Intelligent acquisition, analysis and processing method for transformer cross-cut sheet stock detection data

Abstract

The invention relates to the technical field of data query and analysis, in particular to an intelligent acquisition, analysis and processing method for transformer cross cut sheet stock detection data. The method comprises the following steps: firstly, constructing a cross-cut sheet material production database, and then if a detection result of the cross-cut sheet material is unqualified, performing feature extraction on the unqualified cross-cut sheet material to obtain an unqualified sample feature; obtaining a first cross cut sheet stock set according to the main association retrieval similarity, obtaining a second cross cut sheet stock set according to the secondary association retrieval condition, and constructing an unqualified cross cut sheet stock set according to the first cross cut sheet stock set and the second cross cut sheet stock set; constructing an unqualified factor feature vector set according to the causal strength index; processing elements in the disqualified factor feature vector set according to the effective frequency to obtain main factors of the disqualified cross cut sheet stock; and real-time data of steps or processes where the main factors are located are collected and are compared and analyzed with standard data, and early warning is given out.

Description

Technical Field

[0001] The present invention relates to the technical field of data query and analysis, and in particular to a method for intelligently collecting, analyzing and processing transformer cross-cut sheet material detection data. Background Art

[0002] In the field of transformer cross-cut sheet material inspection, traditional data processing methods rely on offline manual input and single-machine calculations, resulting in insufficient real-time and integrity of inspection data; on the other hand, data storage is scattered in isolated systems and lacks a unified data integration framework, making it impossible to achieve collaborative analysis of multi-source heterogeneous data, and the value of historical data has not been effectively mined.

[0003] Although the existing system has introduced computer-aided analysis, it still has significant defects: first, the cross-protocol integration capability of multimodal data is weak, and a standardized data bus architecture has not been built, resulting in interruptions in real-time data flows; second, the feature extraction and analysis of data rely on static algorithms and lack dynamic weight allocation capabilities such as those based on machine learning models, and cannot adaptively combine multi-source data; third, the causal reasoning logic is rigid, and dynamic reasoning of the root cause of the defect has not been achieved, resulting in low accuracy in solution generation.

[0004] Therefore, a method for intelligent collection, analysis and processing of transformer cross-cut sheet detection data is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a method for intelligently collecting, analyzing and processing detection data of transformer cross-cut sheets, which improves production efficiency and enhances early warning capabilities by analyzing the detection data and production data of transformer cross-cut sheets. First, a cross-cut sheet production database is constructed. Then, if the detection result of the cross-cut sheet is unqualified, feature extraction is performed on the unqualified cross-cut sheet to obtain unqualified sample features; a first cross-cut sheet set is obtained based on the primary association retrieval similarity, a second cross-cut sheet set is obtained based on the secondary association retrieval condition, and an unqualified cross-cut sheet set is constructed based on the first cross-cut sheet set and the second cross-cut sheet set; an unqualified factor feature vector set is constructed based on the causal strength index; elements in the unqualified factor feature vector set are processed according to the effective frequency to obtain the main factors of the unqualified cross-cut sheet; real-time data of the steps or processes where the main factors are located are collected and compared with the standard data for analysis, and an early warning is issued.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for intelligently collecting, analyzing and processing transformer cross-cut sheet material detection data, comprising: Construct a production database of transverse shear sheets based on all production data of transformer transverse shear sheets; Furthermore, the cross-cut sheet production database includes a core detection database, a detection equipment parameter database, a processing equipment parameter database, an order database, a material database and a defect case library; wherein the core detection database includes an image database, a dimension database, a burr database and a deformation database, the image database stores images taken by the detection equipment, the dimension database stores the structural data of the cross-cut sheet, the burr database stores the burr distribution data of the cross-cut sheet, and the deformation database stores the thickness deformation data of the cross-cut sheet; the detection equipment parameter database is used to store the timing operation parameters of the detection equipment when detecting the cross-cut sheet, including speed measuring device parameters, frequency converter parameters and line scan camera parameters; the processing equipment parameter database is used to store the timing processing parameters when processing the cross-cut sheet, including the force-displacement curve, thermal parameters and mold state parameters of the shearing process; the order database is used to store order data; the material database is used to store the material data of the cross-cut sheet; the defect case library is used to store data of cross-cut sheets of unqualified quality.

