Intelligent online transposed conductor detection system

Abstract

The invention belongs to the technical field of transposed conductor detection, and discloses a transposed conductor intelligent on-line detection system, which comprises a sensor module used for collecting surface images, temperature distribution, electrical parameters and internal defect signals of a transposed conductor; and the data acquisition and processing module is connected with the sensor module. According to the invention, the surface defect identification capability breaks through the 10 [mu] m level resolution, the high-frequency eddy current probe adopts the excitation frequency greater than or equal to 5 MHz, accurate positioning of 30 [mu] m level defects in the lead is realized, the active vibration reduction mechanism is utilized to control the imaging offset within 5 [mu] m through real-time vibration compensation, and the detection accuracy is improved. The three parts cooperatively work to form a surface-interior-environment full-dimension detection system, so that the system can still keep the detection precision of + / -5 microns at the production line speed of 2m / s, the false detection rate is reduced to 0.1% or below, and meanwhile, the technical problem that surface and interior detection data are difficult to associate in a traditional method is solved; and the comprehensive performance of the transposed conductor detection system is obviously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of transposed conductor detection, and in particular relates to an intelligent online detection system for transposed conductors. Background Art

[0002] The intelligent online detection system for transposed conductors is an automated monitoring system that combines modern sensing technology, data analysis, and artificial intelligence. It is specifically used to detect quality defects and performance parameters of transposed conductors (winding conductors in power equipment such as transformers and motors) in real time during production or operation.

[0003] Existing intelligent online inspection systems for transposed conductors have a limited ability to detect surface defects due to the use of traditional visual systems that use single-visible light imaging. This has limited ability to identify micron-level insulation damage, especially fine cracks due to interference from metal reflections. For internal defect detection, conventional eddy current testing lacks sensitivity to tiny bubbles, while ultrasonic testing is affected by the twisted structure of the conductors, generating signal interference. Furthermore, the system lacks effective multimodal data fusion capabilities, making it difficult to accurately align surface and internal inspection results. These technical bottlenecks severely restrict inspection accuracy and reliability, necessitating innovative solutions such as multispectral imaging and high-frequency eddy currents to overcome these limitations. Summary of the Invention

[0004] The object of the present invention is to provide an intelligent online detection system for transposed conductors to solve the problems raised in the above background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: an intelligent online detection system for transposed conductors, comprising:

[0006] The sensor module collects the surface image, temperature distribution, electrical parameters and internal defect signals of the transposed conductor;

[0007] The data acquisition and processing module is connected to the sensor module and is used to filter, fuse and extract features from multi-source data;

[0008] The calculation and analysis module, equipped with AI algorithms, analyzes the characteristics of collected and processed data, identifies and determines the defect type of the transposed conductor and outputs the detection results;

[0009] The control and execution module links the production line to perform sorting and process adjustments based on the test results obtained through calculation and analysis;

[0010] Communication module, which completes data interaction between modules and cloud remote monitoring;

[0011] The sensor module includes: a multispectral imaging unit for enhancing the detection of surface defects in the insulation layer; a high-frequency eddy current probe (≥5MHz) for identifying internal micron-level defects; and an active vibration reduction mechanism for suppressing the interference of mechanical vibration on imaging.

[0012] Preferably, the multispectral imaging unit includes a 365nm UV LED light source and a visible light source, the camera pixel size is ≤3.45μm, and it is equipped with a 5X telecentric lens, with a detection accuracy of ±5μm, and the signal-to-noise ratio is improved to ≥30dB through dual-channel image fusion.

[0013] Preferably, the high-frequency eddy current probe has an operating frequency of 5-10 MHz, a probe spacing of ≤1 mm, a detection sensitivity of ≥0.5 mV / μm, and adopts differential eddy current technology, with a common mode rejection ratio of ≥60 dB.

[0014] Preferably, the calculation and analysis module uses a convolutional neural network (CNN) to identify image defects, and is supplemented by a support vector machine (SVM) to determine electrical parameter anomalies.

[0015] Preferably, the control and execution module links the mechanical sorting device via PLC to remove unqualified transposed conductors in real time and feeds back the execution status to the database.

