Automatic powder filling and powder filling method and system for powder high-temperature alloy sleeve based on visual identification

Abstract

This invention belongs to the field of powder superalloy technology, and relates to an automatic powder filling method and system for powder superalloy cladding based on vision recognition. The system includes a powder vibration system, a powder flow control system, a vision recognition system, a comprehensive control system, and a hydraulic vibration system. The vision recognition system collects morphological features such as the powder filling area change rate, contour regularity, and shadow grayscale to predict the filling state in real time, and transmits the signal to the comprehensive control system. Based on the filling state, the system controls the powder vibration system and powder flow control system to automatically replenish powder, and controls the hydraulic vibration system to adjust displacement, frequency, and magnitude to compact the powder within the cladding, ultimately completing the automatic powder filling of the cladding. This invention can automatically identify the powder filling state and simultaneously couple with hydraulic vibration for compaction, improving the stability of powder distribution control during filling and solving the problems of fine powder aggregation, low manual operation efficiency, and high powder loss in existing filling technologies.

Description

Technical Field

[0001] This invention belongs to the field of powder superalloy technology, and particularly relates to an automatic powder filling method and system for powder superalloy cladding based on visual recognition. Background Technology

[0002] As a key component of high-performance aero-engines, powder metallurgy superalloy (PMA) parts endure the combined effects of high temperature and high stress, operating under extremely harsh conditions and requiring complex manufacturing processes. The production of PMA parts typically involves a series of steps, including powder preparation, powder degassing and loading, pre-sintering, and hot isostatic pressing (HIP). To ensure successful HIP, the powder needs to be vacuum-sealed in a cylindrical sleeve with a powder nozzle. Uniform powder distribution, appropriate and precise weighting are prerequisites for obtaining a fully dense, defect-free HIP blank. Uneven powder distribution can lead to fine powder agglomeration, directly affecting the uniformity of the microstructure after isothermal forging and solution aging, and may even result in coarse recrystallized grains, leading to part scrap. Insufficient weighting results in residual porosity, severely impairing the dynamic mechanical properties of the material, especially fatigue and impact resistance, while also affecting the dimensional control accuracy of the subsequent blank. Therefore, stable control of the powder loading process directly determines the microstructure and properties of the HIP blank, thus affecting engine performance, and is therefore extremely important.

[0003] The existing packaging and filling process relies on manual tapping and observation to control the compaction of a mechanical vibrating table. Regarding quality control, this manual operation is heavily influenced by human factors; different people have different visual judgments and control the compaction parameters differently. This not only leads to large fluctuations in filling weight but can also cause frequent excessively long vibration times, resulting in fine powder agglomeration. Furthermore, the mechanical vibrating table can only control the vibration frequency, with limited control parameters, further increasing the difficulty of controlling the packaging weight throughout the process. This results in a current packaging filling weight variation coefficient (standard deviation / mean × 100%) of approximately 1.5%. Regarding cost control, the variability in judgment based on manual visual observation leads to powder overflow and loss of approximately 3% due to failure to promptly determine the full filling status or incorrect adjustment of compaction parameters. In terms of production efficiency, this method is inefficient; completing the powder filling of one high-temperature alloy packaging takes 1 to 6 days, requiring one person to continuously control the operation, making it impossible to implement a one-person-multiple-machine production model, and severely consuming manual labor time.

[0004] Therefore, there is an urgent need for an automatic powder filling method and system based on visual recognition for high-temperature alloy powder packaging. On the one hand, this would reduce the impact of human factors, improve the stability of the packaging filling process, and reduce powder loss. On the other hand, it would reduce the time spent on manual labor, improve production efficiency, reduce costs and increase efficiency, and bring greater benefits. Summary of the Invention

[0005] The purpose of this invention is to provide a visual recognition-based automatic powder filling method and system for high-temperature alloy powder packaging, in order to solve the problems of large fluctuations and low efficiency caused by human influence in the existing powder high-temperature alloy packaging filling process.

[0006] To solve this technical problem, the technical solution of the present invention is as follows: On one hand, this invention provides an automatic powder filling method for high-temperature alloy powder cladding based on visual recognition, including a powder recognition stage and a controlled filling stage. Firstly, in the powder recognition stage, morphological features are extracted through visual recognition, and the image is preprocessed and optimized to improve recognition accuracy. Secondly, an edge detection algorithm is used to extract the features of the circular cladding powder inlet, focusing the image recognition area within the inlet region. A combination of OTU adaptive global thresholding and local sliding thresholding is used to perform threshold segmentation on the preprocessed image, distinguishing between filled and unfilled areas. Based on the calculated filling area, filling area change rate, shadow contour regularity, and grayscale distribution morphological features, the powder filling rate and stacking uniformity are determined. In the controlled filling stage, 80% and 90% are used as the dividing points for powder filling amount, and filling and powder replenishment operations are controlled in different stages according to the powder filling weight.

[0007] The specific steps for the powder identification stage are as follows: Step 1-1: Image preprocessing; Visual recognition requires image preprocessing to improve recognition accuracy for morphological feature extraction, including "Gaussian denoising" (removing ambient light interference) and "grayscale conversion" (converting color images to black and white to reduce computation).

[0008] Step 1-2: Defining Filled and Unfilled Regions; First, an edge detection algorithm is used to extract the features of the circular toner inlet, focusing the image recognition area within the toner inlet region. Second, a combination of OTU adaptive global thresholding and local sliding thresholding is used to perform threshold segmentation on the preprocessed image, thereby distinguishing between filled and unfilled regions; details are as follows: Defining Filled and Unfilled Regions: First, an edge detection algorithm is used to extract the features of the circular powder inlet, focusing the image recognition area within the inlet region. Second, a combination of Otsu adaptive global thresholding and local sliding thresholding is used to perform threshold segmentation on the preprocessed image, thus distinguishing between filled and unfilled regions. Specifically, the adaptive global threshold determines the approximate range, and then fine-tuning is performed within the local range to eliminate the influence of lighting. After threshold segmentation, the system clearly divides the two types of regions using "grayscale value ranges." Unfilled regions include the container inner wall, background environment, and areas directly illuminated; their grayscale values ​​are below the Otsu adaptive global threshold, and there is no material obstruction. Filled regions refer only to the obstructed areas formed by powder material under a fixed light source; their grayscale values ​​are above the Otsu adaptive threshold. This region is the sole target for subsequent feature extraction, completely eliminating background interference.

[0009] Steps 1-3: Fill area calculation; Automatically count the total number N of pixels with "grayscale value ≥ threshold" in the grayscale image. The fill area S = total number of pixels N × conversion factor. The area is recorded as S when the filling status is "full". 满 The area when the filling state is "empty" is denoted as S. 空 The details are as follows: Fill area calculation: The system automatically counts the total number of pixels in the grayscale image with a grayscale value ≥ Otsu threshold (denoted as N). This number directly corresponds to the pixel coverage area of ​​the shadow area. The final fill area S = total number of pixels N × conversion factor, used to determine whether the filling status is "full" or "empty". When the filling status is "full", the area is denoted as S. 满 The area when the filling state is "empty" is denoted as S. 空 This step employs a precise method of "pixel-level statistics + actual size mapping" rather than the traditional rough estimation, achieving the conversion from "pixel-level precision" to "actual physical size," thus providing a quantitative basis for subsequent material injection quantity determination.

