Converter valve voltage-sharing electrode nondestructive testing method based on ray digital imaging

Abstract

The invention discloses a nondestructive testing method for a converter valve voltage-sharing electrode based on ray digital imaging. The nondestructive testing method comprises the steps that a detector wears protective equipment, carries an X-ray emitter, an imaging device and a portable terminal and rises to a detection point; the X-ray emitter and the X-ray imaging device are fixed to the two sides of an electrode respectively, and communication connection among the X-ray emitter, the X-ray imaging device and the portable detection and analysis terminal is established; the portable detection and analysis terminal generates a unique identity code for the electrode, controls X-ray exposure to obtain a digital image, automatically identifies and quantifies the electrode defect by calling a built-in image analysis model, and predicts the electrode health degree and matches a preventive maintenance decision based on a defect quantification result and a pre-stored health archive library. According to the device and the method, the converter valve voltage-sharing electrode is subjected to visual nondestructive testing in a non-dismounting state, the defects of the converter valve voltage-sharing electrode are quantitatively and intelligently evaluated, and safe and stable operation of converter valve equipment is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-destructive testing of power equipment, and particularly relates to a non-destructive testing method for a voltage-sharing electrode of a converter valve based on ray digital imaging. BACKGROUND

[0002] The converter valve is a core device in a high-voltage direct-current power transmission system, and its stable operation is crucial. The voltage-sharing electrode is a key component in the converter valve, and its main function is to balance the electric field distribution and prevent local discharge. During long-term operation, the voltage-sharing electrode may generate defects such as surface fouling, internal damage, and even fracture due to electrical, thermal, and mechanical stresses. These defects can seriously affect the voltage-sharing performance of the electrode and thus threaten the safety of the entire converter valve.

[0003] Currently, the detection of the voltage-sharing electrode of the converter valve mainly relies on periodic power-off manual disassembly visual inspection and simple size measurement. Drainage, disassembly, inspection, installation, water filling, and testing are required, which is a large amount of work and cannot effectively detect hidden defects inside the electrode. Moreover, the disassembly operation of the voltage-sharing electrode is prone to cause water leakage, and the electrode installation strength and sealing performance decrease after multiple disassembly and reinstallation, increasing the risk of water leakage. In addition, the detection result is highly dependent on the experience of the detection personnel, and is highly subjective. It is difficult to achieve accurate quantification of defects, and lacks systematic management of detection data and prediction ability of the future health status of the electrode, which cannot provide a scientific basis for preventive maintenance. SUMMARY

[0004] To solve the above-mentioned problems, the present application provides a non-destructive testing method for a voltage-sharing electrode of a converter valve based on ray digital imaging, comprising: S1, a detection personnel wears radiation protection equipment and wears a radiation dose monitor, carries an X-ray emitter, an X-ray imaging device, and a portable detection and analysis terminal, and ascends to the position of the voltage-sharing electrode of the converter valve to be detected through a lifting platform; S2, the X-ray emitter and the X-ray imaging device are fixed at opposite positions on both sides of the voltage-sharing electrode of the converter valve, and a communication connection is established between the X-ray emitter, the X-ray imaging device, and the portable detection and analysis terminal; S3, the portable detection and analysis terminal is used to uniquely encode the voltage-sharing electrode of the converter valve to be detected; S4, the X-ray emitter is controlled to expose by the portable detection and analysis terminal, and a digital ray image of the voltage-sharing electrode of the converter valve is acquired by the X-ray imaging device; S5, the portable detection and analysis terminal calls an image analysis model built-in to process the digital ray image, identify, and quantify the defect state of the voltage-sharing electrode of the converter valve; S6, according to the quantitative result of the defect state, combining with the pre-constructed converter valve grading electrode health archives, predicting the health degree of the converter valve grading electrode, and matching the corresponding preventive maintenance decision.

[0005] Further, in the step S3, the identity code is a unique identification code generated by a hash algorithm, which contains the device ID, installation position coordinates and detection time information of the converter valve grading electrode, and is stored in association with the digital radiographic image, the quantitative result of the defect state and the prediction result of the health degree.

[0006] Further, during the exposure process of step S4, the radiation dose monitor monitors the radiation dose in real time, and when the radiation dose exceeds the preset safety threshold, the radiation dose monitor triggers an alarm to prompt the operator to interrupt the exposure and adjust the on-site protection configuration.

