round commonality analysis, effectively eliminating the interference of confounding factors, and ultimately locating the root cause fields and specific entities with common impact.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following solution: A method for locating the root causes of production anomalies based on multi-round commonality analysis includes: S1. Obtain the target quality indicator and the set of data records associated with it; locate the attribute fields corresponding to the target quality indicator from the set of data records, and form the set to be analyzed from the remaining attribute fields; S2. Traverse each attribute field in the set to be analyzed and perform the first commonality analysis on each attribute field: perform a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; Based on the results of the first test, several attribute fields whose association significance with the target quality index meets the preset conditions are identified as fields to be verified. S3. Perform a second commonality analysis on each field to be verified: Treat the field to be verified as a fixed variable, traverse the remaining attribute fields in the set to be analyzed, form field combinations with the remaining attribute fields respectively, and conduct multi-dimensional correlation tests on each field combination with the target quality index to obtain the second test results. The results of all second tests for each field to be verified are summarized and analyzed. Fields that maintain a significant association independent of the other attribute fields in different combinations of fields are identified as reliable fields. S4. Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

[0007] Furthermore, in S1, when obtaining the target quality index and the set of data records associated with it, the remaining attribute fields are preprocessed, including sample missing data analysis.

[0008] Furthermore, in S1, when obtaining the target quality indicator and its associated set of data records, the data types of the target quality indicator and the remaining attribute fields are marked. The data types include discrete, continuous, and binary. This is so that in single-dimensional correlation tests and multi-dimensional correlation tests, the corresponding analysis methods can be obtained based on the data type matching, and different analysis methods can be used to obtain the first test result and the second test result.

[0009] Furthermore, the process of single-dimensional correlation testing is as follows: Obtain the data types of the target quality indicator and the attribute field. Match the corresponding univariate analysis method according to the data type. Use the matched univariate analysis method to calculate the correlation between the specific value of the attribute field and the specific value of the target quality indicator. Obtain the significance of the correlation between the attribute field and the target quality indicator, and use it as the first test result of the attribute field.

[0010] Furthermore, the process of multi-dimensional correlation testing is as follows: Obtain the data type of the target quality indicator and the data types of the field to be verified and the other attribute fields in the field combination. Match the corresponding multivariate analysis method according to the data type combination of the field to be verified and the other attribute fields. After eliminating the influence of the other attribute fields on the target quality indicator, use the matched multivariate analysis method to calculate the significance of the independent association between the specific value of the field to be verified and the specific value of the target quality indicator. This is used as the second test result corresponding to the field combination.

[0011] Furthermore, the significance of the association between the attribute field and the target quality indicator meets the preset condition, which means that the significance of the association between the specific value of the target quality indicator and the specific value of the attribute field is greater than the first threshold.

[0012] Furthermore, maintaining independent association significance from other attribute fields in different field combinations means that, in the multi-dimensional association test of each field combination, the independent association significance between the specific value of the field to be verified and the specific value of the target quality indicator is greater than the second threshold.

[0013] Furthermore, the process of identifying fields as trustworthy is as follows: The second test results of the field to be verified in different field combinations are obtained respectively. Based on the second test results, it is analyzed whether the significance of the association between the field to be verified and the target quality indicator meets the preset conditions after the introduction of other attribute fields. If so, it is determined that the field to be verified maintains a significant association independent of other attribute fields in different field combinations, and the field to be verified is determined to be a credible field.

[0014] Furthermore, in S4, the association significance of each attribute field in the credible field in the first test result and the independent association significance in the second test result are ranked. The attribute field with the highest combined ranking of association significance and independent association significance is determined as the attribute field with the highest contribution to the target quality indicator. The root cause conclusion is generated based on the specific value of the attribute field with the highest contribution to the target quality indicator.

