Intelligent dust removal system and method for cutting machine, and electronic device

Abstract

This invention relates to the field of intelligent dust removal solutions for cutting machines, specifically to an intelligent dust removal system, method, and electronic equipment for cutting machines. This invention acquires images of the interior of the cutting impurity containment box and uses a YOLO model for precise identification, accurately determining the actual condition of the cutting impurities. When impurities are detected, the dust removal device is activated promptly, avoiding the ineffective operation of traditional timed dust removal methods when impurities are scarce. This effectively improves dust removal efficiency, ensuring the cutting machine is always in a good working environment, reducing the impact of impurity accumulation on equipment performance, and thus improving the cutting machine's efficiency and cutting quality. Based on the judgment results of the YOLO model, this invention intelligently controls the operation of the dust removal device. This intelligent control method avoids unnecessary energy consumption and significantly reduces energy costs compared to traditional fixed-mode dust removal systems, aligning with the trend of energy conservation and environmental protection.

Description

Technical Field

[0001] This invention relates to the field of intelligent dust removal solutions for cutting machines, specifically to an intelligent dust removal system and method for cutting machines, and electronic equipment. Background Technology

[0002] In numerous industrial sectors, such as machinery manufacturing, construction engineering, and automobile manufacturing, cutting machines are widely used as essential processing equipment. When processing various materials (including metals, stone, wood, and plastics), cutting machines inevitably generate a large amount of cutting impurities. These impurities manifest in various forms, primarily including the following: Metal shavings: When processing metal using metal cutting equipment (such as plasma cutters, laser cutters, CNC lathes, etc.), the high temperature generated by high-speed cutting instantly melts the metal surface, which then cools and solidifies to form tiny metal particles. These small particles are metal shavings. Due to the high speed and high temperature of the cutting process, metal shavings often have a certain shape and hardness, and the particle size varies, some may be only a few micrometers, while others can reach several millimeters.

[0003] Dust clumps: When cutting or grinding brittle materials (such as stone and ceramics) or fine-particle powders (such as carbon powder and coal powder), the materials break and pulverize under the cutting force, forming a large number of dust particles. Because these dust particles have a certain degree of adsorption and agglomeration, they easily adsorb and aggregate to form clump-like mixtures, i.e., dust clumps.

[0004] Fiber filaments: During the processing of materials such as wood and plastics, when the material is subjected to cutting force, its internal fibers or fibrous structures may be broken or torn, forming fibrous debris, i.e., fiber filaments.

[0005] These impurities not only affect the normal operation of the cutting machine and reduce cutting quality, but may also harm the health of operators. Traditional dust removal methods are often performed on a timed basis, lacking real-time monitoring of the actual generation of cutting impurities, resulting in poor dust removal or over-dusting and wasted energy. Therefore, it is necessary to develop an intelligent dust removal system that can intelligently identify cutting impurities and provide real-time dust removal.

[0006] Therefore, existing technologies still need further development. Summary of the Invention

[0007] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide an intelligent dust removal system, dust removal method, and electronic equipment for cutting machines to solve the problems existing in the prior art.

[0008] To achieve the above-mentioned technical objectives, according to a first aspect of the present invention, the present invention provides an intelligent dust removal system for a cutting machine, comprising: A cutting impurity holding box is provided at the bottom of the cutting machine to hold the cutting impurities of the cutting machine; The acquisition module includes a camera, which is used to acquire internal images of the cut impurity containment box to form a dataset; The control module is used to train the YOLO model using the dataset. After the YOLO model is trained, it acquires the internal image of the cutting impurity containment box and inputs it into the trained YOLO model. The trained YOLO model is used to determine whether it can identify the anchor frame used to indicate the cutting impurities, and controls the operation status of the dust removal device based on the judgment result. The dust removal device includes a fan unit connected to the cutting impurity holding box, the fan unit being used to discharge cutting impurities inside the cutting impurity holding box to the outside of the cutting impurity holding box during operation.

[0009] Specifically, the blower unit includes a blower and an exhaust fan, and the cutting impurity holding box is provided with a blower port and an exhaust port. The blower port is connected to the blower through a pipe, and the exhaust port is connected to the exhaust fan through a pipe.

