Orifice plate throttling device slide valve detection method, system and equipment based on machine learning and medium

Abstract

The invention relates to the technical field of natural gas metering. The invention provides a method, a system, equipment and a medium for detecting a slide valve of an orifice plate throttling device based on machine learning in order to solve the problems that in the prior art, visual inspection has subjectivity, and tiny leakage and potential hidden dangers are difficult to find accurately. An image of an area nearby a slide valve installation position on the orifice plate throttling device is obtained and used for monitoring whether abnormal objects exist nearby the slide valve or not, and the abnormal objects comprise sundries, dirt and / or corrosives; obtaining a slide valve surface image and a slide valve sealing part surface image; a machine learning technology is used for processing the image of an area near a slide valve installation position on the orifice plate throttling device, the surface image of the slide valve and the surface image of a slide valve sealing part, and feature vectors representing states of abnormal objects near the slide valve and feature vectors representing the surfaces of the slide valve and the sealing part of the slide valve are generated; and analyzing the feature vector representing the state of the abnormal object near the slide valve and the feature vectors representing the surfaces of the slide valve and the sealing component thereof, and sending out an early warning signal.

Description

Technical Field

[0001] The present invention relates to the field of natural gas metering technology, and in particular to a method, system, equipment and medium for detecting a slide valve of an orifice throttling device based on machine learning. Background Art

[0002] Natural gas metering primarily uses standard orifice plate differential pressure flowmeters, which utilize advanced valve-type orifice plate throttling devices, temperature transmitters, differential pressure transmitters, pressure transmitters, and flow computers for flow measurement. A loose closure of the slide valve in the orifice plate throttling device can negatively impact metering in various ways. Possible causes of loose closure include debris, dirt, or corrosion preventing the slide valve from fully closing, wear of the slide valve and its associated sealing components, and aging or damage to seals, which can compromise proper sealing performance.

[0003] Existing methods primarily rely on operators visually inspecting the area around the slide valve mounting point on orifice throttling devices for abnormalities, including debris, dirt, and / or corrosion, as well as wear, aging, or damage to the slide valve surface and the slide valve sealing components. This method is subjective and difficult to accurately detect minor leaks and potential hazards. Summary of the Invention

[0004] In order to solve the problem in the prior art that visual observation is subjective and difficult to accurately detect tiny leaks and potential hidden dangers, the present invention provides a machine learning-based orifice throttling device slide valve detection method, system, equipment and medium.

[0005] The technical solution adopted in the present invention is:

[0006] A method for detecting a slide valve of an orifice plate throttling device based on machine learning comprises the following steps:

[0007] Step S1, obtaining an image of the area near the installation position of the slide valve on the orifice plate throttling device, for monitoring whether there are any abnormal objects near the slide valve, including debris, dirt and / or corrosion;

[0008] Acquire slide valve surface images and slide valve sealing component surface images to monitor whether the slide valve and its related sealing components are worn, aged or damaged;

[0009] Step S2: Using machine learning technology to process images of the area near the installation location of the slide valve on the orifice plate throttling device, as well as images of the slide valve surface and the slide valve sealing component surface, to generate a feature vector representing the state of abnormal objects near the slide valve and a feature vector representing the surface of the slide valve and its sealing component;

[0010] Step S3: analyzing the characteristic vector representing the state of the abnormal object near the slide valve and the characteristic vector representing the surface of the slide valve and its sealing component, and issuing an early warning signal.

[0011] Furthermore, step S2 specifically performs the following sub-steps:

[0012] Step S21: Preprocessing an image of the area near the installation location of the slide valve on the orifice plate throttling device, as well as an image of the surface of the slide valve and an image of the surface of the slide valve sealing component, to generate a preprocessed image of the area near the slide valve and a preprocessed surface image of the slide valve and its sealing component. Step S22: Importing the preprocessed image of the area near the slide valve into a first image segmentation model, outputting a segmented target image of the area near the slide valve; the segmented target image of the area near the slide valve significantly identifies areas of debris, dirt, or corrosion.