[0010] Steps 1-4: Propose rules for judging powder replenishment rate and packing uniformity: Calculate the rate of change of filling area using the filling area, and then adjust the frequency, displacement, and magnitude of the casing vibration; calculate the outline rectangularity to determine the regularity of the shadow outline; calculate grayscale changes to adjust the vibration frequency to optimize material distribution; details are as follows: Multi-feature collaborative judgment: Based on the calculated filling area, filling area change rate, shadow outline regularity, and gray-scale distribution morphological features, the system completes the judgment of powder replenishment rate and stacking uniformity.

[0011] The rate of change of filling area (ΔS / Δt) is defined as follows: ΔS = current filling area - previous filling area, Δt is the time interval, and ΔS / Δt directly reflects the powder replenishment rate. Based on the different powder replenishment rates, the magnitude, frequency, and displacement of the hydraulic vibration system are set in a coordinated manner. The regularity of the filling contour is determined by using an edge detection algorithm to extract the complete boundary of the filling area and calculating the contour rectangularity = actual area of ​​the filling area / area of ​​the minimum bounding rectangle of the filling area. When the rectangularity is ≥0.9, the filling contour is close to a regular rectangle and is judged as "uniform stacking"; when the rectangularity is <0.9, it is judged as "uneven filling". The grayscale distribution of the filling area involves generating a grayscale histogram of the filling area and calculating the "grayscale mean" and "grayscale variance". The grayscale values ​​of the filling area are statistically analyzed, a grayscale histogram is generated, and the grayscale mean and variance of the filling area are calculated. When the local grayscale mean > 120 or the grayscale variance > 50, the system will adjust the vibration frequency to optimize the material distribution. The filling area change rate (ΔS / Δt) is defined as follows: ΔS = current filling area - previous filling area, Δt is the time interval, and ΔS / Δt directly reflects the powder replenishment rate. Based on the different powder replenishment rates, the magnitude, frequency, and displacement of the hydraulic vibration system are adjusted accordingly. If ΔS / Δt is greater than a certain threshold, it indicates a faster powder replenishment rate, requiring compaction according to the initial parameters. If ΔS / Δt is less than a certain threshold, it indicates a slower powder replenishment rate, requiring parameter adjustment according to rules. This ensures the powder replenishment efficiency and uniformity of the powder compaction process, reducing powder waste.

[0012] The regularity of the filling contour is determined by using an edge detection algorithm to extract the complete boundary of the filling area, and calculating the contour rectangularity as the ratio of the actual area of ​​the filling area to the area of ​​the minimum bounding rectangle. When the rectangularity is ≥0.9, the filling contour is close to a regular rectangle and is judged as "uniformly stacked"; when the rectangularity is <0.9, the contour shows obvious depressions (such as local material shortages) or bulges (such as local excessive stacking) and is judged as "uneven material filling".

[0013] The grayscale distribution of the filling area involves generating a grayscale histogram of the filling region and calculating the "grayscale mean" and "grayscale variance". Grayscale values ​​are statistically analyzed in the filling area to generate a grayscale histogram (grayscale value range 0-255). The grayscale mean and variance of the filling area are calculated (under normal stacking conditions, the grayscale mean is stable between 50-80, and the variance is <20). If the grayscale mean suddenly increases (local grayscale mean >120), it indicates that the material in that area is sparse, and there may be local gaps. If the grayscale variance is too large (grayscale variance >50), it may indicate material bridging. When the local grayscale mean >120 or the grayscale variance >50, the system will adjust the vibration frequency to optimize the material distribution. By using the regularity of the filling contour and grayscale distribution to determine the appropriate thresholds for raw materials with different particle sizes and flowability, including the rate of change of area and the rectangularity of the contour, the system avoids misjudgments caused by the traditional "one-size-fits-all" approach of fixed area thresholds. Simultaneously, it replaces subjective human judgment with quantitative features obtained through visual recognition, ensuring a uniform distribution of powder during the filling process and reducing non-compliant material textures caused by fine agglomeration. If the calculated local grayscale mean is between 80-120 and the grayscale variance is between 20-50, it indicates that the material packing is not perfectly uniform, but neither is it sparse or bridging. In this case, the system continues to monitor without adjusting the vibration frequency.

[0014] In the controlled filling stage, 80% and 90% of the powder filling weight are used as the dividing points for the powder filling amount, and the filling and powder replenishment are controlled in different stages; the specific operation steps are as follows: Step 2-1: The powder loading amount is (0%-80%), and the loading is continuous; a certain amount of high-temperature alloy powder is put into the powder vibration system through the powder tank. After the vacuum process requirements are met, the powder vibration system is automatically turned on. The high-temperature alloy powder climbs in the degassing vibration component and undergoes surface degassing treatment. After climbing to the top, it falls along the pipeline. When the flow meter in the powder flow control system detects that the powder has reached a certain flow rate, its air vibration component is automatically turned on to ensure that the powder passes through the filter device and enters the cladding evenly. Step 2-2: When the powder filling level is between 80% and 90%, start replenishing powder. The visual recognition system identifies the powder filling status by collecting the area morphological features of the powder filling. When the powder filling status is fully detected for the first time, the hydraulic vibration system is turned on according to the initial settings, and the powder vibration component, flow meter, and air vibration component are turned off in sequence. The hydraulic vibration system vibrates for 30-40 seconds and then continues to enter the continuous filling stage. During this stage, the powder filling status will be fully detected 1-3 times, and the parameters of the hydraulic vibration system do not need to be changed. Steps 2-3: When the powder loading is between 90% and 100%, continuously replenish the powder. During this period, the powder loading status will frequently switch between full and empty. When the powder loading status is full for more than 10-15 seconds, the vibration parameters of the hydraulic vibration system need to be adjusted, and the frequency and magnitude are gradually increased until the powder loading status remains full for more than 20-30 seconds, at which point the compaction operation ends.

[0015] The dividing point for the powder loading amount is determined based on the loading weight of the powder. The calculation formula in step 2-1 is (loose packing density × loose packing volume) / (taped density × tapped volume). In step 2-2, based on experience, the weight in step 2-1 is increased by 10%. Finally, 80% to 90% is selected as the dividing point. The distinction of the dividing point classifies the difficulty of overall process control, which helps to reduce the complexity of the overall control process, ensures the stability of control at each stage, and reduces the occurrence of bug-type problems.

[0016] Furthermore, in steps 1-2, the Otsu adaptive threshold for the filled region is 120-255, and the Otsu adaptive threshold for the unfilled region is 0-80.

[0017] Furthermore, in steps 1-2, the threshold for a "empty" state is 10-15%S. 满 .

[0018] Furthermore, in steps 1-3, the conversion factor of "pixel-actual area" needs to be obtained through the previous camera calibration (based on a standard-size calibration board).

