Transverse image distance measuring device and method for transformer replacement traction

Abstract

The invention discloses a transverse image distance measuring device and method for transformer replacement traction, and belongs to the technical field of power equipment maintenance. The device comprises a transverse visual image deviation measuring system, a mounting frame, a quick clamping device and an image processing host, the visual deviation measuring system is arranged above a main transformer transfer trolley track wheel carrier through the mounting frame and the quick clamping device, and the visual angle of the visual deviation measuring system is adjusted to be perpendicular to a track; the system collects images of track wheels and tracks; the method comprises the following steps of: based on an image, extracting a track and a wheel rim edge through a sub-pixel edge detection algorithm which comprises image preprocessing and is based on a Zernike moment, and calculating a transverse deviation amount; according to the method, traditional manual plumb line measurement is replaced with non-contact image measurement, the measurement precision is greatly improved, uninterrupted continuous monitoring is achieved, the rail gnawing risk is effectively eradicated, and the operation safety and universality are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of transformer installation, and particularly relates to a transverse image ranging device and method for transformer replacement traction. BACKGROUND

[0002] In the maintenance and replacement operation of the main transformer of a large hydropower station, the main transformer, a heavy load equipment of several hundred tons, needs to be transferred through a track system. Due to the difficulty in absolute synchronization of the left and right tractions, the main transformer will produce a transverse deviation when moving on the track. If the deviation is too large, it will cause the transfer trolley to "gnaw the track", that is, the track flange and the track side will be subjected to severe friction and extrusion. This not only damages the flange and the track, increases the traction load, but also may cause the equipment to be stuck or even derailed in the most serious case, causing major safety accidents and economic losses. Therefore, real-time and accurate monitoring of the transverse deviation of the main transformer during traction is a key link to ensure the safety and smooth progress of the operation. Industry practice shows that controlling the transverse deviation within 10 millimeters is a safety threshold to avoid the "gnawing track" phenomenon.

[0003] At present, the closest prior art in this field is the artificial plumb line measurement method. The specific implementation is as follows: a nylon plumb line is hung directly above the four track wheels of the main transformer transfer trolley, and the plumb bob points downward to the track surface. The operator needs to stop the machine every 0.5 to 1 meter during the traction of the main transformer, bend down to the space under the main transformer which is usually not more than 500 millimeters high, and manually measure the transverse distance between the plumb line projection point and the track center line or the initial marked point using a steel ruler. By calculating the change of the distance, it is determined whether the transverse deviation is out of limit, and the traction rope is adjusted accordingly.

[0004] However, this prior art has a series of inherent and difficult-to-overcome defects: Low measurement accuracy: the plumb line will produce a swing of 2-5 mm due to the influence of hanging height and on-site equipment vibration, resulting in instability of the reference line itself. At the same time, in a narrow space, the measurement personnel's visual angle is limited, and there is an unavoidable parallax when reading the steel ruler scale, and the comprehensive measurement error often exceeds ±3 mm, which is difficult to reliably approach the safety threshold of 10 mm; low operation efficiency and high risk: the frequent "moving-stopping-measuring" cycle seriously interrupts the continuity of the traction operation, and the cumulative time of a short path may exceed 3 hours. In addition, the operator needs to work in the environment of the bottom of the main transformer, which is limited in space, dim in light, and uneven in ground, with high risk of collision and safety; complete lack of dynamic monitoring capability: this technology belongs to discrete point sampling measurement, and the data update frequency is very low, which cannot capture the instantaneous deviation occurring in the traction interval. This data blind area makes the adjustment operation lag seriously, and cannot realize preventive deviation correction; backward technology: completely relying on manual operation and mechanical measuring tools, the measurement result is significantly affected by the operator's experience and responsibility, and the data cannot be automatically recorded and traced, which is seriously out of line with the development trend of modern industry automation and digitization.

[0005] Therefore, it is necessary to propose a horizontal image ranging device and method for transformer replacement traction to solve the above problems. SUMMARY

[0006] The technical problem to be solved by the present application is to provide a horizontal image ranging device and method for transformer replacement traction, which aims to solve the problems of excessive dependence on manual operation, low measurement accuracy, low operation efficiency and high risk, and complete lack of dynamic monitoring capability, and has the characteristics of automatic realization of traction control and automatic deviation correction in the process of transformer replacement.

