Steel strand quality whole-process accurate detection and tracing system

Abstract

The invention discloses a steel strand quality full-process accurate detection and tracing system, and relates to the technical field of data analysis, and the system comprises the steps: determining the initial performance index of a detection target steel strand based on the steel strand manufacturing material and production technology; acquiring test data of historical steel strands, and establishing a steel strand defect type vector database; the method comprises the following steps: acquiring fluctuation signals of a real-time detected target steel strand, establishing a steel strand state evaluation model, generating defect type vectors of the real-time detected target steel strand, and carrying out matching screening on a defect type vector database of the steel strand, evaluating the defect type influence coefficient of the real-time detected target steel strand, and compensating for the initial performance index of the to-be-detected steel strand; and judging whether the correction performance index of the detected target steel strand meets the corresponding scene performance requirements of the steel strand, and if not, determining deviation production process parameters according to the defect type corresponding to the quality abnormality. The method has the advantage that the recognition accuracy of the steel strand for each abnormal category is improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a full-process precise detection and tracing system for steel strand quality. Background Art

[0002] Accurate detection and traceability of the entire process of steel strand quality refers to the use of digital technologies (such as the Internet of Things, big data, blockchain, etc.) to collect real-time data and conduct multi-dimensional quality analysis (such as tensile strength, relaxation rate, diameter deviation, etc.) on the entire life cycle of steel strands from raw material procurement, production and processing, performance testing to finished product delivery, and establish a unique identification code to ensure that the source of each batch of products can be checked, the process can be traced, and the responsibility can be investigated, ultimately forming a closed-loop quality management system.

[0003] While existing steel strand testing covers multiple aspects, including appearance, mechanics, and chemical composition, some testing items may be overlooked or simplified in practice, resulting in an inability to fully assess the strand's performance. For example, specialized testing for low- and high-temperature performance, which may be required for specific applications, has not received sufficient attention. This can lead to inaccurate test results that fail to truly reflect the strand's quality.

[0004] Secondly, while image-based machine learning AI testing is relatively accurate in inspecting the appearance of steel strands, effectively identifying surface defects such as cracks and rust, the complexity of the strand's internal structure and the hidden nature of its defects make it difficult for this technology to accurately identify internal quality issues such as cracks, inclusions, and uneven material quality directly from external images. This makes the detection of internal defects a weak link in comprehensive steel strand quality inspections.

[0005] To this end, the present invention provides a method for accurate detection and tracing of steel strand quality throughout the entire process. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the present invention provides a method for accurate detection and tracing of the quality of steel strands throughout the entire process, thereby solving the technical problems described in the background technology.

[0007] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0008] A method for accurate detection and traceability of the entire process of steel strand quality, including:

[0009] Based on the manufacturing materials and production processes of steel strands, analyze the corresponding production processes of different manufacturing materials of steel strands and determine the initial performance indicators of the target steel strands to be tested;

[0010] Obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for each scenario of steel strands;

[0011] Based on heterogeneous sensors, the fluctuation signal of the target steel strand in the corresponding scenario of real-time detection is obtained, a steel strand status assessment model is established, and the defect category vector of the target steel strand in the corresponding scenario of real-time detection is generated;

[0012] According to the defect category vector of the target steel strand in the corresponding scene, the defect type vector database of each scene of the steel strand is matched and screened, and the defect type influence coefficient of the target steel strand in the corresponding scene of the real-time detection is evaluated;

[0013] According to the defect type influence coefficient of the target steel strand in the real-time detection scenario, the initial performance index of the target steel strand is fitted and compensated to obtain the corrected performance index of the target steel strand;

[0014] Determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, determine the deviation from the production process parameters.

[0015] Furthermore, based on the known steel strand manufacturing materials and known production processes, a knowledge graph of performance increase and decrease conditions between materials and production processes is established;

[0016] Based on the knowledge graph of performance increase and decrease conditions between materials and production processes, the performance indicators of steel strands were used as independent variables, and the manufacturing materials and production processes of steel strands were used as dependent variables. A machine learning-linear regression algorithm was trained to obtain a material-process performance increase and decrease evaluation model.