[0019] Furthermore, the visual recognition system needs to undergo sample training and error calibration to ensure that the error in morphological feature judgment is controlled within 3% and that the visual recognition accuracy is consistently ≥95%.

[0020] Furthermore, in step 2-1, if the flow rate is greater than 0.5 kg / s, the air vibration assembly is activated.

[0021] Furthermore, in step 2-1, the opening amplitude of the pneumatic valve of the air vibration assembly is controlled between 20% and 40%.

[0022] Furthermore, in step 2-2, the initial vibration parameters of the hydraulic vibration system are displacement of 1.6-2.2 mm, frequency of 18 Hz-20 Hz, and magnitude of 50-70%.

[0023] Further, in steps 2-3, when the powder filling state exceeds 10-15 seconds, the frequency of the hydraulic vibration system increases by 1-3 Hz each time, and the magnitude increases by 10-20% each time, with a maximum frequency not exceeding 25 Hz and a maximum magnitude of 100%. The magnitude of each increase in the frequency and magnitude of the hydraulic vibration system is determined based on the rate of change of the filling area. The specific thresholds corresponding to steps 1-4 are as follows: When 6%S 满 / s≤ΔS / Δt≤9%S 满 / s, only the magnitude of the hydraulic vibration system is adjusted, with each increase in amplitude being 10%, and the maximum magnitude being 100%; When 3%S 满 / s≤ΔS / Δt<6%S 满 / s, the frequency adjustment increment is 1Hz each time, the displacement decreases by 0.2mm, and the minimum displacement is 1.6mm; When the change in ΔS / Δt is less than 3%S 满 / s, initial magnitude 70%, frequency adjustment amplitude increases by 2Hz each time, displacement decreases by 0.4mm, magnitude increases by 20% each time, magnitude maximum 90%, minimum displacement is 1.6mm.

[0024] The formula for calculating the threshold value of the filling area change rate is (S 满 -S 空 The formula is: cosα / powder loading time threshold (i.e., 10-15s and 20-30s in steps 2-3), where α is the angle between the camera and the plane. The quantification of relevant parameters is based on experimental data on reducing fine powder aggregation, which can quantitatively control the stability of the loading process and reduce the problem of poor particle aggregation caused by human influence.

[0025] Furthermore, in steps 2-3 and 1-4, when the contour rectangularity is <0.9, the local grayscale mean is >120, or the grayscale variance is >50, the frequency is increased by 1-3Hz for vibration optimization to avoid uneven distribution.

[0026] Furthermore, in steps 2-3, the vibration waveform of the hydraulic vibration system is controlled by a sine wave, and the difference between the downward acceleration and the upward acceleration is between 0.5g and 1g (g is the acceleration due to gravity).

[0027] On the other hand, the present invention provides an automatic powder filling system for high-temperature alloy powder cladding based on visual recognition, comprising: The powder vibration system uses a high-vacuum environment and a degassing vibration component to perform surface degassing treatment on high-temperature alloy powder. The powder flow control system is used to transfer high-temperature alloy powder that meets the degassing process to the powder high-temperature alloy cladding after passing through a flow meter and an air vibration component. The visual recognition system uses a high-definition image acquisition component to capture the powder stacking status at the upper inlet of the powder high-temperature alloy cladding, and simultaneously transmits the signal to the integrated control system. The integrated control system is used to receive video recognition signals from the vision recognition system, judge the powder stacking status, control the powder vibration system and powder flow control system to carry out powder transfer and filling, and control the hydraulic vibration system to carry out powder compaction. The hydraulic vibration system is used to control the frequency, displacement, and magnitude of the sleeve vibration to achieve the sleeve compaction operation.

[0028] Furthermore, the high-definition image acquisition component of the visual recognition system adopts a dual-camera acquisition method, with the acquisition angles forming angles of 40°-60° and 0°-10° with the vertical powder delivery pipe, respectively. The dual cameras perform feature fusion through weighted averaging, with the top area weighted at 0.6-0.7 and the side area weighted at 0.3-0.4, to comprehensively determine the material injection status.

[0029] Furthermore, the high-definition image acquisition component of the visual recognition system needs to have a pixel count of 5 million or more and a field of view greater than 80mm × 80mm.

[0030] Furthermore, the integrated control system includes a visual analysis module adapted to the visual recognition system, which judges real-time parameters based on morphological features and actual production requirements. Its image processing speed is ≤100ms / frame, and the morphological feature judgment error is <3%.

[0031] The beneficial effects of this invention are: This invention employs a visual recognition-based automatic powder filling system and method for high-temperature alloy powder packaging, replacing existing methods and systems that rely on manual observation, manual tapping, and mechanical vibration table control. On one hand, it accurately locates and identifies filled and unfilled areas within the recognition region by combining the Canny edge detection algorithm with a threshold segmentation algorithm, improving visual recognition accuracy. On the other hand, it constructs a "raw material characteristics - judgment threshold" association library: for raw materials with different particle sizes and flowability, it automatically adjusts judgment thresholds such as area change rate and outline rectangularity, avoiding misjudgments caused by the traditional fixed threshold approach. Simultaneously, by binding vibration parameters with visual recognition features, it enhances process control stability, reduces the impact of human factors, and improves the vibration control stability of the packaging filling process. Its filling weight variance can be controlled within 0.6%-0.7%, a 50% improvement compared to existing coefficients of variation, with no fine powder aggregation and powder loss controlled within 1%. In terms of efficiency, it reduces manual labor time and human intervention by 60%. This invention solves the problems of fine powder agglomeration, poor powder filling weight stability, low manual operation efficiency, and high powder loss in existing filling technologies. Attached Figure Description

[0032] To more clearly illustrate the technical solutions implemented in this invention, the accompanying drawings used in the embodiments of this invention will be briefly explained below. Obviously, the drawings described below are merely some embodiments of this invention. Those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0033] Figure 1 This is a schematic diagram of the automatic powder filling system for high-temperature alloy powder sheathing of the present invention. Figure 2 This is a schematic diagram of the structure of Embodiment 1 of the automatic powder filling system for high-temperature alloy powder packaging of the present invention; Figure 3 shows the judgment image of powder recognition in embodiment 2 of the present invention. In Figure 3(a), "under" means "empty" and in Figure 3(b), "fill" means "full".

[0034] Figure label: 1. Powder vibration system; 2. Powder flow control system; 3. Vision recognition system; 4. Integrated control system; 5. Packaging and filling mechanism; 6. Hydraulic vibration system; 7. Degassing vibration assembly; 8. Flow meter; 9. Pneumatic vibration assembly; 10. High-definition image acquisition assembly; 11. High-temperature alloy packaging for powder. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] The features of various aspects of the embodiments of the present invention will now be described in detail. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can also be practiced without these specific details. The following description of the embodiments is merely intended to provide a better understanding of the invention by illustrating examples. The invention is not limited to any specific setups and methods provided below, but covers all improvements, substitutions, etc., to product structures and methods without departing from the spirit of the invention.