[0007] If the inspection result of the cross-cut sheet material is unqualified, feature extraction is performed on the unqualified cross-cut sheet material to obtain unqualified sample features; the defect case library is retrieved according to the unqualified sample features and the primary associated search similarity to obtain the first cross-cut sheet material set; the core inspection database is retrieved according to the unqualified sample features and the secondary associated search conditions to obtain the second cross-cut sheet material set; and the unqualified cross-cut sheet material set is constructed according to the first cross-cut sheet material set and the second cross-cut sheet material set; Furthermore, the unqualified sample characteristics Including defect image features, process parameter hash values ​​and material features; query the unqualified cross-cut sheets in the defect case library based on the main association retrieval similarity to obtain the first cross-cut sheet set , the formula for the main correlation retrieval similarity is: ; in, Indicates unqualified samples and unqualified samples in the defect case library The main related retrieval similarity of Indicates the defect case library Unqualified samples The unqualified sample characteristics, The formula for calculating the modulus length is: represents the natural base, represents the time attenuation coefficient, Indicates the time difference between the current sample and the historical sample; Query the core inspection database according to the secondary associated search conditions to obtain the second cross-cut sheet material set , the formula for the secondary associated search condition is: ,in Indicates Cross-cut pieces The sample characteristics of represents the standard deviation of historical data, represents the probability of false detection, Represents a logarithmic function.

[0008] Furthermore, the construction formula for constructing the set of unqualified cross-cut sheets is: ; in, Indicates a collection of unqualified cross-cut pieces. Represents the first cross-cut sheet set, Indicates the second cross-cut sheet set, Represents the elements in the set of unqualified cross-cut pieces. represents the intersection operation, represents the union operation, Indicates that the top similarity product is retained A combination of Represents the Cartesian product operation.

[0009] Obtain unqualified cross-cut sheets in the unqualified cross-cut sheet set and obtain unqualified factor feature vectors according to the causal strength index, and further obtain an unqualified factor feature vector set; process the elements in the unqualified factor feature vector set according to the effective frequency to obtain the main factors of the unqualified cross-cut sheets; Furthermore, the calculation formula of the causal strength index is: ; in, Indicates an event Leading to the event The causal strength index, represents the time decay factor, Indicates an event The number of independent occurrences, Indicates an event The number of independent occurrences, represents the conditional entropy, Indicates an event expectations, Indicates an event expectations, represents the logarithmic function with base 2, represents the total event sample size, Represents the time step index and the time decay factor Combined, calculate the weight decay of historical events on current causal relationships, represents the total time step, represents the transfer entropy, Indicates an event In the The expectation of the second Indicates an event In the times expectation.

[0010] Furthermore, the specific steps of obtaining the main factors of unqualified cross-cut sheet materials include: The effective frequency of each factor in the unqualified factor feature vector set is counted and sorted according to the number of occurrences. Factors exceeding the threshold are judged as the main factors causing the unqualified cross-cut sheet. The calculation formula for the effective frequency is: ; in, Indicates The effective frequency of a factor, Indicates The number of times a factor appears in the set, Indicates the total number of times, Expressing the The sum of the causal strength indicators related to the factors, represents the weight coefficient of frequency, represents the weight coefficient of the timeliness correction term, represents the natural base, represents the time attenuation coefficient, Represents the time step index.

[0011] Collect real-time data of the steps where the main factors are located and compare and analyze them with standard data.

[0012] Furthermore, real-time data of the steps corresponding to the main factors are collected and a real-time data vector is constructed. The real-time data vector and the standard data vector are calculated, and the Euclidean distance is calculated. If the threshold is exceeded, it is judged that unqualified factors still exist in the production process and an early warning is issued.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. Construct a cross-cut sheet material production database that includes a core inspection database, an inspection equipment parameter database, a processing equipment parameter database, an order database, a material database, and a defect case database, realizing comprehensive data integration and providing rich data support for subsequent analysis. It can achieve accurate quality traceability and intelligent defect analysis, provide a basis for analyzing key influencing factors in the processing process, and support the optimization and improvement of the production process.