[0016] Preferably, the communication module supports 5G / Industrial Ethernet, uploading the detection data to the cloud platform to achieve remote monitoring and historical data analysis.

[0017] Preferably, the data acquisition and processing module integrates an adaptive filtering algorithm to eliminate industrial noise in real time. The adaptive filtering algorithm adopts db4 wavelet basis and Kalman filtering, and the signal-to-noise ratio after composite filtering is ≥40dB.

[0018] Preferably, the calculation and analysis module further includes: a motion compensation algorithm for correcting dynamic detection errors; an adaptive eddy current parameter adjustment unit for automatically optimizing detection parameters according to the wire material; and an aging model compensation unit for dynamically adjusting the insulation layer defect determination threshold.

[0019] An intelligent online detection method for transposed conductors includes the following detection steps:

[0020] S1. Collecting the surface image, temperature and electrical parameters of the transposed conductor through the sensor module;

[0021] S2. Filter and fuse the collected data in real time to extract defect features;

[0022] S3, classify and locate defects based on deep learning models;

[0023] S4. Trigger production line sorting or alarm based on the detection results.

[0024] The beneficial effects of the present invention are as follows:

[0025] (1) The present invention introduces a multispectral imaging unit into the sensor module and cooperates with dual-band collaborative detection to enable the surface defect recognition capability to exceed 10μm resolution. By using a high-frequency eddy current probe with an excitation frequency of ≥5MHz, the precise positioning of 30μm defects inside the conductor is achieved. The active vibration reduction mechanism is used to control the imaging offset within 5μm through real-time vibration compensation. The three work together to form a full-dimensional detection system of "surface-interior-environment", so that the system can still maintain a detection accuracy of ±5μm at a production line speed of 2m / s, and the false detection rate is reduced to below 0.1%. At the same time, it solves the technical problem of the difficulty in associating surface and internal detection data in traditional methods, and significantly improves the comprehensive performance of the transposed conductor detection system.

[0026] (2) This invention uses a CNN neural network to perform deep feature extraction on surface images, achieving a micro-defect recognition accuracy exceeding 10μm, and a detection efficiency five times higher than that of traditional algorithms. Combined with the SVM's strict pattern determination of electrical parameters, a "visual-electrical" dual-dimensional judgment system is constructed, controlling the misjudgment rate to below 0.1%. This hybrid intelligent algorithm architecture leverages the advantages of CNN in image processing while retaining the high reliability of SVM in analyzing small sample electrical data. This enables the system to simultaneously possess sub-millimeter defect localization capabilities and millisecond-level real-time response performance, perfectly adapting to the detection needs of high-speed production lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A schematic diagram of the system of the present invention;

[0028] Figure 2 Schematic diagram of the specific contents of the sensor module of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] like Figures 1 to 2 As shown, an embodiment of the present invention provides an intelligent online detection system for transposed conductors, 1. An intelligent online detection system for transposed conductors, comprising:

[0031] The sensor module collects the surface image, temperature distribution, electrical parameters and internal defect signals of the transposed conductor;

[0032] The data acquisition and processing module is connected to the sensor module and is used to filter, fuse and extract features from multi-source data;

[0033] The calculation and analysis module, equipped with AI algorithms, identifies the defect types of transposed conductors and outputs the detection results;

[0034] The control and execution module links the production line to perform sorting and process adjustments based on the test results;

[0035] Communication module, which completes data interaction between modules and cloud remote monitoring;

[0036] Among them, the sensor module includes: a multispectral imaging unit for enhancing the detection of surface defects in the insulation layer; a high-frequency eddy current probe (≥5MHz) for identifying internal micron-level defects; and an active vibration reduction mechanism to suppress the interference of mechanical vibration on imaging.

[0037] Among them, the multispectral imaging unit includes a 365nm UV LED light source and a visible light source. The camera pixel size is ≤3.45μm. Combined with a 5X telecentric lens, the detection accuracy is ±5μm, and the signal-to-noise ratio is improved to ≥30dB through dual-channel image fusion.

