Array pulsed eddy current based equipment inspection method and system

Abstract

The application provides an array pulse eddy current-based equipment flaw detection method and system. Adjacent sample data belong to different categories, and adjacent sample data have a distinguishing manner. When array pulse eddy current data is divided, a description field division manner and a description field category of the array pulse eddy current data are obtained to determine the distinguishing manner in the array pulse eddy current data and the sample data category corresponding to each data node. The description field of the array pulse eddy current data is enriched. In a plurality of data ranges obtained by dividing according to the distinguishing manner and the category corresponding to each data node, the categories corresponding to each two adjacent data ranges are different, the accuracy of the data range is improved, and therefore, flaw detection can be more accurately performed.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method and system for equipment flaw detection based on array pulse eddy currents. Background Technology

[0002] Pulsed eddy current technology is widely used in modern industrial production. It is a non-destructive testing method for detecting defects in objects. Due to its high sensitivity, accurate positioning, and non-contact nature, it is widely used in the field of defect detection of materials such as metals, non-metals, and composite materials.

[0003] Currently, in the process of using array pulse eddy current technology for equipment flaw detection, interference may occur, making it difficult to guarantee the accuracy of flaw detection. Therefore, there is an urgent need for a technical solution to improve the above-mentioned technical problems. Summary of the Invention

[0004] To address the technical problems existing in related technologies, this application provides a device flaw detection method and system based on array pulse eddy currents.

[0005] Firstly, a device flaw detection method based on array pulse eddy currents is provided. The method includes: extracting descriptive fields from array pulse eddy current data to obtain descriptive field information of the array pulse eddy current data. The descriptive field information includes a descriptive field division method and a descriptive field type. The descriptive field division method is used to indicate the distinction between different sample data in the array pulse eddy current data. The descriptive field type is used to indicate the type corresponding to each data node in the array pulse eddy current data, and the type is the type to which the sample data where the data node is located belongs. Based on the distinction method indicated by the descriptive field division method, the array pulse eddy current data is divided into several reference data ranges. Based on the descriptive field type, the several reference data ranges are optimized to obtain several target data ranges of the array pulse eddy current data, so that the sample data corresponding to each two adjacent target data ranges belong to different types. The existence of a defect in the device is determined by the difference in the types of the sample data.

[0006] In one standalone embodiment, dividing the array pulse eddy current data into several reference data ranges according to the differentiation method indicated by the description field division method includes: determining the location of multiple differentiation methods in the array pulse eddy current data according to the description field division method; and determining the range between every two adjacent differentiation methods as a reference data range according to the location of the multiple differentiation methods, thereby obtaining the several reference data ranges.

[0007] In one independently implemented embodiment, optimizing the plurality of reference data ranges based on the description field type to obtain a plurality of target data ranges of the array pulse eddy current data includes: determining the type corresponding to each data node in the array pulse eddy current data based on the description field type; determining the type corresponding to each reference data range based on the data nodes in each reference data range and the type corresponding to each data node; and optimizing the plurality of reference data ranges based on the types corresponding to the plurality of reference data ranges to obtain the plurality of target data ranges.

[0008] In one independently implemented embodiment, optimizing the plurality of reference data ranges based on the categories corresponding to the plurality of reference data ranges to obtain the plurality of target data ranges includes: merging any two adjacent reference data ranges that correspond to the same category to obtain a target data range; and determining the random reference data range as the target data range in response to a random reference data range that corresponds to a different category than the adjacent reference data ranges.

[0009] In one standalone embodiment, determining the category corresponding to each reference data range based on the data nodes in each reference data range and the category corresponding to each data node includes: determining the category corresponding to each reference data range and the confidence weight of the category based on the data nodes in each reference data range and the category corresponding to each data node.

[0010] In one independent embodiment, the description field information further includes an importance description field, which is used to indicate the importance of the array pulse eddy current data in terms of the differentiation method; the step of optimizing the plurality of reference data ranges according to the categories corresponding to the plurality of reference data ranges to obtain the plurality of target data ranges includes: in response to a random reference data range corresponding to a confidence weight being less than a reference confidence weight, determining at least two target importances located within the reference data range according to the importance description field; and dividing the reference data range according to the differentiation method constituted by the at least two importances to obtain the target data range.

[0011] In one standalone embodiment, determining the importance of at least two targets within the reference data range based on the importance description field includes: determining the importance of at least two targets within the reference data range based on several important locations in the importance description field and the location of the reference data range in the array pulse eddy current data.

[0012] In one independently implemented embodiment, determining the category corresponding to each reference data range based on the data nodes in each reference data range and the category corresponding to each data node includes: for a random reference data range, determining the category corresponding to each data node in the reference data range based on the data nodes in the reference data range and the category corresponding to each data node; and determining the category corresponding to the reference data range based on the category corresponding to each data node in the reference data range.

[0013] In one standalone embodiment, determining the type corresponding to the reference data range based on the type corresponding to each data node in the reference data range includes: determining the type with the largest number of corresponding data nodes as the type corresponding to the reference data range.