[0010] According to a second aspect of the present invention, a method for intelligent dust removal in a cutting machine is provided, comprising: S100: Obtain an internal image of the cutting impurity retention box to form a dataset; S200. Train the YOLO model using the dataset. After the YOLO model is trained, obtain the internal image of the cut impurity containment box and input it into the trained YOLO model. Use the trained YOLO model to determine whether the anchor box used to indicate the cut impurities can be identified. S300: Control the operating status of the dust removal device based on the judgment result.

[0011] Specifically, controlling the operating state of the cutting machine based on the judgment result includes: If the anchor frame used to indicate the cutting of impurities can be identified, the dust removal device can be controlled to operate. If the anchor frame used to indicate the cutting of impurities cannot be identified, the dust removal device will stop operating.

[0012] Specifically, controlling the operating status of the cutting machine based on the judgment result includes: If the anchor frame used to indicate the cutting of impurities can be identified, the operation of the blower and exhaust fan can be controlled. If the anchor frame used to indicate the cutting of impurities cannot be identified, control the blower and exhaust fan to stop operating.

[0013] Specifically, training the YOLO model using the dataset includes: S1. Create a dataset; S2, Labelme marks the predicted bounding box for cutting impurities; S3. Train the YOLO model; S4. Save the optimal YOLO model.

[0014] Specifically, step S1 includes: S11. Collect raw image data containing cutting impurities during the operation of the cutting machine. The raw image data comes from multiple cutting machine working scenarios with different perspectives, different lighting conditions, and different cutting conditions. S12. Filter the collected raw image data to remove blurry, poor image quality, and images that do not contain effective information about cutting impurities. S13. Perform preprocessing operations on the filtered image data before labeling. The preprocessing operations include adjusting the image size to a uniform size, converting the image to grayscale or color space, and normalizing the image pixel values ​​so that the image data meets the requirements of subsequent model training. S14. Divide the preprocessed image data into training set, validation set and test set according to a certain ratio to form a dataset for training the YOLO model.

[0015] Specifically, step S2 includes: S21. Start the Labelme annotation tool and import the images from the dataset created in step S1 into the Labelme software interface in sequence. S22. For each image, carefully observe the specific shape, location and distribution of the cutting impurities in the image, and determine different labeling categories based on the type of cutting impurities, including metal fragments, dust clumps and fiber filaments. S23. Using the drawing function in Labelme tool, accurately draw the prediction box along the edge of the cut impurity on the image to ensure that the prediction box completely covers the cut impurity area, while minimizing the wrong selection of non-impurity areas. For cut impurities with irregular shapes, the prediction box should fit its outer contour as closely as possible. S24. Assign a corresponding category label to each drawn prediction box, and accurately fill in the determined cut impurity category name in the Labelme annotation information column to ensure that the category label is consistent with the pre-set category system. S25. Conduct a preliminary check on the labeled images to see if the prediction boxes are drawn accurately and if the category labels are correct. For images that are questionable or inaccurately labeled, go back and observe and label them again to ensure the labeling quality. S26. After all images are labeled, export the data file with predicted bounding box annotations and category labels in the format specified by Labelme, so that it can be used for subsequent training of the YOLO model.

[0016] Specifically, step S3 includes: S31. Set the initial training parameters of the YOLO model, including the learning rate, momentum, weight decay coefficient, batch size, and maximum number of iterations. S32. Input the dataset with labeled information obtained after cutting the impurity prediction boxes by Labelme in step S2 into the YOLO model and start the training process according to the set initial training parameters. S33. During training, monitor the model's loss function value, accuracy, recall, and mAP evaluation metrics on the validation set in real time. S34. Based on the monitored changes in evaluation metrics, dynamically adjust the training parameters. For example, when the loss function value stops decreasing or the evaluation metrics fluctuate, reduce the learning rate. S35. Continue training and parameter adjustment until the model's evaluation metrics on the validation set reach the preset optimal threshold, or the number of training iterations reaches the maximum number of iterations.

[0017] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, which, when executed by the processor, implement the above-described intelligent dust removal method for a cutting machine.