[0038] The multispectral imaging unit also includes a pulse-modulated light source (1-10kHz) and a linear polarizer (extinction ratio ≥100:1); a dual-telecentric optical path design (object / image telecentricity ≤0.01°); FPGA-triggered HDR imaging (dynamic range ≥90dB); and a dual-channel adaptive fusion algorithm based on wavelet transform and U-Net.

[0039] Pulse modulation technology uses high-frequency pulse drive (1-10kHz) to make the UV LED and visible light source flash alternately to avoid light pollution interference; and a linear polarizer (extinction ratio ≥100:1) is added to the visible light channel to suppress metal surface reflections; dynamic light intensity adjustment is performed to automatically adjust the light source power (10-100%) based on the surface reflectivity of the wire (0.1-0.9) to avoid overexposure or underexposure, comprehensively reducing ambient light interference, improving the signal-to-noise ratio by ≥5dB, and eliminating reflection interference. Polarized light can effectively suppress metal reflections, increasing the detection rate of surface defects (such as scratches) by 20%, and ensuring that clear images of wires of different materials (copper / aluminum) can be obtained through dynamic light intensity adjustment.

[0040] Among them, the high-frequency eddy current probe has an operating frequency of 5-10MHz, a probe spacing of ≤1mm, a detection sensitivity of ≥0.5mV / μm, and adopts differential eddy current technology with a common mode rejection ratio of ≥60dB.

[0041] Specifically, a high-frequency oscillation circuit is used, and an LC resonant circuit is combined with a low-noise amplifier to ensure high-frequency signal stability. Within the 5-10MHz frequency adjustable range, continuous adjustment of 5-10MHz is achieved through a digital potentiometer or varactor diode. When operating in the high-frequency band, the depth is controlled within the range of 0.02-0.1mm, which can effectively detect tiny defects (≥30μm) under the insulation layer. Compared with traditional 1-2MHz eddy current testing, the resolution is improved by 5 times;

[0042] By controlling the probe distance to ≤1mm, and using ceramic guide rails and air bearings to ensure motion accuracy of ±0.01mm, and integrating eddy current displacement sensors to monitor the distance in real time (sampling rate 1kHz), a micro cylinder is used to maintain a constant contact pressure (0.5N±0.05N), and the distance is strictly controlled to stabilize the detection sensitivity at 0.5mV / μm±3%, avoiding signal attenuation caused by distance changes (traditional systems can fluctuate up to 20%).

[0043] The minimum identifiable defect size is reduced from 100μm to 30μm, and the operating temperature range is expanded to -20℃~70℃ (originally 0-50℃), improving environmental adaptability. This technical solution achieves a technological leap from "defect judgment" to "micron-level defect quantitative analysis" in transposed conductor detection through the systematic integration of core technologies such as high-frequency excitation, precision spacing control, and differential detection.

[0044] Among them, the calculation and analysis module uses convolutional neural network (CNN) to identify image defects, and is supplemented by support vector machine (SVM) to determine electrical parameter abnormalities.

[0045] Technical improvements to this computing and analysis module have enabled an intelligent upgrade to the inspection system: Using a CNN neural network to extract deep features from surface images, the system achieves a micro-defect recognition accuracy exceeding 10μm, increasing inspection efficiency fivefold compared to traditional algorithms. Combined with SVM's rigorous pattern determination of electrical parameters, a dual-dimensional "visual-electrical" judgment system has been constructed, keeping the error rate below 0.1%. This hybrid intelligent algorithm architecture leverages the advantages of CNN in image processing while retaining the high reliability of SVM in analyzing small-sample electrical data. This gives the system both submillimeter defect localization capabilities and millisecond-level real-time response performance, making it a perfect match for the inspection needs of high-speed production lines.

[0046] Among them, the control and execution module uses PLC to link the mechanical sorting device to remove unqualified transposed conductors in real time and feed back the execution status to the database.

[0047] Technical improvements to this control and execution module achieved a key breakthrough in detection-execution closed-loop control: through PLC high-precision synchronous control (response time ≤10ms), the mechanical sorting device can accurately remove defective wires from high-speed production lines (≥2m / s), and the execution positioning error is controlled to ±0.1mm. At the same time, the real-time data feedback mechanism establishes a complete quality traceability chain, strictly binding the detection data of each wire to the sorting action (timestamp alignment accuracy ≤1μs). This integration of hard real-time control and digital traceability not only guarantees a defective product interception rate of more than 99.9%, but also provides full-process data support for process improvement, upgrading quality control from passive detection to active prevention.