[0014] In one standalone embodiment, the step of extracting descriptive fields from the array pulse eddy current data to obtain descriptive field information of the array pulse eddy current data includes: calling a descriptive field extraction network to extract descriptive fields from the array pulse eddy current data to obtain the descriptive field information.

[0015] In one independently implemented embodiment, the description field extraction network includes a description field extraction sub-network, a data scaling sub-network, and a description field detection sub-network. The step of invoking the description field extraction network to extract description fields from the array pulse eddy current data to obtain the description field information includes: invoking the description field extraction sub-network to extract description fields from the array pulse eddy current data to obtain a first set of description fields for the array pulse eddy current data; invoking the data scaling sub-network to scale the first set of description fields to obtain a second set of description fields for the array pulse eddy current data; and invoking the description field detection sub-network to detect description fields in the second set of description fields to obtain the description field information.

[0016] In one standalone embodiment, the step of invoking the data scale reorganization subnetwork to perform data scale reorganization on the first description field set to obtain the second description field set of the array pulse eddy current data includes: invoking the data scale reorganization subnetwork to perform data scale reorganization processing on the first description field set to obtain a plurality of reference description field sets of data scales corresponding to the first description field set; and concatenating the plurality of reference description field sets of data scales to obtain the second description field set of the array pulse eddy current data.

[0017] Secondly, a device flaw detection system based on array pulsed eddy currents is provided, including a processor and a memory that communicate with each other. The processor is used to read a computer program from the memory and execute it to implement the above-described method.

[0018] The device flaw detection method and system based on array pulse eddy current provided in this application have adjacent sample data of different types and different ways of distinguishing between them. When dividing the array pulse eddy current data, the method of dividing the array pulse eddy current data and the type of the description field are obtained to determine the distinguishing method in the array pulse eddy current data and the type of sample data corresponding to each data node. This enriches the description field of the array pulse eddy current data and makes each pair of adjacent data ranges have different types in the several data ranges divided according to the distinguishing method and the type corresponding to each data node. This improves the accuracy of the data range and enables more accurate flaw detection. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a device flaw detection method based on array pulsed eddy currents provided in an embodiment of this application. Detailed Implementation

[0021] To better understand the above technical solutions, the technical solutions of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this application and the specific features in the embodiments are detailed descriptions of the technical solutions of this application, rather than limitations on the technical solutions of this application. In the absence of conflict, the embodiments of this application and the technical features in the embodiments can be combined with each other.

[0022] Please see Figure 1 This paper presents a device flaw detection method based on array pulse eddy currents, which may include the technical solutions described in steps 201-203.

[0023] 201. Extract the descriptive fields from the array pulse eddy current data to obtain the descriptive field information of the array pulse eddy current data. The descriptive field information includes the descriptive field division method and the descriptive field type.

[0024] The description field partitioning method is used to indicate the way different sample data are distinguished in the array pulse eddy current data, and the description field type is used to indicate the type corresponding to each data node in the array pulse eddy current data. The type is the type to which the sample data of the data node belongs.

[0025] 202. Based on the differentiation method indicated by the description field, divide the array pulse eddy current data into several reference data ranges.

[0026] Adjacent sample data have a distinguishing method. By describing the distinguishing method indicated by the field division method, the array pulse eddy current data is divided into several reference data ranges, each of which may include one sample data.

[0027] 203. Based on the description field type, optimize several reference data ranges to obtain several target data ranges of array pulse eddy current data, so that the sample data corresponding to each two adjacent target data ranges belong to different types, and determine the existence of defects in the equipment by the different types of the sample data.

[0028] If two adjacent sample data belong to different categories, the categories of sample data for each data node are specified in the description field. Several reference data ranges are then optimized so that the sample data for each adjacent data range belongs to different categories.

[0029] The method provided in this application embodiment involves adjacent sample data belonging to different categories and having a distinguishing method between them. When dividing array pulse eddy current data, the method of dividing the array pulse eddy current data and the category of the description field are obtained to determine the distinguishing method in the array pulse eddy current data and the category of sample data corresponding to each data node. This enriches the description field of the array pulse eddy current data, so that in the several data ranges divided according to the distinguishing method and the category corresponding to each data node, every two adjacent data ranges correspond to different categories, which can improve the accuracy of the data range. In this way, more accurate flaw detection can be performed.

[0030] In one possible implementation, the method includes the following steps.

[0031] 301. Call the description field extraction network to extract description fields from the array pulse eddy current data and obtain description field information, which includes the description field division method and description field type.

[0032] For example, the array pulsed eddy current data is obtained through a signal receiver. A descriptive field extraction network is used to extract descriptive field information from the array pulsed eddy current data. This descriptive field information describes the descriptive fields of the sample data within the array pulsed eddy current data. The descriptive field partitioning method indicates how different sample data in the array pulsed eddy current data are distinguished; each partitioning method is the intersection line between two adjacent sample data. The descriptive field category indicates the category corresponding to each data node in the array pulsed eddy current data; this category is the category to which the sample data containing the data node belongs.

[0033] It includes several sample data, and adjacent sample data have different distinguishing methods. The two adjacent sample data belong to different categories. Therefore, by extracting the descriptive field division method and descriptive field type from the array pulse eddy current data, the array pulse eddy current data can be divided according to the division descriptive field and descriptive field type.