[0014] 2. Through the primary association retrieval similarity, similar unqualified samples in the defect case library can be accurately matched to form the first cross-cut sheet material set, which helps to quickly locate similar defect types and facilitates analysis and processing; the secondary association retrieval conditions comprehensively search the cross-cut sheets to form the second cross-cut sheet material set, which helps to capture potential quality problems; combining the intersection and union of the first and second cross-cut sheet material sets to construct a set of unqualified cross-cut sheet materials can explore the potential association patterns under different parameter combinations, which is conducive to multi-dimensional analysis and in-depth search for the root causes of quality problems.

[0015] 3. By calculating the causal strength index of the set of unqualified cross-cut sheets, the influence of each factor on the unqualified product quality can be quantified, and then the main unqualified factors can be identified; by counting the effective frequency of each factor in the unqualified factor feature vector set and sorting them according to the number of occurrences, the main factors leading to the unqualified cross-cut sheets can be determined, which is helpful for optimizing the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a method for intelligently collecting, analyzing and processing transformer cross-cut sheet material detection data provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a cross-cut sheet material production database provided in an embodiment of the present invention; Figure 3 A flow chart for constructing a collection of unqualified cross-cut sheets of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1 A company introduced a method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data provided by the present invention, hoping to determine the factors of unqualified quality and issue early warnings by analyzing all the data of cross-cut sheet materials. Figure 1 As shown, the specific implementation is as follows: First, a production database of transverse shear sheet materials is constructed based on all the production data of the transformer transverse shear sheet materials. The structure is as follows: Figure 2 shown.

[0019] Table 1. Order database example

[0020] Furthermore, the cross-cut sheet production database includes a core detection database, a detection equipment parameter database, a processing equipment parameter database, an order database, a material database and a defect case library; wherein the core detection database includes an image database, a dimension database, a burr database, and a deformation database, the image database stores images taken by the detection equipment; the dimension database stores the structural data of the cross-cut sheet, the burr database stores the burr distribution data of the cross-cut sheet, and the deformation database stores the thickness deformation data of the cross-cut sheet; the detection equipment parameter database is used to store the timing operation parameters of the detection equipment when each cross-cut sheet is detected, including speed measuring device parameters, frequency converter parameters and line scan camera parameters; the processing equipment parameter database is used to store the timing processing parameters when processing the cross-cut sheet, including the force-displacement curve, thermal parameters and mold state parameters of the shearing process; the order database is used to store order data, as shown in Table 1, which is an example of a partial order database; the material database is used to store the material data of the cross-cut sheet; the defect case library is used to store the data of cross-cut sheets of unqualified quality, as shown in Table 2, which is an example of a partial defect case library.

[0021] Table 2. Example of defect case library

[0022] Build a cross-cut sheet material production database that includes a core detection database, detection equipment parameter database, processing equipment parameter database, order database, material database and defect case database, fully integrate the data to form a complete production data chain, and provide rich data support for subsequent intelligent analysis.