[0007] A horizontal image ranging device for transformer replacement traction, comprising: A lateral visual image deviation measurement system configured to be installed at the track wheel frame of the main transformer transfer trolley, for non-contact measurement of the lateral distance between the track wheel and the track center line, and output of the lateral deviation; A lateral deviation measurement mounting frame for fixing the lateral visual image deviation measurement system above the track wheel frame; A quick clamping device for quickly fixing the lateral visual image deviation measurement system to the lateral deviation measurement mounting frame; An image processing host computer in communication connection with the lateral visual image deviation measurement system, configured to process the collected images and calculate the lateral deviation; The lateral visual image deviation measurement system includes an industrial camera, and the optical axis of the industrial camera is perpendicular to the extension direction of the track, so as to capture an image containing the edges of the track wheel and the track.

[0008] Preferably, the lateral deviation measuring mounting frame is fixed by a hand wheel fixing mechanism, which comprises a U-shaped clamping seat clamped with the skirt of the main transformer; a locking wheel is threadedly connected to the top of the clamping seat, and a pad is connected to the bottom of the locking wheel and the inner wall of the clamping seat for clamping and locking.

[0009] Preferably, the bottom of the hand wheel fixing mechanism is slidingly connected with an adjusting support of the cross slide structure to realize position adjustment in the longitudinal and lateral directions.

[0010] Preferably, a gimbal is connected to the bottom of the adjusting support, and the image visual lateral deviation measuring detector is mounted on the quick clamping device through the gimbal.

[0011] Preferably, the image processing host performs machine vision-based lateral deviation calculation, which is specifically configured to: preprocess the collected image to obtain a gray image; accurately extract the track wheel edge and the track edge by using a sub-pixel edge detection algorithm; calculate the pixel distance between the track wheel and the track center line based on the extracted edge; convert the pixel distance into an actual physical lateral deviation through camera calibration parameters.

[0012] Preferably, a lateral image ranging method for transformer replacement traction comprises the following steps: deploy the lateral visual image deviation measuring system above the track wheel frame of the main transformer transfer trolley through the lateral deviation measuring mounting frame and the quick clamping device; adjust the viewing angle of the lateral visual image deviation measuring system so that its optical axis is perpendicular to the track and its field of view covers the track wheel and the track; collect images containing the track wheel and the track through the lateral visual image deviation measuring system; calculate the lateral deviation between the track wheel and the track center line in real time through an image processing algorithm based on the collected images; output the lateral deviation to the main control system for deviation correction control in the main transformer traction process.

[0013] Preferably, the step of calculating the lateral deviation in real time based on the collected images through the image processing algorithm specifically comprises: (1) image preprocessing: convert the collected color image into a gray image , wherein is the pixel coordinate; smooth and denoise the gray image using a Gaussian filter to obtain the preprocessed image ; (2) sub-pixel level extraction of edge features: Defining the region of interest of the track wheel edge and the track edge in the image; In the region of interest, pixel-level edge rough positioning is performed by using a Canny operator; For each pixel-level edge point A neighborhood of is taken as the center; The Zernike moments of the neighborhood are calculated ; The sub-pixel level accurate position of the edge is solved by using the formula : ; ; ; Where S is a scale normalization factor; Obtaining the track wheel sub-pixel edge profile And the track sub-pixel edge profile ; (3) Lateral displacement calculation and output: Based on the profile , the position of the track center line in the image is determined by linear fitting ; Based on the profile , the position of the track wheel flange in the image is determined ; Calculate the pixel lateral displacement: ; By using the pre-calibrated camera parameters, the pixel lateral displacement is converted into the physical lateral displacement by using the coordinate conversion formula ; ; Where M is the pixel equivalent; Finally, the physical lateral displacement is output .

[0014] Preferably, the track wheel edge is the inner side edge of the rim, and the track edge is the inner side working edge of the rail head.

[0015] Preferably, after deploying the lateral visual image deviation measurement system, the viewing angle is fine-tuned by its micro gimbal to minimize the interference of the complex ground scene.

[0016] Preferably, the data update frequency of the lateral displacement is not less than 10Hz, realizing continuous and dynamic monitoring of the traction lateral displacement of the main transformer.

[0017] The beneficial effects of the present application are: 1, The prior art has low and unstable measurement precision due to plumb line swing and artificial reading parallax, and the present application eliminates errors caused by mechanical contact and reference line swing by deploying a lateral visual image deviation measurement system and using a non-contact image measurement method; the method uses a sub-pixel edge detection algorithm to calculate the complex matrix of the image neighborhood and calculate the positioning precision of the track wheel and the track edge from the pixel level to the sub-pixel level below 0.1 pixel. This makes the final lateral deviation measurement precision stable within ±1mm, which is much higher than the traditional method; such high-precision real-time data enables the operator to detect the trend of slight deviation in the first time in the complex vibration environment of the hydropower station, and to intervene accurately before the deviation approaches the safety threshold, thereby completely eliminating the risk of rail biting and providing unprecedented technical support for the safe traction of hundreds of tons of equipment.