[0017] The manufacturing materials and production process parameters of the target steel strand to be tested are determined and substituted into the material-process performance increase and decrease evaluation model to evaluate the initial performance indicators of the target steel strand to be tested.

[0018] Furthermore, multiple groups of arrayed ultrasonic guided wave sensors and magnetostrictive sensors are deployed at several locations of the target steel strand for real-time detection to obtain the fluctuation signals under the corresponding scenarios of real-time detection of the target steel strand;

[0019] Based on the real-time detection of the fluctuation signal of the target steel strand in the corresponding scene, the wavelet threshold noise reduction algorithm and Kalman filter algorithm are used to eliminate the environmental noise and improve the signal-to-noise ratio, and the fluctuation signal is pre-processed;

[0020] According to the fluctuation signal in the corresponding scenario of real-time detection of target steel strands, the Wigner-Ville distribution function is used to perform time domain conversion according to the sampling rate of the fluctuation signal to obtain the time-frequency diagram in the corresponding scenario of real-time detection of target steel strands;

[0021] According to the time-frequency diagram under the corresponding scenario of real-time detection of target steel strands, the energy distribution of each frequency of the fluctuation signal per unit time is marked, and the dispersion curve in the fluctuation signal under the corresponding scenario of real-time detection of target steel strands is extracted.

[0022] Furthermore, the mean square error between the dispersion curve in the fluctuation signal under the corresponding scenario of the real-time detection target steel strand and the reference dispersion curve is calculated, and the abnormal deviation curve is marked as the abnormal fluctuation signal under the corresponding scenario of the real-time detection target steel strand;

[0023] According to the abnormal fluctuation signal in the corresponding scene of real-time detection of target steel strands, the abnormal ultrasonic signal characteristics and abnormal magnetic signal characteristics in the corresponding scene of real-time detection of target steel strands are extracted;

[0024] Based on the abnormal ultrasonic signal characteristics and abnormal magnetic signal characteristics in the corresponding scenario of real-time detection of target steel strands, the DS evidence theory method is used to fuse the ultrasonic signal characteristics and magnetic signal characteristics, and the abnormal fusion decision feature set in the corresponding scenario of real-time detection of target steel strands is obtained.

[0025] Furthermore, based on the abnormal fusion decision feature set in the corresponding scenario of real-time detection of target steel strands, the machine learning SVM support vector machine is trained to generate a hyperplane between the steel strand abnormality category decision edges, and a steel strand state assessment model is obtained; the fluctuation signal in the corresponding scenario of real-time detection of target steel strands is used as input, and the defect category vector in the corresponding scenario of real-time detection of target steel strands is used as output.

[0026] Furthermore, based on the defect type vector database under various scenarios of the steel strand, the defect category vector data stored in association with each scenario of the steel strand is divided to obtain an array of defect category vector data under each scenario of the steel strand;

[0027] The PCA principal component analysis method is used to reduce the dimension of the defect category vector data array under various scenarios of steel strands, and the defect category vector matrix A under various scenarios of steel strands is constructed;

[0028]

[0029] Among them, q j′k is the kth defect type vector data array of the steel strand in the j′th scenario, n′ is the total number of scenarios, and m′ is the total number of defect category vectors;

[0030] Based on each element in the defect category vector matrix under each scenario of the steel strand, the similarity between the defect category vector under the corresponding scenario of the real-time detection target steel strand is calculated.

