A data management system for a wire automated detection platform
By constructing a data platform management system for an automated stranded wire testing platform, the problems of identifying and analyzing the causes of periodic defects in stranded wire production have been solved, and the quality traceability and intelligent management of stranded wire products throughout their entire life cycle have been realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGYIN ELECTRICAL ALLOY
- Filing Date
- 2025-12-19
- Publication Date
- 2026-07-28
AI Technical Summary
Existing technologies struggle to identify periodic defects during the stranding process, and the root cause analysis process cannot correlate defects with production process parameters or raw material batches, making it difficult to improve quality issues and achieve intelligent management.
By constructing a data platform management system, we can achieve vertical mapping of multi-source heterogeneous detection data, establishment of a spectral master data model, construction of a hierarchical data warehouse, and defect spectrum analysis, identify the dominant defect frequency, and match potential causes in the spectral model.
It enables full lifecycle quality traceability of stranded wire products, and can identify the periodic characteristics of defects caused by equipment or process factors in the frequency domain, providing data-driven decision-making to improve processes and control quality.
Smart Images
Figure CN121524253B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of middleware management, specifically relating to a data middleware management system for an automated stranded wire testing platform. Background Technology
[0002] Stranded wire is a fundamental material in power transmission, communication, and precision manufacturing. Modern production lines are typically equipped with various online monitoring systems that generate inspection data during production. However, existing technologies have some limitations: data from different inspection stages and sensors are stored and analyzed independently, lacking a unified spatiotemporal reference for correlation between data sources. This makes it difficult to map data collected at different times to the physical location of the stranded wire product itself, ultimately resulting in the inability to form a complete and continuous quality data profile along the length of the stranded wire.
[0003] When analyzing collected quality data, existing technologies primarily rely on detecting and alarming isolated defect events exceeding preset thresholds. While this can identify product defects, it falls short in recognizing the inherent patterns in the spatial distribution of defects, especially periodic defects. Periodic defects are related to the periodic movement or fixed damage of production equipment and the periodic fluctuations of specific process parameters. Due to the lack of advanced data mining techniques for spectral analysis of defect spatial distribution sequences, periodic features cannot be extracted. Therefore, when quality problems are discovered, the root cause analysis process often relies on engineers' manual experience and offline troubleshooting, making it difficult to correlate defects with specific production process parameters, equipment status, or upstream single-line raw material batches, thus hindering production quality improvement and intelligent management. Summary of the Invention
[0004] This invention provides a data platform management system for an automated stranded wire testing platform to solve the technical problems of existing technologies that make it difficult to identify periodically occurring defects and to correlate defects in the cause analysis process.
[0005] A data platform management system for an automated stranded wire testing platform includes: The data acquisition and synchronization unit is used to acquire multi-source heterogeneous detection data during the stranded wire production process, and to map the multi-source heterogeneous detection data to the longitudinal physical coordinates of the stranded wire products based on the acquisition time and production line speed, forming a longitudinal data profile. The master data modeling unit is used to construct a hierarchical master data model containing data on stranded wire products, single wire raw materials and production processes, and associate the vertical data profile with stranded wire product instances in the hierarchical master data model through the vertical physical coordinates. The data warehouse layered processing unit is used to establish a layered data warehouse: based on the vertical data profile, a data detail layer containing full standardized indicators is constructed; The identification unit is used to construct a data theme layer on top of the data detail layer: extracting the defect spatial distribution sequence along the length direction of the stranded wire product, performing a discrete Fourier transform on the defect spatial distribution sequence to obtain defect spectrum data representing the periodic characteristics of defects, and identifying one or more dominant defect frequencies whose spectrum amplitude exceeds a preset threshold from the defect spectrum data. The application service unit is used to build a data application layer on top of the data theme layer: based on the identified dominant defect frequency or the corresponding physical wavelength, it retrieves and matches production process data or single-line raw material batches that may lead to the periodic characteristics of the defect in the spectral master data model, and performs cause-based annotation.
[0006] Furthermore, the multi-source heterogeneous testing data includes surface images, diameter, and conductivity of the stranded wire products.