[0018] 2, The present application completely abandons the discontinuous operation mode of traditional technology. Through the high-frequency data acquisition capability of the lateral image ranging device ≥10Hz, the lateral deviation of the main transformer during the entire traction process is monitored continuously. This means that the traction equipment no longer needs to be frequently started and stopped for measurement, and the equipment idle waiting time is reduced to nearly zero compared with traditional technology, and the continuity of the main transformer movement is greatly improved; at the same time, the modular design cooperates with the quick clamping device and the miniature gimbal, so that the deployment and debugging of the system can be completed by a single person within ten minutes, and the efficiency is greatly improved compared with the traditional method.

[0019] 3, The present application introduces modern machine vision technology into the traditional power maintenance field, reduces the proportion of manual participation, and greatly improves the operation safety; the operator no longer needs to enter the narrow space and dangerous main transformer bottom for dangerous manual measurement, but only needs to monitor the real-time data remotely through the mobile display panel in the safe area, which fundamentally eliminates the personal safety risk; the device has high versatility: on the hardware, through the adjustable mounting bracket and the quick clamping mechanism, it can quickly adapt to different power stations, different types of transfer trolleys and tracks; on the software, based on image recognition, it is not limited by specific mechanical size; and the continuous and high-frequency deviation data stream produced by the present application is a valuable data basis for building a main transformer traction digital twin model and training AI predictive maintenance algorithms, which can promote the overall digitalization of the river basin hydropower station operation and maintenance management. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a device structure schematic diagram of the present application; Figure 2 is an installation schematic diagram of the present application. DETAILED DESCRIPTION

[0021] Example 1: As shown in Figure 1 , a lateral image ranging device for transformer replacement traction, comprising: A lateral vision image deviation system is configured to be installed at the track wheel frame of the main transformer transfer trolley, used for non-contact measurement of the lateral distance between the track wheel and the track center line, and output of the lateral deviation value; A lateral deviation mounting frame is used to fix the lateral vision image deviation system above the track wheel frame; A quick clamping device is used to quickly fix the lateral vision image deviation system on the lateral deviation mounting frame; An image processing host is in communication connection with the lateral vision image deviation system, configured to process the collected images and calculate the lateral deviation value; The lateral vision image deviation system includes an industrial camera, and the optical axis of the camera is perpendicular to the track extension direction to capture images containing the track wheel edge and the track edge.

[0022] Preferably, the lateral deviation mounting frame is fixed by a hand wheel fixing mechanism, which includes a U-shaped clamping seat clamped with the skirt of the main transformer, and a locking wheel is threadedly connected to the top of the clamping seat, and a pad is connected to the bottom of the locking wheel and the inner wall of the clamping seat for clamping and locking.

[0023] Preferably, the bottom of the hand wheel fixing mechanism is in sliding connection with the adjusting support of the cross slide structure, realizing longitudinal and lateral position adjustment.

[0024] Preferably, the bottom of the adjusting support is connected with a gimbal through a universal joint, and the image vision lateral deviation measurement probe is installed on the quick clamping device through the gimbal.

[0025] Preferably, the image processing host performs machine vision-based lateral deviation value calculation, which is specifically configured to: Pretreat the collected images to obtain a gray image; Use a sub-pixel edge detection algorithm to accurately extract the track wheel edge and the track edge; Based on the extracted edges, calculate the pixel distance between the track wheel and the track center line; Convert the pixel distance into the actual physical lateral deviation value through the camera calibration parameters.

[0026] As shown in Figure 2 Preferably, a lateral image ranging method for transformer replacement traction includes the following steps: Deploy the lateral vision image deviation system above the track wheel frame of the main transformer transfer trolley through the lateral deviation mounting frame and the quick clamping device; Adjust the viewing angle of the lateral vision image deviation system so that its optical axis is perpendicular to the track, and ensure that its field of view covers the track wheel and the track; Collect images containing the track wheel and the track through the lateral vision image deviation system; Based on the collected image, the lateral deviation between the track wheel and the track center line is calculated in real time by an image processing algorithm; The lateral deviation is output to the master control system for deviation correction control during the main transformer traction process.