[0048] Among them, the communication module supports 5G / Industrial Ethernet, uploading detection data to the cloud platform to realize remote monitoring and historical data analysis.

[0049] The technological upgrade of this communication module has realized the full-area digital management of the detection system: through 5G / Industrial Ethernet dual-mode transmission (latency ≤ 20ms, reliability ≥ 99.99%), a high-speed transmission channel for detection data is built, which enables the real-time upload rate of raw data (including multispectral images, eddy current signals, etc.) to reach 1Gbps; cloud platform integration not only realizes cross-regional remote collaborative monitoring (response delay ≤ 500ms), but also uses the big data analysis engine to conduct in-depth mining of historical detection data (processing capacity ≥ 10^6 items / second), thereby establishing a quality prediction model and identifying potential process defects in advance. This "end-edge-cloud" collaborative architecture not only meets the strict real-time requirements of industrial sites, but also provides intelligent decision-making support for the continuous optimization of quality management, upgrading quality control from single-point detection to full life cycle management.

[0050] Among them, the data acquisition and processing module integrates an adaptive filtering algorithm (wavelet transform + Kalman filter) to eliminate industrial noise in real time. The adaptive filtering algorithm uses the db4 wavelet basis (decomposition layer number ≥ 5 layers) and Kalman filter (state estimation error ≤ 1%). After composite filtering, the signal-to-noise ratio is ≥ 40dB.

[0051] The technological breakthrough of this data acquisition and processing module realizes accurate signal extraction in complex industrial environments: through the multi-scale decomposition of db4 wavelet basis (≥5 layers), defect features and background noise are effectively separated. Combined with the dynamic state estimation of Kalman filtering (error ≤1%), a composite adaptive filtering system is constructed, which can still maintain a signal-to-noise ratio of ≥40dB in strong electromagnetic interference (≥30dB) and mechanical vibration environments. This intelligent filtering architecture enables the system to accurately extract 0.1mV-level effective features from the original signal with a signal-to-noise ratio of ≤10dB, reducing the false detection rate of micro defects (≥20μm) to below 0.5%. At the same time, through real-time processing (delay ≤5ms), it meets the online detection needs of high-speed production lines and provides a high-fidelity data foundation for subsequent intelligent analysis.

[0052] Among them, the calculation and analysis module also includes: a motion compensation algorithm to correct dynamic detection errors; an adaptive eddy current parameter adjustment unit to automatically optimize detection parameters according to the wire material; and an aging model compensation unit to dynamically adjust the insulation layer defect judgment threshold.

[0053] The technological innovation of this calculation and analysis module realizes the adaptive optimization and dynamic compensation of the detection system: the motion compensation algorithm (based on the optical flow method) is used to correct the image offset caused by high-speed motion in real time (compensation accuracy ±2μm), reducing the dynamic detection error by 80%; the adaptive eddy current parameter adjustment unit automatically matches the optimal detection frequency (adjustment step ≤0.1MHz) according to the material impedance characteristics (copper / aluminum), improving the defect detection consistency of wires of different materials; the aging model compensation unit dynamically adjusts the defect judgment threshold (update cycle ≤1h) by analyzing the insulation layer performance attenuation curve, enabling the system to adapt to the service time. The three intelligent compensation mechanisms work together to solve the three major technical pain points of motion blur, material difference and aging misjudgment in traditional detection, and stably control the comprehensive detection accuracy of the system under complex working conditions within the range of ±5μm, realizing a technological leap from "fixed threshold detection" to "dynamic intelligent diagnosis".

[0054] An intelligent online detection method for transposed conductors includes the following detection steps:

[0055] S1. Collecting the surface image, temperature and electrical parameters of the transposed conductor through the sensor module;

[0056] Specifically, it uses a 500W pixel global shutter CMOS sensor (IMX541) with a pixel size of 2.74μm×2.74μm. It is equipped with a three-band optical system, including a visible light channel: 400-700nm, using an LED array light source (color temperature 6500K±200K), an ultraviolet channel: 365nm, a light source power of 15W±5%, and an infrared channel: 850nm. It is used to detect surface temperature distribution and complete surface image and electrical data collection of transposed conductors.

