A method and system for identifying weak magnetic signals of ferromagnetic materials

By establishing a sample feature library and deep recognition network, extracting and integrating weak magnetic signal characteristics of ferromagnetic materials, the problems of low identification efficiency and unstable results in the prior art are solved, and efficient identification and analysis of ferromagnetic materials are achieved. Improved identification accuracy and stability.

CN118553344BActive Publication Date: 2024-11-01XUZHOU NANFANG YONGCI MATERIAL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410332326.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-22
Publication Date
2024-11-01
Estimated Expiration
2044-03-22

AI Technical Summary

Technical Problem

In the prior art, the weak magnetic signal recognition analysis of ferromagnetic materials is inefficient and the results are unstable, making it difficult to accurately identify defects and stress distributions inside complex materials.

Method used

By establishing a sample feature library, multiple features of the signal, such as peak-peak, maximum, average, gradient maximum, etc., perform test source analysis and aggregation, and use the deep recognition hidden layer identification network for optimization search and integration The weak magnetic signal results collected multiple times generate the final identification results.

Benefits of technology

It realizes efficient identification and analysis of weak magnetic signals of ferromagnetic materials, improves the stability and accuracy of identification results, and can promptly detect defects and stress distribution changes in the material, improves product quality and production efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118553344B_ABST
    Figure CN118553344B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of data processing technology, and provides a method and system for identifying weak magnetic signals of ferromagnetic materials, including: establishing a sample feature library, extracting multiple features of weak magnetic signals of ferromagnetic materials; performing test source analysis, and assigning a deep identification to each sample source; aggregating the sample feature library according to the recognition result, and training it using a recognition network; optimizing network parameters through optimal search, and establishing a recognition network including a deep recognition hidden layer; collecting new weak magnetic signals and integrating previous recognition results to generate a final recognition result. The method solves the technical problems of low efficiency of weak magnetic signal recognition and analysis and unstable weak magnetic signal recognition results, and realizes efficient recognition and analysis of weak magnetic signals of ferromagnetic materials through rapid detection of ferromagnetic materials, thereby ensuring the technical effect of the stability of weak magnetic signal recognition results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field related to data processing, and in particular to a method and system for identifying weak magnetic signals of ferromagnetic materials. Background Art

[0002] Ferromagnetic materials are widely used in industrial production and daily life, such as steel, magnetic materials, etc. During the manufacturing and use of these materials, they may be affected by various factors, such as defects, stress, temperature, etc., causing changes in their magnetic properties. Therefore, identifying and detecting weak magnetic signals of ferromagnetic materials is of great significance for ensuring product quality and improving production efficiency.

[0003] Existing weak magnetic signal recognition methods mainly include magnetic measurement and magnetic imaging. The magnetic measurement method mainly measures the magnetic properties of magnetic materials, such as magnetic permeability and magnetization, to infer the defects and stress distribution inside the material. However, this method requires a lot of time and computing resources, and the accuracy is affected by many factors, such as the measurement environment and measurement equipment. The magnetic imaging method infers the defects and stress distribution inside the material through changes in the magnetic field. However, this method requires the use of magnetic imaging equipment, and it is difficult to obtain accurate imaging results for some complex internal structures of materials.

[0004] In summary, the prior art has the technical problems of low efficiency of weak magnetic signal recognition and analysis and unstable weak magnetic signal recognition results. Summary of the invention

[0005] The present application provides a weak magnetic signal recognition method and system for ferromagnetic materials, aiming to solve the technical problems in the prior art of low efficiency of weak magnetic signal recognition and analysis and unstable weak magnetic signal recognition results.

[0006] In view of the above problems, the present application provides a method and system for identifying weak magnetic signals of ferromagnetic materials.

[0007] The first aspect disclosed in the present application provides a method for identifying weak magnetic signals of ferromagnetic materials, wherein the method comprises: establishing a sample feature library, wherein the sample feature library is constructed by extracting data features of samples, wherein the extracted features include peak-to-peak value of the signal, maximum value of the signal, minimum value of the signal, average value of the signal, maximum value of the signal gradient, minimum value of the signal gradient, average value of the signal gradient, and energy of the signal wavelet packet, wherein the sample is a weak magnetic signal of a ferromagnetic material; performing a test source analysis on the sample, and establishing a source depth identifier of the test source; aggregating the sample feature library according to the feature recognition result, wherein the aggregation result carries a source depth identifier; using the aggregation result as basic data to execute recognition network establishment; and using the aggregation result as the basic data to generate a recognition network. The extracted features are used as the solution space, and an optimal search is performed in the solution space, the search results are evaluated by a fitness function, and the optimal search direction is updated; the recognition network is established according to the optimal search results, wherein the recognition network includes a deep recognition hidden layer; weak magnetic signals are collected from ferromagnetic materials, and the collection results are transmitted to the recognition network to generate a first weak magnetic signal recognition result; the first weak magnetic signal recognition result is distance parsed, a newly added collection direction is selected, a newly added weak magnetic signal collection is performed, and the newly added weak magnetic signal collection result is synchronized to the recognition network to generate a second weak magnetic signal recognition result; the first weak magnetic signal recognition result and the second weak magnetic signal recognition result are integrated to generate an integrated weak magnetic signal recognition result.