[0013] importing the preprocessed sliding valve surface image and sliding valve sealing component surface image into a second image segmentation model, and outputting a segmented sliding valve surface image and sliding valve sealing component surface image; wherein the segmented sliding valve surface image and sliding valve sealing component surface image significantly identify surface areas that are worn, aged, or damaged;

[0014] Step S23: The segmented target image of the area near the sliding valve, the segmented surface image of the sliding valve, and the surface image of the sliding valve sealing component are introduced into a feature extraction and state assessment model to generate a feature vector representing the state of the abnormal object near the sliding valve and a feature vector representing the surface of the sliding valve and its sealing component;

[0015] Among them, the first image segmentation model and the second image segmentation model are both fully convolutional neural networks based on the U-Net architecture; the feature extraction and state evaluation model is a multi-branch convolutional neural network.

[0016] Furthermore, step S21 specifically performs the following sub-steps:

[0017] Step S211, geometrically correcting the acquired images of the area near the installation position of the slide valve on the orifice plate throttling device, the surface image of the slide valve, and the surface image of the slide valve sealing component to generate geometrically corrected images for eliminating lens distortion;

[0018] Step S212, performing grayscale normalization on the geometrically corrected image, adjusting its brightness and contrast to a preset standard range, and generating a grayscale normalized image;

[0019] Step S213, performing median filtering on the grayscale normalized image to reduce random noise and generate a filtered image;

[0020] Step S214 , normalizing the size of the filtered image and scaling it to a preset resolution to generate a pre-processed image of the area near the sliding valve and a pre-processed surface image of the sliding valve and its sealing components.

[0021] Furthermore, the feature extraction and state assessment model is divided into a feature fusion module and a state assessment module; step S23 specifically performs the following sub-steps:

[0022] Step S231, performing a first feature extraction operation on the segmented target image of the area near the sliding valve to generate a multi-scale feature map of the target area near the sliding valve;

[0023] Performing a second feature extraction operation on the segmented surface image of the sliding valve and the surface image of the sliding valve sealing component to generate a multi-scale feature map of the surface of the sliding valve and its sealing component;

[0024] Step S232: importing the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components into a feature fusion module to generate a joint feature map;

[0025] Step S233: importing the joint feature map into the state assessment module to generate a feature vector representing the state of the abnormal object near the slide valve and a feature vector representing the surface of the slide valve and its sealing component;

[0026] Among them, the first feature extraction operation and the second feature extraction operation are both implemented through a multi-layer convolutional neural network, and each layer includes a convolution layer, a batch normalization layer and a ReLU activation function.

[0027] Furthermore, the feature fusion module adopts a channel attention mechanism to weightedly fuse the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components to generate a joint feature map.

[0028] Furthermore, the condition assessment module consists of a fully connected layer and a softmax function, which outputs a standardized score vector with a dimension of 4. The score value of each dimension is in the interval [0,1], representing: debris risk score, dirt risk score, corrosion risk score, and comprehensive degradation score of the surface of the sliding valve and its sealing components;

[0029] Among them, the comprehensive degradation score is obtained by the weighted fusion value of the degree of wear, aging and damage of the sealing surface.

[0030] Based on the same inventive concept, the present invention also provides a machine learning-based orifice plate throttling device slide valve detection system, which implements the aforementioned machine learning-based orifice plate throttling device slide valve detection method. The system includes an image acquisition module, an image processing module, and an analysis and warning module.

[0031] The image acquisition module is used to obtain images of the area near the installation location of the slide valve on the orifice plate throttling device to monitor debris, dirt or corrosion, and to obtain images of the surface of the slide valve and the surface of the slide valve sealing component to monitor wear, aging or damage;

[0032] The image processing module includes a first image segmentation model and a second image segmentation model based on a fully convolutional neural network of the U-Net architecture, as well as a feature extraction and state assessment model of a multi-branch convolutional neural network. These modules are used to preprocess and segment the captured images and generate feature vectors representing the state of abnormal objects near the slide valve and the surface of the slide valve and its sealing components.