[0031] Furthermore, the defect category vectors with positive correlation in similarity under the corresponding scenes of the real-time detection target steel strand are screened, the defect types under the corresponding scenes of the real-time detection target steel strand are marked, and weights are assigned according to the degree of influence of the defect types on the corresponding scenes of the real-time detection target steel strand;

[0032] Determine the defect type of the target steel strand corresponding to the scene in real-time detection and focus on external influencing factors; the defect type focus on external influencing factors include: environmental influencing factors, load status;

[0033] Calculate and evaluate the defect impact coefficient of the corresponding scenario of the real-time detection target steel strand according to the defect type and the defect type weight of the corresponding scenario of the real-time detection target steel strand;

[0034] The similarity between the defect category vectors in the corresponding scenario of real-time detection of target steel strands is normalized, and the defect type influence weight in the corresponding scenario of real-time detection of target steel strands and the defect type attention external influencing factors in the corresponding scenario of real-time detection of target steel strands are used to calculate the defect type influence coefficient in the corresponding scenario of real-time detection of target steel strands.

[0035] A full-process accurate detection and traceability system for steel strand quality, including:

[0036] Initial performance module, comparison database, defect category module, defect impact coefficient module, performance correction module, and abnormal decision module;

[0037] The initial performance module is used to analyze the production processes corresponding to different steel strand manufacturing materials based on the steel strand manufacturing materials and production processes, and determine the initial performance indicators of the target steel strand to be tested;

[0038] The reference database is used to obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for each scenario of steel strands;

[0039] The defect classification module is used to obtain the fluctuation signal of the corresponding scene of the real-time detection target steel strand based on heterogeneous sensors, establish a steel strand state assessment model, and generate the defect classification vector of the corresponding scene of the real-time detection target steel strand;

[0040] The defect influence coefficient module is electrically connected to the defect category module and the reference database. The defect influence coefficient module is used to match and screen the defect type vector database under various scenarios of the steel strand according to the defect category vector under the corresponding scenario of the real-time steel strand being detected, and evaluate the defect type influence coefficient under the corresponding scenario of the real-time target steel strand being detected;

[0041] The performance correction module is electrically connected to the defect influence coefficient module and the initial performance module. The performance correction module is used to perform fitting compensation on the initial performance index of the steel strand to be detected according to the defect type influence coefficient in the corresponding scenario of the real-time detection target steel strand, so as to obtain the corrected performance index of the detection target steel strand;

[0042] The abnormal decision module is electrically connected to the performance correction module. The abnormal decision module is used to determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, the production process is traced.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention proposes a full-process precise detection and tracing solution for steel strand quality. Compared with the traditional single-modal detection which has a low recognition rate for inner layer damage of steel strand, the invention automatically triggers magnetic domain deflection detection when an abnormality is detected by ultrasound, and uses the ultrasonic signal and magnetic signal components to assemble the multi-dimensional abnormal feature data of the steel strand to verify the defect type, thereby improving the recognition accuracy of the steel strand for each abnormality category. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flow chart of a method for accurate detection and traceability of the entire process of steel strand quality;

[0046] Figure 2 This is a framework diagram of a full-process accurate detection and traceability system for steel strand quality; DETAILED DESCRIPTION

[0047] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0048] Reference Figure 1 As shown, the present invention provides a method for accurate detection and tracing of the quality of steel strands throughout the entire process, including:

[0049] Step 1: Based on the manufacturing materials and production processes of steel strands, analyze the corresponding production processes of different manufacturing materials of steel strands and determine the initial performance indicators of the target steel strands to be tested;

[0050] The step 1 includes the following:

[0051] Step 101: Based on known steel strand manufacturing materials and known production processes, a knowledge graph of performance increase and decrease conditions between materials and production processes is established;

[0052] Step 102: Based on the knowledge graph of performance increase and decrease conditions between materials and production processes, the performance indicators of the steel strand are used as independent variables, and the manufacturing materials and production processes of the steel strand are used as dependent variables. A machine learning-linear regression algorithm is trained to obtain a material-process performance increase and decrease evaluation model.

[0053] As a further content, in order to ensure that the output results of the material-process performance increase and decrease evaluation model can achieve the optimal fit, during the model training process, the optimal parameters of the model can be estimated using the least squares method or the gradient descent method based on the performance increase and decrease condition data between the material and the production process to ensure that the error function of the model reaches the minimum value.