[0007] Furthermore, based on the acquisition time and production line speed, the multi-source heterogeneous detection data is mapped to the longitudinal physical coordinates of the stranded wire product, forming a longitudinal data profile, including: Record the acquisition timestamp of each frame of surface defect image or each diameter measurement data. ; Obtain the real-time production line speed provided by the PLC of the stranded wire production equipment. ; Set production start time as By monitoring real-time production line speed From the moment production begins up to the current collection timestamp Integrate to calculate the longitudinal physical coordinate L of the stranded wire product corresponding to the current detection data point, using the following formula: ; The collected multi-source heterogeneous detection data are associated and stored with the calculated longitudinal physical coordinates L to form a longitudinal data profile indexed by the physical coordinates.
[0008] Furthermore, a hierarchical master data model is constructed, including data on stranded wire products, single-wire raw materials, and production processes, comprising: Create a stranded wire product instance entity and assign a unique identifier to each stranded wire product; Create a single-line raw material batch entity and record the supplier, batch number, diameter, and characteristic defect wavelength information of the single-line raw material batch entity; Create a production process parameter entity to record the strand pitch, traction wheel speed, and wire tension parameters used in the production of this stranded wire. Establish relationships: Create bidirectional relationships between the stranded wire product instance entity and the multiple single-wire raw material batch entities used to produce the stranded wire product, as well as the production process parameter entities.
[0009] Furthermore, a data detail layer containing all standardized metrics is constructed, including: For the collected surface defect image data, the type, length, width, and area of the defects are extracted as standardized indicators using image segmentation algorithms; For the collected online diameter measurement data, the average, maximum, minimum, and standard deviation of the online diameter measurement data within the length interval are calculated as standardized indicators; The test data from different sources and the standardized indicators are combined with the unique identifier of the stranded wire product, the vertical physical coordinates, and the collection timestamp, and stored in the data table of the data details layer.
[0010] Furthermore, the spatial distribution sequence of defects is extracted along the length direction of the stranded wire product, including: Set a fixed spatial sampling interval ; The total length of the stranded wire product is divided into N consecutive length units along the longitudinal axis; The sum of the areas of all defects within each length cell is used as the defect signal value for that length cell. ,in, The cell index is i = 1, 2, ..., N; Defect signal values of all length units Arranged in sequence, they form a discrete spatial distribution sequence of defects.
[0011] Further, identifying one or more dominant defect frequencies from the defect spectrum data whose spectral amplitude exceeds a preset threshold includes: Calculate the average value of all spectral amplitudes in the defect spectral data. and standard deviation ; The preset threshold for: ,in, Statistical factor; Traverse the defect spectrum data and set the spectrum amplitude. Greater than the preset threshold frequency Identify the dominant defect frequency.
[0012] Furthermore, the statistical factor The value is 3.
[0013] Further, retrieving and matching production process parameters or single-line raw material batches that may potentially cause the periodic characteristics of the defects in the phylogenetic master data model includes: The dominant defect frequency Through formula Converted to the corresponding physical wavelength ; The circumference information of each rotating component is pre-stored in the production process knowledge base, and the physical wavelength is... Compare with the circumference of the rotating component; if the physical wavelength... If the error between the value and the circumference of a certain rotating component is within the preset tolerance range, then the rotating component and its related process parameters are identified as potential causes. In the spectral master data model, the batches of single-wire raw materials used in stranded wire products are retrieved, and the calculated physical wavelengths are obtained. If the wavelength of the characteristic defect recorded in the single-line raw material batch is compared with the wavelength of the single-line raw material batch, and the error between the two values is within the preset tolerance range, then the single-line raw material batch is matched as a potential cause.
[0014] Furthermore, causal annotation is performed, including: In the hierarchical master data model, a defect event record is created for each stranded wire product instance; The identified dominant defect frequency, the matched potential cause production process parameters or single-line raw material batch information, and the time and vertical coordinate location of the defect event are all written into the defect event record to complete the cause identification.