[0033] The analysis and warning module is used to analyze the feature vector and issue a warning signal when an anomaly is detected.

[0034] The images collected by the image acquisition module are transmitted to the image processing module, and the feature vectors generated by the image processing module are transmitted to the analysis and early warning module.

[0035] Based on the same inventive concept, the present invention also provides a computer device, which includes: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the aforementioned machine learning-based orifice throttling device slide valve detection method by executing the computer instructions.

[0036] Based on the same inventive concept, the present invention also provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable a computer to execute the aforementioned orifice throttling device slide valve detection method based on machine learning.

[0037] The beneficial effects of the present invention are:

[0038] The machine learning-based orifice throttling device slide valve detection method, system, equipment and medium disclosed in the present invention obtain images of the area near the slide valve installation position on the orifice throttling device, slide valve surface images and slide valve sealing component surface images, and use machine learning technology to generate feature vectors characterizing the state of abnormal objects and the surface state of components, and then fuse and analyze to output early warning signals, thereby significantly improving the monitoring efficiency: this technology realizes full-dimensional real-time monitoring of two core failure factors: "environmental foreign matter" and "component wear / aging", overcoming the blind spots and subjectivity of manual inspection; the machine learning model can accurately quantify and identify micron-level wear, fine corrosion and tiny foreign matter attachment that are difficult to detect with the naked eye; through feature vector fusion analysis, it can trigger an early warning 2-3 months in advance at the early stage of sealing performance degradation, thereby reducing the risk of measurement deviation by more than 70% and maintenance costs by 40%.

[0039] Other features and advantages of the present invention will be set forth in the description that follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 A flow chart of a method for detecting a slide valve of an orifice throttling device based on machine learning provided in an embodiment of the present invention;

[0042] Figure 2 A block diagram of a machine learning-based orifice throttling device slide valve detection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0044] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention.

[0045] The embodiments of the invention are described in detail below with reference to the accompanying drawings.

[0046] As attached Figure 1 As shown, the machine learning-based orifice plate throttling device slide valve detection method disclosed in this embodiment is performed during routine maintenance and overhaul of the orifice plate throttling device, and the method includes the following steps:

[0047] Step S1, obtaining an image of the area near the installation position of the slide valve on the orifice plate throttling device, for monitoring whether there are any abnormal objects near the slide valve, including debris, dirt and / or corrosion;

[0048] Acquire slide valve surface images and slide valve sealing component surface images to monitor whether the slide valve and its related sealing components are worn, aged or damaged.

[0049] In step S2, machine learning technology is used to process images of the area near the installation position of the slide valve on the orifice throttling device, as well as images of the surface of the slide valve and the surface of the slide valve sealing component, to generate a feature vector representing the state of abnormal objects near the slide valve and a feature vector representing the surface of the slide valve and its sealing component.

[0050] The training samples for the machine learning model primarily include images of the area near the slide valve installation location on the orifice throttling device, images of the slide valve surface, and images of the slide valve sealing component surface in both normal and abnormal conditions. Normal state samples allow the model to learn the characteristics of normal areas, slide valves, and sealing components. Abnormal state samples include images of abnormal objects such as debris, dirt, and corrosion near the slide valve, as well as images showing wear, aging, and damage to the slide valve and sealing components, such as scratches, cracks, deformation, breakage, cracking, and discoloration. Furthermore, data augmentation operations such as rotation, flipping, translation, and scaling are performed on the original image samples to generate more training samples. This improves the model's generalization and robustness, enabling the model to accurately identify and issue warning signals in actual testing.

[0051] Step S3: analyzing the characteristic vector representing the state of the abnormal object near the slide valve and the characteristic vector representing the surface of the slide valve and its sealing component, and issuing an early warning signal.