[0008] Another aspect disclosed in the present application provides a weak magnetic signal recognition system for ferromagnetic materials, wherein the system includes: a feature library establishment module, used to establish a sample feature library, the sample feature library is constructed by extracting data features of the sample, the extracted features include the peak-to-peak value of the signal, the maximum value of the signal, the minimum value of the signal, the average value of the signal, the maximum value of the signal gradient, the minimum value of the signal gradient, the average value of the signal gradient, and the energy of the signal wavelet packet, wherein the sample is a weak magnetic signal of the ferromagnetic material; a test source analysis module, used to perform test source analysis on the sample and establish a source depth identification of the test source; a feature library aggregation module, used to aggregate the sample feature library according to the feature recognition result, wherein the aggregation result carries a source depth identification; an identification network establishment module, used to use the aggregation result as basic data to execute identification network establishment; an optimization search module, used to aggregate the aggregation result; The extracted features in the result are used as the solution space, and an optimal search is performed in the solution space, the search results are evaluated by the fitness function, and the optimal search direction is updated; an identification network establishment module is used to complete the establishment of the identification network according to the optimal search results, wherein the identification network includes a deep identification hidden layer; a first signal acquisition module is used to perform weak magnetic signal acquisition on ferromagnetic materials, and transmit the acquisition results to the identification network to generate a first weak magnetic signal recognition result; a second signal acquisition module is used to perform distance analysis on the first weak magnetic signal recognition result, select a newly added acquisition direction, perform newly added weak magnetic signal acquisition, synchronize the newly added weak magnetic signal acquisition results to the identification network, and generate a second weak magnetic signal recognition result; a recognition result output module is used to integrate the first weak magnetic signal recognition result and the second weak magnetic signal recognition result to generate an integrated weak magnetic signal recognition result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] By establishing a sample feature library, multiple features of weak magnetic signals of ferromagnetic materials are extracted, including peak-to-peak value, maximum value, minimum value, average value, gradient maximum value, gradient minimum value, gradient average value and wavelet packet energy; test source analysis is performed to assign a deep identification to each sample source; the sample feature library is aggregated according to the recognition results and trained using the recognition network; the network parameters are optimized through optimal search, and a recognition network including a deep recognition hidden layer is established; new weak magnetic signals are collected and previous recognition results are integrated to generate the final recognition results, thereby achieving efficient recognition and analysis of weak magnetic signals of ferromagnetic materials and ensuring the stability of weak magnetic signal recognition results.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart of a weak magnetic signal recognition method for ferromagnetic materials is provided for an embodiment of the present application.

[0013] Figure 2 A schematic flow chart of updating characteristic data in a weak magnetic signal identification method for ferromagnetic materials is provided for an embodiment of the present application.

[0014] Figure 3 A structural schematic diagram of a weak magnetic signal recognition system for ferromagnetic materials is provided for an embodiment of the present application.

[0015] Explanation of the accompanying drawings: feature library establishment module 100, test source analysis module 200, feature library aggregation module 300, identification network establishment module 400, optimization search module 500, identification network establishment module 600, first signal acquisition module 700, second signal acquisition module 800, identification result output module 900. DETAILED DESCRIPTION

[0016] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0017] Embodiment 1

[0018] like Figure 1 As shown, an embodiment of the present application provides a method for identifying weak magnetic signals of ferromagnetic materials, wherein the method comprises:

[0019] Step-1: Establish a sample feature library, which is constructed by extracting data features of samples. The extracted features include the peak-to-peak value of the signal, the maximum value of the signal, the minimum value of the signal, the average value of the signal, the maximum value of the signal gradient, the minimum value of the signal gradient, the average value of the signal gradient, and the energy of the signal wavelet packet. The sample is a weak magnetic signal of a ferromagnetic material;

[0020] Step-2: Analyze the test source of the sample and establish the source depth identification of the test source;

[0021] Weak magnetic signal samples of ferromagnetic materials are collected, and the samples can be obtained by on-site collection or other methods; for each sample, its data features are extracted, and the data features include but are not limited to the peak-to-peak value, maximum value, minimum value, average value, gradient maximum value, gradient minimum value, gradient average value and wavelet packet energy of the signal, which reflect some key characteristics of the weak magnetic signal, such as amplitude distribution, fluctuation and frequency distribution; after the features of all samples are extracted, a sample feature library is constructed, which contains the feature data of various weak magnetic signals and can be used for subsequent analysis and processing.