[0007] Further, in the step S5, the image analysis model adopts a defect detection model based on a deep convolutional neural network, and the specific method is: The digital radiographic image of the converter valve grading electrode is preprocessed, and the preprocessing operation includes image denoising using Gaussian filtering, contrast enhancement using histogram equalization, and image size normalization to a preset standard size; The preprocessed image is input into the pre-trained defect detection model, which extracts features from the image through multiple convolutional layers and pooling layers to obtain a feature map representing the defects of the converter valve grading electrode; Based on the feature map, the full connection layer and the classifier of the defect detection model are used to identify and judge the defect category, which includes electrode surface fouling, electrode internal defect and electrode fracture; For the identified defects, the regression positioning layer of the defect detection model is used to determine the pixel-level position coordinates and the boundary box of the defects; Based on the pixel-level position coordinates and the boundary box of the defects, the pixel-millimeter level size conversion model of the digital radiographic image is used to calculate the geometric parameters of the defects, including area, length, width and physical position coordinates.

[0008] Further, the pixel-millimeter level size conversion model of the digital radiographic image is based on the imaging geometric parameters and image resolution of the X-ray imaging device to establish the mapping relationship between pixel coordinates and physical size, and through the pre-calibrated imaging scale factor, the pixel-level position coordinates and the boundary box of the defects are converted into actual physical size.

[0009] Further, in the step S6, the converter valve grading electrode health profile is constructed based on historical defect data, health degree change trend, and corresponding maintenance record and maintenance feedback record, the portable detection and analysis terminal stores the health profile, and after each detection is completed, the portable detection and analysis terminal adds the quantitative result of the defect state, the health degree prediction result and the maintenance feedback obtained this time to the converter valve grading electrode health profile.

[0010] Further, in the step S6, the health degree of the converter valve grading electrode is predicted and a corresponding preventive maintenance decision is matched, specifically including: Based on the historical defect data and the health degree change trend stored in the converter valve grading electrode health profile, a nonlinear regression analysis is performed on the quantitative result of the current defect state to predict the evolution rate of the current defect. The health degree score of the converter valve grading electrode is generated by weighted fusion calculation on the quantitative result of the current defect state and the evolution rate. The health degree score is compared with a preset maintenance threshold, and a corresponding preventive maintenance decision is matched and output from a predefined maintenance strategy library according to the comparison result, the maintenance strategy library is constructed based on the historical maintenance record and the maintenance feedback record in the health profile, and the maintenance decision includes maintenance timing, maintenance level and maintenance measures.

[0011] The beneficial effects of the present application are: The present application uses digital X-ray imaging technology to realize visual nondestructive detection of the converter valve grading electrode in a non-disassembly state, and combines a deep learning model to automatically and accurately identify and quantify defects, realize efficient detection without draining water, eliminate leakage risks that may be caused by electrode disassembly and seal ring failure of the converter valve equipment, and improve the objectivity and efficiency of detection; complete equipment "health profile" is formed by the detection data each time through the unique identity code, realizing traceability and systematic management of the detection data; based on the health profile and trend analysis, the health state of the electrode can be predicted and maintenance decisions can be matched, providing strong data support for condition-based maintenance and full life cycle management of power equipment, and ensuring safe and stable operation of the converter valve equipment. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a flowchart of a nondestructive detection method for a converter valve grading electrode based on ray digital imaging provided by an embodiment; Figure 2 is a nondestructive detection result diagram for a converter valve grading electrode based on ray digital imaging provided by an embodiment. DETAILED DESCRIPTION

[0013] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0014] Example 1 like Figure 1 As shown, a non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging includes: S1. The testing personnel wear radiation protection equipment and a radiation dose monitor, and carry an X-ray emitter, an X-ray imaging device and a portable testing and analysis terminal. They are then lifted to the location of the equalizing electrode of the converter valve to be tested via a lifting platform. S2. Fix the X-ray emitter and the X-ray imaging device to opposite positions on both sides of the equalizing electrode of the converter valve, and establish a communication connection between the X-ray emitter, the X-ray imaging device and the portable detection and analysis terminal. S3. Use the portable detection and analysis terminal to uniquely identify the equalizing electrode of the converter valve to be tested; S4. The portable detection and analysis terminal controls the X-ray emitter to perform exposure, and the X-ray imaging device is used to acquire a digital X-ray image of the equalizing electrode of the converter valve. S5. The portable detection and analysis terminal calls the built-in image analysis model to process the digital X-ray image, identify and quantify the defect state of the equalizing electrode of the converter valve; S6. Based on the quantification results of the defect status, combined with the pre-constructed health archive of the converter valve equalizing electrode, predict the health status of the converter valve equalizing electrode and match the corresponding preventive maintenance decision.