[0018] Beneficial effects: This invention acquires images of the interior of the cutting impurity containment box and uses a YOLO model for precise identification, enabling accurate judgment of the actual situation of cutting impurities. When impurities are detected, the dust removal device is activated promptly, avoiding the ineffective operation of traditional timed dust removal methods when impurities are few. This effectively improves dust removal efficiency, ensuring the cutting machine is always in a good working environment, reducing the impact of impurity accumulation on equipment performance, and thus improving the cutting machine's efficiency and cutting quality. Based on the judgment results of the YOLO model, this invention intelligently controls the operation of the dust removal device; that is, it activates the blower and exhaust fan for dust removal when cutting impurity anchor frames are detected, and stops operation when impurities are not detected. This intelligent control method avoids unnecessary energy consumption, significantly reducing energy costs compared to traditional fixed-mode dust removal systems, and aligns with the trend of energy conservation and environmental protection. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the camera installation position provided in a specific embodiment of the present invention; Figure 2This is a schematic diagram of the intelligent dust removal system for cutting machines provided in a specific embodiment of the present invention; Figure 3 This is a flowchart illustrating the intelligent dust removal method for a cutting machine provided in a specific embodiment of the present invention; Figure 4 This is a schematic diagram of the system composition of the intelligent dust removal system for cutting machines provided in a specific embodiment of the present invention; The above figures contain the following reference numerals: 1. Cutting machine; 2. Cutting impurity collection box; 3. Inclined plate; 4. Air blowing port; 5. Air extraction port; 6. Branch air blowing pipe; 7. Main air blowing pipe; 8. Air blower; 9. Suction pipe; 10. Suction port; 11. Branch exhaust pipe; 12. Main exhaust pipe; 13. Exhaust fan; 14. Dust discharge pipe; 15. Dust box; 16. Fan bracket; 17. Camera; 100. Acquisition module; 200. Control module; 300. Dust removal device. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Based on the embodiments in this application, other similar embodiments obtained by those skilled in the art without creative effort should all fall within the scope of protection of this application. Furthermore, directional terms mentioned in the following embodiments, such as "up," "down," "left," and "right," are only for reference to the directions in the accompanying drawings; therefore, the directional terms used are for illustrative purposes and not for limiting the invention.

[0021] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0022] Please see Figures 1-4 This invention provides an intelligent dust removal system for cutting machines, comprising: Cutting impurity retention box 2, which is located at the bottom of the cutting machine, is used to retain cutting impurities from the cutting machine; The acquisition module 100 includes a camera 17, which is used to acquire internal images of the cut impurity containment box 2 to form a dataset. Specifically, the camera 17 is fixedly installed above the inside of the cutting impurity holding box 2, and is used to take pictures of the inside of the cutting impurity holding box 2.

[0023] The control module 200 is used to train a YOLO model using a dataset. After the YOLO model is trained, it acquires the internal image of the cutting impurity containment box 2 and inputs it into the trained YOLO model. The trained YOLO model is used to determine whether the anchor frame used to indicate the cutting impurities can be identified, and the operating status of the dust removal device 300 is controlled according to the judgment result. The dust removal device 300 includes a fan unit connected to the cutting impurity holding box 2. The fan unit is used to discharge the cutting impurities in the cutting impurity holding box 2 to the outside of the cutting impurity holding box 2 during operation.

[0024] Specifically, the blower unit includes a blower 8 and an exhaust fan 13. The cutting impurity holding box 2 is provided with a blower port 4 and an exhaust port 5. The blower port 4 is connected to the blower 8 through a pipe, and the exhaust port 5 is connected to the exhaust fan 13 through a pipe.

[0025] The device includes a cutting machine 1, which has several cutting impurity holding boxes 2 in the middle. The bottom of each cutting impurity holding box 2 is provided with an inclined plate 3. The two inner side plates of each cutting impurity holding box 2 are symmetrically provided with air blowing ports 4 and air extraction ports 5. Each air blowing port 4 is connected to a main air blowing pipe 7 through a branch air blowing pipe 6. One end of the main air blowing pipe 7 is provided with a blower 8. The blower 8 is connected to an air extraction port 10 through an air extraction pipe 9. The air extraction port 5 is connected to a main air extraction pipe 12 through a branch air extraction pipe 11. One end of the main air extraction pipe 12 is provided with an exhaust fan 13. The exhaust fan 13 is connected to a dust box 15 through a dust exhaust pipe 14. The exhaust fan 13 and the blower 8 are fixedly installed on the side surface of the cutting machine 1 through a fan bracket 16.