[0022] After obtaining the sample feature library, it is necessary to perform test source analysis on the samples. The test source refers to the source of the weak magnetic signal. Different sources correspond to different physical phenomena or process. By analyzing the characteristics of each sample, its possible test source can be inferred. For example, if the peak-to-peak value of a sample is higher, the sample comes from an object with greater magnetism; if the wavelet packet energy of a sample is lower, the sample comes from a rougher surface, etc. After determining the test source, a source depth mark of the test source can be established. The source depth mark includes the position distance of the weak magnetic signal (defect) in the sample, indicating the degree of correlation between the test source and the weak magnetic signal of the ferromagnetic material or the depth of the source. For example, if a test source corresponds to multiple samples, then its source depth mark is higher; otherwise, it is lower. Establishing a source depth mark can reflect the complexity or hierarchical relationship between different sources, and provide a reference for subsequent data analysis and decision-making. For example, if there are obvious differences or hierarchical relationships between data from different sources, such differences and hierarchical relationships can be considered in data analysis to improve the accuracy of recognition.

[0023] Step-3: Aggregate the sample feature library according to the feature recognition result, wherein the aggregation result is marked with a source depth mark;

[0024] Step-4: Use the aggregation results as basic data to establish a recognition network;

[0025] Step-5: Take the extracted features in the aggregation result as the solution space, perform optimal search in the solution space, evaluate the search results through the fitness function, and update the optimal search direction;

[0026] The sample feature library is aggregated according to the feature recognition results. Aggregation is a process of merging the features of multiple samples, which can be used to reduce the size of the feature space and improve search efficiency. During the aggregation process, attention is paid to retaining the source depth identifier of each sample. The aggregation result is a feature set with a source depth identifier, which can be used to indicate the degree of correlation or source depth between the test source to which the sample belongs and the weak magnetic signal of the ferromagnetic material.

[0027] The aggregation results are used as basic data to establish a recognition network, which can be used to classify and identify weak magnetic signals. The method of establishing the recognition network may vary depending on the machine learning algorithm used, such as support vector machine (SVM), neural network, decision tree, etc. By using the aggregation results as basic data, a more accurate and reliable recognition network can be constructed.

[0028] The extracted features in the aggregation results are used as the solution space, and the optimization search is performed in the solution space. The optimization search is a process of finding the optimal solution, which can be used to find a position in the solution space where the objective function (such as the fitness function) reaches the minimum or maximum value; in the optimization search process, the fitness function is needed to evaluate the quality of the search results. The fitness function is a function used to measure the quality of the solution, which can be designed according to the specific problem. In the embodiment of the present application, the fitness function can be used to measure indicators such as the classification accuracy or robustness of the recognition network; by continuously updating the search direction and search range, the optimal solution is gradually approached. The method of updating the search direction may vary depending on the optimization algorithm used, such as genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, etc. By evaluating the search results with the fitness function, the current optimal solution and search direction can be determined, so as to continuously update the optimization search direction, which can improve the search efficiency and the search accuracy for the optimal solution, and finally obtain a more accurate and reliable recognition result.

[0029] Step-6: Complete the establishment of the recognition network according to the optimization search results, wherein the recognition network includes a deep recognition hidden layer;

[0030] Step-7: Collect weak magnetic signals from the ferromagnetic material, and transmit the collected results to the recognition network to generate a first weak magnetic signal recognition result;

[0031] Step-8: Perform distance analysis on the first weak magnetic signal recognition result, select a new collection direction, perform new weak magnetic signal collection, synchronize the new weak magnetic signal collection result to the recognition network, and generate a second weak magnetic signal recognition result;

[0032] Step-9: Integrate the first weak magnetic signal recognition result and the second weak magnetic signal recognition result to generate an integrated weak magnetic signal recognition result.

[0033] The final recognition network is established according to the results of the optimal search, and the recognition network contains a deep recognition hidden layer to achieve in-depth analysis and recognition of weak magnetic signals; the process of establishing the recognition network includes determining the network structure (such as the number of layers, the number of neurons in each layer, etc.), initializing network parameters (such as weights and biases, etc.), selecting activation functions, etc.; once the recognition network is established, the recognition network is used to classify and identify weak magnetic signals. By completing the establishment of the recognition network according to the optimal search results, a more accurate and reliable recognition network can be constructed, which can better represent and identify the characteristics of weak magnetic signals and improve the accuracy of classification and recognition.

[0034] Weak magnetic signals of ferromagnetic materials are collected, and the collected results are transmitted to the established recognition network; the process of weak magnetic signal collection includes the steps of setting up collection equipment, adjusting collection parameters, and performing signal preprocessing; after the recognition network receives the collected weak magnetic signal, it will output a corresponding recognition result and use it as the first weak magnetic signal recognition result.