[0054] Step 103: Determine the manufacturing material and production process parameters of the target steel strand to be tested and substitute them into the material-process performance increase and decrease evaluation model to evaluate the initial performance indicators of the target steel strand to be tested in the following manner:

[0055]

[0056] Among them, P is the initial performance index of the target steel strand, R0 is the benchmark tensile strength, N i To detect the i-th manufacturing material parameter of the target steel strand, M j To detect the jth production process parameter of the target steel strand, α i is the contribution regression coefficient of the i-th manufacturing material to the tensile strength, α j is the regression coefficient of the contribution of the j-th production process to the tensile strength, β NM is the interaction coefficient between material and process, N is the total number of manufacturing materials, and M is the total number of production processes.

[0057] Combining the contents from 101 to 103:

[0058] Furthermore, by constructing a knowledge graph of the performance increase and decrease conditions between materials and production processes and using a machine learning-linear regression algorithm to train a material-process performance increase and decrease assessment model, the initial performance indicators of the target steel strand can be accurately assessed. This step not only improves the accuracy of the assessment but also provides key parameters for steel strand quality testing.

[0059] Step 2: Obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for steel strands in each scenario.

[0060] Step 3: Based on heterogeneous sensors, obtain the fluctuation signal of the target steel strand in the corresponding scenario of real-time detection, establish a steel strand state assessment model, and generate the defect category vector of the target steel strand in the corresponding scenario of real-time detection;

[0061] The step three includes the following:

[0062] Step 301: deploy multiple groups of arrayed ultrasonic guided wave sensors and magnetostrictive sensors based on several locations of the target steel strand for real-time detection, and obtain fluctuation signals in corresponding scenarios of the target steel strand for real-time detection;

[0063] Step 302: Based on the real-time detection of the fluctuation signal of the target steel strand in the corresponding scene, the environmental noise is eliminated and the signal-to-noise ratio is improved according to the wavelet threshold noise reduction algorithm and the Kalman filter algorithm, and the fluctuation signal is pre-processed;

[0064] Step 303: Based on the fluctuation signal in the corresponding scenario of real-time detection of the target steel strand, a Wigner-Ville distribution function is used to perform time domain conversion according to the sampling rate of the fluctuation signal to obtain a time-frequency diagram in the corresponding scenario of real-time detection of the target steel strand;

[0065] According to the time-frequency diagram under the corresponding scenario of real-time detection of target steel strands, the energy distribution of each frequency of the fluctuation signal per unit time is marked, and the dispersion curve in the fluctuation signal under the corresponding scenario of real-time detection of target steel strands is extracted as follows:

[0066]

[0067] Among them, WVD(t,f) is the energy distribution value of frequency f at the t-th unit time in the corresponding scenario of real-time detection of target steel strand, x(t) is the fluctuation signal value at the t-th unit time, x * (t) is the complex conjugate operation of the fluctuation signal value under the t-th unit time, τ is the delay variable, e -j2πfτ is a complex exponential function, and dτ is the integral increment of the time variable.

[0068] Step 304: Calculate the mean square error between the dispersion curve in the fluctuation signal in the corresponding scenario of the real-time detection target steel strand and the reference dispersion curve, mark the abnormal deviation curve, and record it as the abnormal fluctuation signal in the corresponding scenario of the real-time detection target steel strand;

[0069] As a further development, a baseline dispersion curve is developed based on the initial testing phase of the target steel strand. An array of ultrasonic guided wave sensors and magnetostrictive sensors are deployed to obtain the target steel strand's healthy state fluctuation signal. This is used as the baseline dispersion curve. During real-time monitoring, the dispersion curve is continuously extracted and compared with the dispersion curve in a healthy state. Based on the changes in the dispersion curve, changes in the corresponding frequency range are marked to identify the damaged area.