[0015] The beneficial effects are as follows: Compared with existing technologies, this invention maps multi-source heterogeneous detection data to the longitudinal physical coordinates of the stranded wire based on acquisition time and production line speed. By constructing a spectralized master data model, it associates the longitudinal data profile with specific stranded wire products, single-wire raw material batches, and production process data, establishing a full lifecycle quality traceability chain from raw materials to finished products. By performing a discrete Fourier transform on the defect spatial distribution sequence extracted along the length of the stranded wire product, the periodic characteristics of defects caused by equipment or process factors can be identified from a frequency domain perspective. By matching and retrieving the identified dominant defect frequencies with the spectralized data, the cause of quality problems can be traced back to production, providing a data-driven decision-making basis for process improvement and quality control. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the data detail layer; Figure 2 This is a schematic diagram of the defect spectrum. Figure 3 This is a schematic diagram of the spatial distribution sequence of defects. Detailed Implementation
[0017] An embodiment of the data platform management system for an automated stranded wire testing platform provided by this invention: A data platform management system for an automated stranded wire testing platform includes: The data acquisition and synchronization unit is used to acquire multi-source heterogeneous detection data during the stranded wire production process, and to map the multi-source heterogeneous detection data to the longitudinal physical coordinates of the stranded wire products based on the acquisition time and production line speed, forming a longitudinal data profile.
[0018] Specifically, sensors such as visual inspection cameras, laser diameter gauges, and eddy current flaw detectors deployed on the production line collect surface images, diameter, conductivity, and other inspection data in real time, while simultaneously recording the timestamps of the collected data. Real-time production line speed is obtained from the production control system. By integrating or multiplying the difference between the production line speed and the timestamp since production began, the longitudinal physical coordinates of the stranded wire product corresponding to each inspection data point are calculated. Data from different sensors within the same longitudinal physical coordinate point or a very small coordinate range are aggregated to form a data record with physical coordinates as the primary key and containing multiple quality indicators. All data records are arranged in coordinate order to form a longitudinal data profile.
[0019] In an optional embodiment, multi-source heterogeneous testing data is mapped to the longitudinal physical coordinates of the stranded wire product based on the acquisition time and production line speed to form a longitudinal data profile, including: Record the acquisition timestamp of each frame of surface defect image or each diameter measurement data. ; Obtain the real-time production line speed provided by the PLC of the stranded wire production equipment. ; Set production start time as By monitoring real-time production line speed From the moment production begins up to the current collection timestamp Integrate to calculate the longitudinal physical coordinate L of the stranded wire product corresponding to the current detection data point, using the following formula: ; The collected multi-source heterogeneous detection data are associated and stored with the calculated longitudinal physical coordinates L to form a longitudinal data profile indexed by the physical coordinates.
[0020] Specifically, assuming that stranding production begins at 8:00 AM, at this time... The time recorded is 08:00:00. The production line speed is not constant; the PLC of the stranding production equipment reports the real-time production line speed at a frequency of once per second. At 08:05:30, the vision inspection camera captured a scratch, and the timestamp t at this moment was recorded as 08:05:30. To calculate the physical location of this scratch on the stranded wire, all real-time production line speeds during the 330 seconds from 08:00:00 to 08:05:30 were retrieved, for example, the production line speed gradually increased from an initial 100 m / min and stabilized at 120 m / min.
[0021] By numerically integrating the velocity curve over this period, the longitudinal physical coordinate L is calculated. For example, a simple calculation shows that the average velocity for the first 60 seconds is 105 m / min, and the average velocity for the next 270 seconds is 120 m / min, then the longitudinal physical coordinate L is 645 m. The image data and area size information of the scratch are associated with and stored with the longitudinal physical coordinate 645 m. Similarly, the laser diameter gauge detects an anomaly in diameter at 08:06:00, and this anomaly is calculated by integrating from... At a speed of 08:06:00, the corresponding longitudinal physical coordinate is calculated to be approximately 690m. The diameter data is then correlated with this longitudinal physical coordinate to form a quality data profile along the entire length of the stranded wire product.
[0022] The master data modeling unit is used to construct a spectral master data model containing data on stranded wire products, single-wire raw materials, and production processes, and to associate the vertical data profile with stranded wire product instances in the spectral master data model through the vertical physical coordinates.
[0023] Specifically, the master data modeling unit defines key data entities and their relationships in the database, including stranded wire products, single-wire raw materials, production work orders, and process parameters. For example, the master data modeling unit creates a stranded wire product instance. This instance not only includes its own product batch number and specifications but also links to the batch numbers of multiple single-wire raw materials used to produce the stranded wire product. It also links to the work order number that performed the production and the specific process parameters such as the stranding pitch and traction tension corresponding to that work order. In this way, a traceable, hierarchical data chain is formed.
[0024] Subsequently, the longitudinal data profile generated by the data acquisition and synchronization unit is treated as a whole as a data object. It is then associated with the corresponding stranded wire product instance in the spectral master data model through the product batch number to which it belongs, so that any longitudinal physical coordinate point in the longitudinal data profile can be traced back to the production background information.