[0035] The first weak magnetic signal recognition result is subjected to distance analysis to determine which directions of weak magnetic signals have the greatest impact on the recognition result; the distance analysis process includes the steps of calculating the distance between each feature and the target and weighting the distance; according to the results of the distance analysis, a new acquisition direction is selected and the newly added weak magnetic signal acquisition is performed; the newly acquired weak magnetic signal results are synchronized to the recognition network and the second weak magnetic signal recognition result is generated. By performing distance analysis on the first weak magnetic signal recognition result, a new acquisition direction can be determined to improve the efficiency and accuracy of the acquisition.

[0036] The first weak magnetic signal identification result and the second weak magnetic signal identification result are integrated to generate a more comprehensive integrated weak magnetic signal identification result, wherein the integration process includes a step of weighted averaging the two results and a step of selecting the optimal result; the integrated weak magnetic signal identification result can more comprehensively reflect the properties and state of the ferromagnetic material, and provide strong support for subsequent analysis and processing.

[0037] Furthermore, the present application method also includes:

[0038] Collecting material data of the ferromagnetic material to establish a material data set, wherein the material data set is a dimension data set of the ferromagnetic material;

[0039] Obtaining a distance analysis result of the first weak magnetic signal recognition result, and locating the defect position based on the initial signal acquisition position and the distance analysis result to generate a fuzzy positioning result;

[0040] Optimizing the acquisition position based on the fuzzy positioning result and the material data set;

[0041] Based on the optimal collection position results, a new collection direction is selected to perform new weak magnetic signal collection.

[0042] The first weak magnetic signal recognition result is subjected to distance analysis, a new collection direction is selected, and the new weak magnetic signal collection is executed. Prior to this, material data collection of the ferromagnetic material is also included: the type and specification of the ferromagnetic material to be collected is determined, different ferromagnetic materials are selected, and the characteristics of each material, such as size, shape, and weight, are determined; appropriate measuring tools, such as laser scanners, microscopes, etc., are used to measure the selected ferromagnetic material, including various dimensional information of the material, such as length, width, height, etc., and other related characteristics, such as density and magnetic permeability of the material; the measured data is organized into a material data set to obtain various properties and dimensional information of the ferromagnetic material, so as to provide basic data for subsequent weak magnetic signal recognition and defect location.

[0043] The distance analysis result refers to the distance information between the signal source and the ferromagnetic material obtained by analyzing the weak magnetic signal; the initial signal acquisition position refers to the acquisition position of the weak magnetic signal in the initial state; the defect position is located based on the distance analysis result and the initial signal acquisition position, and the approximate position of the defect is inferred; the weak magnetic signal recognition algorithm is used to analyze the collected weak magnetic signal, and the information related to the defect, such as the position, size, and shape of the defect, is extracted from the signal, and the defect information is fuzzily located through the analyzed defect information to obtain an approximate defect position range (not an exact position). The defect position is located based on the initial signal acquisition position and the distance analysis result, and the approximate position of the defect can be determined quickly and accurately.

[0044] The fuzzy positioning results and material data sets are combined, and optimization algorithms such as simulated annealing and genetic algorithms are used to determine the best weak magnetic signal acquisition position. By optimizing the acquisition position, the optimal acquisition position can be selected while ensuring acquisition accuracy and stability, which can improve acquisition efficiency and reduce the impact of interference factors on acquisition results.

[0045] The newly added acquisition direction is selected based on the acquisition position optimization result to obtain more and more comprehensive weak magnetic signal data; the newly added acquisition direction can be determined based on the best acquisition position obtained in the previous step; once the new acquisition direction is determined, a new weak magnetic signal acquisition operation can be performed. The newly added acquisition direction is selected based on the acquisition position optimization result, which can further expand the acquisition range of weak magnetic signals and obtain more and more comprehensive weak magnetic signal data. By performing the newly added weak magnetic signal acquisition, new defects or stress concentration areas can be discovered in a timely manner.

[0046] Furthermore, the present application method also includes:

[0047] Based on the fuzzy positioning result, a signal closer to the first weak magnetic signal identification result and the second weak magnetic signal is selected as a reference signal;

[0048] performing attenuation compensation calculation at a corresponding unit distance based on the first weak magnetic signal recognition result and the second weak magnetic signal recognition result respectively;

[0049] Calculating an average value based on the attenuation compensation as a compensation result and performing compensation on the reference signal;

[0050] The integrated weak magnetic signal identification result is generated according to the compensation result.

[0051] The fuzzy positioning result provides a possible defect location range, from which two possible defect locations are selected, namely the first weak magnetic signal identification result and the second weak magnetic signal identification result; the distances between these two locations and the fuzzy positioning result are compared, and the location with a closer distance is selected as the reference signal. The weak magnetic signal with a closer distance is selected as the reference signal based on the fuzzy positioning result, which helps to improve the accuracy and reliability of subsequent calculations.