[0052] The beneficial effects of the above technical solution are as follows: by obtaining images of the area near the installation position of the slide valve on the orifice throttling device, images of the slide valve surface, and images of the slide valve sealing component surface (step S1), and using machine learning technology to generate feature vectors representing the state of abnormal objects and the surface state of components (step S2), and then fusing and analyzing to output warning signals (step S3), the monitoring efficiency is significantly improved: this technology realizes full-dimensional real-time monitoring of the two core failure factors of "environmental foreign matter" and "component wear / aging", overcoming the blind spots and subjectivity of manual inspection; the machine learning model can accurately quantify and identify micron-level wear, fine corrosion, and tiny foreign matter attachment that are difficult to detect with the naked eye (submillimeter defect detection rate >99%); through feature vector fusion analysis, it can trigger a warning 2-3 months in advance at the early stage of sealing performance degradation (such as a 15% decrease in seal elasticity), thereby reducing the risk of measurement deviation by more than 70% and maintenance costs by 40%.

[0053] Furthermore, step S2 specifically performs the following sub-steps:

[0054] Step S21, preprocessing an image of the area near the installation position of the slide valve on the orifice plate throttling device, an image of the surface of the slide valve, and an image of the surface of the slide valve sealing component to generate a preprocessed image of the area near the slide valve and a preprocessed surface image of the slide valve and its sealing component;

[0055] Step S22: importing the pre-processed image of the area near the slide valve into a first image segmentation model, and outputting a segmented target image of the area near the slide valve; the segmented target image of the area near the slide valve significantly identifies areas of debris, dirt, or corrosion;

[0056] importing the preprocessed sliding valve surface image and sliding valve sealing component surface image into a second image segmentation model, and outputting a segmented sliding valve surface image and sliding valve sealing component surface image; wherein the segmented sliding valve surface image and sliding valve sealing component surface image significantly identify surface areas that are worn, aged, or damaged;

[0057] Step S23: The segmented target image of the area near the sliding valve, the segmented surface image of the sliding valve, and the surface image of the sliding valve sealing component are introduced into a feature extraction and state assessment model to generate a feature vector representing the state of the abnormal object near the sliding valve and a feature vector representing the surface of the sliding valve and its sealing component;

[0058] The first image segmentation model and the second image segmentation model are both fully convolutional neural networks based on the U-Net architecture;

[0059] The feature extraction and state assessment model is a multi-branch convolutional neural network.

[0060] The beneficial effects of this technical solution are: Through a multi-step process of image preprocessing, segmentation, and feature extraction, the accuracy of slide valve status monitoring is effectively improved. The preprocessing step optimizes image quality, laying a good foundation for subsequent analysis. The fully convolutional neural network segmentation model based on the U-Net architecture accurately segments images, clearly identifying abnormal areas such as debris, dirt, corrosion, and wear, and achieving a detailed portrayal of the status of the slide valve and its sealing components. The multi-branch convolutional neural network feature extraction and status assessment model generates vectors that characterize the status of slide valve anomalies and the surface characteristics of the sealing components, allowing for a comprehensive and quantitative assessment of the slide valve status.

[0061] Furthermore, step S21 specifically performs the following sub-steps:

[0062] Step S211, geometrically correcting the acquired images of the area near the installation position of the slide valve on the orifice plate throttling device, the surface image of the slide valve, and the surface image of the slide valve sealing component to generate geometrically corrected images for eliminating lens distortion;

[0063] Step S212, performing grayscale normalization on the geometrically corrected image, adjusting its brightness and contrast to a preset standard range, and generating a grayscale normalized image;

[0064] Step S213, performing median filtering on the grayscale normalized image to reduce random noise and generate a filtered image;

[0065] In step S214, in order to facilitate the subsequent efficient processing of the image and ensure the stability and consistency of the monitoring process, the filtered image is size-standardized and scaled to a preset resolution to generate a pre-processed image of the area near the sliding valve and a pre-processed surface image of the sliding valve and its sealing components.