[0034] In one possible implementation, the descriptive field partitioning method includes the probability that each data node in the array pulse eddy current data falls within the partitioning method. A higher probability indicates a greater likelihood that the corresponding data node falls within the partitioning method, and a lower probability indicates a less likely likelihood that the corresponding data node falls within the partitioning method.

[0035] The greater the probability that the sample data of a data node belongs to a random category, the greater the probability that the sample data of that data node belongs to that category; the smaller the probability that the sample data of a data node belongs to a random category, the smaller the probability that the sample data of that data node belongs to that category.

[0036] In one possible implementation, the description field extraction network includes a description field extraction sub-network, a data scale preparation sub-network, and a description field detection sub-network. In this case, step 301 includes the following steps 3011-3013.

[0037] 3011. Call the description field extraction sub-network to extract description fields from the array pulse eddy current data, and obtain the first description field set of the array pulse eddy current data.

[0038] The first description field set is used to describe the description fields of the corrugated plate in the array pulse eddy current data. The description field extraction sub-network is used to obtain the description field set of the array pulse eddy current data, which is used to describe the information of the sample data in the array pulse eddy current data. Optionally, the description field extraction sub-network is a convolutional network.

[0039] 3012. Call the data scale reorganization sub-network to perform data scale reorganization on the first description field set to obtain the second description field set of array pulse eddy current data.

[0040] The data scaling subnetwork is used to scale the descriptive field set of the array pulse eddy current data to obtain another descriptive field set for the array pulse eddy current data. This second descriptive field set has several descriptive fields corresponding to the data scales of the array pulse eddy current data. The data scale of the second descriptive field set obtained by data scaling may be the same as or different from that of the first descriptive field set.

[0041] By using a data scale reorganization subnetwork, the first description field set is reorganized to include description field information of several data scales in the second description field set, thereby enhancing the description fields of the sample data in the array pulse eddy current data and improving the accuracy of the description field set.

[0042] In one possible implementation, step 3012 includes: calling the data scale reorganization sub-network to perform data scale reorganization processing on the first description field set to obtain several reference description field sets of data scales corresponding to the first description field set; and concatenating the several reference description field sets of data scales to obtain the second description field set of the array pulse eddy current data.

[0043] Since the descriptive field information contained in the descriptive field sets of different data sizes may be different, the data size sorting subnetwork is used to obtain several reference descriptive field sets corresponding to the first descriptive field set. These reference descriptive field sets of several data sizes are then concatenated to make the second descriptive field set contain descriptive field information of several data sizes, thereby improving the accuracy of the descriptive field set.

[0044] 3013. Call the description field detection sub-network to perform description field detection on the second description field set and obtain description field information.

[0045] The description field detection subnetwork is used to detect description field information in the second description field set.

[0046] By using the description field extraction subnetwork and the data scale organization subnetwork, the ripple description field in the second description field set is enhanced. Then, by using the description field detection subnetwork to detect the description field information in the second description field set, accurate description field information can be obtained.

[0047] In one possible implementation, the description field detection subnetwork includes a differentiation method detection subnetwork and a category detection subnetwork; then step 3013 includes: calling the differentiation method detection subnetwork to perform differentiation method detection on the second description field set to obtain the description field partitioning method, and calling the category detection subnetwork to perform category detection on the second description field set to obtain the description field category.

[0048] The method of differentiation detection subnetwork is used to detect the classification method of description fields in the description field set, and the category detection subnetwork is used to detect the category of description fields in the description field set. By using the method of differentiation detection subnetwork and the category detection subnetwork, description field detection is performed on the second description field set to obtain the classification method and category of description fields in the description field set, thereby improving the accuracy of the classification method and category of description fields.

[0049] In one possible implementation, the description field information also includes important description fields, and the description field detection subnetwork also includes an important detection subnetwork; then step 3013 further includes: calling the important detection subnetwork to perform important detection on the second description field set to obtain important description fields.

[0050] The importance detection subnetwork is used to detect important descriptive fields in the descriptive field set. These important descriptive fields indicate the importance of the array pulse eddy current data in terms of differentiation. By obtaining important descriptive fields from the descriptive field set through the importance detection subnetwork, the array pulse eddy current data can be segmented and processed based on these important descriptive fields.

[0051] The array pulse eddy current data is extracted by two convolutional modules to obtain the first set of descriptive fields. The first set of descriptive fields is then resized by three high-resolution networks to obtain the second set of descriptive fields. Subsequently, the second set of descriptive fields is detected by an important detection subnetwork, a differentiation method detection subnetwork, and a category detection subnetwork to obtain the important descriptive fields, the descriptive field classification method, and the descriptive field category, respectively.

[0052] Optionally, the important detection sub-network is invoked to perform important detection on the second description field set to obtain the important initial description field. The data nodes in the important initial description field are then selected to obtain the important description field.

[0053] The important initial description field includes the confidence weight corresponding to each data node in the array pulse eddy current data, and also includes the confidence weight corresponding to some data nodes in the array pulse eddy current data. By selecting data nodes in the important initial description field, the accuracy of the important description field is improved.