[0057] S2. Filter and fuse the collected data in real time to extract defect features;

[0058] Also included are:

[0059] Perform non-uniformity correction on the multispectral images (visible light + ultraviolet light) acquired by the visual acquisition unit to eliminate the impact of uneven lighting; use adaptive median filtering to process eddy current signals to effectively suppress electromagnetic interference in industrial environments; and use temperature compensation algorithms to eliminate the impact of ambient temperature fluctuations on electrical parameter measurements.

[0060] A unified time base is established based on the encoder signal to ensure strict synchronization of data from different sensors; coordinate transformation is used to map the data of each sensor to the same spatial coordinate system, and an interpolation algorithm is used to compensate for data timing differences caused by different sampling rates.

[0061] S3, classify and locate defects based on deep learning models;

[0062] It also includes: using an improved ResNet-18 network for defect classification; the network input is a 256×256 pixel multispectral image block; the output includes 8 common defects such as insulation damage and metal burrs; and multi-criteria fusion decision-making.

[0063] S4. Trigger production line sorting or alarm based on the detection results.

[0064] It also includes: execution control and quality traceability, graded response mechanism and high-precision sorting control.

[0065] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent online detection system for transposed conductors, characterized in that: include: The sensor module collects the surface image, temperature distribution, electrical parameters and internal defect signals of the transposed conductor; The data acquisition and processing module is connected to the sensor module and is used to filter, fuse and extract features from multi-source data; The calculation and analysis module, equipped with AI algorithms, analyzes the characteristics of collected and processed data, identifies and determines the defect type of the transposed conductor and outputs the detection results; The control and execution module links the production line to perform sorting and process adjustments based on the test results obtained through calculation and analysis; Communication module, which completes data interaction between modules and cloud remote monitoring; The sensor module includes: a multispectral imaging unit for enhancing the detection of surface defects in the insulation layer; a high-frequency eddy current probe (≥5MHz) for identifying internal micron-level defects; and an active vibration reduction mechanism for suppressing the interference of mechanical vibration on imaging.

2. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The multispectral imaging unit includes a 365nm UV LED light source and a visible light source. The camera pixel size is ≤3.45μm. Combined with a 5X telecentric lens, the detection accuracy is ±5μm, and the signal-to-noise ratio is improved to ≥30dB through dual-channel image fusion.

3. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The high-frequency eddy current probe has an operating frequency of 5-10 MHz, a probe spacing of ≤1 mm, a detection sensitivity of ≥0.5 mV / μm, and adopts differential eddy current technology with a common mode rejection ratio of ≥60 dB.

4. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The calculation and analysis module uses a convolutional neural network (CNN) to identify image defects and uses a support vector machine (SVM) to determine electrical parameter anomalies.

5. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The control and execution module links the mechanical sorting device through the PLC to remove unqualified transposed conductors in real time and feeds back the execution status to the database.

6. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The communication module supports 5G / industrial Ethernet, uploading detection data to the cloud platform to achieve remote monitoring and historical data analysis.

7. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The data acquisition and processing module integrates an adaptive filtering algorithm to eliminate industrial noise in real time. The adaptive filtering algorithm adopts db4 wavelet basis and Kalman filtering. The signal-to-noise ratio after composite filtering is ≥40dB.

8. The intelligent online detection system for transposed conductors according to claim 1, characterized in that: The calculation and analysis module also includes: a motion compensation algorithm to correct dynamic detection errors; an adaptive eddy current parameter adjustment unit to automatically optimize detection parameters according to the wire material; and an aging model compensation unit to dynamically adjust the insulation layer defect judgment threshold.

9. The intelligent online detection method for transposed conductors according to claim 1, characterized in that: The following detection steps are included: S1. Collecting the surface image, temperature and electrical parameters of the transposed conductor through the sensor module; S2. Filter and fuse the collected data in real time to extract defect features; S3, classify and locate defects based on deep learning models; S4. Trigger production line sorting or alarm based on the detection results.

Images