[0015] Specifically, this invention employs an X-ray source and a digital imaging ionization chamber for rapid installation and deployment. In terms of safety, the valve tower piping is made of PVDF, which has a high X-ray penetration rate, making it suitable for using a low-power, low-radiation X-ray source. Operators can operate remotely wirelessly, resulting in a high safety factor and reducing radiation safety hazards. A portable detection and analysis terminal is equipped with a X-ray image analysis method to analyze the digital image of the equalizing electrode, identify and determine the defect type, and solve the problem of effectively detecting the equalizing electrode of the converter valve without disassembly, and using the detection data to achieve predictive maintenance.

[0016] In a preferred embodiment, in step S3, the identity code is a unique identifier generated by a hash algorithm. The identifier includes the device ID, installation location coordinates, and detection time information of the equalizing electrode of the converter valve, and is stored in association with the digital X-ray image, the quantification result of the defect state, and the health prediction result.

[0017] Specifically, the operator uses dedicated software on the portable detection and analysis terminal to select the device ID of the equalizing electrode of the converter valve currently being detected from the preset device list, performs spatial positioning based on the installation location of the valve tower (e.g., high-end valve tower of pole I, third layer, B phase inlet), and automatically obtains the current detection time. A fixed-length, unique and irreversible string, i.e., a unique identification code, is generated through a hash algorithm.

[0018] This unique identification code is bound to the digital radiographic images, defect status quantification results, and health prediction results obtained in subsequent steps, forming a complete and traceable detection record with the identification code as the primary key, thus solving the problem of "data silos" in traditional detection.

[0019] In a preferred embodiment, during the exposure process in step S4, the radiation dose monitor monitors the radiation dose in real time. When the radiation dose exceeds a preset safety threshold, the radiation dose monitor triggers an alarm, prompting the operator to interrupt the exposure and adjust the on-site protection configuration.

[0020] Specifically, before the exposure begins, the operator sets a preset safety threshold on the radiation dose monitor. Throughout the exposure process, the monitor continuously and in real-time monitors the radiation dose rate level of the work area. The monitoring data is displayed on the instrument's screen in real time, and the internal microprocessor compares the real-time detection values ​​with the preset safety threshold. Once the detection value exceeds the preset threshold, the monitor triggers a high-frequency buzzer alarm and flashes a red warning light, while simultaneously displaying an over-limit warning on the screen. Upon receiving the alarm, the operator immediately interrupts the X-ray exposure using the "Stop" button on the terminal and then adjusts the protective configuration according to the situation on site. As an independent "sensor," the radiation dose monitor transforms imperceptible radiation risks into quantifiable electrical signals and data, allowing operators to have a clear understanding of the safety status of the working environment. This reduces the psychological burden of working in high-pressure, high-radiation-risk environments and helps improve work efficiency and stability.

[0021] In a preferred embodiment, in step S5, the image analysis model employs a defect detection model based on a deep convolutional neural network, specifically as follows: The digital radiographic image of the converter valve equalizing electrode is subjected to a preprocessing operation, the preprocessing operation comprising image denoising by Gaussian filtering, contrast enhancement by histogram equalization, and image size normalization to a preset standard size; The preprocessed image is input into a pre-trained defect detection model, the defect detection model extracting features of the image through multiple convolution layers and pooling layers to obtain a feature map representing defects of the converter valve equalizing electrode; Based on the feature map, a full connection layer and a classifier of the defect detection model are used to identify and determine a defect category, the defect category including electrode surface fouling, electrode internal defect and electrode fracture; For the identified defect, a regression positioning layer of the defect detection model is used to determine a pixel-level position coordinate and a bounding box of the defect; Based on the pixel-level position coordinate and the bounding box of the defect, a pixel-millimeter level size conversion model of the digital radiographic image is used to calculate geometric parameters of the defect, the geometric parameters including area, length, width and physical position coordinate.

[0022] Specifically, the obtained original digital radiographic image first enters a preprocessing stage, Gaussian filtering denoising effectively suppresses random noise such as X-ray quantum noise, and improves the signal-to-noise ratio; histogram equalization makes the gray difference of different regions such as electrodes, fouling and background pipes more obvious, and the details are more prominent; size normalization ensures the consistency of subsequent neural network input.