[0054] There are three ways to select and process data nodes in important initial description fields.

[0055] The first method: Based on the probability that each data node in the important initial description field is important, select the data nodes with a probability greater than the reference probability to generate the important description field.

[0056] The reference probability can be any value, such as 0.4 or 0.45. This initial importance description field includes the probability that each data node is important; this probability indicates the likelihood that the corresponding data node is important. By using this initial importance description field, data nodes with a high probability are selected as important, thereby improving the accuracy of the importance description field.

[0057] The second method is to select several data nodes in the important initial description field based on the location of each data node in the important initial description field, so that the difference between any two data nodes in the selected data nodes is greater than the reference difference, and generate the important description field based on the selected data nodes.

[0058] The third method: Based on the probability that each data node in the important initial description field is important, select data nodes with a probability greater than the reference probability to generate an important reference description field. Based on the location of each data node in the important reference description field, select several data nodes in the important reference description field to make the difference between any two data nodes in the selected data nodes greater than the reference difference. Based on the selected data nodes, generate the important description field.

[0059] This important description field includes the probability corresponding to each selected data node and the location corresponding to each data node.

[0060] Optionally, when selecting several data nodes in the important reference description field, the difference between each pair of data nodes is determined based on the location of the data nodes. In response to the fact that the difference between any two data nodes is less than the reference difference, the data node with the lower probability between the two data nodes is deleted. The important description field is generated based on the probability and location of the remaining data nodes.

[0061] 302. Based on the differentiation method indicated by the description field, divide the array pulse eddy current data into several reference data ranges.

[0062] In the embodiments of this application, there is a distinction between each pair of adjacent sample data. Therefore, the array pulse eddy current data is divided by the distinction indicated by the description field division method to obtain several reference data ranges. Each reference data range may represent a sample data.

[0063] In one possible implementation, step 302 includes: determining the location of multiple distinction methods in the array pulse eddy current data according to the description field division method; and determining the range between every two adjacent distinction methods as a reference data range according to the location of the multiple distinction methods, thereby obtaining several reference data ranges.

[0064] The location of multiple distinction methods in the array pulse eddy current data is determined by describing the field division method. According to the location of multiple distinction methods, the range between each pair of adjacent distinction methods is determined as a reference data range, thereby obtaining several reference data ranges.

[0065] In one possible implementation, the descriptive field partitioning method includes the possibility that each data node in the array pulse eddy current data is located in the differentiation method. Then, the possibility that each data node in the descriptive field partitioning method is located in the differentiation method is transformed to obtain multiple differentiation methods in the array pulse eddy current data.

[0066] Optionally, the probability of each data node in the partitioning method of the description field being located on the differentiation method is transformed to obtain several differentiation method segments. In response to the fact that the difference between the two ends of any two differentiation method segments is less than the reference difference, the two differentiation method segments are connected to obtain a differentiation method.

[0067] 303. Based on the description field type, determine the type corresponding to each data node in the array pulse eddy current data.

[0068] In one possible implementation, the description field category includes the confidence weight of the sample data of each data node in the array pulse eddy current data belonging to each category; step 303 includes: for a random data node, based on the confidence weight of the sample data of the data node belonging to each category, determining the category corresponding to the largest confidence weight as the category to which the sample data of the data node belongs.

[0069] The confidence weight for a particular category represents the probability that the sample data containing the data node belongs to that category. A higher confidence weight indicates a greater probability that the sample data containing the data node belongs to that category, while a lower confidence weight indicates a lower probability that the sample data containing the data node belongs to that category. Optionally, the confidence weight can be represented by probability.

[0070] Since the array pulse eddy current data includes sample data of several types, the category description field can represent the confidence weight of the sample data of a random data node belonging to each category. Based on these confidence weights, the category corresponding to the data node can be determined.

[0071] 304. Based on the data nodes in each reference data range and the type corresponding to each data node, determine the type corresponding to each reference data range.

[0072] For a random reference data range corresponding to array pulse eddy current data, the type of each data node in the reference data range can be determined, and the type of the reference data range can be determined based on the type of each reference data range.

[0073] In one possible implementation, for a random reference data range, each data node belonging to the reference data range is determined based on the coordinate information of the reference data range, the category corresponding to each data node in the reference data range is determined, and the category corresponding to the reference data range is determined based on the category corresponding to each data node.

[0074] The coordinate information of the reference data range is used to indicate the location of the reference data range. Optionally, the coordinate information of the reference data range includes the coordinates of two distinction methods that constitute the reference data range. These two distinction methods are two adjacent distinction methods among multiple distinction methods, and these two distinction methods constitute the reference data range.

[0075] In one possible implementation, step 304 includes: determining the type and the confidence weight of each type corresponding to each reference data range based on the data nodes in each reference data range and the type corresponding to each data node in the array pulse eddy current data.

[0076] The confidence weight of a category is used to represent the probability that the reference data range is sample data of that category. The higher the confidence weight of a category, the greater the probability that the reference data range is sample data of that category, and the lower the confidence weight of a category, the less likely the reference data range is sample data of that category.