[0023] If the inspection result of the cross-cut sheet material is unqualified, feature extraction is performed on the unqualified cross-cut sheet material to obtain unqualified sample features; the defect case library is retrieved according to the unqualified sample features and the primary associated search similarity to obtain the first cross-cut sheet material set; the core inspection database is retrieved according to the unqualified sample features and the secondary associated search conditions to obtain the second cross-cut sheet material set; and the unqualified cross-cut sheet material set is constructed according to the first cross-cut sheet material set and the second cross-cut sheet material set; Furthermore, the process of constructing the set of unqualified cross-cut pieces is as follows: Figure 3 As shown, the specific process includes: If the test result of the cross-cut sheet material is unqualified, the data of the unqualified cross-cut sheet material stored in the core test database is obtained and feature extraction is performed to obtain the unqualified sample features , including defect image features, process parameter hash values ​​and material features; query the unqualified cross-cut sheets in the defect case library according to the main association retrieval similarity to obtain the first cross-cut sheet set , the formula for the main correlation retrieval similarity is: ; in, Indicates unqualified samples and unqualified samples in the defect case library The main related retrieval similarity of Indicates the defect case library Unqualified samples The unqualified sample characteristics, The formula for calculating the modulus length is: represents the natural base, represents the time attenuation coefficient, Indicates the time difference between the current sample and the historical sample; According to the unqualified sample characteristics and secondary related search conditions, the data in the core test database is queried to obtain the second cross-cut sheet material set , the formula for the secondary associated search condition is: ; in, Indicates Cross-cut pieces The sample characteristics of Indicates unqualified samples The unqualified sample characteristics, represents the standard deviation of historical data, represents the allowed false detection probability, represents the logarithmic function; By extracting features from unqualified cross-cut sheets and matching similar samples from the defect case library using primary association retrieval similarity, it is possible to quickly and accurately locate the historical case that is most similar to the current unqualified sample, providing a clear reference for subsequent cause analysis; using secondary association retrieval conditions to query all cross-cut sheet data, construct a second cross-cut sheet set, and form comprehensive data coverage.

[0024] Furthermore, the construction formula for constructing the set of unqualified cross-cut sheets is: ; in, Indicates a collection of unqualified cross-cut pieces. Represents the first cross-cut sheet set, Indicates the second cross-cut sheet set, Represents the elements in the set of unqualified cross-cut pieces. represents the intersection operation, represents the union operation, Indicates retaining the top cross-parameter combinations, Represents the Cartesian product operation, which describes all possible combinations of the first set of cross-cut sheets and the second set of cross-cut sheets, and is used to explore potential correlation patterns across parameter combinations.

[0025] By performing intersection, union and Cartesian product operations on the first set of cross-cut sheets and the second set of cross-cut sheets, and retaining the part with the highest similarity product among all possible combinations, comprehensive data capture is achieved, key samples are accurately screened, and data support is provided for subsequent analysis of factors that lead to unqualified cross-cut sheets, providing support for data-driven quality analysis and management.

[0026] Obtain unqualified cross-cut sheets from the unqualified cross-cut sheet set and obtain unqualified factor feature vectors according to the causal strength index, and further obtain an unqualified factor feature vector set; process the unqualified factor feature vector set according to the effective frequency to obtain the main factors of the unqualified cross-cut sheets; Furthermore, the calculation formula of the causal strength index is: ; in, Indicates an event Leading to the event The causal strength index, represents the time decay factor, emphasizing the importance of recent events, Indicates an event The number of independent occurrences, Indicates an event The number of independent occurrences, represents conditional entropy, which measures the uncertainty of causal relationships. Indicates an event expectations, Indicates an event expectations, represents the logarithmic function with base 2, Represents the total event sample size. The formula is valid only when it is greater than 1000. Represents the time step index, which is used to measure the relative time interval between events and the time decay factor Combined, calculate the weight decay of historical events on current causal relationships, Indicates the total time step, limits the effective time range of causal relationships, and avoids infinite tracing of history. Indicates an event In the The expectation of the second Indicates an event In the The expectation of the second Represents transfer entropy, calculating events To the event The information flow is: ; in, Indicates an event The Shannon entropy of Indicates an event The conditional entropy of .

[0027] The causal strength index integrates the time decay factor and transfer entropy, comprehensively considers the uncertainty within the system and the information flow between events, accurately characterizes and quantifies the causal strength between events, reduces the ambiguity in causal analysis, achieves accurate matching and data quantification of unqualified cross-cut sheets and the factors that cause them to be unqualified, and quickly determines the specific scope of the unqualified factors.

[0028] Further, unqualified cross-cut sheets are obtained The specific steps of the disqualification factors include: The effective frequency of each factor in the unqualified factor feature vector set is counted and sorted according to the number of occurrences. Factors exceeding the threshold are judged as the main factors causing the unqualified cross-cut sheet. The calculation formula for the effective frequency is: ; in, Indicates The effective frequency of a factor, Indicates The number of times a factor appears in the set, Indicates the total number of times, Expressing the The sum of the causal strength indicators related to the factors, represents the weight coefficient of frequency, represents the weight coefficient of the timeliness correction term, represents the natural base, represents the time attenuation coefficient, Indicates the time step index. Table 3 shows the unqualified factors and their effective frequencies in this query. If the effective frequency is greater than 0.7, the unqualified factor is identified as the main factor of unqualified cross-cut sheet material.