[0025] In an optional embodiment, a hierarchical master data model is constructed, including data on stranded wire products, single-wire raw materials, and production processes, comprising: Create a stranded wire product instance entity and assign a unique identifier to each stranded wire product; Create a single-line raw material batch entity and record the supplier, batch number, diameter, and characteristic defect wavelength information of the single-line raw material batch entity; Create a production process parameter entity to record the strand pitch, traction wheel speed, and wire tension parameters used in the production of this stranded wire. Establish relationships: Create bidirectional relationships between the stranded wire product instance entity and the multiple single-wire raw material batch entities used to produce the stranded wire product, as well as the production process parameter entities.
[0026] Specifically, when production of a new spool of seven-core stranded wire begins, a unique identifier is generated, such as PROD20240401A01, and a stranded wire product instance is created based on this identifier. The source information of the seven individual wires that make up the spool of stranded wire is recorded: the operator creates a single-wire raw material batch entity for each single wire by scanning a barcode or manually entering the information. For example, the supplier of the first single wire is Company A, batch number A-SN20240315, nominal diameter 1.5mm; the supplier of the second single wire is Company B, batch number B-SN20240320, and historical data indicates that this batch of single-wire raw materials exhibits periodic indentations with a characteristic wavelength of 5m.
[0027] The process parameters set for this production task are recorded, such as a strand pitch of 120mm, a target traction wheel speed of 300rpm, and a tension of 50N for each of the seven pay-off reels. These data constitute a production process parameter entity. Through database operations, links are established between these entities, linking the stranded wire product instance entity PROD20240401A01 to the seven single-wire raw material batch entities and the created production process parameter entity. By querying the identifier PROD20240401A01, it is possible to trace which batches of raw materials were used and under what process conditions it was produced.
[0028] In a preferred embodiment, the longitudinal physical coordinates are calculated based on time integration. The method assumes that the stranded wire remains at a constant length during the traction process. However, in actual production, due to factors such as tension fluctuations, temperature changes, or material creep, metal stranded wires (such as copper and aluminum) may undergo elastic or plastic elongation, resulting in a deviation between the theoretical coordinates and the actual physical position.
[0029] To correct this error, this system further introduces an online length calibration mechanism: a high-precision encoder length measuring device (such as a laser interferometer or rotating encoder wheel) is installed at the end of the production line to measure the actual output length of the stranded wire in real time. The system periodically (e.g., every 100 meters or every 5 minutes) calculates the theoretical length using integrals. and Compare and calculate the scaling factor. And this factor is used to dynamically correct the historical and subsequent vertical coordinates: ,in, This is the corrected length. This compensation mechanism significantly improves defect location accuracy, ensuring that spectral analysis and cause matching are based on accurate physical coordinates.
[0030] The data warehouse layered processing unit is used to establish a layered data warehouse: based on the vertical data profile, a data detail layer containing full standardized indicators is constructed.
[0031] Specifically, the data warehouse layered processing unit cleans and standardizes heterogeneous data from different sensors. For example, it converts pixel defects in image data into standard defect type codes and size values, and normalizes measurement values of different dimensions to a unified range. Using the vertical physical coordinates of the stranded wire product as the core dimension, a wide-table structure is created for the data detail layer, such as... Figure 1 As shown in the table, each row represents a discrete physical location on the stranded wire product, and the columns contain all available quality inspection indicators for that discrete physical location, such as wire diameter, surface defect level, and ellipticity, ensuring the integrity and consistency of the data.
[0032] In an optional embodiment, a data detail layer containing all standardized metrics is constructed, including: For the collected surface defect image data, the type, length, width, and area of the defects are extracted as standardized indicators using image segmentation algorithms; For the collected online diameter measurement data, the average, maximum, minimum, and standard deviation of the online diameter measurement data within the length interval are calculated as standardized indicators; The test data from different sources and the standardized indicators are combined with the unique identifier of the stranded wire product, the vertical physical coordinates, and the collection timestamp, and stored in the data table of the data details layer.
[0033] Specifically, when an online camera detects a defect in an image, the image processing module immediately runs. It segments the defect from the background using edge detection and region growing algorithms, and calculates the defect's geometric properties. For example, if a scratch-type defect is identified, with a longest axis length of 12mm, a maximum vertical width of 0.8mm, and a total pixel count translating to an actual area of 9.6... These metrics, namely defect type, length, width, and area, are extracted as standardized structured data.