[0052] The attenuation compensation calculation refers to the compensation calculation of the recognition result according to the attenuation characteristics of the weak magnetic signal to reduce the influence of the signal attenuation on the recognition result; for the selected first weak magnetic signal recognition result and the second weak magnetic signal recognition result, the attenuation situation at the corresponding unit distance; according to the material characteristics and parameters such as magnetic permeability, the attenuation degree of the magnetic signal at a specific distance can be calculated, wherein the corresponding unit distance refers to the corresponding attenuation degree or compensation value at a specific distance;

[0053] The attenuation compensation calculation results of the first weak magnetic signal identification result and the second weak magnetic signal identification result are averaged as the final compensation result, and the compensation result is a specific value when compensating the reference signal; the reference signal is compensated using the compensation result to correct the signal attenuation effect caused by distance.

[0054] The compensated reference signal is integrated with other possible weak magnetic signal identification results to generate the final integrated weak magnetic signal identification result; the integrated result will include multiple possible weak magnetic signal identification results, and multiple possible weak magnetic signal identification results have undergone the same attenuation compensation processing. The integrated weak magnetic signal identification result is generated based on the compensation result, providing more accurate and stable data support for subsequent analysis and decision-making.

[0055] Furthermore, the present application method also includes:

[0056] Configure an initial space size, and randomly select features that meet the initial space size in the solution space as initial space features;

[0057] The initial spatial feature data is updated according to the sample feature values ​​in the aggregation result, and the search result is evaluated by the fitness function;

[0058] Perform feature iteration according to the search evaluation results, and obtain optimal search results based on the feature iteration results.

[0059] Configure the initial space size, which can be a numerical value, such as the number of iterations or population size commonly seen in optimization problems; randomly select some features in the solution space, the number or distribution of these features conforms to the set initial space size, and these features constitute the initial space features, the initial space size refers to the maximum number of features or feature range allowed when randomly selecting features in the solution space; the initial space features refer to the set of features randomly selected in the solution space that conform to the initial space size limit, and random selection can ensure the diversity of the search. By configuring the initial space size and randomly selecting features that conform to the initial space size in the solution space as the initial space features, the scope of the search space can be effectively controlled and the search efficiency can be improved.

[0060] Use the sample feature values ​​in the aggregation results to update the initial spatial features, such as mapping the sample feature values ​​to a specific feature space; set a fitness function to evaluate the quality of the search results. The fitness function should be associated with the objective function. For example, in the optimization problem, the objective function can be a function that seeks to minimize or maximize, while the fitness function can be the negative value of the objective function or other functions that can represent the quality; use the fitness function to evaluate the updated initial spatial features and obtain the evaluation results of the search results. By updating the data of the initial spatial features based on the sample feature values ​​in the aggregation results, the feature set of the search space can be dynamically adjusted to make it closer to the needs of the actual problem.

[0061] Based on the evaluation of the search results obtained in the previous step, decide whether feature iteration is needed: if the evaluation result meets the preset conditions (such as reaching the preset accuracy or number of iterations), the search is terminated; otherwise, feature iteration is performed; during the iteration process, the initial spatial features need to be updated, such as gradient descent, genetic algorithm, etc., and the value or distribution of the feature is adjusted according to the evaluation result of the fitness function; during the iteration process, the value of the fitness function needs to be recalculated after each update, and the next feature update direction and strategy are determined based on the new evaluation results, and this process continues until the end conditions are met. The features obtained at this time are the results of the optimal search.

[0062] Furthermore, the present application method also includes:

[0063] Setting a search evaluation result threshold, wherein the search evaluation result threshold is marked with an iteration number;

[0064] If the feature iteration count satisfies the iteration count identifier, and the search evaluation result cannot satisfy the search evaluation result threshold, then a new instruction is generated;

[0065] Adding spatial features to the initial space according to the adding instruction;

[0066] Complete the optimal search by adding results through spatial features.

[0067] Setting a search evaluation result threshold is usually a preset value used to judge the quality of search results. The search evaluation result threshold is usually associated with a fitness function. For example, if the goal is to find an optimal solution, the threshold can be set to a value that is better than the currently known optimal solution. The search evaluation result threshold is marked with an iteration count, which means that when setting the threshold, a limit on the number of iterations is also set, which can prevent too many iterations from being performed before a solution that meets the threshold is found, thereby improving search efficiency.

[0068] After each feature iteration, check whether the current search evaluation result meets the threshold: if it does, stop the search and return the current solution; if it does not, check whether the number of feature iterations has reached the preset iteration limit; if the number of feature iterations has reached the preset iteration limit, but the search evaluation result still does not meet the threshold, then it can be considered that the current search space may not be able to find a solution that meets the threshold. At this time, it is necessary to generate a new instruction to add new features or change the original features based on the original search space. The new instruction refers to the instruction generated to guide the addition of spatial features when the number of feature iterations reaches the preset iteration number identifier and the search evaluation result fails to meet the search evaluation result threshold. It is helpful to jump out of the local optimum and continue the search by adding new instructions when the search process falls into a local optimal solution.