[0037] In the various accompanying drawings and the following description, well-known structures and techniques are not shown in order to avoid unnecessarily obscuring the invention.

[0038] As a key component of high-performance aero-engines, powder metallurgy superalloy (PMA) parts endure the combined effects of high temperature and high stress, operating under extremely harsh conditions and requiring complex manufacturing processes. The production of PMA parts typically involves a series of steps, including powder preparation, powder degassing and loading, pre-sintering, and hot isostatic pressing (HIP). To ensure successful HIP, the powder needs to be vacuum-sealed in a cylindrical sleeve with a powder nozzle. Uniform powder distribution, appropriate and precise weighting are prerequisites for obtaining a fully dense, defect-free HIP blank. Uneven powder distribution can lead to fine powder agglomeration, directly affecting the uniformity of the microstructure after isothermal forging and solution aging, and may even result in coarse recrystallized grains, leading to part scrap. Insufficient weighting results in residual porosity, severely impairing the dynamic mechanical properties of the material, especially fatigue and impact resistance, while also affecting the dimensional control accuracy of the subsequent blank. Therefore, stable control of the powder loading process directly determines the microstructure and properties of the HIP blank, thus affecting engine performance, and is therefore extremely important.

[0039] The existing packaging and filling process relies on manual tapping and observation to control the compaction of a mechanical vibrating table. This method is inefficient, taking 1 to 6 days to complete the powder filling of a single high-temperature alloy powder packaging, requiring continuous operation by one person and making a multi-machine production model impossible, thus consuming significant manpower. Furthermore, regarding quality control, human operation is highly susceptible to human factors. Different people have different visual judgments and control the compaction parameters differently, leading to large fluctuations in filling weight and even frequent excessively long vibration times, resulting in fine powder agglomeration. The mechanical vibrating table can only control the vibration frequency, with limited control parameters, further increasing the difficulty of controlling the packaging weight throughout the process. This results in a current packaging filling weight variation coefficient (standard deviation / mean × 100%) of approximately 1.5%. In terms of cost control, the variability in judgment based on manual visual observation leads to powder overflow and loss of approximately 3% due to failure to promptly determine the full filling status or incorrect adjustment of compaction parameters. Therefore, the present invention provides an automatic powder filling method for powder superalloy cladding based on visual recognition, including a powder recognition stage and a filling control stage; The specific steps for the powder identification stage are as follows: Step 1-1: Image preprocessing; Visual recognition requires image preprocessing to improve recognition accuracy for morphological feature extraction, including "Gaussian denoising" (removing ambient light interference) and "grayscale conversion" (converting color images to black and white to reduce computation).

[0040] Steps 1-2: Defining Filled and Unfilled Regions; First, the Canny edge detection algorithm is used to extract the features of the circular powder inlet, focusing the image recognition area within the powder inlet region. Second, a combination of Otsu adaptive global thresholding and local sliding thresholding is used to perform threshold segmentation on the preprocessed image, thus distinguishing between filled and unfilled regions. Specifically, the adaptive global threshold is used to determine the approximate range, and then fine-tuning is performed within the local range to eliminate the influence of lighting. After threshold segmentation, the system clearly divides the two types of regions using "grayscale value ranges." Unfilled regions include the container inner wall, background environment, and areas directly illuminated by light; their grayscale values ​​are lower than the Otsu adaptive global threshold, and there is no material obstruction. Filled regions refer only to the obstructed areas formed by powder material under a fixed light source; their grayscale values ​​are higher than the Otsu adaptive threshold. This region is the sole object for subsequent feature extraction, completely eliminating background interference.

[0041] Steps 1-3: Fill Area Calculation; The system automatically counts the total number of pixels in the grayscale image with a grayscale value ≥ Otsu threshold (denoted as N). This number directly corresponds to the pixel coverage area of ​​the shadow region; the final fill area S = total number of pixels N × conversion factor, used to determine whether the filling status is "full" or "empty". If the filling status is "full", the area is denoted as S. 满 The area when the filling state is "empty" is denoted as S. 空 This step employs a precise method of "pixel-level statistics + actual size mapping" rather than the traditional rough estimation, achieving the conversion from "pixel-level precision" to "actual physical size," thus providing a quantitative basis for subsequent material injection quantity determination.

[0042] Steps 1-4: Multi-feature collaborative judgment; Based on the calculated filling area, filling area change rate, shadow outline regularity, and gray-scale distribution morphological features, the system completes the judgment of powder replenishment rate and stacking uniformity.

[0043] The rate of change of filling area (ΔS / Δt), where ΔS = current filling area - previous filling area, and Δt is the time interval, directly reflects the powder replenishment rate. Based on the different powder replenishment rates, the magnitude, frequency, and displacement of the hydraulic vibration system are adjusted accordingly. If ΔS / Δt is greater than a certain threshold, it indicates a faster powder replenishment rate, requiring compaction according to the initial parameters; if ΔS / Δt is less than a certain threshold, it indicates a slower powder replenishment rate, requiring parameter adjustment according to rules. This ensures the powder replenishment efficiency and uniformity of distribution during the compaction process, reducing powder waste.

[0044] The regularity of the filling contour is determined by using the Canny edge detection algorithm to extract the complete boundary of the filling area, and the contour rectangularity is calculated as: actual area of ​​the filling area / area of ​​the minimum bounding rectangle of the filling. When the rectangularity is ≥0.9, the filling contour is close to a regular rectangle and is judged as "uniform stacking"; when the rectangularity is <0.9, the contour shows obvious depressions (such as local material shortages) or bulges (such as local excessive stacking) and is judged as "uneven material filling".

[0045] The grayscale distribution of the filling area involves generating a grayscale histogram of the filling region and calculating the "grayscale mean" and "grayscale variance". Grayscale values ​​are statistically analyzed in the filling area to generate a grayscale histogram (grayscale value range 0-255). The grayscale mean and variance of the filling area are calculated (under normal stacking conditions, the grayscale mean is stable between 50-80, and the variance is <20). If the grayscale mean suddenly increases (local grayscale mean >120), it indicates that the material in that area is sparse, and there may be local gaps. If the grayscale variance is too large (grayscale variance >50), it may indicate material bridging. When the local grayscale mean >120 or the grayscale variance >50, the system will adjust the vibration frequency to optimize the material distribution. By using the regularity of the filling contour and the grayscale distribution to determine the appropriate thresholds for raw materials with different particle sizes and flowability, such as the rate of change of area and the rectangularity of the contour, the system avoids misjudgments caused by the traditional fixed area threshold. At the same time, it replaces the subjective judgment of the human eye with the quantitative feature judgment of visual recognition, ensuring the uniform distribution of powder during the filling process and reducing the problem of unqualified structure caused by fine agglomeration.