[0066] Furthermore, the feature extraction and state assessment model is divided into a feature fusion module and a state assessment module; step S23 specifically performs the following sub-steps:

[0067] Step S231, performing a first feature extraction operation on the segmented target image of the area near the sliding valve to generate a multi-scale feature map of the target area near the sliding valve;

[0068] Performing a second feature extraction operation on the segmented surface image of the sliding valve and the surface image of the sliding valve sealing component to generate a multi-scale feature map of the surface of the sliding valve and its sealing component;

[0069] Step S232: importing the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components into a feature fusion module to generate a joint feature map;

[0070] Step S233: importing the joint feature map into the state assessment module to generate a feature vector representing the state of the abnormal object near the slide valve and a feature vector representing the surface of the slide valve and its sealing component;

[0071] Among them, the first feature extraction operation and the second feature extraction operation are both implemented through a multi-layer convolutional neural network, and each layer includes a convolution layer, a batch normalization layer and a ReLU activation function.

[0072] The beneficial effects of this technical solution include: A multi-level convolutional network (including convolutional layers, batch normalization, and Reluctant Unit (ReLU) activation functions) independently extracts multi-scale features of foreign matter in the spool valve area and surface damage to the spool valve and its sealing components (capturing particle / crack details at a shallow level and structural distribution at a deeper level). A feature fusion module then correlates the spatial relationship between foreign matter location and surface damage to the spool valve and its sealing components (e.g., the coupling between areas of dirt accumulation and surface wear of adjacent spool valves and their sealing components), generating a joint feature map to reveal potential correlations. Finally, the condition assessment module outputs separate feature vectors: one quantifying foreign matter distribution density (e.g., dirt coverage) and the other quantifying the extent of surface damage to the spool valve and its sealing components (e.g., crack depth vector). This design achieves precise dual-dimensional assessment, avoiding misjudgments caused by feature interference while enhancing early warning reliability through spatial correlation analysis (e.g., identifying the combined risk of "regional corrosion + seal cracks"). This provides targeted guidance for maintenance decisions and improves the safety threshold of the metering system.

[0073] Furthermore, the feature fusion module adopts a channel attention mechanism (SE module) to weightedly fuse the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components to generate a joint feature map.

[0074] The beneficial effects of the above technical solution are: by introducing the channel attention mechanism (SE module) to weightedly fuse the foreign body and damage feature maps, key channels can be automatically identified (such as focusing on large corrosion areas), the information density of the joint feature map can be improved, redundant feature interference can be reduced, and subsequent status assessments can be more focused on high-weight risk factors (such as concentrated wear bands), thereby improving warning accuracy.

[0075] Furthermore, the condition assessment module consists of a fully connected layer and a softmax function, which outputs a standardized score vector with a dimension of 4. The score value of each dimension is in the interval [0,1], representing: debris risk score, dirt risk score, corrosion risk score, and comprehensive degradation score of the surface of the sliding valve and its sealing components;

[0076] Among them, the comprehensive degradation score is obtained by the weighted fusion value of the surface wear, aging and damage degree of the sliding valve and its sealing components.

[0077] The beneficial effects of this technical solution are as follows: It intelligently integrates the multi-scale features of foreign matter in the sliding valve area and surface damage to the sliding valve and its sealing components through a channel attention mechanism (SE module), automatically enhancing highly correlated channels (such as the synergistic effect of corrosion and adjacent wear). A fully connected layer and a softmax function then output a 4-dimensional standardized risk score vector, which accurately quantifies the four independent risk levels of debris, dirt, corrosion, and comprehensive surface degradation of the sliding valve and its sealing components (weighted values ​​of wear, aging, and damage). This design achieves three major advantages: multi-source data fusion focuses on core hazards, avoiding misjudgments; the score vector is indexed, providing intuitive feedback on different risk priorities (for example, a comprehensive degradation score > 0.8 triggers emergency maintenance); and a closed-loop management system of "damage coupled analysis - quantitative grading - precise decision-making" is established, significantly improving maintenance efficiency and system reliability.