[0034] Statistical calculations were performed every 5 meters. For example, for a product measuring 100m to 105m, all diameter measurements were collected, resulting in 1000 data points. The average diameter of these 1000 data points was calculated to be 7.02mm, the maximum to be 7.05mm, the minimum to be 6.99mm, and the standard deviation to be 0.015mm. These statistical values constituted the standardized diameter index for that length. Whether the defect indicators came from images or the statistical indicators from the diameter gauge, they were all appended with the same unique product identifier, their corresponding physical coordinates, and a timestamp, and stored in a unified database table, forming a comprehensive and consistently formatted layer of quality data.
[0035] The identification unit is used to construct a data theme layer on top of the data detail layer: extracting the spatial distribution sequence of defects along the length direction of the stranded wire product, performing a discrete Fourier transform on the spatial distribution sequence of defects to obtain defect spectrum data representing the periodic characteristics of defects, and identifying one or more dominant defect frequencies whose spectral amplitude exceeds a preset threshold from the defect spectrum data.
[0036] Specifically, from the data detail layer, for a specific defect type (such as surface scratches), the identification unit extracts a binary sequence of 0s and 1s sequentially along the longitudinal physical coordinates of the strand, where 1 represents the presence of a defect at that coordinate point, and 0 represents the absence of a defect. The Fast Fourier Transform algorithm is applied to the defect spatial distribution sequence to transform it from the spatial domain to the frequency domain, resulting in a defect spectrum. The horizontal axis of the defect spectrum represents spatial frequency, and the vertical axis represents the amplitude of the corresponding frequency component, such as... Figure 2 As shown, the identification unit scans the defect spectrum and identifies frequency points whose amplitude exceeds a preset statistical threshold as the dominant defect frequencies representing periodic defects. Frequency points exceeding the preset statistical threshold can be frequency points that exceed three standard deviations of the average amplitude.
[0037] In an optional embodiment, extracting the spatial distribution sequence of defects along the length direction of the stranded wire product includes: Set a fixed spatial sampling interval ; The total length of the stranded wire product is divided into N consecutive length units along the longitudinal axis; The sum of the areas of all defects within each length cell is used as the defect signal value for that length cell. ,in, The cell index is i = 1, 2, ..., N; Defect signal values of all length units Arranged in sequence, they form a discrete spatial distribution sequence of defects.
[0038] Specifically, for a stranded wire product with a total length of 1000m, a fixed spatial sampling interval is set, such as 0.1m. The entire stranded wire product is divided into 10,000 consecutive length units. The first length unit covers the range from 0 to 0.1m, the second length unit covers the range from 0.1 to 0.2m, and so on, until the ten thousandth length unit covers the range from 999.9 to 1000m.
[0039] Traverse all defect data that have been mapped to the vertical physical coordinates, such as Figure 3 As shown, suppose two defects are found within the 53rd length unit, i.e., between 5.2m and 5.3m, one with an area of 2... The other is 3.5. Then the defect signal value of that length unit The calculation is 5.5. If no defects are found within the 54th length unit, then... The value is 0. The same calculation is performed on all 10,000 units to obtain an ordered list containing 10,000 values, in the form of s(1), s(2), ..., s(10,000).
[0040] In an optional embodiment, identifying one or more dominant defect frequencies from the defect spectrum data whose spectral amplitude exceeds a preset threshold includes: Calculate the average value of all spectral amplitudes in the defect spectral data. and standard deviation ; The preset threshold for: ,in, Statistical factor; Traverse the defect spectrum data and set the spectrum amplitude. Greater than the preset threshold frequency Identify the dominant defect frequency.
[0041] Specifically, a Fast Fourier Transform (FFT) is performed on the generated defect spatial distribution sequence to obtain a set of frequencies and their corresponding spectral amplitude data. Assuming the transform yields 1024 discrete frequency points, the average value of these 1024 amplitudes is calculated. The result is 0.8. Calculate the standard deviation of the amplitude. The result was 0.3.