[0046] The specific operation steps for the controlled loading stage are as follows: Step 2-1: Continuous loading (powder loading amount between 0% and 80%): A certain amount of high-temperature alloy powder is placed into the powder vibration system through the powder tank. After the vacuum process requirements are met, the powder vibration system is automatically turned on. The high-temperature alloy powder rises in the degassing vibration component and undergoes surface degassing treatment. After rising to the top, it falls along the pipeline. When the flow meter in the powder flow control system detects that the powder has reached a certain flow rate, its air vibration component is automatically turned on to ensure that the powder passes through the filter device and enters the cladding evenly. Step 2-2: Begin powder replenishment (powder filling level between 80% and 90%). The visual recognition system identifies the powder filling status by collecting the area morphological features of the powder shadow. When the powder filling status is first detected as full, the hydraulic vibration system is activated according to the initial settings, and the powder vibration component, flow meter, and pneumatic vibration component are sequentially deactivated. The hydraulic vibration system vibrates for 30-40 seconds before continuing to enter the continuous filling stage. During this stage, the powder filling status will be full 1-3 times, and the hydraulic vibration system parameters do not need to be changed. Steps 2-3: Continuously replenish powder (powder loading amount is 90%-100%). During this period, the powder loading state will frequently switch between full and empty. If the powder loading state is full for more than 10-15 seconds, the vibration parameters of the hydraulic vibration system need to be adjusted, and the frequency and magnitude should be gradually increased until the powder loading state remains full for more than 20-30 seconds, at which point the compaction operation ends.

[0047] The dividing point for the powder loading amount is distinguished based on the loading weight of the powder. The calculation formula in step 2-1 is (loose packing density × loose packing volume) / (taped density × tapped volume). Step 2-2 adds 10% to the weight in step 2-1 based on experience. The distinction of the dividing point classifies the difficulty of overall process control, which helps to reduce the complexity of the overall control process, ensures the stability of control at each stage, and reduces the occurrence of bug-type problems.

[0048] It should be noted that the difficulty in controlling the continuous powder replenishment stage lies in the fact that the interval between empty and full is about 10 seconds. It is necessary to control the start and stop of the powder vibration component and the air vibration component in a timely manner according to the visual recognition status. This requires the visual recognition system to have an accuracy of more than 95%, otherwise it will cause huge powder loss and waste.

[0049] Specifically, in steps 1-2, the Otsu adaptive threshold for the filled region is 120-255, and the Otsu adaptive threshold for the unfilled region is 0-80.

[0050] Specifically, in steps 1-2, the threshold for a "empty" state is 10-15%S. 满 .

[0051] Specifically, in steps 1-3, the conversion factor of "pixel-actual area" needs to be obtained through the initial camera calibration (based on a standard-size calibration board).

[0052] Specifically, the visual recognition system requires training with 30 sets of full-process video samples and comparison with manual annotation to ensure that the error in morphological feature judgment is controlled within 3% and that the visual recognition accuracy is stable at ≥95%.

[0053] Specifically, visual recognition uses pre-trained convolutional neural networks (such as ResNet and MobileNet) for transfer learning to train the accuracy of morphological feature recognition.

[0054] Specifically, in step 2-1, if the flow rate is greater than 0.5 kg / s, the air vibration component will be activated.

[0055] Specifically, in step 2-2, the opening amplitude of the pneumatic valve of the air vibration assembly is controlled between 20% and 40%.

[0056] It should be noted that some high-temperature alloy powders have poor flowability and are prone to clogging. Therefore, it is necessary to turn on the air vibration component to assist in powder transmission. However, the air vibration component acts directly on the bellows that transmit the powder. Therefore, if the opening range is too large, it will greatly reduce the life of the bellows and cause air leakage, thereby affecting the powder degassing effect.

[0057] Specifically, in step 2-2, the initial vibration parameters of the hydraulic vibration system are displacement of 1.6-2.2 mm, frequency of 18Hz-20Hz, and magnitude of 50-70%. Preferably, in step 2-2, the initial vibration parameters of the hydraulic vibration system are displacement of 1.7-2.1 mm, frequency of 18Hz-19Hz, and magnitude of 50-55%. It should be noted that the threshold time for determining full material is selected based on the length of the transport channel of the powder flow control system. This is to prevent powder from continuing to fall into the pipeline after the vibration lifting component and the air vibration component are turned off when the material is full, which would cause powder overflow and powder loss.

[0058] Specifically, in steps 2-3, when the powder filling state exceeds 10-15 seconds, the frequency of the hydraulic vibration system increases by 1-3 Hz each time, and the magnitude increases by 10-20% each time, with a maximum frequency not exceeding 25 Hz and a maximum magnitude of 100%. The magnitude of each increase in the frequency and magnitude of the hydraulic vibration system is determined based on the rate of change of the filling area. The specific threshold corresponding to steps 1-4 is that when 6%S 满 / s≤ΔS / Δt≤9%S 满 / s, only the magnitude of the hydraulic vibration system is adjusted, with each increase in amplitude being 10%, and the maximum magnitude being 100%; when 3%S 满 / s≤ΔS / Δt<6%S 满 / s, the frequency adjustment increment is 1Hz, and the displacement decreases by 0.2mm with each increment, with a minimum displacement of 1.6mm; when the change in ΔS / Δt is <3%S 满 / s, initial magnitude 70%, frequency adjustment increment of 2Hz, displacement decreases by 0.4mm with each increase in magnitude, magnitude increases by 20% with each increase, maximum magnitude 90%, minimum displacement 1.6mm. The formula for calculating the threshold value of the filling area change rate is (S 满 -S 空 The formula is: cosα / powder loading time threshold (see step 2-3), where α is the angle between the camera and the plane. The quantification of relevant parameters is based on experimental data on reducing fine powder aggregation, which enables quantitative control of the stability of the loading process and reduces the problem of poor particle aggregation caused by human influence.

[0059] It should be noted that the selection of the frequency and magnitude of the hydraulic vibration system directly affects the compaction effect. If the amplitude is too large when the filling volume is less than 98%, it will cause powder overflow and fine powder agglomeration, resulting in powder loss and uneven structure.

[0060] Specifically, in steps 2-3 and 1-4, when the contour rectangularity is <0.9, the local grayscale mean is >120, or the grayscale variance is >50, the frequency is increased by 1-3Hz for vibration optimization to avoid uneven distribution.

[0061] Specifically, in step S3, the vibration waveform of the hydraulic vibration system is controlled by a sine wave, and the difference between the downward acceleration and the upward acceleration is between 0.5g and 1g (g is the acceleration due to gravity).

[0062] It should be noted that if the acceleration difference is too small, it will easily lead to increased powder flow, which will affect the compaction effect on the one hand, and cause fine powder to agglomerate on the other hand, thus affecting the microstructure after isothermal forging and solution aging.

[0063] To achieve the method described in this invention, this invention provides an automatic powder filling system for high-temperature alloy powder cladding based on visual recognition, comprising: The powder vibration system uses a high-vacuum environment and a degassing vibration component to perform surface degassing treatment on high-temperature alloy powder. The powder flow control system is used to transfer high-temperature alloy powder that meets the degassing process to the powder high-temperature alloy cladding after passing through a flow meter and an air vibration component. The visual recognition system uses a high-definition image acquisition component to capture the powder stacking status at the upper inlet of the powder high-temperature alloy cladding, and simultaneously transmits the signal to the integrated control system. The integrated control system is used to receive video recognition signals from the vision recognition system, judge the powder stacking status, control the powder vibration system and powder flow control system to carry out powder transfer and filling, and control the hydraulic vibration system to carry out powder compaction. The hydraulic vibration system is used to control the frequency, displacement, and magnitude of the sleeve vibration to achieve the sleeve compaction operation.