[0069] Features in the initial space are added according to the newly added instructions; by adding new features, the search space is expanded, thereby increasing the chance of finding a better solution; after the spatial features are added, the search is performed again through the new feature set, and if a solution that meets the threshold is found, the search is stopped and the solution is returned; if not found, it is necessary to generate new instructions again and repeat the above process until a solution that meets the threshold is found or the preset maximum number of iterations is reached.

[0070] Furthermore, if Figure 2 As shown, the present application method also includes:

[0071] Set the extreme variation space of features;

[0072] Configure fine-tuning probability and feature update probability;

[0073] When performing feature iteration, the initial spatial feature data is updated within the extreme variation space based on the fine-tuning probability and the feature update probability.

[0074] The limit change space of the feature is set, and the limit change space is usually a preset range used to limit the change of the feature; the setting of the limit change space can ensure that the change of the feature does not exceed the preset range, thereby ensuring the feasibility of the search. By setting the limit change space of the feature, the change range of each feature can be effectively controlled to avoid abnormal changes in the feature value and affect the accuracy of the search results.

[0075] Configure the fine-tuning probability and feature update probability. The fine-tuning probability refers to the probability of fine-tuning a feature, while the feature update probability refers to the probability of updating a feature according to the result of the fitness function. The configuration of the fine-tuning probability and feature update probability can affect the efficiency and accuracy of the search. If the fine-tuning probability is too high, the search will be too scattered and the efficiency will be reduced; if the fine-tuning probability is too low, the search will be too concentrated and the accuracy will be reduced. Similarly, if the feature update probability is too high, the search will be too dependent on the result of the fitness function; if the feature update probability is too low, the search will be too conservative and it will be difficult to find a better solution. By configuring the fine-tuning probability and feature update probability, you can control the adjustment method and frequency of the feature value during the feature iteration process, thereby affecting the stability and efficiency of the search process.

[0076] When performing feature iteration, the decision on whether to update the feature is based on the fine-tuning probability and the feature update probability: if an update is required, a new feature value is randomly selected in the extreme change space to replace the original feature value; the initial spatial feature data in the extreme change space is updated through feature updates based on the fine-tuning probability and the feature update probability. By performing the initial spatial feature data update in the extreme change space based on the fine-tuning probability and the feature update probability, the speed and direction of feature iteration can be controlled, ensuring the stability of the search process, making the search process more controllable and efficient.

[0077] Furthermore, the present application method also includes:

[0078] Performing identification verification on the integrated weak magnetic signal identification result to generate a verification result;

[0079] Extract verification features of verification results and generate compensation data;

[0080] The recognition network is optimized based on the compensation data.

[0081] There are many ways to verify the integrated weak magnetic signal recognition results, such as comparing with known reference answers or using other recognition algorithms for verification; based on the verification results, the accuracy rate, false recognition rate and other indicators of the integrated weak magnetic signal recognition results can be obtained as verification results. By verifying the integrated weak magnetic signal recognition results, its accuracy and reliability can be evaluated.

[0082] According to the verification results obtained in the previous step, it is necessary to extract key features as verification features, which include information such as signal amplitude, frequency, phase, etc.; based on the verification features, generate compensation data, which includes correction values ​​for certain features, or new training data;

[0083] According to the compensation data obtained in the previous step, the recognition network can be adjusted or optimized, including correcting the network parameters, adding or deleting certain layers, changing the activation function, etc. The optimized recognition network should be able to better identify weak magnetic signals and improve the accuracy and robustness of recognition. By optimizing the recognition network based on the compensation data, its performance and accuracy can be improved, and its recognition ability and reliability for weak magnetic signals can be improved.

[0084] In summary, the weak magnetic signal recognition method for ferromagnetic materials provided in the embodiments of the present application has the following technical effects:

[0085] 1. By establishing a sample feature library and using a deep learning algorithm to build a recognition network, high-precision weak magnetic signal recognition can be achieved, improving recognition accuracy and stability.

[0086] 2. Real-time online monitoring of weak magnetic signals of ferromagnetic materials can timely detect defects and stress distribution changes inside the materials, which helps to improve product quality and production efficiency.

[0087] 3. The modular design makes operation and use easier, reduces the difficulty of use, and can realize automatic and intelligent identification, thus improving work efficiency and measurement accuracy.

[0088] 4. The recognition and verification of the integrated weak magnetic signal recognition results is adopted to generate the verification results; the verification features of the verification results are extracted to generate compensation data; and the recognition network is optimized based on the compensation data. By optimizing the recognition network based on the compensation data, its performance and accuracy can be improved, and its recognition ability and reliability for weak magnetic signals can be improved.