[0042] According to the preset statistical algorithm, the statistical factors are... Set to 3, preset threshold Set to average Add three standard deviations ,Right now The value is 1.7. (Preset threshold) This represents a statistically significant outlier, far exceeding the level of most random noise. The amplitude of each frequency point is examined individually. For example, a frequency of 0.25 Hz with an amplitude of 2.1, greater than 1.7, is identified and marked as a dominant defect frequency. Another frequency point, 0.5 Hz, with an amplitude of 1.2, less than 1.7, is ignored. By iterating through all frequency points, all frequencies with amplitudes exceeding 1.7 are identified as the characteristic frequencies of the periodic defects requiring attention.
[0043] Among them, statistical factors It is a key parameter for defining the statistical significance of periodic defects, which is usually based on experience in industrial statistical process control. =3 was used as the initial value, and tests were conducted based on the actual noise level and quality requirements of the production line. If there were too many false alarms, the value was increased. The value decreases if a critical defect is missed. value.
[0044] In a preferred embodiment, statistical factors The value is determined based on the distribution characteristics analysis of historical defect spectrum data. By performing discrete Fourier transform on the defect spatial distribution sequence of multiple batches of stranded wire products and statistically analyzing the distribution pattern of their spectral amplitudes, it was found that the amplitudes of most non-periodic noise components approximately follow a normal or quasi-normal distribution. Based on this, the "3" method commonly used in industrial statistical process control is adopted. "Principles" Using 3 as the initial threshold setting can effectively distinguish random noise from defect signals with significant periodicity.
[0045] In addition, the system supports dynamic adjustment based on the actual noise level of the production line. Value: Increase if the false alarm rate is too high. (e.g., set to 3.5 or 4); if important periodic defects are missed, reduce the value appropriately. (e.g., set to 2.5). This adjustment mechanism can be configured through the human-machine interface or automatically optimized by the system based on historical feedback of misjudgments / missed judgments.
[0046] The application service unit is used to build a data application layer on top of the data theme layer: based on the identified dominant defect frequency or the corresponding physical wavelength, it retrieves and matches production process data or single-line raw material batches that may lead to the periodic characteristics of the defect in the spectral master data model, and performs cause-based annotation.
[0047] Specifically, the identified dominant defect frequency is converted into its corresponding physical wavelength, for example, by calculating the reciprocal of the dominant defect frequency. A pre-built production equipment and process knowledge base is queried. This knowledge base stores the circumference of rotating components (such as drums and guide wheels) and the physical dimensions of process parameters (such as hinge pitch) that may have a periodic effect. The calculated physical wavelength of the defect is compared with the physical dimensions in the knowledge base to find matching or multiple-related items. For example, if the physical wavelength of a defect matches the circumference of a guide wheel, that guide wheel is identified as a possible source of the defect, and the periodic characteristic event of this defect is associated with the asset information of that guide wheel in the data application layer.
[0048] In an optional embodiment, based on the identified dominant defect frequency or corresponding physical wavelength, the system retrieves and matches in the spectralized master data model production process parameters or single-line raw material batches that may lead to the periodicity of the defects, including: The dominant defect frequency Through formula Converted to the corresponding physical wavelength ; The circumference information of each rotating component is pre-stored in the production process knowledge base, and the physical wavelength is... Compare with the circumference of the rotating component; if the physical wavelength... If the error between the value and the circumference of a certain rotating component is within the preset tolerance range, then the rotating component and its related process parameters are identified as potential causes. In the spectral master data model, the batches of single-wire raw materials used in stranded wire products are retrieved, and the calculated physical wavelengths are obtained. If the wavelength of the characteristic defect recorded in the single-line raw material batch is compared with the wavelength of the single-line raw material batch, and the error between the two values is within the preset tolerance range, then the single-line raw material batch is matched as a potential cause.
[0049] Specifically, assuming the analysis identifies a dominant defect frequency of 0.5 times / m, the corresponding physical wavelength of this defect is calculated to be 2m. This indicates that a similar defect occurs every 2m on the stranded wire product. A pre-set equipment knowledge base is consulted, which records the circumference of all rotating parts on the production line. For example, the circumference of traction wheel A is 2.01m, and the circumference of guide wheel B is 0.5m. Comparing the calculated wavelength of 2m with the circumference, the difference between 2m and the circumference of traction wheel A (2.01m) is only 0.5%, far less than the preset tolerance of 2%. Therefore, it is preliminarily determined that traction wheel A is likely the source of the defect.