[0064] Specifically, the high-definition image acquisition component of the visual recognition structure adopts a dual-camera acquisition method, with the acquisition angles forming angles of 40°-60° and 0°-10° with the vertical powder delivery pipe, respectively. The dual cameras perform feature fusion through weighted averaging, with the top area weighted at 0.6-0.7 and the side area weighted at 0.3-0.4, to comprehensively determine the material injection status.

[0065] Specifically, the high-definition image acquisition component of the visual recognition structure needs to have a pixel count of 5 million or more and a field of view greater than 80mm × 80mm.

[0066] Preferably, the high-definition image acquisition component of the visual recognition structure has a pixel count of 8 million pixels or more.

[0067] Specifically, the integrated control system includes a visual analysis module adapted to the visual recognition system, which judges real-time parameters based on morphological features and actual production requirements. Its image processing speed is ≤100ms / frame, and the morphological feature judgment error is <3%.

[0068] To more clearly describe the present invention, the following embodiments and comparative examples are provided for further illustration.

[0069] Example 1 According to one aspect of the present invention, this embodiment provides an automatic powder filling system for powder superalloy cladding based on vision recognition, such as... Figure 2 As shown, it includes: The powder vibration system 1 performs surface degassing treatment on high-temperature alloy powder through a high-vacuum environment and a degassing vibration component 7. The powder flow control system 2 is used to transfer the high-temperature alloy powder that meets the degassing process to the powder high-temperature alloy cladding 11 after passing through the flow meter 8 and the air vibration component 9. The visual recognition system 3 acquires the powder stacking state at the upper inlet of the powder high-temperature alloy cladding 11 through the high-definition image acquisition component 10, and transmits the signal to the integrated control system 4 at the same time. The integrated control system 4 is used to receive video recognition signals from the vision recognition system 3, judge the powder stacking status, control the switching status of the degassing vibration component 7, flow meter 8, and air vibration component 9 to carry out powder transfer and filling, and control the hydraulic vibration system 6 to carry out powder compaction operations. Packaging and filling mechanism 5 is used to provide a high vacuum environment to ensure the filling process of high-temperature alloy powder; The hydraulic vibration system 6 is used to control the frequency, displacement, and magnitude of the sleeve vibration to achieve the sleeve compaction operation.

[0070] Specifically, the high-definition image acquisition component of the visual recognition system adopts a dual-camera acquisition method, with the acquisition angles forming angles of 45° and 3° with the vertical powder delivery pipe, respectively. The dual cameras perform feature fusion through weighted averaging, with the top area weighted at 0.6 and the side area weighted at 0.4, to comprehensively determine the material injection status.

[0071] Specifically, the high-definition image acquisition component of the visual recognition system has 5 million pixels and a field of view of 80mm×80mm.

[0072] Specifically, the integrated control system includes a visual analysis module adapted to the visual recognition system. By collecting morphological features such as the contour regularity, area change rate, and grayscale distribution of material shadows, a "volume-shadow" mapping model is established. The system judges the powder filling status in real time based on morphological features and real-time parameters required for actual production. Its image processing speed is ≤100ms / frame, and the morphological feature judgment error is <3%.

[0073] Example 2 According to one aspect of the present invention, this embodiment provides an automatic powder filling method for powder superalloy cladding based on visual recognition, and the automatic filling system described in Embodiment 1 is used to complete the cladding filling. The specific steps are as follows: Step 1: Sample collection. Collect 30 sets of full-process video samples, including key frames of each video recording "material feeding → filtering → vibration → filling", such as material shortage, full material, and material blockage, focusing on vibration frequency, lighting, and material filling speed, covering more than 90% of the working conditions.

[0074] Step 2: Perform powder identification, capturing frames at 0.5 seconds per frame (2400 frames per group), labeling each frame with "shading parameters + measured injection amount (0.1g accuracy) + bulk density". The specific identification steps are as follows: Step 2-1: Image preprocessing; Visual recognition requires image preprocessing to improve recognition accuracy for morphological feature extraction, including "Gaussian denoising" (3×3 filter kernel + pixel gradient weighting) and "grayscale conversion" (converted according to the formula R=0.299, G=0.587, B=0.114, reducing the data volume by 2 / 3).

[0075] Step 2-2: Defining Filled and Unfilled Regions; First, the Canny edge detection algorithm is used to extract the features of the circular powder inlet, focusing the image recognition area on the powder inlet region. Then, a combination of Otsu adaptive global thresholding and local sliding thresholding (Otsu global threshold (120 during the day / 80 at night) + 5×5 local threshold correction, outputting a binary image) is used to perform threshold segmentation on the preprocessed image, thus distinguishing between filled and unfilled regions. That is, the adaptive global threshold is used to determine the approximate range, and then fine-tuning is performed within the local range to eliminate the influence of lighting. After threshold segmentation, the system clearly divides the two types of regions by "grayscale value range". Unfilled regions include the container inner wall, background environment, and areas directly illuminated by light; their grayscale values ​​are lower than the Otsu adaptive global threshold, and there is no material obstruction. Filled regions refer only to the obstructed areas formed by powder material under a fixed light source; their grayscale values ​​are higher than the Otsu adaptive threshold. This region is the sole object for subsequent feature extraction, completely eliminating background interference.

[0076] Steps 2-3: Fill Area Calculation; The system automatically counts the total number of pixels in the grayscale image with grayscale values ​​between 120 and 255 (denoted as N). This number directly corresponds to the pixel coverage range of the shadow area. The final fill area S = total number of pixels N × conversion factor (1 pixel corresponds to 0.01 mm²), realizing the conversion from pixel to actual physical size, used to determine whether the filling status is "full" or "empty". The threshold for the "empty" status is 13%S. 满 .

[0077] Steps 2-4: Multi-feature collaborative judgment; Based on the calculated filling area, filling area change rate, shadow outline regularity, and gray-scale distribution morphological features, the system completes the judgment of powder replenishment rate and stacking uniformity.

[0078] The rate of change of filling area (ΔS / Δt), where ΔS = current filling area - previous filling area, and Δt is the time interval, directly reflects the powder replenishment rate. Based on different powder replenishment rates, the magnitude, frequency, and displacement of the hydraulic vibration system are adjusted accordingly. When 50mm... 2 / s≤ΔS / Δt≤100mm 2 / s, only the magnitude is adjusted, increasing by 10% each time, with a maximum magnitude of 100%; when ΔS / Δt < 50mm 2 / s, the frequency adjustment increment is 1Hz, and the displacement decreases by 0.2mm with each increment, with a minimum displacement of 1.6mm; when the change in ΔS / Δt is <33 mm 2 / s, initial magnitude 70%, frequency adjustment range increases by 2Hz each time, displacement decreases by 0.4mm, magnitude increases by 10% each time, magnitude maximum 90%, minimum displacement is 1.6mm.