[0027] Preferably, the step of calculating the lateral deviation in real time based on the collected image by the image processing algorithm specifically comprises: (1) Image preprocessing: Convert the collected color image to a grayscale image , wherein is the pixel coordinate; Smooth and denoise the grayscale image using Gaussian filtering to obtain the preprocessed image ; (2) Sub-pixel level extraction of edge features: Define the region of interest of the track wheel edge and the track edge in the image; Within the region of interest, use the Canny operator for pixel-level edge rough positioning; For each pixel-level edge point , take a neighborhood centered on it; Calculate the Zernike moment of the neighborhood ; Use the formula to solve the sub-pixel level accurate position of the edge : ; ; ; Where S is the scale normalization factor; Obtain the track wheel sub-pixel edge profile and the track sub-pixel edge profile ; (3) Lateral deviation calculation and output: Based on the profile , determine the position of the track center line in the image by linear fitting ; Based on the profile , determine the position of the track wheel flange in the image ; Calculate the pixel lateral deviation: ; Use the pre-calibrated camera parameters to convert the pixel lateral deviation to physical lateral deviation by the coordinate conversion formula ; ; Wherein M is pixel equivalent; Finally output the physical lateral deviation .

[0028] Preferably, the track wheel edge is the inner edge of the rim, and the track edge is the inner working edge of the rail head.

[0029] Preferably, after deploying the lateral vision image deviation measurement system, the viewing angle is fine-tuned through its miniature pan-tilt to minimize the interference of complex ground scenes.

[0030] Preferably, the data update frequency of the lateral deviation is not less than 10Hz, realizing continuous and dynamic monitoring of the lateral deviation of the main transformer traction.

[0031] Embodiment two: The embodiment provides a technical scheme for outputting the lateral deviation to a main control system for deviation correction control in the main transformer traction process, which is as follows: 1. Data encapsulation and standardized transmission: Data packet structure: the image processing host or local processing unit encapsulates the calculated physical lateral deviation, corresponding timestamp, data quality identifier such as signal-to-noise ratio, confidence, and sensor ID, i.e., identifier of a specific track wheel, into a standard data structure. Communication protocol: the data packet is pushed to the positioning device main control system at a fixed frequency of ≥10Hz through an industrial Ethernet or real-time Ethernet protocol. The protocol stack ensures the real-time and certainty of data transmission, and the network delay is usually required to be less than 50ms.

[0032] 2. Main control system data receiving and preprocessing: Data verification and filtering: after receiving the data packet, the main control system first performs CRC verification to ensure data integrity. Subsequently, a digital filter such as a first-order low-pass filter or a Kalman filter is applied to the continuous lateral deviation data stream to suppress high-frequency measurement noise, smooth the data curve, and extract the true trend of the main transformer lateral deviation; Data fusion: the main control system performs space-time alignment and fusion of the smoothed lateral deviation, real-time position information from the longitudinal laser ranging device, and data from other modules; through the fusion algorithm, the main control system can calculate the overall lateral deviation and yaw angle of the main transformer.

[0033] 3. Deviation correction decision generation based on AI control algorithm: State judgment and trend prediction: the main control system compares the fused real-time attitude with the preset safety threshold by using the built-in AI auxiliary analysis and prediction algorithm. The algorithm analyzes the historical trend and speed of the deviation and predicts whether the deviation will exceed the limit or cause “rail gnawing” in a short period of time in the future if no intervention is taken.

[0034] Quantitative correction instruction calculation: Manual / semi-automatic mode: the algorithm generates understandable quantitative operation prompts based on the prediction results.

[0035] For example: if the main variable is detected to be 2mm right offset and the trend is increasing, it is recommended to increase the left side traction speed by 5% in the next 2 seconds and maintain the right side unchanged.

[0036] This provides precise and forward-looking operation guidance for operators, replacing traditional experience-based judgment.

[0037] Full-automatic mode: the algorithm directly outputs a closed-loop control quantity. This control quantity is usually calculated based on classical control theory such as PID control or more advanced model predictive control MPC. The core is to calculate a correction angular velocity or a direct left-right side speed difference.

[0038] 4. Instruction output and execution feedback: Human-machine interface presentation: the generated correction instructions are sent to the mobile display tablet system in real time. The mobile display tablet system clearly displays in a graphical way, accompanied by sound and light prompts, to ensure that the operator can understand intuitively.

[0039] Control system interface: in full-automatic mode, the main control system sends speed instructions directly to the traction motor driver or hydraulic proportional valve controller through standard industrial field bus or analog / digital output modules.

[0040] Closed-loop verification: after the execution of the instructions, the lateral image monitoring system continuously monitors the change of the lateral deviation and feeds back the new data to the main control system. The main control system verifies the effectiveness of the control instructions by comparing the expected correction effect with the actual feedback data. If the deviation does not converge as expected, the AI algorithm will dynamically adjust the control parameters or re-evaluate the system model to generate new correction instructions, forming a complete and adaptive intelligent control closed loop of "perception - decision - execution - verification".