[0023] The preprocessed image is sent into a pre-trained defect detection model, the image is sequentially generated through multiple convolution layers and pooling layers to generate a set of highly abstract feature maps representing the essence of the image, the feature maps are flattened and input into a full connection layer and a classifier, the probability of the image belonging to each predefined defect category is calculated, and the category with the highest probability is taken as the recognition result, at the same time, a regression positioning layer predicts a minimum rectangular bounding box containing the defect in the image and gives the corresponding pixel-level coordinate.

[0024] For the located defect, the pixel size of the bounding box is converted into actual physical size by calling a size conversion model, the area, length and width of the defect are automatically calculated, and the accurate physical position of the defect on the electrode is determined.

[0025] The deep learning model of the present application replaces the traditional image processing algorithm which relies on artificial setting rules, automatically learns the deep and abstract features of the defects from a large amount of labeled data, realizes the full-flow automation from image to defect recognition and positioning, and the accuracy and consistency far exceed the judgment of the human eye, reducing missed detection and misjudgment.

[0026] As a preferred embodiment, the pixel-millimeter level size conversion model of the digitized radiographic image is based on the imaging geometry parameters and image resolution of the X-ray imaging device, establishes a mapping relationship between pixel coordinates and physical size, and converts the pixel level position coordinates and bounding box of the defect into actual physical size through a pre-calibrated imaging scale factor.

[0027] Specifically, a one-time calibration is performed before detection. A standard test block with a known accurate size is placed at a typical imaging position of the equalizing electrode. The test block is imaged using the same imaging geometry parameters as the actual detection. In the obtained digital image, the pixel size of the standard hole on the image is measured. The conversion scale under the current imaging system is calculated to determine the scale factor of the current imaging. In the field detection, when the pixel level bounding box of the defect is output by the defect detection model, the actual physical size is automatically converted through calculation to realize the quantification of the defect, avoiding the parallax and subjective error of manual measurement.

[0028] As a preferred embodiment, in the step S6, the health profile library of the converter valve equalizing electrode is constructed based on historical defect data, health degree change trend, and corresponding maintenance record and maintenance feedback record. The portable detection and analysis terminal stores the health profile library, and after each detection is completed, the portable detection and analysis terminal adds the quantification result of the defect state, the health degree prediction result, and the maintenance feedback obtained in this detection to the health profile library of the converter valve equalizing electrode.

[0029] Specifically, after each detection is completed, the portable detection and analysis terminal automatically adds the quantification result of this detection, the health degree prediction result, and the maintenance feedback after subsequent maintenance as a new data entry to the health profile library of the electrode, so that the health profile library becomes a continuously growing dynamic database containing the whole process of "disease diagnosis-prognosis judgment-treatment feedback", and provides extremely valuable data assets for equipment reliability analysis, life assessment, spare parts management, and new equipment selection.

[0030] As a preferred embodiment, in the step S6, the health degree of the converter valve equalizing electrode is predicted and a corresponding preventive maintenance decision is matched, specifically including: Based on the historical defect data and health degree change trend stored in the health profile library of the converter valve equalizing electrode, a nonlinear regression analysis is performed on the quantification result of the current defect state to predict the evolution rate of the current defect; The health degree score of the converter valve equalizing electrode is generated by weighted fusion calculation on the quantification result of the current defect state and the evolution rate; The health score is compared with a preset maintenance threshold, and a corresponding preventive maintenance decision is matched and output from a predefined maintenance strategy library according to a comparison result, the maintenance strategy library being constructed based on historical maintenance records and maintenance feedback records in the health record library, and the maintenance decision including a maintenance time, a maintenance level and a maintenance measure.

[0031] Specifically, when the portable detection and analysis terminal performs health degree prediction, historical defect data and health degree change trend stored in the health record library of the electrode are called, nonlinear regression algorithm is used to perform curve fitting on the data points, the evolution rate of the current defect is predicted, a weighted formula is used to fuse and calculate the quantization result of the current defect and the evolution rate of the defect, and finally a comprehensive health degree score is generated, so that an electrode with a current problem but rapid deterioration can obtain a lower health score than an electrode with a stable or even improved problem.

[0032] Based on big data analysis of historical maintenance records and feedback, a maintenance strategy library is constructed, and different health degree intervals are predefined in the maintenance strategy library. According to the calculated health degree score, specific maintenance time, maintenance level and maintenance measures are matched from the strategy library to accurately predict the development of the defect and make maintenance plans in advance when the defect is still in the early stage, thereby ensuring the reliability of the converter station.