[0079] When the material inside the package is evenly piled up (rectangularity ≥ 0.9), the shadow outline is close to a rectangle; if the outline shows obvious depression (rectangularity < 0.9), it is judged as "uneven material filling". At this time, the system will combine the vibration table control logic to increase the vibration frequency by 2Hz to optimize the material distribution and avoid "false full material" (full surface material but empty inside) caused by uneven pile-up.

[0080] If the grayscale mean suddenly increases (local grayscale mean > 120), it indicates that the material in that area is sparse and there may be local gaps; if the grayscale variance is too large (grayscale variance > 50), it may be that the material is bridging. In this case, the system will increase the vibration frequency by 2Hz to optimize the material distribution.

[0081] Step 3, MobileNet transfer learning The model was initialized using a MobileNetV2 pre-trained model (3.5M parameters) on ImageNet, with the first 10 layers frozen and the last 5 layers fine-tuned (outputting shadow parameters and injection status). The training / validation / test sets were divided into 7:2:1 ratios based on 81,000 frames. The training set was enhanced with brightness jitter and density compensation logic. Finally, the model was validated using the Adam optimizer (1e-4 initial learning rate) for 20 epochs (total time 5 hours). The validation set was stopped when the error was less than 3%. Based on the above powder recognition algorithm, the following steps are performed to fill the package with a diameter of 600mm and a height of 1200mm: Step 1: Continuous loading (powder loading amount between 0% and 80%): A certain amount of high-temperature alloy powder is put into the powder vibration system through the powder tank. The main control switch is turned on. After the vacuum process requirements are met, the powder vibration system is automatically turned on. The high-temperature alloy powder rises in the degassing vibration component and undergoes surface degassing treatment. After rising to the top, it falls along the pipeline. When the flow meter in the powder flow control system detects a flow rate greater than 0.5 kg / s, the air vibration component is automatically turned on. The opening amplitude of the pneumatic valve of the air vibration component is controlled at 30% to ensure that the powder passes through the filter device and enters the cladding evenly. Step 2: Begin replenishing powder (powder filling level between 80%-90%); the visual recognition system identifies the powder filling status by collecting the area morphological features of the powder shadow. When the first instance of powder filling is detected, it is considered full (the threshold for full material shadow is 10%S). 空 S 空 =339mm 2 When the powder filling reaches full capacity, the hydraulic vibration system is activated according to the initial settings (displacement of 1.7mm, frequency of 19Hz, and magnitude of 50%), and the powder vibration component, flow meter, and pneumatic vibration component are shut down sequentially. After the hydraulic vibration system vibrates for 30 seconds, it continues to enter the continuous filling stage. During this stage, there will be two instances where the powder filling state is full, and the parameters of the hydraulic vibration system do not need to be changed. Step 3: Continuous powder replenishment (powder loading 90%-100%); During this stage, the powder loading status will frequently switch between full and empty. To cope with the complexity of recognition at this stage, the visual recognition system needs to incorporate area transformation rate, contour regularity, and shadow grayscale morphological features in addition to the filled area for auxiliary judgment. If the powder loading status remains full for more than 13 seconds, the vibration parameters of the hydraulic vibration system need to be adjusted, with the frequency and magnitude gradually increased. The compaction operation ends when the powder loading status remains full for more than 20 seconds.

[0082] Comparative Example 1 Comparative Example 1 uses a package with a diameter of 600mm and a height of 1200mm for powder transfer and filling, employing manual observation and assisted tapping, followed by a mechanical vibration table for compaction. The process includes the following steps: S1: Continuous loading (powder loading amount between 0% and 80%): A certain amount of high-temperature alloy powder is put into the powder vibration system through the powder tank. The main control switch is turned on. After the vacuum process requirements are met, the powder vibration system is manually turned on. The high-temperature alloy powder rises in the degassing vibration component and undergoes surface degassing treatment. After rising to the top, it falls along the pipeline. At the same time, the air vibration is turned on. The air vibration is based on the manual operation of the pneumatic valve switch to help ensure that the powder passes through the filter device and enters the package evenly. S2: Start replenishing powder (powder filling amount is between 80% and 90%); through manual observation, when the powder is found to be full for the first time, turn on the mechanical vibration table to compact it, with a vibration frequency of 17Hz. During this stage, there will be two instances where the powder filling state is full, and the vibration frequency does not need to be changed. S3: Continuous powder replenishment (powder filling amount is 90%-100%); if the compaction effect is slow through manual observation, adjust the frequency converter to increase the vibration frequency to 20Hz, increasing it by 1Hz each time, until the powder cannot be filled.

[0083] Performance testing The time, weight, and tissue uniformity of Example 2 and Comparative Example 1 were tested, and the tissue uniformity was detected by ultrasonic flaw detection.

[0084] Table 1 Performance Test Results

[0085] The above data are statistical data. Each set of examples or comparative examples produced 20 powder high-temperature alloy parts.

[0086] As can be seen from Examples 1-2 and Comparative Example 1 of the present invention and Table 1, the automatic filling method and system provided by the present invention improves the stability of powder distribution control during filling, reduces the problem of fine powder agglomeration, controls the powder filling weight variation coefficient at 0.6%, eliminates fine powder agglomeration, and controls powder loss to within 1%. At the same time, it reduces manual labor time, reduces manual intervention by 75%, and improves filling efficiency, meeting the requirements.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for automatically filling powder into high-temperature alloy claddings based on visual recognition, characterized in that, This includes the powder identification stage and the controlled filling stage; The powder identification stage involves the following steps: Step 1-1: Preprocess and optimize the image to improve the accuracy of visual recognition, including "Gaussian denoising" and "grayscale conversion"; Step 1-2: Defining filled and unfilled regions; First, the features of the circular packaging powder inlet are extracted by using an edge detection algorithm, and the image recognition area is focused on the powder inlet area. Second, the preprocessed image is thresholded by combining Otsu adaptive global threshold and local sliding threshold, thereby distinguishing between filled and unfilled regions. Steps 1-3: Fill area calculation; automatically count the total number N of pixels with "grayscale value ≥ threshold" in the grayscale image, and calculate the fill area S = total number of pixels N × conversion factor. The area when the filling status is "full" is recorded as S. 满 The area when the filling state is "empty" is denoted as S. 空 The threshold for an "empty" filling status is 10-15%S. 满 ; Steps 1-4: Propose judgment rules for powder replenishment rate and stacking uniformity: calculate the rate of change of filling area using the filling area, and then adjust the frequency, displacement, and magnitude of the casing vibration; calculate the outline rectangularity to determine the outline regularity; calculate the gray scale change to adjust the vibration frequency to optimize material distribution; In the controlled filling stage, 80% and 90% of the powder filling weight are used as the dividing points for different filling and replenishment operations; the operation steps are as follows: Step 2-1: The powder loading amount is (0%-80%), and the loading is continuous. The high-temperature alloy powder is put into the powder vibration system through the powder tank. After the vacuum process requirements are met, the powder vibration system is automatically turned on. The high-temperature alloy powder rises in the degassing vibration component and undergoes surface degassing treatment. After rising to the top, it falls along the pipeline. When the flow meter in the powder flow control system detects that the powder flow rate is greater than 0.5Kg / s, its air vibration component is automatically turned on to ensure that the powder passes through the filter device and enters the cladding evenly. Step 2-2: When the powder filling level is between 80% and 90%, start replenishing powder. The visual recognition system identifies the powder filling status by collecting the area morphological features of the powder filling. When the powder filling status is detected as full for the first time, the hydraulic vibration system is turned on according to the initial settings, and the powder vibration component, flow meter, and air vibration component are turned off in sequence. The hydraulic vibration system vibrates for 30-40 seconds and then continues to enter the continuous filling stage. During this stage, the powder filling status will be full 1-3 times, and the parameters of the hydraulic vibration system do not need to be changed. Steps 2-3: When the powder loading is between 90% and 100%, continuously replenish the powder. During this period, the powder loading status will frequently switch between full and empty. When the powder loading status is full for more than 10-15 seconds, the vibration parameters of the hydraulic vibration system need to be adjusted, and the frequency and magnitude are gradually increased until the powder loading status remains full for more than 20-30 seconds, at which point the compaction operation ends.