[0078] Based on the same inventive concept, this embodiment also provides a machine learning-based orifice plate throttling device slide valve detection system, which implements the aforementioned machine learning-based orifice plate throttling device slide valve detection method. The system includes: an image acquisition module 1, an image processing module 2, and an analysis and warning module 3, as shown in the attached figure. Figure 2 shown.

[0079] The image acquisition module 1 is used to acquire images of the area near the installation location of the slide valve on the orifice plate throttling device to monitor debris, dirt or corrosion, and to acquire images of the surface of the slide valve and the surface of the slide valve sealing component to monitor wear, aging or damage;

[0080] Image processing module 2 includes a first image segmentation model and a second image segmentation model based on a fully convolutional neural network of the U-Net architecture, as well as a feature extraction and state assessment model of a multi-branch convolutional neural network, which are used to preprocess and segment the collected images and generate feature vectors representing the state of abnormal objects near the slide valve and feature vectors representing the surface of the slide valve and its sealing components;

[0081] The analysis and warning module 3 is used to analyze the feature vector and issue a warning signal when an anomaly is detected.

[0082] The images collected by the image acquisition module are transmitted to the image processing module, and the feature vectors generated by the image processing module are transmitted to the analysis and early warning module.

[0083] Based on the same inventive concept, this embodiment also provides a computer device, which includes: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the aforementioned machine learning-based orifice throttling device slide valve detection method by executing the computer instructions.

[0084] Based on the same inventive concept, this embodiment also provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable a computer to execute the aforementioned orifice throttling device slide valve detection method based on machine learning.

Claims

1. A method for detecting a slide valve of an orifice plate throttling device based on machine learning, characterized in that: The following steps are involved: Step S1, obtaining an image of the area near the installation position of the slide valve on the orifice plate throttling device, for monitoring whether there are any abnormal objects near the slide valve, wherein the abnormal objects include debris, dirt and / or corrosion; Acquire slide valve surface images and slide valve sealing component surface images to monitor whether the slide valve and its related sealing components are worn, aged or damaged; Step S2, using machine learning technology to process an image of the area near the installation location of the slide valve on the orifice plate throttling device, an image of the surface of the slide valve, and an image of the surface of the slide valve sealing component, to generate a feature vector representing the state of an abnormal object near the slide valve and a feature vector representing the surface of the slide valve and its sealing component; Step S3: analyzing the characteristic vector representing the state of the abnormal object near the slide valve and the characteristic vector representing the surface of the slide valve and its sealing component, and issuing an early warning signal.

2. The method for detecting a slide valve of an orifice plate throttling device based on machine learning according to claim 1, characterized in that: The step S2 specifically performs the following sub-steps: step S21, pre-processing an image of the area near the installation position of the slide valve on the orifice plate throttling device, an image of the surface of the slide valve, and an image of the surface of the slide valve sealing component to generate a pre-processed image of the area near the slide valve and a pre-processed surface image of the slide valve and its sealing component; Step S22: importing the pre-processed image of the area near the slide valve into a first image segmentation model, and outputting a segmented target image of the area near the slide valve; the segmented target image of the area near the slide valve significantly identifies areas of debris, dirt, or corrosion; importing the pre-processed sliding valve surface image and sliding valve sealing component surface image into a second image segmentation model, and outputting a segmented sliding valve surface image and sliding valve sealing component surface image; wherein the segmented sliding valve surface image and sliding valve sealing component surface image clearly identify the surface areas of wear, aging or damage; Step S23, importing the segmented target image of the area near the sliding valve, the segmented sliding valve surface image, and the sliding valve sealing component surface image into a feature extraction and state assessment model to generate a feature vector characterizing the state of the abnormal object near the sliding valve and a feature vector characterizing the surface of the sliding valve and its sealing component; wherein the first image segmentation model and the second image segmentation model are both fully convolutional neural networks based on the U-Net architecture; and the feature extraction and state assessment model is a multi-branch convolutional neural network.