[0029] Table 3. Some unqualified factors and their effective frequencies

[0030] By constructing effective frequencies based on the number of occurrences, causal strength indicators and timeliness correction items, and sorting each unqualified factor according to the effective frequency, the impact of each unqualified factor on the cross-cut sheet material can be accurately quantified. The main unqualified factors can be identified through thresholds, which helps to prioritize and resolve the factors that have the greatest impact on the quality of cross-cut materials and improve the efficiency of quality improvement.

[0031] Collect real-time data of the steps or processes where the unqualified factors are located and compare and analyze them with standard data. If problems still exist, issue an early warning.

[0032] Furthermore, real-time data of the steps or projects corresponding to the unqualified factors are collected and a real-time data vector is constructed. The real-time data vector and the standard data vector are calculated, and the Euclidean distance is calculated. If the threshold is exceeded, it is determined that unqualified factors still exist in the production process and an early warning is issued.

[0033] By collecting and analyzing real-time data on the steps or locations where the unqualified factors are located, a closed loop is constructed to trace back the factors that caused the unqualified cross-cut sheets. This can achieve real-time detection and early warning of key parameters in the production process, and realize analysis and optimization of the production links.

[0034] Through the intelligent collection, analysis and processing method for transformer cross-cut sheet material detection data provided by the present invention, it is possible to achieve comprehensive integration and intelligent analysis of the full-process production data of cross-cut sheets, and then accurately extract the characteristics of unqualified samples, and construct a collection of unqualified cross-cut sheets through the main association retrieval similarity formula and the secondary association retrieval conditions, and use the causal strength index and effective frequency analysis to clarify the main factors affecting product quality. Finally, the data of the key steps corresponding to the main factors are collected in real time and compared with the standard data, which can timely discover anomalies in the production process and issue early warnings, thereby providing data-driven quality monitoring and continuous improvement support for enterprises.

[0035] Embodiment 2 A certain company applied a method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data provided by the present invention to improve the query and analysis capabilities of cross-cut sheet material detection data and optimize the production process. The specific implementation methods are as follows: Construct a production database of transverse shear sheets based on all production data of transformer transverse shear sheets; Furthermore, the cross-cut sheet production database includes a core detection database, a detection equipment parameter database, a processing equipment parameter database, an order database, a material database and a defect case library; wherein the core detection database includes an image database, a dimension database, a burr database and a deformation database, the image database stores images taken by the detection equipment, the dimension database stores the structural data of the cross-cut sheet, the burr database stores the burr distribution data of the cross-cut sheet, and the deformation database stores the thickness deformation data of the cross-cut sheet; the detection equipment parameter database is used to store the timing operation parameters of the detection equipment when detecting the cross-cut sheet, including speed measuring device parameters, frequency converter parameters and line scan camera parameters; the processing equipment parameter database is used to store the timing processing parameters when processing the cross-cut sheet, including the force-displacement curve, thermal parameters and mold state parameters of the shearing process; the order database is used to store order data; the material database is used to store the material data of the cross-cut sheet; the defect case library is used to store data of cross-cut sheets of unqualified quality.