2. The method according to claim 1, characterized in that, In steps 1-4, The rate of change of filling area ΔS / Δt, where ΔS = current filling area - previous filling area, Δt is the time interval, and ΔS / Δt directly reflects the powder replenishment rate; according to the different powder replenishment rates, the magnitude, frequency and displacement of the hydraulic vibration system are set in conjunction. The contour regularity is determined by using an edge detection algorithm to extract the complete boundary of the filling area and calculating the contour rectangularity as the ratio of the actual area of ​​the filling area to the area of ​​the minimum bounding rectangle. When the rectangularity is ≥0.9, the filling contour is close to a regular rectangle and is judged as "uniformly stacked". When the rectangularity is <0.9, it is judged as "uneven filling". The grayscale values ​​of the filled area are statistically analyzed to generate a grayscale histogram. The mean and variance of the grayscale values ​​of the filled area are calculated. When the local mean grayscale value is greater than 120 or the variance of the grayscale value is greater than 50, the system will adjust the vibration frequency to optimize the material distribution.

3. The method according to claim 1, characterized in that, In steps 1-2, the Otsu adaptive global threshold for filled regions is 120-255, and the Otsu adaptive global threshold for unfilled regions is 0-80.

4. The method according to claim 1, characterized in that, In steps 1-3, the conversion factor of "pixel-actual area" needs to be obtained through the initial camera calibration based on the standard size calibration board.

5. The method according to claim 1, characterized in that, The recognition of a visual recognition system requires sample training and error calibration to ensure that the error in morphological feature judgment is controlled within 3% and that the visual recognition accuracy is stable at ≥95%.

6. The method according to claim 1, characterized in that, In step 2-1, the opening amplitude of the pneumatic valve of the air vibration assembly is controlled between 20% and 40%.

7. The method according to claim 1, characterized in that, In step 2-2, the initial vibration parameters of the hydraulic vibration system are displacement of 1.6-2.2 mm, frequency of 18 Hz-20 Hz, and magnitude of 50-70%.

8. The method according to claim 1, characterized in that, In steps 2-3, when the powder filling state exceeds 10-15 seconds, the frequency of the hydraulic vibration system increases by 1-3 Hz each time, and the magnitude increases by 10-20% each time, with a maximum frequency not exceeding 25 Hz and a maximum magnitude of 100%. The magnitude of each increase in the frequency and magnitude of the hydraulic vibration system is determined based on the rate of change of the filling area. The specific thresholds corresponding to steps 1-4 are as follows: When 6%S 满 / s≤ΔS / Δt≤9%S 满 / s, only the magnitude of the hydraulic vibration system is adjusted, with each increase in amplitude being 10%, and the maximum magnitude being 100%; When 3%S 满 / s≤ΔS / Δt<6%S 满 / s, the frequency adjustment increment is 1Hz each time, the displacement decreases by 0.2mm, and the minimum displacement is 1.6mm; When ΔS / Δt < 3%S 满 / s, initial magnitude 70%, frequency adjustment amplitude increases by 2Hz each time, displacement decreases by 0.4mm, magnitude increases by 20% each time, magnitude maximum 90%, minimum displacement is 1.6mm.

9. The method according to claim 2, characterized in that, When the outline rectangularity is less than 0.9, the local grayscale mean is greater than 120, or the grayscale variance is greater than 50, the frequency is increased by 1-3 Hz for vibration optimization to avoid uneven distribution.

10. The method according to claim 1, characterized in that, In steps 2-3, the vibration waveform of the hydraulic vibration system is controlled by a sine wave, and the difference between the downward acceleration and the upward acceleration is between 0.5g and 1g, where g is the acceleration due to gravity.

11. An automatic powder filling system for high-temperature alloy powder cladding based on vision recognition, used to implement the method as described in claim 1, characterized in that, include: The powder vibration system uses a high-vacuum environment and a degassing vibration component to perform surface degassing treatment on high-temperature alloy powder. The powder flow control system is used to transfer high-temperature alloy powder that meets the degassing process to the powder high-temperature alloy cladding after passing through a flow meter and an air vibration component. The visual recognition system uses a high-definition image acquisition component to capture the powder stacking status at the upper inlet of the powder high-temperature alloy cladding, and simultaneously transmits the signal to the integrated control system. The integrated control system is used to receive video recognition signals from the vision recognition system, judge the powder stacking status, control the powder vibration system and powder flow control system to carry out powder transfer and filling, and control the hydraulic vibration system to carry out powder compaction. The hydraulic vibration system is used to control the frequency, displacement, and magnitude of the sleeve vibration to achieve the sleeve compaction operation.

12. The automatic powder filling system for high-temperature alloy cladding according to claim 11, characterized in that, The high-definition image acquisition component of the visual recognition system adopts a dual-camera acquisition method. The acquisition angle is 40°-60° and 0°-10° with the vertical powder inlet pipe, respectively. The dual cameras perform feature fusion by weighted averaging, with the top area weighted at 0.6-0.7 and the side area weighted at 0.3-0.4, to comprehensively determine the material injection status.

13. The automatic powder filling system for high-temperature alloy cladding according to claim 11, characterized in that, The high-definition image acquisition component of the visual recognition system must have a pixel count of 5 million or more and a field of view greater than 80mm × 80mm.

14. The automatic powder filling system for high-temperature alloy cladding according to claim 11, characterized in that, The integrated control system includes a visual analysis module adapted to the visual recognition system. Based on morphological features and real-time parameter judgment results of actual production requirements, its image processing speed is ≤100ms / frame and the morphological feature judgment error is <3%.

Images