[0070] According to the abnormal fluctuation signal in the corresponding scene of real-time detection of target steel strands, the abnormal ultrasonic signal characteristics and abnormal magnetic signal characteristics in the corresponding scene of real-time detection of target steel strands are extracted;

[0071] Based on the abnormal ultrasonic signal features and abnormal magnetic signal features in the corresponding scenario of real-time detection of target steel strands, the DS evidence theory method is used to fuse the ultrasonic signal features and magnetic signal features to obtain the abnormal fusion decision feature set in the corresponding scenario of real-time detection of target steel strands. The method is as follows:

[0072]

[0073] Among them, l k (A) is the confidence of the abnormal fusion decision feature set A for the k-th defect category in the corresponding scenario of real-time detection of the target steel strand, l 1,k (B) is the confidence level of the abnormal ultrasonic signal feature B of the ultrasonic guided wave sensor for the k-th defect category, l 1,k (C) is the confidence level of the abnormal magnetic signal feature C of the magnetostrictive sensor for the kth defect category.

[0074] Step 305: Based on the abnormal fusion decision feature set in the corresponding scenario of real-time detection of the target steel strand, a machine learning SVM support vector machine is trained to generate a hyperplane between the steel strand abnormality category decision edges, and a steel strand state assessment model is obtained; the fluctuation signal in the corresponding scenario of real-time detection of the target steel strand is used as input, and the defect category vector in the corresponding scenario of real-time detection of the target steel strand is used as output, in the following manner:

[0075]

[0076] Among them, G k To detect the kth defect category vector of the target steel strand in real time, f k (x(t)) is the classification decision function sign outputting the k-th defect category label in the corresponding scenario of real-time detection of the target steel strand given the fluctuation signal value at the t-th unit time. is the weight coefficient of the i′th support vector of the kth defect category, is the label data of the i′th support vector of the kth defect category, is the i′th support vector of the kth defect category, b k is the bias term of the k-th defect category, K(·) is the kernel function, S k is the total number of support vectors for the k-th defect category.

[0077] Combined with the content in 301-305:

[0078] As a further content, when multiple groups of array-type ultrasonic guided wave sensors are deployed at several positions of the target steel strand and the pulse echo method is used to transmit the guided wave signal during the propagation process, when the reflected wave changes, it indicates that the current steel strand may have broken wires / corrosion / stress abnormalities, etc., and thus the magnetostrictive sensor deflection detection will be triggered to capture the measured magnetic permeability changes and Barkhausen noise signals in the corresponding ultrasonic detection abnormal area. Compared with the traditional single-mode detection problem of low recognition rate of inner layer damage of steel strands, when the ultrasonic detection detects an abnormality, the magnetic domain deflection detection is automatically triggered, and the ultrasonic signal and the magnetic signal component steel strand's multi-dimensional abnormal feature data are used to verify the defect type, so as to improve the steel strand's recognition accuracy for various abnormal categories.

[0079] Step 4: Match and screen the defect type vector database of each steel strand scenario according to the defect category vector of the target real-time steel strand in the corresponding scenario, and evaluate the defect type influence coefficient of the target real-time steel strand in the corresponding scenario;

[0080] The step 4 includes the following:

[0081] Step 401: Based on the defect type vector database for each scenario of the steel strand, the defect category vector data associated with each scenario of the steel strand is divided to obtain an array of defect category vector data for each scenario of the steel strand;

[0082] The PCA principal component analysis method is used to reduce the dimension of the defect category vector data array under various scenarios of steel strands, and the defect category vector matrix A under various scenarios of steel strands is constructed;

[0083]

[0084] Among them, q j′k is the kth defect type vector data array of the steel strand in the j′th scenario, n′ is the total number of scenarios, and m′ is the total number of defect category vectors;

[0085] Based on each element in the defect category vector matrix of each steel strand scenario, the similarity between the defect category vector and the corresponding defect category vector of the target steel strand in real-time detection is calculated as follows:

[0086]

[0087] Among them, θ j′k To detect the similarity between the k-th defect category vectors of the target steel strand corresponding to the j′-th scenario in real time;

[0088] Step 402: Filter defect category vectors with positive correlation in similarity under the scene corresponding to the real-time detection target steel strand, mark the defect type under the scene corresponding to the real-time detection target steel strand, and assign weights according to the degree of influence of the defect type on the scene corresponding to the real-time detection target steel strand;