[0015] A production anomaly root cause localization system based on multi-round commonality analysis, employing a production anomaly root cause localization method based on multi-round commonality analysis, including: Data acquisition module: Acquires the target quality indicator and its associated set of data records; locates the attribute fields corresponding to the target quality indicator from the set of data records, and constructs the set to be analyzed from the remaining attribute fields; First Commonality Analysis Module: Traverses each attribute field in the set to be analyzed and performs the first commonality analysis on each attribute field: performs a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; based on the first test result, determines several attribute fields whose correlation significance with the target quality indicator meets the preset conditions as fields to be verified. The second commonality analysis module performs a second commonality analysis on each field to be verified: the field to be verified is treated as a fixed variable, and the remaining attribute fields in the set to be analyzed are traversed. Each field combination is combined with the remaining attribute fields to form a field combination. Each field combination is then subjected to a multi-dimensional correlation test with the target quality indicator to obtain the second test results. All the second test results of each field to be verified are summarized and analyzed. Fields to be verified that maintain a significant correlation independent of the remaining attribute fields in different field combinations are identified as credible fields. Root cause localization module: Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

[0016] The beneficial effects of this invention are: This invention provides a method for locating the root cause of production anomalies based on multi-round commonality analysis, applicable to industrial quality expert platforms, and incorporating numerous scientific analysis algorithms and AI business models. This method transforms the manual troubleshooting process—where engineers painstakingly group, calculate, and compare data in tables containing a large number of fields—into an automated multi-round analysis workflow. The computer automatically traverses all attribute fields in the analysis set, performing single-dimensional and multi-dimensional correlation tests, completing the screening and verification of dozens of attribute fields in a short time, significantly shortening the response cycle from the occurrence of a quality anomaly to root cause location.

[0017] Based on this, the multi-round commonality analysis employed in this invention utilizes a two-stage progressive screening and verification architecture, with each commonality analysis performed through a full traversal. The first commonality analysis traverses all candidate attribute fields, performing single-dimensional tests one by one to identify fields with significant correlations for verification. The second commonality analysis uses each field to be verified as a fixed variable, again traversing all remaining attribute fields, combining them one by one for multi-dimensional testing. After eliminating the influence of other attribute fields, the significance of independent association is calculated to determine whether the association of the field to be verified is independent of other variables. Therefore, the meaning of multi-round commonality analysis in this invention refers not only to the progressive relationship between the two commonality analyses but also to the complete traversal of all attribute fields within each commonality analysis. This mechanism, combining full traversal with two-stage verification, ensures that all possible influencing factors are thoroughly investigated, effectively identifying and eliminating false positive associations caused by collinearity or other variable-driven factors, and avoiding erroneous attribution leading to unnecessary material freezes or equipment adjustments.

[0018] In summary, this method not only identifies the attribute field that contributes the most to the quality indicators, but also drills down to the specific abnormal entities within that field, outputting root cause conclusions that can directly guide production containment and process verification. This achieves a fully automated closed loop from data filtering and obfuscation elimination to precise location. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the method for locating the root cause of production anomalies in Embodiment 1 of the present invention.

[0020] Figure 2 This is a schematic diagram of the multi-round commonality analysis in Embodiment 1 of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0023] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0024] Furthermore, for clarity and brevity, descriptions of well-known structures, functions, and configurations may have been omitted. Those skilled in the art will recognize that various changes and modifications can be made to the examples described herein without departing from the spirit and scope of this disclosure.

[0025] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0026] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0027] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: Example 1 In this embodiment, a method for locating the root causes of production anomalies based on multi-round commonality analysis is provided, such as... Figure 1 As shown, this method for generating root cause localization of anomalies includes the following steps: S1. Obtain the target quality indicator and the set of data records associated with it; locate the attribute fields corresponding to the target quality indicator from the set of data records, and form the set to be analyzed from the remaining attribute fields; S2. Traverse each attribute field in the set to be analyzed and perform the first commonality analysis on each attribute field: perform a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; Based on the results of the first test, several attribute fields whose association significance with the target quality index meets the preset conditions are identified as fields to be verified. S3. Perform a second commonality analysis on each field to be verified: Treat the field to be verified as a fixed variable, traverse the remaining attribute fields in the set to be analyzed, form field combinations with the remaining attribute fields respectively, and conduct multi-dimensional correlation tests on each field combination with the target quality index to obtain the second test results. The results of all second tests for each field to be verified are summarized and analyzed. Fields that maintain a significant association independent of the other attribute fields in different combinations of fields are identified as reliable fields. S4. Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

[0028] The above-mentioned method for locating the root causes of production anomalies is explained below: This method is suitable for automated root cause analysis of quality indicator anomalies occurring in industrial production processes. For example, in semiconductor packaging or surface mount technology processes, it can be used to analyze the cause of a high percentage of open-circuit failures. The execution device for this method can be a server, an industrial computer, or any electronic device with data processing capabilities.