[0077] By identifying the types corresponding to the data nodes within a reference data range, the types corresponding to that reference data range can be determined. However, since different data nodes within the reference data range may correspond to different types, meaning that the types corresponding to each data node within the reference data range may not be completely identical, the types corresponding to that reference data range may be inaccurate. To accurately reflect whether the types corresponding to that reference data range are accurate or not, a confidence weight for that type is used to represent its accuracy.

[0078] In one possible implementation, step 304 includes the following steps 3041-3044.

[0079] 3041. For a random reference data range, determine the type of each data node in the reference data range based on the data nodes in the reference data range and the type of each data node in the array pulse eddy current data.

[0080] By describing the field type, we can obtain the type corresponding to each data node in the array pulse eddy current data. By referring to the data range, we can determine the data nodes located in the reference data range, and thus determine the type corresponding to each data node in the reference data range.

[0081] 3042. Determine the type corresponding to the reference data range based on the type corresponding to each data node in the reference data range.

[0082] After determining the category corresponding to each data node in the reference data range, the category corresponding to the reference data range can be determined by the category corresponding to each data node.

[0083] In one possible implementation, step 3042 includes: determining the category with the largest number of corresponding data nodes as the category corresponding to the reference data range, based on the category corresponding to each data node in the reference data range.

[0084] Within this reference data range, each data node may correspond to a different category. Therefore, by determining the number of data nodes corresponding to each category, the category with the largest number of corresponding data nodes is selected as the category corresponding to this reference data range.

[0085] 3043. Obtain the number of data nodes in the reference data range whose type is the same as the type corresponding to the reference data range.

[0086] By identifying the category corresponding to each data node in the reference data range, the number of data nodes whose categories are the same as those corresponding to the reference data range can be determined.

[0087] 3044. The ratio between the number of data nodes and the total number of data nodes in the reference data range is used as the confidence weight of the category corresponding to the reference data range.

[0088] The confidence weight represents the accuracy of the category corresponding to the reference data range. The larger the confidence weight, the higher the accuracy, and the smaller the confidence weight, the lower the accuracy.

[0089] 305. Based on the categories corresponding to several reference data ranges, optimize several reference data ranges to obtain several target data ranges.

[0090] Since the distinction method indicated by the description field division method may result in inaccurate data ranges among the several reference data ranges, in order to improve the accuracy of the data ranges, the several reference data ranges are optimized by the categories corresponding to the several reference data ranges, so that the sample data corresponding to each two adjacent target data ranges belong to different categories. The difference in the categories of the sample data is used to determine the defects in the equipment, thereby improving the accuracy of the target data ranges.

[0091] In one possible implementation, step 305 includes: merging any two adjacent reference data ranges to obtain a target data range in response to the fact that any two adjacent reference data ranges correspond to the same category; and determining a random reference data range as the target data range in response to the fact that a random reference data range corresponds to a different category than the adjacent reference data ranges.

[0092] In one possible implementation, the description field information also includes an important description field, which is used to indicate the importance of the array pulse eddy current data in terms of differentiation; step 305 includes the following steps 3051-3053.

[0093] 3051. In response to a random reference data range having a confidence weight less than a reference confidence weight, determine that at least two targets within the reference data range are important based on the importance description field.

[0094] In this embodiment of the application, if the confidence weight corresponding to any reference data range is less than the reference confidence weight, it indicates that the accuracy of the category corresponding to the reference data range is low. In this case, the reference data range needs to be optimized. Therefore, at least two targets within the reference data range are determined to be important so that the reference data range can be optimized based on the importance of the at least two targets.

[0095] In one possible implementation, step 3051 includes: determining at least two target importances located within the reference data range based on several important locations in the important description field and the location of the reference data range in the array pulse eddy current data.

[0096] Optionally, the importance description field is used to indicate the importance and the confidence weight of importance in the array pulse eddy current data in terms of the differentiation method. Then, based on the location of several important points in the importance description field and the location of the reference data range in the array pulse eddy current data, several reference importances located within the reference data range are determined. Based on the confidence weight of several reference importances in the importance description field, at least two target importances are selected from several reference importances.

[0097] The importance of the target lies in the significance of the differentiating methods that may exist within the reference data range.

[0098] 3052. Based on at least two important distinguishing methods, divide the reference data range to obtain the target data range.

[0099] By using at least two important elements, at least one distinction method is formed, and the reference data range is divided using the at least one distinction method to obtain at least two target data ranges.

[0100] In one possible implementation, step 3052 includes: dividing the reference data range according to at least two important constituent distinction methods to obtain at least two divided data ranges; determining the type and confidence weight of each divided data range; in response to a random divided data range having a confidence weight greater than the reference confidence weight, taking the divided data range as a target data range; and in response to a random divided data range having a confidence weight less than the reference confidence weight, repeating the above steps to continue dividing the divided data range.

[0101] 3053. In response to a random reference data range having a higher confidence weight than the reference confidence weight, a random reference data range is selected as the target data range.

[0102] If the confidence weight of a random reference data range is greater than the reference confidence weight, it means that the category corresponding to the reference data range is accurate, and the reference data range can be used as a target data range.