[0050] Using the unique identifier of the current stranded wire product, the batch information of all single-wire raw materials used was searched in the spectral master data model. The query results showed that one single-wire raw material used in this stranded wire product came from supplier C, with batch number C-SN20240322. This batch of single-wire raw material had a characteristic defect with a period of approximately 1.98m noted in its historical records. Comparing the wavelength of 2m with the characteristic defect wavelength of 1.98m of the single-wire raw material, the two were found to be very close. Therefore, the raw material from supplier C was also listed as another potential cause of the defect.
[0051] In an optional embodiment, the cause annotation includes: In the hierarchical master data model, a defect event record is created for each stranded wire product instance; The identified dominant defect frequency, the matched potential cause production process parameters or single-line raw material batch information, and the time and vertical coordinate location of the defect event are all written into the defect event record to complete the cause identification.
[0052] Specifically, after confirming a defect with a 2-meter cycle on the stranded wire product PROD20240401A01, a new associated object, namely a defect event record, was created under the data model of this product instance and assigned a unique ID, such as DEFEVENT-001. This defect event record is specifically used to archive the entire defect analysis process.
[0053] The key information derived from the analysis is entered into the various fields of the defect event record. For example, it records the dominant defect frequency of 0.5 times / m and the calculated physical wavelength of 2m. In the potential cause field, two entries are added: the first is for equipment reasons, pointing to traction wheel A, and the second is for material reasons, pointing to single-line raw material batch C-SN20240322. Additionally, the physical location range where the periodic characteristic of this defect mainly occurs is recorded, such as from 350m to 820m of the stranded wire product, along with the corresponding production time period. After completing and saving the information, the cause annotation for this periodic defect characteristic is complete. This information is stored in the digital archive of the stranded wire product, allowing quality engineers to review and analyze it at any time.
[0054] In a preferred embodiment, a preset tolerance range is used to determine whether the calculated physical wavelength of the defect matches the perimeter of the equipment component or the characteristic wavelength of the raw material. This tolerance range can be determined comprehensively based on equipment manufacturing tolerances, sensor accuracy, and process stability.
[0055] In a typical implementation scenario, the preset tolerance range is set to... 2%, for example, if a traction wheel has a nominal circumference of 2.00 m, its actual machining and installation tolerance is usually within ±1 cm (i.e., ±0.5%). Considering measurement noise and slight slippage, the total tolerance is relaxed to ±2% (i.e., ±0.04 m). Therefore, when the calculated defect wavelength... satisfy: ,in If the wavelength is the circumference of the rotating component or the characteristic defect wavelength of the raw material, then the match is considered successful.
[0056] This tolerance value can be adjusted by process engineers in the system configuration interface according to the specific production line conditions, in order to balance matching sensitivity and mismatch risk.
[0057] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.
Claims
1. A data hub management system for a strand automated detection platform, characterized by, include: The data acquisition and synchronization unit is used to acquire multi-source heterogeneous detection data during the stranded wire production process, and to map the multi-source heterogeneous detection data to the longitudinal physical coordinates of the stranded wire products based on the acquisition time and production line speed, forming a longitudinal data profile. The master data modeling unit is used to construct a hierarchical master data model containing data on stranded wire products, single wire raw materials and production processes, and associate the vertical data profile with stranded wire product instances in the hierarchical master data model through the vertical physical coordinates. The data warehouse layered processing unit is used to establish a layered data warehouse: based on the vertical data profile, a data detail layer containing full standardized indicators is constructed; The identification unit is used to construct a data theme layer on top of the data detail layer: extracting the defect spatial distribution sequence along the length direction of the stranded wire product, performing a discrete Fourier transform on the defect spatial distribution sequence to obtain defect spectrum data representing the periodic characteristics of defects, and identifying one or more dominant defect frequencies whose spectrum amplitude exceeds a preset threshold from the defect spectrum data. The application service unit is used to build a data application layer on top of the data theme layer: based on the identified dominant defect frequency or the corresponding physical wavelength, it retrieves and matches production process data or single-line raw material batches that may lead to the periodic characteristics of the defect in the spectral master data model, and performs cause-based annotation.
2. The data hub management system for the strand automated detection platform of claim 1, wherein, Multi-source heterogeneous testing data includes surface images, diameter, and conductivity of stranded wire products.