[0089] Determine the defect type of the target steel strand corresponding to the scene in real-time detection and focus on external influencing factors; the defect type focus on external influencing factors include: environmental influencing factors, load status;

[0090] Calculate and evaluate the defect impact coefficient of the corresponding scenario of the real-time detection target steel strand according to the defect type and the defect type weight of the corresponding scenario of the real-time detection target steel strand;

[0091] As a further aspect, the weighting of the impact of defect types on the target steel strands to be detected in real time can be based on historical experience and assigned according to AHP hierarchical analysis, or it can be assigned according to fuzzy comprehensive evaluation. For example, the defect type of the target steel strand to be detected in real time at site A is a broken wire, and it is located at a high stress position of the steel strand. Therefore, when assigning values according to AHP hierarchical analysis, the weight assigned to this defect type is higher.

[0092] Step 403: normalize the similarity between the defect category vectors in the scene corresponding to the real-time detection target steel strand, the defect type influence weight in the scene corresponding to the real-time detection target steel strand, and the defect type attention external influencing factor in the scene corresponding to the real-time detection target steel strand, and calculate the defect type influence coefficient in the scene corresponding to the real-time detection target steel strand in the following manner:

[0093]

[0094] Among them, D j′k To detect the kth defect type influence coefficient of the target steel strand in real time under the j′th scenario, θ j′k To detect the similarity normalized value between the k-th defect category vectors of the target steel strand corresponding to the j′-th scene in real time, F j′k Focus on external influencing factors for real-time detection of the kth defect type in the j′th scenario of the target steel strand;

[0095] Combined with the content of 401 to 403:

[0096] As a further development, through scenario-based defect vector matching (PCA dimensionality reduction + dynamic similarity calculation), dynamic multi-source weighting (AHP / fuzzy comprehensive evaluation) and environment-load coupling correction, the precise positioning and risk quantification of steel strand defects are achieved, providing a quantitative basis for intelligent maintenance and traceability decisions.

[0097] Step 5: Perform fitting compensation on the initial performance index of the target steel strand according to the defect type influence coefficient in the corresponding scenario of the target steel strand to be detected in real time, and obtain the corrected performance index of the target steel strand to be detected, as follows:

[0098] P 修 =p·D j ' k

[0099] Among them, P 修 It is the modified performance index for detecting the target steel strand.

[0100] Step 6: Determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, determine the deviation from the production process parameters.

[0101] Reference Figure 2 As shown, the present invention provides a full-process accurate detection and tracing system for steel strand quality, including:

[0102] Initial performance module, comparison database, defect category module, defect impact coefficient module, performance correction module, and abnormal decision module;

[0103] The initial performance module is used to analyze the production processes corresponding to different steel strand manufacturing materials based on the steel strand manufacturing materials and production processes, and determine the initial performance indicators of the target steel strand to be tested;

[0104] The reference database is used to obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for each scenario of steel strands;

[0105] The defect classification module is used to obtain the fluctuation signal of the corresponding scene of the real-time detection target steel strand based on heterogeneous sensors, establish a steel strand state assessment model, and generate the defect classification vector of the corresponding scene of the real-time detection target steel strand;

[0106] The defect influence coefficient module is electrically connected to the defect category module and the reference database. The defect influence coefficient module is used to match and screen the defect type vector database under various scenarios of the steel strand according to the defect category vector under the corresponding scenario of the real-time steel strand being detected, and evaluate the defect type influence coefficient under the corresponding scenario of the real-time target steel strand being detected;

[0107] The performance correction module is electrically connected to the defect influence coefficient module and the initial performance module. The performance correction module is used to perform fitting compensation on the initial performance index of the steel strand to be detected according to the defect type influence coefficient in the corresponding scenario of the real-time detection target steel strand, so as to obtain the corrected performance index of the detection target steel strand;

[0108] The abnormal decision module is electrically connected to the performance correction module. The abnormal decision module is used to determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, the production process is traced.