[0103] It is understood that the embodiments of this application are described by optimizing several reference data ranges by determining the type of each data node and the type of each reference data range. However, in another embodiment, it is not necessary to perform steps 303-305. Other methods can be adopted to optimize several reference data ranges according to the type of the description field to obtain several target data ranges of array pulse eddy current data.

[0104] It is understood that the embodiments of this application are only described by dividing the array pulse eddy current data to obtain several data ranges of the array pulse eddy current data. In another embodiment, the point cloud data of the array pulse eddy current data can also be extracted, and the point cloud data of the array pulse eddy current data can be analyzed by a description field extraction network to obtain the description field information of the array pulse eddy current data, thereby obtaining several data ranges of the array pulse eddy current data.

[0105] The method provided in this application utilizes a partitioning technique to locate the data range of the acquired array pulse eddy current data, improving the accuracy and robustness of the partitioning. Furthermore, the description field extraction network includes a high-resolution network, a differentiation detection subnetwork, a category detection subnetwork, and an importance detection subnetwork. The description field extraction network is trained using a multi-task learning approach, enabling it to accurately output important description fields, description field partitioning methods, and description field categories in the array pulse eddy current data. It also leverages the synergistic effect among the three description fields, combining important description fields, description field partitioning methods, and description field categories to avoid range location errors caused by using only the description field partitioning method, thus improving the accuracy of the data range.

[0106] The method provided in this application embodiment involves adjacent sample data belonging to different categories and having a distinguishing method between them. When dividing array pulse eddy current data, the method of dividing the array pulse eddy current data and the category of the description field are obtained to determine the distinguishing method in the array pulse eddy current data and the category of sample data corresponding to each data node. This enriches the description field of the array pulse eddy current data, so that in the several data ranges divided according to the distinguishing method and the category corresponding to each data node, every two adjacent data ranges correspond to different categories, which can improve the accuracy of the data range. In this way, more accurate flaw detection can be performed.

[0107] Furthermore, by dividing the array pulse eddy current data according to the division method, types, and important descriptive fields, the descriptive fields in the array pulse eddy current data are enriched, which can improve the accuracy of the data range. In this way, more accurate flaw detection can be performed.

[0108] Based on the above embodiments, the process includes...

[0109] 1. The corrugated plate is photographed by a line scan camera to obtain array pulse eddy current data.

[0110] 2. Input the obtained array pulse eddy current data into the description field extraction network. Obtain important description fields by performing importance detection on the array pulse eddy current data; obtain the description field division method by performing differentiation detection on the array pulse eddy current data; obtain the description field types of the array pulse eddy current data by performing type detection on the array pulse eddy current data.

[0111] 3. By describing the multiple differentiation methods indicated by the field division method, the array pulse eddy current data is divided into several reference data ranges. Based on the important description field and the type of the description field, the reference data ranges are optimized to obtain several target data ranges in the array pulse eddy current data and the type of each target data range.

[0112] It should be noted that the embodiments in this application only illustrate the use of a differentiation detection subnetwork, a category detection subnetwork, and an importance detection subnetwork. Before invoking these subnetworks, they need to be trained. In this embodiment, a multi-task approach is used to simultaneously train the differentiation detection subnetwork, category detection subnetwork, and importance detection subnetwork, so that these three subnetworks mutually promote each other, thereby obtaining a correlated differentiation detection subnetwork, category detection subnetwork, and importance detection subnetwork. The training process for these differentiation detection subnetworks, category detection subnetwork, and importance detection subnetwork is as follows.

[0113] 1. Obtain the sample description field set of the pulse eddy current data of the sample array, as well as the sample description field division method, sample description field type, and important sample description fields corresponding to the pulse eddy current data of the sample array.

[0114] Among them, the sample description field division method, sample description field type and sample importance description field are all obtained by labeling the sample array pulse eddy current data. The sample description field division method is used to indicate the description field of the true distinction method in the sample array pulse eddy current data. The sample description field type is used to indicate the true type corresponding to each data node in the sample array pulse eddy current data. The sample importance description field is used to indicate the true importance in the sample array pulse eddy current data.

[0115] 2. Call the discrimination method detection sub-network to perform discrimination method detection on the sample description field set to obtain the predicted description field partitioning method. Based on the predicted description field partitioning method and the sample description field partitioning method, train the discrimination method detection sub-network.

[0116] In one possible implementation, when training the discrimination detection subnetwork, a first loss value for the discrimination detection subnetwork is determined based on the prediction description field partitioning method and the sample description field partitioning method, and the discrimination detection subnetwork is trained based on the first loss value.

[0117] 3. Call the category detection subnetwork to perform category detection on the sample description field set, obtain the predicted description field category, and train the category detection subnetwork based on the predicted description field category and the sample description field category.

[0118] In one possible implementation, when training the category detection subnetwork, a second loss value for the category detection subnetwork is determined based on the predicted description field category and the sample description field category, and the category detection subnetwork is trained based on the second loss value.

[0119] 4. Call the importance detection subnetwork to perform importance detection on the second description field set of the sample to obtain the predicted important description field. Based on the predicted important description field and the important description field of the sample, train the importance detection subnetwork.