[0029] In step S1, the user can specify one of the data record sets as the target quality indicator. The target quality indicator refers to the quality indicator that needs to be analyzed for root cause analysis, such as the percentage of open-circuit failures. The remaining attribute fields constitute the set to be analyzed, which serves as the candidate dimension set for subsequent commonality analysis. Specifically, the data record set is usually stored in tabular form.

[0030] While acquiring the data record set, this invention also proposes preprocessing each attribute field. Preprocessing includes sample missing data analysis, checking the completeness of each attribute field's values ​​in the data record set, and marking or excluding attribute fields with missing proportions exceeding a preset value. Furthermore, preprocessing also includes marking the data type of each attribute field, including discrete, continuous, and dichotomous data types, so that appropriate analysis methods can be automatically matched based on the data type in subsequent correlation tests.

[0031] In step S2, a first commonality analysis is performed on each attribute field. The core actions of the first commonality analysis are full traversal and preliminary screening. By traversing all attribute fields in the set to be analyzed, each attribute field is subjected to a single-dimensional correlation test with the target quality indicator, which is recorded as a multi-round commonality analysis to obtain the first test result for each attribute field.

[0032] The specific process of unidimensional correlation testing is as follows: First, obtain the data types of the target quality indicator and the current attribute field, and then match the corresponding univariate analysis method according to the data types. At this point, the default data type of the target quality indicator is continuous. If the attribute field's data type is discrete, ANOVA (one-way ANOVA) can be used, specifically analyzing the variance difference between discrete and continuous variables using box plots. If the attribute field's data type is continuous, correlation analysis can be used, specifically analyzing the association strength between continuous variables using scatter plots. If the attribute field's data type is dichotomous, a T-test can be used, specifically analyzing the mean difference between binary and continuous variables using box plots. Then, using the matched univariate analysis method, the correlation between the specific value of the attribute field and the specific value of the target quality indicator is calculated to obtain the significance of the association between the attribute field and the target quality indicator, which serves as the first test result for that attribute field.

[0033] After obtaining the initial test results for all attribute fields, attribute fields with a significance level greater than the first threshold are identified as fields to be validated. Attribute fields with a significance level below the first threshold are considered to have no significant correlation or a weak correlation with the target quality indicator and are excluded at this stage, not proceeding to subsequent validation. If no attribute field meets the preset conditions, a message indicating no valid signal at this stage can be generated, and the user can be advised to supplement more data records or change the target quality indicator. In step S3, a second commonality analysis is performed on each field to be verified. The core action of the second commonality analysis is to perform a full traversal and independent verification again. For each field to be verified identified in step S2, it is used as a fixed variable. All other attribute fields in the set to be analyzed are traversed again. Each time, one other attribute field is taken out and combined with the fixed variable to form a field combination. This field combination is then subjected to a multi-dimensional correlation test with the target quality indicator to obtain a second test result corresponding to the field combination. After the traversal is completed, a field to be verified will generate multiple second test results, which are recorded as a multi-round commonality analysis, corresponding to the test results when the field to be verified is combined with different other attribute fields.

[0034] The specific process of multidimensional association testing is as follows: First, obtain the data type of the target quality indicator and the data types of the fixed variable and other attribute fields in the current field combination. Then, match the corresponding multivariate analysis method based on the data type combination of the two attribute fields. Next, using the matched multivariate analysis method, after eliminating the influence of the other attribute fields on the target quality indicator, calculate the independent association significance between the fixed variable and the target quality indicator, which serves as the second test result for this field combination. The key to this process is that it calculates the independent association significance of the fixed variable after excluding the interference of other attribute fields, rather than simply inputting the values ​​of both fields into the model simultaneously.

[0035] The results of all second tests for each field to be validated are summarized and analyzed. If the significance of the independent association between a field to be validated and all other attribute fields is greater than the second threshold in every combination, it indicates that the significant association between the field and the target quality indicator is independent and does not depend on the presence or absence of any other attribute fields; that is, the association remains stable in every combination of multidimensional tests. Such a field to be validated is identified as a credible field. Conversely, if the significance of the independent association between a field to be validated and some other attribute fields drops below the second threshold, it indicates that the significant association shown by the field in the first commonality analysis is actually a false positive association driven by these other attribute fields, and the field to be validated is excluded at this stage.