[0089] Embodiment 2

[0090] Based on the same inventive concept as the weak magnetic signal identification method for ferromagnetic materials in the aforementioned embodiment, Figure 3As shown, an embodiment of the present application provides a weak magnetic signal recognition system for ferromagnetic materials, wherein the system includes:

[0091] The feature library building module 100 is used to build a sample feature library, wherein the sample feature library is constructed by extracting data features of the sample, wherein the extracted features include the peak-to-peak value of the signal, the maximum value of the signal, the minimum value of the signal, the average value of the signal, the maximum value of the signal gradient, the minimum value of the signal gradient, the average value of the signal gradient, and the energy of the signal wavelet packet, wherein the sample is a weak magnetic signal of a ferromagnetic material;

[0092] The test source analysis module 200 is used to perform test source analysis on the sample and establish a source depth identification of the test source;

[0093] A feature library aggregation module 300 is used to aggregate the sample feature library according to the feature recognition result, wherein the aggregation result is marked with a source depth mark;

[0094] An identification network establishment module 400 is used to use the aggregation result as basic data to perform identification network establishment;

[0095] The optimization search module 500 is used to use the extracted features in the aggregation results as a solution space, perform optimization search in the solution space, evaluate the search results through a fitness function, and update the optimization search direction;

[0096] The recognition network establishment module 600 is used to complete the establishment of the recognition network according to the optimization search results, wherein the recognition network includes a deep recognition hidden layer;

[0097] The first signal acquisition module 700 is used to acquire weak magnetic signals from ferromagnetic materials and transmit the acquisition results to the recognition network to generate a first weak magnetic signal recognition result;

[0098] The second signal acquisition module 800 is used to perform distance analysis on the first weak magnetic signal recognition result, select a newly added collection direction, perform newly added weak magnetic signal collection, synchronize the newly added weak magnetic signal collection result to the recognition network, and generate a second weak magnetic signal recognition result;

[0099] The recognition result output module 900 is used to integrate the first weak magnetic signal recognition result and the second weak magnetic signal recognition result to generate an integrated weak magnetic signal recognition result.

[0100] Furthermore, the system also includes:

[0101] Collecting material data of the ferromagnetic material to establish a material data set, wherein the material data set is a dimension data set of the ferromagnetic material;

[0102] Obtaining a distance analysis result of the first weak magnetic signal recognition result, and locating the defect position based on the initial signal acquisition position and the distance analysis result to generate a fuzzy positioning result;

[0103] Optimizing the acquisition position based on the fuzzy positioning result and the material data set;

[0104] Based on the optimal collection position results, a new collection direction is selected to perform new weak magnetic signal collection.

[0105] Furthermore, the system also includes:

[0106] Based on the fuzzy positioning result, a signal closer to the first weak magnetic signal identification result and the second weak magnetic signal is selected as a reference signal;

[0107] performing attenuation compensation calculation at a corresponding unit distance based on the first weak magnetic signal recognition result and the second weak magnetic signal recognition result respectively;

[0108] Calculating an average value based on the attenuation compensation as a compensation result and performing compensation on the reference signal;

[0109] The integrated weak magnetic signal identification result is generated according to the compensation result.

[0110] Furthermore, the system also includes:

[0111] Configure an initial space size, and randomly select features that meet the initial space size in the solution space as initial space features;

[0112] The initial spatial feature data is updated according to the sample feature values ​​in the aggregation result, and the search result is evaluated by the fitness function;

[0113] Perform feature iteration according to the search evaluation results, and obtain optimal search results based on the feature iteration results.

[0114] Furthermore, the system also includes:

[0115] Setting a search evaluation result threshold, wherein the search evaluation result threshold is marked with an iteration number;

[0116] If the feature iteration count satisfies the iteration count identifier, and the search evaluation result cannot satisfy the search evaluation result threshold, then a new instruction is generated;

[0117] Adding spatial features to the initial space according to the adding instruction;

[0118] Complete the optimal search by adding results through spatial features.

[0119] Furthermore, the system also includes:

[0120] Set the extreme variation space of features;

[0121] Configure fine-tuning probability and feature update probability;

[0122] When performing feature iteration, the initial spatial feature data is updated within the extreme variation space based on the fine-tuning probability and the feature update probability.

[0123] Furthermore, the system also includes:

[0124] Performing identification verification on the integrated weak magnetic signal identification result to generate a verification result;

[0125] Extract verification features of verification results and generate compensation data;

[0126] The recognition network is optimized based on the compensation data.