[0033] Embodiment two During the annual maintenance of the pole I high-end valve tower of a certain ultra-high voltage converter station in Henan Province, the detection personnel wear a complete set of radiation protection equipment and wear a radiation dose monitor, use a special lifting platform, safely transport the detection equipment (including an X-ray emitter, an X-ray imaging device and a portable detection and analysis terminal) to the detection area about 15 meters away from the ground inside the valve tower. The safety officer checks according to the risk points and preventive measures table to ensure the safety of the operation, and the risk points and preventive measures table is shown in Table 1: Table 1 Risk points and preventive measures table Serial number Risk point Preventive measures 1 Prevent falling from high altitude during high-altitude operation Wearing full-body safety belt is required for operation above 2 meters 2 Prevent dropping objects during valve tower operation Record the articles carried by the personnel entering and leaving the valve tower 3 Prevent injury caused by falling objects during high-altitude operation Set up a dedicated safety officer to patrol and remind the area below the high-altitude operation area to prevent unrelated personnel from entering the area 4 Prevent equipment damage caused by inexperienced personnel during valve tower operation Valve tower operators must have more than 3 years of experience in thyristor operation to avoid equipment damage caused by lack of experience 5 Prevent equipment damage caused by improper use of elevating platform The elevating platform must be operated by a dedicated person and the operator must be a qualified personnel trained by the converter station 6 Prevent failure to restore after disassembly during ten-step processing Strictly follow the record table during ten-step processing to ensure record synchronization 7 Prevent electric shock during live testing of the valve tower The testing instrument should be grounded according to the requirements, and mutual reminders should be given during testing and power-on to ensure safety measures 8 Clean the work site and check the equipment for any remaining objects after work The work supervisor should check and verify the site one by one 9 Radiation protection Operators should be equipped with radiation protection clothing (radiation protection clothing includes: protective lead hat, protective lead glasses, protective lead neck, protective lead clothing) The detection target is the equalizing electrode at the flange of the main water pipe inlet pipe of the pole I high-end valve tower. The operator fixes an X-ray emitter on one side of the water pipe about 30 cm away from the electrode through a magnetic base, and fixes an X-ray imaging device on the other side corresponding to the water pipe and opposite to the emitter. Wireless communication connection of the X-ray machine, the X-ray imaging device and the portable detection and analysis terminal is established through Wi-Fi.

[0034] In the portable detection and analysis terminal, a unique identity code is generated for the electrode through SHA-256 hash algorithm, and the code is bound with all data obtained thereafter.

[0035] The operator sets the exposure parameters and remotely triggers the exposure through the portable detection and analysis terminal in the ionization chamber. At the moment of exposure, the radiation dose monitor displays a transient increase in the surrounding dose rate, but it is far below the preset safety threshold, and no alarm is triggered. The X-ray imaging device transmits the collected digital radiographic image (such as the one shown in FIG. 8) to the portable detection and analysis terminal in real time, and the image clearly shows the outline of the electrode needle and the surrounding scaling condition. Figure 2

[0036] The portable detection and analysis terminal automatically calls the built-in image analysis model to process the image, identifies that there is uniform scaling on the electrode surface, classifies it as a "surface scaling" defect, and frames the scaling area. Through the pre-calibrated pixel-millimeter conversion model, the average thickness of the scaling layer is calculated.

[0037] Since this is the first detection of the electrode, there is no historical data in the health archive. Based on the quantitative results of this time, the portable detection and analysis terminal determines the health of the electrode as "good" according to the preset initial scoring rules of health. The preventive maintenance decision matched by the system is "continue to run and include in the next detection cycle for re-inspection". All information of this detection, including the identity code, the original image, the defect report, the health, and the maintenance suggestion, are automatically packaged as the first health record of the electrode and stored in the database.

[0038] This embodiment successfully completes the visual non-destructive detection of the high-end valve tower equalizing electrode of pole I in the converter station without draining water and disassembling the electrode. The detection process is safe and efficient, and accurately quantifies the scaling condition of the electrode. The unique identity code and health record established lay a solid data foundation for tracking the state change of the electrode and realizing predictive maintenance, and effectively avoid the risk of sealing failure caused by repeated disassembly.

[0039] Those skilled in the art will appreciate that embodiments of the application can be provided as methods, systems, or computer program products. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer-usable program code embodied in the medium. The solutions in the embodiments of the application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0040] ​The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0041] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0042] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0043] While the preferred embodiments of the application have been described, additional variations and modifications can be employed by those skilled in the art. Therefore, the appended claims intend to cover all such modifications and variations as fall within the true spirit and scope of the application.