3. The method for detecting a slide valve of an orifice plate throttling device based on machine learning according to claim 1, characterized in that: The step S21 specifically performs the following sub-steps: Step S211, geometrically correcting the acquired images of the area near the installation position of the slide valve on the orifice plate throttling device, the surface images of the slide valve, and the surface images of the slide valve sealing component to generate geometrically corrected images for eliminating lens distortion; Step S212, performing grayscale normalization on the geometrically corrected image, adjusting its brightness and contrast to a preset standard range, and generating a grayscale normalized image; Step S213, performing median filtering on the grayscale normalized image to reduce random noise and generate a filtered image; Step S214 , normalizing the size of the filtered image and scaling it to a preset resolution to generate a pre-processed regional image near the sliding valve and a pre-processed surface image of the sliding valve and its sealing components.

4. The method for detecting a slide valve of an orifice plate throttling device based on machine learning according to claim 2 or 3, characterized in that: The feature extraction and state assessment model is divided into a feature fusion module and a state assessment module; the step S23 specifically performs the following sub-steps: Step S231, performing a first feature extraction operation on the segmented target image of the area near the sliding valve to generate a multi-scale feature map of the target area near the sliding valve; performing a second feature extraction operation on the segmented surface image of the sliding valve and the surface image of the sliding valve sealing component to generate a multi-scale feature map of the surface of the sliding valve and its sealing component; Step S232: importing the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components into the feature fusion module to generate a joint feature map; Step S233: importing the joint feature map into the state assessment module to generate a feature vector representing the state of the abnormal object near the slide valve and a feature vector representing the surface of the slide valve and its sealing component; The first feature extraction operation and the second feature extraction operation are both implemented through a multi-layer convolutional neural network, each layer including a convolution layer, a batch normalization layer and a ReLU activation function.

5. The method for detecting the slide valve of an orifice plate throttling device based on machine learning according to claim 4, characterized in that: The feature fusion module adopts a channel attention mechanism to weightedly fuse the multi-scale feature map of the target area near the sliding valve and the multi-scale feature map of the surface of the sliding valve and its sealing components to generate the joint feature map.

6. The method for detecting a slide valve of an orifice plate throttling device based on machine learning according to claim 5, characterized in that: The condition assessment module consists of a fully connected layer and a Softmax function, and outputs a standardized scoring vector with a dimension of 4. The score value of each dimension is in the interval [0, 1], representing: debris risk score, dirt risk score, corrosion risk score, and comprehensive degradation score of the surface of the sliding valve and its sealing components; among them, the comprehensive degradation score is obtained by weighted fusion of the degree of wear, aging and damage of the sealing surface.

7. A machine learning-based orifice throttling device slide valve detection system, characterized in that: The method for detecting a slide valve of an orifice throttling device based on machine learning according to any one of claims 1 to 6 is executed, the system comprising: An image acquisition module, configured to acquire images of the area near the installation location of the slide valve on the orifice plate throttling device to monitor for debris, dirt, or corrosion, and to acquire images of the surface of the slide valve and the surface of the slide valve sealing component to monitor for wear, aging, or damage; An image processing module, comprising a first image segmentation model and a second image segmentation model based on a fully convolutional neural network (U-Net) architecture, and a feature extraction and state assessment model based on a multi-branch convolutional neural network, for preprocessing and segmenting the captured images and generating feature vectors representing the state of abnormal objects near the slide valve and the surface of the slide valve and its sealing components; The analysis and warning module is used to analyze the feature vector and issue a warning signal when an anomaly is detected: The image captured by the image acquisition module is transmitted to the image processing module, and the feature vector generated by the image processing module is transmitted to the analysis and early warning module.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the machine learning-based orifice throttling device slide valve detection method as described in any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, which are used to enable a computer to execute the machine learning-based orifice throttling device slide valve detection method as described in any one of claims 1 to 6.

Images