Claims

1. A transverse image ranging device for transformer replacement traction, characterized in that, include: The lateral visual image deviation measurement system is configured to be installed on the track wheel frame of the main transformer transfer trolley for non-contact measurement of the lateral distance between the track wheel and the track centerline, and outputs the lateral deviation. A lateral deviation measurement mounting bracket is used to fix the lateral visual image deviation measurement system above the track wheel frame; A quick-fix device is used to quickly fix the lateral visual image deviation measurement system onto the lateral deviation measurement mounting frame; The image processing host is communicatively connected to the lateral visual image deviation measurement system and is configured to process the acquired images and calculate the lateral deviation. The lateral visual image measurement system includes an industrial camera whose optical axis is perpendicular to the track extension direction to capture images including the edge of the track wheel and the edge of the track.

2. The transverse image ranging device for transformer traction replacement according to claim 1, characterized in that, The lateral deviation mounting bracket is fixed by a handwheel fixing mechanism, which includes a U-shaped bracket that engages with the skirt of the main transformer; a locking wheel is threaded to the top of the bracket, and pads are connected to the bottom of the locking wheel and the inner wall of the bracket for clamping and locking.

3. The transverse image ranging device for transformer replacement traction according to claim 2, characterized in that, The bottom of the handwheel fixing mechanism is slidably connected to the adjustment bracket of the cross slide structure to achieve position adjustment in the longitudinal and transverse directions.

4. The transverse image ranging device for transformer replacement traction according to claim 3, characterized in that, The bottom of the adjustment bracket is connected to a gimbal via a universal joint, and the image visual lateral deviation measurement detector is mounted on the fast card device via the gimbal.

5. The transverse image ranging device for transformer traction replacement according to claim 1, characterized in that, The image processing host performs machine vision-based lateral deviation calculation, specifically configured as follows: The acquired images are preprocessed to obtain grayscale images; Using a sub-pixel edge detection algorithm, the edges of the track wheel and the track are accurately extracted; Based on the extracted edges, calculate the pixel distance between the track wheel and the track centerline; The pixel distance is converted into an actual physical lateral deviation by using camera calibration parameters.

6. A transverse image ranging method for transformer traction replacement according to any one of claims 1-5, characterized in that, Includes the following steps: The lateral visual image deviation measurement system is deployed above the track wheel frame of the main transformer transfer trolley using a lateral deviation measurement mounting bracket and a quick-clamp device. Adjust the viewing angle of the lateral visual image measurement system so that its optical axis is perpendicular to the track and its field of view covers the track wheel and the track; The lateral visual image deviation measurement system acquires images containing the track wheel and track; Based on the acquired images, the lateral deviation between the track wheel and the track centerline is calculated in real time using image processing algorithms. The lateral deviation is output to the main control system for correction control during the main transformer traction process.

7. The transverse image ranging method for transformer traction replacement according to claim 6, characterized in that, The step of calculating the lateral deviation in real time based on the acquired image using an image processing algorithm specifically includes: (1) Image preprocessing: Convert the acquired color image to a grayscale image. ,in These are pixel coordinates; Gaussian filtering is used to smooth and denoise the grayscale image, resulting in the preprocessed image. ; (2) Sub-pixel level extraction of edge features: Define the region of interest (ROI) for the wheel edge and the track edge in the image; Within the region of interest, the Canny operator is used for coarse pixel-level edge localization. For each pixel-level edge point Take one as the center The neighborhood; Calculate the Zernike moments of this neighborhood. ; Calculate the sub-pixel precise position of the edge using a formula. : ； ； ； Where S is the scale normalization factor; Obtain the subpixel edge contour of the orbit wheel and orbital subpixel edge contour ; (3) Calculation and output of lateral deviation: Based on contour The position of the orbital centerline in the image is determined by linear fitting. ; Based on contour Determine the position of the track wheel flange in the image. ; Calculate pixel lateral offset: ； Using pre-calibrated camera parameters, pixel lateral offset is converted into physical lateral offset using a coordinate transformation formula. ; ； Where M is the pixel equivalent; Finally, the physical lateral deviation is output. .

8. The transverse image ranging method for transformer traction replacement according to claim 7, characterized in that, The edge of the track wheel is the inner edge of the wheel rim, and the edge of the track is the inner working edge of the track head.

9. A transverse image ranging method for transformer traction replacement according to claim 6, characterized in that, After deploying the lateral visual image deviation measurement system, the viewing angle is fine-tuned using its miniature gimbal.

10. A transverse image ranging method for transformer traction replacement according to claim 6, characterized in that, The data update frequency for the lateral deviation is no less than 10Hz.

Images