[0120] In one possible implementation, when training the importance detection subnetwork, a third loss value for the importance detection subnetwork is determined based on the predicted importance description field and the sample importance description field, and the importance detection subnetwork is trained based on the third loss value.

[0121] In the embodiments of this application, during the training of the differentiation method detection subnetwork, the category detection subnetwork and the importance detection subnetwork, the importance, differentiation method and sample data in the array pulse eddy current data are used simultaneously to train the differentiation method detection subnetwork, the category detection subnetwork and the importance detection subnetwork, which improves the robustness of the description field extraction network and avoids the performance instability of the description field extraction network caused by using a single piece of information.

[0122] Based on the above, a device flaw detection apparatus based on arrayed pulsed eddy currents is provided, the apparatus comprising:

[0123] The information acquisition module is used to extract descriptive fields from the array pulse eddy current data to obtain descriptive field information of the array pulse eddy current data. The descriptive field information includes a descriptive field division method and a descriptive field type. The descriptive field division method is used to indicate the way different sample data in the array pulse eddy current data are distinguished. The descriptive field type is used to indicate the type corresponding to each data node in the array pulse eddy current data. The type is the type to which the sample data where the data node is located belongs.

[0124] The range division module is used to divide the array pulse eddy current data into several reference data ranges according to the distinction method indicated by the division method of the description field;

[0125] The defect determination module is used to optimize the several reference data ranges according to the description field type to obtain several target data ranges of the array pulse eddy current data, so that the sample data corresponding to each two adjacent target data ranges belong to different types, and the existence of defects in the equipment is determined by the different types of the sample data.

[0126] Based on the above, a device flaw detection system based on array pulsed eddy currents is shown, including a processor and a memory that communicate with each other. The processor is used to read a computer program from the memory and execute it to implement the above method.

[0127] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method during runtime.

[0128] In summary, based on the above scheme, since adjacent sample data belong to different categories and there is a way to distinguish between adjacent sample data, when dividing the array pulse eddy current data, by obtaining the division method and category of the description field of the array pulse eddy current data, the distinction method in the array pulse eddy current data and the sample data category corresponding to each data node are determined. This enriches the description field of the array pulse eddy current data, so that in the several data ranges divided according to the distinction method and the category corresponding to each data node, every two adjacent data ranges correspond to different categories, which can improve the accuracy of the data range. In this way, more accurate flaw detection can be performed.

[0129] It should be understood that the systems and modules described above can be implemented in various ways. For example, in some embodiments, the systems and modules can be implemented by hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the methods and systems described above can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and modules of this application can be implemented not only by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also by software executed by various types of processors, or by a combination of the aforementioned hardware circuits and software (e.g., firmware).

[0130] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects may be any one or a combination of the above, or any other possible beneficial effects.

[0131] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore remain within the spirit and scope of the exemplary embodiments of this application.

[0132] Furthermore, this application uses specific terms to describe embodiments of the application. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of the application. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different locations in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of the application can be appropriately combined.

[0133] Furthermore, those skilled in the art will understand that aspects of this application can be described and illustrated through several patentable types or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, aspects of this application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, aspects of this application may manifest as a computer product located on one or more computer-readable media, the product including computer-readable program code.

[0134] Computer storage media may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and suitable combinations thereof. Computer storage media can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer storage medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0135] The computer program code required for the operation of each part of this application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby, and Groovy, or other programming languages. This program code can run entirely on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as Software as a Service (SaaS).

[0136] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this application are not intended to limit the order of the processes and methods of this application. Although the foregoing disclosure has discussed some currently considered useful embodiments of the invention through various examples, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the substance and scope of the embodiments of this application. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely through software solutions, such as installing the described system on existing servers or mobile devices.

[0137] Similarly, it should be noted that, in order to simplify the description of the present application and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of the embodiments of the present application sometimes combines multiple features into a single embodiment, drawing, or description thereof. However, this disclosure method does not imply that the subject matter of the application requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of the single embodiments disclosed above.

[0138] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are open to adaptive variation. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters are taken into account a specified number of significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of application in some embodiments of this application are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0139] For each patent, patent application, patent application publication, and other material such as articles, books, specifications, publications, and documents referenced in this application, the entire contents of that patent are incorporated herein by reference. This excludes historical application documents that are inconsistent with or conflict with the content of this application, as well as documents that limit the broadest scope of the claims in this application (currently or subsequently appended to this application). It should be noted that if there are any inconsistencies or conflicts between the descriptions, definitions, and / or terminology used in the supplementary materials of this application and the content of this application, the descriptions, definitions, and / or terminology used in this application shall prevail.

[0140] Finally, it should be understood that the embodiments described in this application are merely illustrative of the principles of the embodiments of this application. Other modifications may also fall within the scope of this application. Therefore, alternative configurations of the embodiments of this application are considered as examples and not limitations, and are regarded as consistent with the teachings of this application. Accordingly, the embodiments of this application are not limited to the embodiments explicitly described and illustrated in this application.