[0036] Both of the above commonality analyses were performed using a full traversal approach, such as... Figure 2As shown, the first commonality analysis traverses all attribute fields in the set to be analyzed, performing a single-dimensional correlation test on each attribute field, i.e., performing one round of analysis for each candidate attribute field. The second commonality analysis uses each field to be verified as a fixed variable, and again traverses all the remaining attribute fields in the set to be analyzed, performing a multi-dimensional correlation test on each combination of the remaining attribute fields, i.e., performing one round of analysis for each field combination. Neither analysis focuses solely on the few fields selected; rather, each analysis covers all candidate dimensions, with multiple rounds of traversal, ensuring that no possible influencing factors or confounding factors are overlooked.

[0037] In step S4, root cause conclusions are located from the credible fields based on the results of the first and second tests. After the second commonality analysis is completed, there may still be multiple credible fields. Each credible field undergoes a dual screening process of full-scale single-dimensional testing and multi-dimensional validation. Step S4 then further identifies the attribute field that contributes the most to the target quality indicator from among the credible fields.

[0038] Specifically, the association significance of each attribute field in the confidence field in the first test result and the independent association significance in the second test result are comprehensively ranked. The attribute field with the highest comprehensive ranking is determined as the attribute field that contributes the most to the target quality indicator. Then, the root cause conclusion is generated based on the specific value of the attribute field that contributes the most to the target quality indicator.

[0039] For example, in an open circuit failure analysis scenario, if the Fab batch number is determined to be the attribute field with the highest contribution, then the open circuit failure percentage corresponding to each Fab batch number is further analyzed to locate the specific batch numbers with the highest failure percentage. Furthermore, the root cause conclusion directly indicates that these specific batch numbers are the root cause of the high open circuit failure rate.

[0040] After generating root cause conclusions, a commonality analysis report can also be output. This report includes a list of attribute fields excluded in the first commonality analysis and the basis for exclusion, records of the independent association significance of each field to be verified in combination with each of the remaining attribute fields in the second commonality analysis, and the finally determined target association fields and root cause conclusions. Furthermore, the report may also include containment measures and process parameter verification plans generated based on the root cause conclusions, such as implementing full inspection of high-risk batches and tracing the production process parameter records of that batch.

[0041] Example 2 A production anomaly root cause localization system based on multi-round commonality analysis, employing a production anomaly root cause localization method based on multi-round commonality analysis, including: Data acquisition module: Acquires the target quality indicator and its associated set of data records; locates the attribute fields corresponding to the target quality indicator from the set of data records, and constructs the set to be analyzed from the remaining attribute fields; First Commonality Analysis Module: Traverses each attribute field in the set to be analyzed and performs the first commonality analysis on each attribute field: performs a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; based on the first test result, determines several attribute fields whose correlation significance with the target quality indicator meets the preset conditions as fields to be verified. The second commonality analysis module performs a second commonality analysis on each field to be verified: the field to be verified is treated as a fixed variable, and the remaining attribute fields in the set to be analyzed are traversed. Each field combination is combined with the remaining attribute fields to form a field combination. Each field combination is then subjected to a multi-dimensional correlation test with the target quality indicator to obtain the second test results. All the second test results of each field to be verified are summarized and analyzed. Fields to be verified that maintain a significant correlation independent of the remaining attribute fields in different field combinations are identified as credible fields. Root cause localization module: Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for locating the root causes of production anomalies based on multi-round commonality analysis, characterized in that, include: S1. Obtain the target quality indicators and the set of data records associated with them; Locate the attribute fields corresponding to the target quality indicators that cause anomalies in the industrial production process from the data record set. The data types of the attribute fields include discrete, continuous and binary types. The remaining attribute fields are then used to form the set to be analyzed. S2. Traverse each attribute field in the set to be analyzed and perform the first commonality analysis on each attribute field: perform a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; Based on the results of the first test, several attribute fields whose association significance with the target quality index meets the preset conditions are identified as fields to be verified. S3. Perform a second commonality analysis on each field to be verified: Treat the field to be verified as a fixed variable, traverse the remaining attribute fields in the set to be analyzed, form field combinations with the remaining attribute fields respectively, and conduct multi-dimensional correlation tests on each field combination with the target quality index to obtain the second test results. The results of all second tests for each field to be verified are summarized and analyzed. Fields that maintain a significant association independent of the other attribute fields in different combinations of fields are identified as reliable fields. The process of multidimensional correlation testing is as follows: Obtain the data type of the target quality indicator and the data types of the field to be verified and the other attribute fields in the field combination. Match the corresponding multivariate analysis method according to the data type combination of the field to be verified and the other attribute fields. After eliminating the influence of the other attribute fields on the target quality indicator, use the matched multivariate analysis method to calculate the significance of the independent association between the specific value of the field to be verified and the specific value of the target quality indicator. This is used as the second test result corresponding to the field combination. S4. Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

2. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, In S1, when obtaining the target quality index and the set of data records associated with it, the remaining attribute fields are preprocessed, including sample missing data analysis.

3. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, In S1, when obtaining the target quality indicator and the set of data records associated with it, the data type of the target quality indicator and the remaining attribute fields is marked. The data types include discrete, continuous, and binary. In order to obtain the corresponding analysis method based on data type matching in single-dimensional correlation tests and multi-dimensional correlation tests, different analysis methods are used to obtain the first test result and the second test result.

4. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 3, characterized in that, The process of single-dimensional correlation testing is as follows: Obtain the data types of the target quality indicator and the attribute field. Match the corresponding univariate analysis method according to the data type. Use the matched univariate analysis method to calculate the correlation between the specific value of the attribute field and the specific value of the target quality indicator. Obtain the significance of the correlation between the attribute field and the target quality indicator, and use it as the first test result of the attribute field.

5. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, The significance of the association between the attribute field and the target quality indicator meets the preset condition, which means that the significance of the association between the specific value of the target quality indicator and the specific value of the attribute field is greater than the first threshold.

6. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, Maintaining independent association significance across different field combinations means that, in the multi-dimensional association test of each field combination, the independent association significance between the specific value of the field to be verified and the specific value of the target quality indicator is greater than the second threshold.

7. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, The process of determining a field as a trustworthy field is as follows: The second test results of the field to be verified in different field combinations are obtained respectively. Based on the second test results, it is analyzed whether the significance of the association between the field to be verified and the target quality indicator meets the preset conditions after the introduction of other attribute fields. If so, it is determined that the field to be verified maintains a significant association independent of other attribute fields in different field combinations, and the field to be verified is determined to be a credible field.

8. The method for locating the root cause of production anomalies based on multi-round commonality analysis according to claim 1, characterized in that, In S4, the association significance of each attribute field in the confidence field in the first test result and the independent association significance in the second test result are ranked. The attribute field with the highest combined association significance and independent association significance is determined as the attribute field with the highest contribution to the target quality indicator. The root cause conclusion is generated based on the specific value of the attribute field with the highest contribution to the target quality indicator.

9. A production anomaly root cause localization system based on multi-round commonality analysis, characterized in that, The method for locating the root causes of production anomalies based on multi-round commonality analysis as described in any one of claims 1-8 includes: Data acquisition module: Acquires the target quality indicator and its associated set of data records; locates the attribute fields corresponding to the target quality indicator from the set of data records, and constructs the set to be analyzed from the remaining attribute fields; First Commonality Analysis Module: Traverses each attribute field in the set to be analyzed and performs the first commonality analysis on each attribute field: performs a one-dimensional correlation test between the attribute field and the target quality indicator to obtain the first test result; based on the first test result, determines several attribute fields whose correlation significance with the target quality indicator meets the preset conditions as fields to be verified. The second commonality analysis module performs a second commonality analysis on each field to be verified: the field to be verified is treated as a fixed variable, and the remaining attribute fields in the set to be analyzed are traversed. Each field combination is combined with the remaining attribute fields to form a field combination. Each field combination is then subjected to a multi-dimensional correlation test with the target quality indicator to obtain the second test results. All the second test results of each field to be verified are summarized and analyzed. Fields to be verified that maintain a significant correlation independent of the remaining attribute fields in different field combinations are identified as credible fields. Root cause localization module: Based on the results of the first and second tests, locate the attribute field that contributes the most to the target quality index from the credible fields, and generate root cause conclusions based on the specific values ​​of the attribute field.

Images