[0036] If the inspection result of the cross-cut sheet material is unqualified, feature extraction is performed on the unqualified cross-cut sheet material to obtain unqualified sample features; the defect case library is retrieved according to the unqualified sample features and the primary associated search similarity to obtain the first cross-cut sheet material set; the core inspection database is retrieved according to the unqualified sample features and the secondary associated search conditions to obtain the second cross-cut sheet material set; and the unqualified cross-cut sheet material set is constructed according to the first cross-cut sheet material set and the second cross-cut sheet material set; Furthermore, the unqualified sample characteristics Including defect image features, process parameter hash values ​​and material features; query the unqualified cross-cut sheets in the defect case library based on the main association retrieval similarity to obtain the first cross-cut sheet set , the formula for the main correlation retrieval similarity is: ; in, Indicates unqualified samples and unqualified samples in the defect case library The main related retrieval similarity of Indicates the defect case library Unqualified samples The unqualified sample characteristics, The formula for calculating the modulus length is: represents the natural base, represents the time attenuation coefficient, Indicates the time difference between the current sample and the historical sample; Query the core inspection database according to the secondary associated search conditions to obtain the second cross-cut sheet material set , the formula for the secondary associated search condition is: ,in Indicates Cross-cut pieces The sample characteristics of represents the standard deviation of historical data, represents the probability of false detection, Represents a logarithmic function.

[0037] Furthermore, the construction formula for constructing the set of unqualified cross-cut sheets is: ; in, Indicates a collection of unqualified cross-cut pieces. Represents the first cross-cut sheet set, Indicates the second cross-cut sheet set, Represents the elements in the set of unqualified cross-cut pieces. represents the intersection operation, represents the union operation, Indicates that the top similarity product is retained A combination of Represents the Cartesian product operation.

[0038] Obtain unqualified cross-cut sheets in the unqualified cross-cut sheet set and obtain unqualified factor feature vectors according to the causal strength index, and further obtain an unqualified factor feature vector set; process the elements in the unqualified factor feature vector set according to the effective frequency to obtain the main factors of the unqualified cross-cut sheets; Furthermore, the calculation formula of the causal strength index is: ; in, Indicates an event Leading to the event The causal strength index, represents the time decay factor, Indicates an event The number of independent occurrences, Indicates an event The number of independent occurrences, represents the conditional entropy, Indicates an event expectations, Indicates an event expectations, represents the logarithmic function with base 2, represents the total event sample size, Represents the time step index and the time decay factor Combined, calculate the weight decay of historical events on current causal relationships, represents the total time step, represents the transfer entropy, Indicates an event In the The expectation of the second Indicates an event In the times expectation.

[0039] Furthermore, the specific steps of obtaining the main factors of unqualified cross-cut sheet materials include: The effective frequency of each factor in the unqualified factor feature vector set is counted and sorted according to the number of occurrences. Factors exceeding the threshold are judged as the main factors causing the unqualified cross-cut sheet. The calculation formula for the effective frequency is: ; in, Indicates The effective frequency of a factor, Indicates The number of times a factor appears in the set, Indicates the total number of times, Expressing the The sum of the causal strength indicators related to the factors, represents the weight coefficient of frequency, represents the weight coefficient of the timeliness correction term, represents the natural base, represents the time attenuation coefficient, Represents the time step index.

[0040] Collect real-time data of the steps where the main factors are located and compare and analyze them with standard data.

[0041] Furthermore, the real-time data of the steps corresponding to the main factors are collected and the real-time data vector is constructed. The real-time data vector and the standard data vector are calculated, and the Euclidean distance is calculated. If the threshold is exceeded, it is judged that there are still unqualified factors in the production process, and an early warning is issued. As shown in Table 4, the real-time data monitoring results of the key steps corresponding to the main factors can be seen. There may be problems in the shearing link, and the early warning of the corresponding steps is issued in time.

[0042] Table 4. Real-time data monitoring results

[0043] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data, characterized in that: include: Construct a production database of transverse shear sheets based on all production data of transformer transverse shear sheets; If the test result of the cross-cut sheet material is unqualified, feature extraction is performed on the unqualified cross-cut sheet material to obtain unqualified sample features; According to the unqualified sample characteristics and the main associated search similarity, the defect case library is retrieved to obtain the first cross-cut sheet material set; according to the unqualified sample characteristics and the secondary associated search conditions, the core inspection database is retrieved to obtain the second cross-cut sheet material set; and the unqualified cross-cut sheet material set is constructed based on the first cross-cut sheet material set and the second cross-cut sheet material set; Obtain unqualified cross-cut sheets in the unqualified cross-cut sheet set and obtain unqualified factor feature vectors according to the causal strength index, and further obtain an unqualified factor feature vector set; process the elements in the unqualified factor feature vector set according to the effective frequency to obtain the main factors of the unqualified cross-cut sheets; Collect real-time data of the steps where the main factors are located and compare and analyze them with standard data.

2. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 1 is characterized in that: The cross-cut sheet production database includes a core detection database, a detection equipment parameter database, a processing equipment parameter database, an order database, a material database and a defect case library; wherein the core detection database includes an image database, a dimension database, a burr database and a deformation database, the image database stores images taken by the detection equipment, the dimension database stores the structural data of the cross-cut sheet, the burr database stores the burr distribution data of the cross-cut sheet, and the deformation database stores the thickness deformation data of the cross-cut sheet; the detection equipment parameter database is used to store the timing operation parameters of the detection equipment when detecting the cross-cut sheet, including the speed measuring device parameters, the frequency converter parameters and the line scan camera parameters; the processing equipment parameter database is used to store the timing processing parameters when processing the cross-cut sheet, including the force-displacement curve, thermal parameters and mold state parameters of the shearing process; the order database is used to store order data; the material database is used to store the material data of the cross-cut sheet; the defect case library is used to store the data of the cross-cut sheet with unqualified quality.

3. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 1 is characterized in that: Unqualified sample characteristics Including defect image features, process parameter hash values ​​and material features; query the unqualified cross-cut sheets in the defect case library based on the main association retrieval similarity to obtain the first cross-cut sheet set , the formula for the main correlation retrieval similarity is: ； in, Indicates unqualified samples and unqualified samples in the defect case library The main related retrieval similarity of Indicates the defect case library Unqualified samples The unqualified sample characteristics, The formula for calculating the modulus length is: represents the natural base, represents the time attenuation coefficient, Indicates the time difference between the current sample and the historical sample; Query the core inspection database according to the secondary associated search conditions to obtain the second cross-cut sheet material set , the formula for the secondary associated search condition is: ,in Indicates Cross-cut pieces The sample characteristics of represents the standard deviation of historical data, represents the probability of false detection, Represents a logarithmic function.

4. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 3 is characterized in that: The construction formula for constructing the set of unqualified cross-cut pieces is: ； in, Indicates a collection of unqualified cross-cut pieces. Indicates the first cross-cut sheet set, Indicates the second cross-cut sheet set, Represents the elements in the set of unqualified cross-cut pieces. represents the intersection operation, represents the union operation, Indicates retaining the top A combination of Represents the Cartesian product operation.

5. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 1 is characterized in that: The calculation formula of the causal strength index is: ； in, Indicates an event Leading to the event The causal strength index, represents the time decay factor, Indicates an event The number of independent occurrences, Indicates an event The number of independent occurrences, represents the conditional entropy, Indicates an event expectations, Indicates an event expectations, represents the logarithmic function with base 2, represents the total event sample size, Represents the time step index and the time decay factor Combined, calculate the weight decay of historical events on current causal relationships, represents the total time step, represents the transfer entropy, Indicates an event In the The expectation of the second Indicates an event In the times expectation.

6. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 1 is characterized in that: The specific steps of the main factors for obtaining unqualified cross-cut sheets include: The effective frequency of each factor in the unqualified factor feature vector set is counted and sorted according to the number of occurrences. Factors exceeding the threshold are judged as the main factors causing the unqualified cross-cut sheet. The calculation formula for the effective frequency is: ； in, Indicates The effective frequency of a factor, Indicates The number of times a factor appears in the set, Indicates the total number of times, Expressing the The sum of the causal strength indicators related to the factors, represents the frequency weight coefficient, represents the weight coefficient of the timeliness correction term, represents the natural base, represents the time attenuation coefficient, Represents the time step index.

7. The method for intelligent collection, analysis and processing of transformer cross-cut sheet material detection data according to claim 1 is characterized in that: Collect real-time data of the steps corresponding to the main factors and construct real-time data vectors, calculate the real-time data vectors and standard data vectors, and calculate the Euclidean distance. If it exceeds the threshold, it is judged that there are still unqualified factors in the production process and an early warning is issued.

Images