[0141] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A device flaw detection method based on array pulsed eddy currents, characterized in that, The method includes: Descriptive fields are extracted from the array pulse eddy current data to obtain the descriptive field information of the array pulse eddy current data. The descriptive field information includes the descriptive field division method and the descriptive field type. The descriptive field division method is used to indicate the way to distinguish different sample data in the array pulse eddy current data. The descriptive field type is used to indicate the type corresponding to each data node in the array pulse eddy current data. The type is the type to which the sample data where the data node is located belongs. Based on the differentiation method indicated by the description field division method, the array pulse eddy current data is divided into several reference data ranges; Based on the description field type, the several reference data ranges are optimized to obtain several target data ranges of the array pulse eddy current data, so that the sample data corresponding to each two adjacent target data ranges belong to different types, and the existence of equipment defects is determined by the different types of the sample data.

2. The method according to claim 1, characterized in that, The array pulse eddy current data is divided into several reference data ranges according to the distinction method indicated by the description field division method, including: Based on the description field division method, the location of multiple differentiation methods in the array pulse eddy current data is determined; Based on the positioning of the multiple differentiation methods, the range between each pair of adjacent differentiation methods is determined as a reference data range, thus obtaining the multiple reference data ranges.

3. The method according to claim 1, characterized in that, Based on the description field type, the optimization of the several reference data ranges yields several target data ranges for the array pulse eddy current data, including: Based on the described field type, determine the type corresponding to each data node in the array pulse eddy current data; Based on the data nodes in each reference data range and the type corresponding to each data node, the type corresponding to each reference data range is determined respectively; Based on the categories corresponding to the aforementioned reference data ranges, the aforementioned reference data ranges are optimized to obtain the aforementioned target data ranges.

4. The method according to claim 3, characterized in that, The step of optimizing the plurality of reference data ranges based on the categories corresponding to the plurality of reference data ranges to obtain the plurality of target data ranges includes: In response to the fact that any two adjacent reference data ranges correspond to the same category, the two reference data ranges are merged to obtain a target data range; In response to the fact that a random reference data range corresponds to a different category than an adjacent reference data range, the random reference data range is determined as the target data range.

5. The method according to claim 3, characterized in that, The step of determining the category corresponding to each reference data range based on the data nodes in each reference data range and the category corresponding to each data node includes: Based on the data nodes in each reference data range and the category corresponding to each data node, the category corresponding to each reference data range and the confidence weight of the category are determined respectively.

6. The method according to claim 5, characterized in that, The description field information also includes an importance description field, which is used to indicate the importance of the array pulse eddy current data in terms of the differentiation method; the optimization of the several reference data ranges according to the categories corresponding to the several reference data ranges to obtain the several target data ranges includes: in response to the fact that the confidence weight corresponding to a random reference data range is less than the reference confidence weight, determining at least two target importances located in the reference data range according to the importance description field; and dividing the reference data range according to the differentiation method constituted by the at least two importances to obtain the target data range.

7. The method according to claim 6, characterized in that, The step of determining the importance of at least two targets within the reference data range based on the important description field includes: determining the importance of at least two targets within the reference data range based on the location of several important targets in the important description field and the location of the reference data range in the array pulse eddy current data.

8. The method according to claim 3, characterized in that, The step of determining the category corresponding to each reference data range based on the data nodes in each reference data range and the category corresponding to each data node includes: For a random reference data range, the type corresponding to each data node in the reference data range is determined based on the data nodes in the reference data range and the type corresponding to each data node; Based on the type corresponding to each data node in the reference data range, determine the type corresponding to the reference data range; The step of determining the category corresponding to the reference data range based on the category corresponding to each data node in the reference data range includes: determining the category with the largest number of corresponding data nodes as the category corresponding to the reference data range.

9. The method according to claim 1, characterized in that, The step of extracting descriptive fields from the array pulse eddy current data to obtain the descriptive field information of the array pulse eddy current data includes: calling a descriptive field extraction network to extract descriptive fields from the array pulse eddy current data to obtain the descriptive field information; The description field extraction network includes a description field extraction sub-network, a data scaling sub-network, and a description field detection sub-network. The step of calling the description field extraction network to extract description fields from the array pulse eddy current data to obtain the description field information includes: The description field extraction sub-network is invoked to extract description fields from the array pulse eddy current data, thereby obtaining the first description field set of the array pulse eddy current data. The data scale reorganization subnetwork is invoked to perform data scale reorganization on the first description field set to obtain the second description field set of the array pulse eddy current data. The description field detection sub-network is invoked to perform description field detection on the second description field set to obtain the description field information. The step of calling the data scaling sub-network to scale the first description field set to obtain the second description field set of the array pulse eddy current data includes: The data scale reorganization subnetwork is invoked to perform data scale reorganization processing on the first description field set, thereby obtaining several reference description field sets of data scales corresponding to the first description field set. The reference description field sets of the several data scales are concatenated to obtain the second description field set of the array pulse eddy current data.

10. A device flaw detection system based on arrayed pulsed eddy currents, characterized in that, The method includes a processor and a memory that communicate with each other, the processor being configured to read a computer program from the memory and execute it to implement the method of any one of claims 1-9.

Images