[0109] This solution achieves accurate detection and traceability of steel strand quality throughout the entire process by constructing a steel strand defect type vector database and combining it with real-time detection data. It can effectively evaluate and compensate for the impact of defects, ensure that steel strand performance meets scenario requirements, promptly detect quality anomalies and trace the source of the production process, and improve the level of steel strand quality control.

[0110] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only for some logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0113] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0114] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for accurate detection and tracing of the quality of steel strands throughout the entire process, characterized in that: include: Based on the manufacturing materials and production processes of steel strands, analyze the corresponding production processes of different manufacturing materials of steel strands and determine the initial performance indicators of the target steel strands to be tested; Obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for each scenario of steel strands; Based on heterogeneous sensors, the fluctuation signal of the target steel strand in the corresponding scenario of real-time detection is obtained, a steel strand status assessment model is established, and the defect category vector of the target steel strand in the corresponding scenario of real-time detection is generated; According to the defect category vector of the target steel strand in the corresponding scene, the defect type vector database of each scene of the steel strand is matched and screened, and the defect type influence coefficient of the target steel strand in the corresponding scene of the real-time detection is evaluated; According to the defect type influence coefficient of the target steel strand in the real-time detection scenario, the initial performance index of the target steel strand is fitted and compensated to obtain the corrected performance index of the target steel strand; Determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, determine the deviation from the production process parameters.

2. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 1, characterized in that: Based on the known steel strand manufacturing materials and known production processes, a knowledge graph of performance increase and decrease conditions between materials and production processes is established; Based on the knowledge graph of performance increase and decrease conditions between materials and production processes, the performance indicators of steel strands were used as independent variables, and the manufacturing materials and production processes of steel strands were used as dependent variables. A machine learning-linear regression algorithm was trained to obtain a material-process performance increase and decrease evaluation model. The manufacturing materials and production process parameters of the target steel strand to be tested are determined and substituted into the material-process performance increase and decrease evaluation model to evaluate the initial performance indicators of the target steel strand to be tested.

3. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 2, characterized in that: Deploy multiple arrays of ultrasonic guided wave sensors and magnetostrictive sensors at several locations of the target steel strands for real-time detection to obtain fluctuation signals corresponding to the target steel strands under real-time detection scenarios. Based on the real-time detection of the fluctuation signal of the target steel strand in the corresponding scene, the wavelet threshold noise reduction algorithm and Kalman filter algorithm are used to eliminate the environmental noise and improve the signal-to-noise ratio, and the fluctuation signal is pre-processed; According to the fluctuation signal in the corresponding scenario of real-time detection of target steel strands, the Wigner-Ville distribution function is used to perform time domain conversion according to the sampling rate of the fluctuation signal to obtain the time-frequency diagram in the corresponding scenario of real-time detection of target steel strands; According to the time-frequency diagram under the corresponding scenario of real-time detection of target steel strands, the energy distribution of each frequency of the fluctuation signal per unit time is marked, and the dispersion curve in the fluctuation signal under the corresponding scenario of real-time detection of target steel strands is extracted.

4. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 3, characterized in that: Calculate the mean square error between the dispersion curve in the fluctuation signal under the corresponding scenario of real-time detection of the target steel strand and the reference dispersion curve, mark the abnormal deviation curve, and record it as the abnormal fluctuation signal under the corresponding scenario of real-time detection of the target steel strand; According to the abnormal fluctuation signal in the corresponding scene of real-time detection of target steel strands, the abnormal ultrasonic signal characteristics and abnormal magnetic signal characteristics in the corresponding scene of real-time detection of target steel strands are extracted; Based on the abnormal ultrasonic signal characteristics and abnormal magnetic signal characteristics in the corresponding scenario of real-time detection of target steel strands, the DS evidence theory method is used to fuse the ultrasonic signal characteristics and magnetic signal characteristics, and the abnormal fusion decision feature set in the corresponding scenario of real-time detection of target steel strands is obtained.

5. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 4, characterized in that: According to the abnormal fusion decision feature set in the corresponding scenario of real-time detection of target steel strands, the machine learning SVM support vector machine is trained to generate the hyperplane between the steel strand abnormality category decision edges, and the steel strand status assessment model is obtained; the fluctuation signal in the corresponding scenario of real-time detection of target steel strands is used as input, and the defect category vector in the corresponding scenario of real-time detection of target steel strands is used as output.

6. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 5, characterized in that: Based on the defect type vector database of each steel strand scenario, the defect category vector data stored in association with each steel strand scenario is divided to obtain an array of defect category vector data of each steel strand scenario; The PCA principal component analysis method is used to reduce the dimension of the defect category vector data array under various scenarios of steel strands, and the defect category vector matrix A under various scenarios of steel strands is constructed; Among them, q j′k is the kth defect type vector data array of the steel strand in the j′th scenario, n′ is the total number of scenarios, and m′ is the total number of defect category vectors; Based on each element in the defect category vector matrix under each scenario of the steel strand, the similarity between the defect category vector under the corresponding scenario of the real-time detection target steel strand is calculated.

7. A method for accurate detection and tracing of the quality of steel strands throughout the entire process according to claim 6, characterized in that: Filter the defect category vectors with positive correlation in the similarity of the corresponding scene of the real-time detection target steel strand, mark the defect type of the corresponding scene of the real-time detection target steel strand, and assign weights according to the degree of influence of the defect type on the corresponding scene of the real-time detection target steel strand; Determine the defect type of the target steel strand corresponding to the scenario in real-time detection and focus on external influencing factors; The defect type focuses on external influencing factors including: environmental factors, load conditions; Calculate and evaluate the defect impact coefficient of the corresponding scenario of the real-time detection target steel strand according to the defect type and the defect type weight of the corresponding scenario of the real-time detection target steel strand; The similarity between the defect category vectors in the corresponding scenario of real-time detection of target steel strands is normalized, and the defect type influence weight in the corresponding scenario of real-time detection of target steel strands and the defect type attention external influencing factors in the corresponding scenario of real-time detection of target steel strands are used to calculate the defect type influence coefficient in the corresponding scenario of real-time detection of target steel strands.

8. A full-process accurate detection and traceability system for steel strand quality, including: Initial performance module, comparison database, defect category module, defect impact coefficient module, performance correction module, and abnormal decision module; The initial performance module is used to analyze the production processes corresponding to different steel strand manufacturing materials based on the steel strand manufacturing materials and production processes, and determine the initial performance indicators of the target steel strand to be tested; The reference database is used to obtain historical test data of steel strands applied in different scenarios, mark the defect categories in each scenario, and establish a defect type vector database for each scenario of steel strands; The defect classification module is used to obtain the fluctuation signal of the corresponding scene of the real-time detection target steel strand based on heterogeneous sensors, establish a steel strand state assessment model, and generate the defect classification vector of the corresponding scene of the real-time detection target steel strand; The defect influence coefficient module is electrically connected to the defect category module and the reference database. The defect influence coefficient module is used to match and screen the defect type vector database under various scenarios of the steel strand according to the defect category vector under the corresponding scenario of the real-time steel strand being detected, and evaluate the defect type influence coefficient under the corresponding scenario of the real-time target steel strand being detected; The performance correction module is electrically connected to the defect influence coefficient module and the initial performance module. The performance correction module is used to perform fitting compensation on the initial performance index of the steel strand to be detected according to the defect type influence coefficient in the corresponding scenario of the real-time detection target steel strand, so as to obtain the corrected performance index of the detection target steel strand; The abnormal decision module is electrically connected to the performance correction module. The abnormal decision module is used to determine whether the corrected performance indicators of the target steel strand meet the performance requirements of the corresponding scenario of the steel strand. If so, the quality of the steel strand is determined to be normal. If not, the quality of the steel strand is determined to be abnormal. According to the defect type corresponding to the quality abnormality, the production process is traced.

Images