[0044] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging, characterized in that, include: S1. The testing personnel wear radiation protection equipment and a radiation dose monitor, and carry an X-ray emitter, an X-ray imaging device and a portable testing and analysis terminal. They are then lifted to the location of the equalizing electrode of the converter valve to be tested via a lifting platform. S2. Fix the X-ray emitter and the X-ray imaging device to opposite positions on both sides of the equalizing electrode of the converter valve, and establish a communication connection between the X-ray emitter, the X-ray imaging device and the portable detection and analysis terminal. S3. Use the portable detection and analysis terminal to uniquely identify the equalizing electrode of the converter valve to be tested; S4. The portable detection and analysis terminal controls the X-ray emitter to perform exposure, and the X-ray imaging device is used to acquire a digital X-ray image of the equalizing electrode of the converter valve. S5. The portable detection and analysis terminal calls the built-in image analysis model to process the digital X-ray image, identify and quantify the defect state of the equalizing electrode of the converter valve; S6. Based on the quantification results of the defect status, combined with the pre-constructed health archive of the converter valve equalizing electrode, predict the health status of the converter valve equalizing electrode and match the corresponding preventive maintenance decision.

2. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 1, characterized in that: In step S3, the identity code is a unique identifier generated by a hash algorithm. The identifier includes the device ID, installation location coordinates, and detection time information of the equalizing electrode of the converter valve, and is stored in association with the digital X-ray image, the quantification result of the defect status, and the health prediction result.

3. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 1, characterized in that: During the exposure process in step S4, the radiation dose monitor monitors the radiation dose in real time. When the radiation dose exceeds the preset safety threshold, the radiation dose monitor triggers an alarm, prompting the operator to interrupt the exposure and adjust the on-site protection configuration.

4. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 1, characterized in that: In step S5, the image analysis model employs a defect detection model based on a deep convolutional neural network, specifically as follows: The digital X-ray image of the equalizing electrode of the converter valve is preprocessed. The preprocessing operation includes denoising the image using Gaussian filtering, enhancing the contrast using histogram equalization, and normalizing the image size to a preset standard size. The preprocessed image is input into a pre-trained defect detection model, which extracts features from the image through multiple convolutional and pooling layers to obtain a feature map characterizing the defects of the equalizing electrode of the converter valve. Based on the feature map, the defect categories are identified and determined through the fully connected layer and classifier of the defect detection model. The defect categories include scale on the electrode surface, internal defects in the electrode, and electrode breakage. For the identified defects, the pixel-level location coordinates and bounding boxes of the defects are determined through the regression localization layer of the defect detection model. Based on the pixel-level location coordinates and bounding box of the defect, the geometric parameters of the defect are calculated using the pixel-to-millimeter size conversion model of the digital ray image. The geometric parameters include area, length, width, and physical location coordinates.

5. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 3, characterized in that: The pixel-to-millimeter size conversion model of the digital X-ray image is based on the imaging geometry parameters and image resolution of the X-ray imaging device. It establishes a mapping relationship between pixel coordinates and physical size, and converts the pixel-level position coordinates and bounding boxes of defects into actual physical sizes through a pre-calibrated imaging scaling factor.

6. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 1, characterized in that: In step S6, the health record library of the converter valve equalizing electrode is constructed based on historical defect data, health status change trends, and corresponding maintenance records and maintenance feedback records. The portable detection and analysis terminal stores the health record library, and after each test, the portable detection and analysis terminal adds the quantitative results of the defect status, the health status prediction results, and the maintenance feedback obtained in this test to the health record library of the converter valve equalizing electrode.

7. The non-destructive testing method for the equalizing electrode of a converter valve based on X-ray digital imaging according to claim 1, characterized in that: In step S6, predicting the health of the equalizing electrode of the converter valve and matching it with corresponding preventative maintenance decisions specifically includes: Based on the historical defect data and health status change trends stored in the health archive of the equalizing electrode of the converter valve, nonlinear regression analysis is performed on the quantitative results of the current defect status to predict the evolution rate of the current defect. A health score for the equalizing electrode of the converter valve is generated by weighted fusion calculation of the quantification results of the current defect state and the evolution rate. The health score is compared with a preset maintenance threshold. Based on the comparison result, a corresponding preventive maintenance decision is matched from a predefined maintenance strategy library and output. The maintenance strategy library is constructed based on historical maintenance records and maintenance feedback records in the health record library. The maintenance decision includes maintenance timing, maintenance level, and maintenance measures.

Images