[0127] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0128] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for identifying weak magnetic signals of ferromagnetic materials, characterized in that: The method comprises: Establishing a sample feature library, wherein the sample feature library is constructed by extracting data features of the sample, wherein the extracted features include a peak-to-peak value of the signal, a maximum value of the signal, a minimum value of the signal, an average value of the signal, a maximum value of the signal gradient, a minimum value of the signal gradient, an average value of the signal gradient, and a signal wavelet packet energy, wherein the sample is a weak magnetic signal of a ferromagnetic material; Conduct test source analysis on samples and establish source depth identification of test sources; Aggregating the sample feature library according to the feature recognition result, wherein the aggregation result is marked with a source depth mark; Using the aggregation results as basic data, establishing a recognition network; Taking the extracted features in the aggregation results as a solution space, performing an optimal search in the solution space, evaluating the search results through a fitness function, and updating the optimal search direction; Establishing a recognition network according to the optimization search results, wherein the recognition network includes a deep recognition hidden layer; Collecting weak magnetic signals from the ferromagnetic material, and transmitting the collected results to the identification network to generate a first weak magnetic signal identification result; Performing distance analysis on the first weak magnetic signal recognition result, selecting a new collection direction, executing new weak magnetic signal collection, synchronizing the new weak magnetic signal collection result to the recognition network, and generating a second weak magnetic signal recognition result; The first weak magnetic signal recognition result and the second weak magnetic signal recognition result are integrated to generate an integrated weak magnetic signal recognition result.

2. The method according to claim 1, characterized in that The method further comprises: Collecting material data of the ferromagnetic material to establish a material data set, wherein the material data set is a dimension data set of the ferromagnetic material; Obtaining a distance analysis result of the first weak magnetic signal recognition result, and locating the defect position based on the initial signal acquisition position and the distance analysis result to generate a fuzzy positioning result; Optimizing the acquisition position based on the fuzzy positioning result and the material data set; Based on the optimal collection position results, a new collection direction is selected to perform new weak magnetic signal collection.

3. The method according to claim 2, characterized in that The method further comprises: Based on the fuzzy positioning result, a signal closer to the first weak magnetic signal identification result and the second weak magnetic signal is selected as a reference signal; performing attenuation compensation calculation at a corresponding unit distance based on the first weak magnetic signal recognition result and the second weak magnetic signal recognition result respectively; Calculating an average value based on the attenuation compensation as a compensation result and performing compensation on the reference signal; The integrated weak magnetic signal identification result is generated according to the compensation result.

4. The method according to claim 1, characterized in that The method further comprises: Configure an initial space size, and randomly select features that meet the initial space size in the solution space as initial space features; The initial spatial feature data is updated according to the sample feature values ​​in the aggregation result, and the search result is evaluated by the fitness function; Perform feature iteration according to the search evaluation results, and obtain optimal search results based on the feature iteration results.

5. The method according to claim 4, characterized in that The method further comprises: Setting a search evaluation result threshold, wherein the search evaluation result threshold is marked with an iteration number; If the feature iteration count satisfies the iteration count identifier, and the search evaluation result cannot satisfy the search evaluation result threshold, then a new instruction is generated; Adding spatial features to the initial space according to the adding instruction; Complete the optimal search by adding results through spatial features.

6. The method according to claim 5, characterized in that The method further comprises: Set the extreme variation space of features; Configure fine-tuning probability and feature update probability; When performing feature iteration, the initial spatial feature data is updated within the extreme variation space based on the fine-tuning probability and the feature update probability.

7. The method according to claim 1, characterized in that The method further comprises: Performing identification verification on the integrated weak magnetic signal identification result to generate a verification result; Extract verification features of verification results and generate compensation data; The recognition network is optimized based on the compensation data.

8. A weak magnetic signal recognition system for ferromagnetic materials, characterized in that: A method for identifying weak magnetic signals of ferromagnetic materials according to any one of claims 1 to 7, comprising: A feature library building module is used to build a sample feature library, wherein the sample feature library is constructed by extracting data features of the sample, wherein the extracted features include the peak-to-peak value of the signal, the maximum value of the signal, the minimum value of the signal, the average value of the signal, the maximum value of the signal gradient, the minimum value of the signal gradient, the average value of the signal gradient, and the energy of the signal wavelet packet, wherein the sample is a weak magnetic signal of a ferromagnetic material; The test source analysis module is used to analyze the test source of the sample and establish the source depth identification of the test source; A feature library aggregation module, used to aggregate the sample feature library according to the feature recognition result, wherein the aggregation result carries a source depth mark; An identification network establishment module, used to use the aggregation result as basic data to perform identification network establishment; An optimization search module, used to use the extracted features in the aggregation results as a solution space, perform optimization search in the solution space, evaluate search results through a fitness function, and update the optimization search direction; A recognition network establishment module, used to complete the establishment of a recognition network according to the optimization search results, wherein the recognition network includes a deep recognition hidden layer; A first signal acquisition module is used to acquire weak magnetic signals from ferromagnetic materials and transmit the acquisition results to the recognition network to generate a first weak magnetic signal recognition result; A second signal acquisition module is used to perform distance analysis on the first weak magnetic signal recognition result, select a newly added acquisition direction, execute newly added weak magnetic signal acquisition, synchronize the newly added weak magnetic signal acquisition result to the recognition network, and generate a second weak magnetic signal recognition result; The recognition result output module is used to integrate the first weak magnetic signal recognition result and the second weak magnetic signal recognition result to generate an integrated weak magnetic signal recognition result.