3. The data hub management system for the strand automated detection platform of claim 1, wherein, Based on the acquisition time and production line speed, multi-source heterogeneous testing data is mapped to the longitudinal physical coordinates of stranded wire products, forming a longitudinal data profile, including: recording a time stamp of the acquisition of each frame of surface defect image or each diameter measurement data ; Obtain the real-time production line speed provided by the PLC of the stranded wire production equipment. ; Set production start time as By monitoring real-time production line speed From the moment production begins up to the current collection timestamp Integrate to calculate the longitudinal physical coordinate L of the stranded wire product corresponding to the current detection data point, using the following formula: ; The collected multi-source heterogeneous detection data are associated and stored with the calculated longitudinal physical coordinates L to form a longitudinal data profile indexed by the physical coordinates.
4. The data platform management system for an automated stranded wire testing platform according to claim 1, characterized in that, Construct a hierarchical master data model that includes data on stranded wire products, single-wire raw materials, and production processes, including: Create a stranded wire product instance entity and assign a unique identifier to each stranded wire product; Create a single-line raw material batch entity and record the supplier, batch number, diameter, and characteristic defect wavelength information of the single-line raw material batch entity; Create a production process parameter entity to record the strand pitch, traction wheel speed, and wire tension parameters used in the production of this stranded wire. Establish relationships: Create bidirectional relationships between the stranded wire product instance entity and the multiple single-wire raw material batch entities used to produce the stranded wire product, as well as the production process parameter entities.
5. The data platform management system for an automated stranded wire testing platform according to claim 4, characterized in that, Construct a data detail layer containing all standardized metrics, including: For the collected surface defect image data, the type, length, width, and area of the defects are extracted as standardized indicators using image segmentation algorithms; For the collected online diameter measurement data, the average, maximum, minimum, and standard deviation of the online diameter measurement data within the length interval are calculated as standardized indicators; The test data from different sources and the standardized indicators are combined with the unique identifier of the stranded wire product, the vertical physical coordinates, and the collection timestamp, and stored in the data table of the data details layer.
6. The data platform management system for an automated stranded wire testing platform according to claim 1, characterized in that, Extract the spatial distribution sequence of defects along the length of the stranded wire product, including: Set a fixed spatial sampling interval ; The total length of the stranded wire product is divided into N consecutive length units along the longitudinal axis; The sum of the areas of all defects within each length cell is used as the defect signal value for that length cell. ,in, The cell index is i = 1, 2, ..., N; Defect signal values of all length units Arranged in sequence, they form a discrete spatial distribution sequence of defects.
7. The data platform management system for an automated stranded wire testing platform according to claim 6, characterized in that, Identifying one or more dominant defect frequencies from the defect spectrum data whose spectral amplitude exceeds a preset threshold includes: Calculate the average value of all spectral amplitudes in the defect spectral data. and standard deviation ; The preset threshold for: ,in, Statistical factor; Traverse the defect spectrum data and set the spectrum amplitude. Greater than the preset threshold frequency Identify the dominant defect frequency.
8. The data platform management system for an automated stranded wire testing platform according to claim 7, characterized in that, The statistical factors The value is 3.
9. The data platform management system for an automated stranded wire testing platform according to any one of claims 1-8, characterized in that, Retrieving and matching production process parameters or single-line raw material batches that potentially cause the periodicity of the defects in the phylogenetic master data model includes: The dominant defect frequency Through formula Converted to the corresponding physical wavelength ; The circumference information of each rotating component is pre-stored in the production process knowledge base, and the physical wavelength is... Compare with the circumference of the rotating component; if the physical wavelength... If the error between the value and the circumference of a certain rotating component is within the preset tolerance range, then the rotating component and its related process parameters are identified as potential causes. In the spectral master data model, the batches of single-wire raw materials used in stranded wire products are retrieved, and the calculated physical wavelengths are obtained. If the wavelength of the characteristic defect recorded in the single-line raw material batch is compared with the wavelength of the single-line raw material batch, and the error between the two values is within the preset tolerance range, then the single-line raw material batch is matched as a potential cause.
10. The data platform management system for an automated stranded wire testing platform according to claim 9, characterized in that, Cause annotation includes: In the hierarchical master data model, a defect event record is created for each stranded wire product instance; The identified dominant defect frequency, the matched potential cause production process parameters or single-line raw material batch information, and the time and vertical coordinate location of the defect event are all written into the defect event record to complete the cause identification.