[0026] Preferably, the number of blowing ports 4 and exhaust ports 5 is the same, and there are six of each; the exhaust port 5 is semi-circular in shape, and its diameter is half the side length of the impurity collection box 2, which can improve the dust removal effect; the main blowing pipe 7 and the main exhaust pipe 12 are both made of random copolymer polypropylene pipe, which has the advantages of energy saving and material saving, environmental protection, lightweight and high strength, corrosion resistance, smooth inner wall without scaling, simple construction and maintenance, and long service life.

[0027] The main blower, located on the side of the main body of the cutting machine 1, draws air from the environment through the suction port and delivers it to the main blower pipe via the suction pipe. The main blower pipe then delivers the air to the branch blower pipe, which in turn delivers the ambient air to the cutting impurity collection box 2, thus accelerating dust removal efficiency. The inclined plate 3 at the bottom of the cutting impurity collection box 2 effectively prevents dust from being blown up, thus avoiding affecting the user's working environment and facilitating long-term use and operation. The exhaust port on the inner side of the cutting impurity collection box 2 further enhances dust removal, delivering the dust to the branch exhaust pipe 11 and the main exhaust pipe 12. The exhaust fan 13 then delivers the dust to the dust collection box 15. The main exhaust pipe 12 reduces the number of exhaust fans 13 required, thus reducing resource waste and lowering costs. The dust collection box 15 at one end of the dust discharge pipe 14 facilitates dust disposal and prevents dust pollution.

[0028] It is understood that this invention, by acquiring images of the interior of the cutting impurity containment box 2 and using the YOLO model for precise identification, can accurately determine the actual situation of the cutting impurities. When impurities are identified, the dust removal device 300 is activated promptly, avoiding the ineffective operation of traditional timed dust removal methods when impurities are few. This effectively improves dust removal efficiency, ensures the cutting machine is always in a good working environment, reduces the impact of impurity accumulation on equipment performance, and thus improves the cutting machine's working efficiency and cutting quality. Based on the judgment results of the YOLO model, this invention intelligently controls the operation of the dust removal device 300; that is, it activates the blower and exhaust fan for dust removal when the cutting impurity anchor frame is identified, and stops operation when it cannot be identified. This intelligent control method avoids unnecessary energy consumption, significantly reducing energy costs compared to traditional fixed-mode dust removal systems, and conforms to the development trend of energy conservation and environmental protection.

[0029] This invention provides another embodiment, which provides an intelligent dust removal method for a cutting machine, the intelligent dust removal method for a cutting machine comprising: S100: Obtain the internal image of the cutting impurity retention box 2 to form a dataset.

[0030] Specifically, S200. Train the YOLO model using the dataset. After the YOLO model is trained, obtain the internal image of the cut impurity containment box 2 and input it into the trained YOLO model. Use the trained YOLO model to determine whether the anchor box used to indicate the cut impurities can be identified.

[0031] S300: Control the operating status of the dust removal device 300 based on the judgment result.

[0032] Specifically, controlling the operating state of the cutting machine based on the judgment result includes: If the anchor frame used to indicate the cutting of impurities can be identified, control the dust removal device 300 to operate; If the anchor frame used to indicate the cutting of impurities cannot be identified, the dust removal device 300 will stop operating.

[0033] Specifically, controlling the operating status of the cutting machine based on the judgment result includes: If the anchor frame used to indicate the cutting of impurities can be identified, control the operation of the blower 8 and the exhaust fan 13; If the anchor frame used to indicate the cutting of impurities cannot be identified, control the blower 8 and the exhaust fan 13 to stop operating.

[0034] Specifically, training the YOLO model using the dataset includes: S1. Create a dataset.

[0035] Specifically, step S1 includes: S11. Collect raw image data containing cutting impurities during the operation of the cutting machine. The raw image data comes from multiple cutting machine working scenarios with different perspectives, different lighting conditions, and different cutting conditions.

[0036] S12. Filter the collected raw image data to remove blurry, poor-quality images and images that do not contain effective information about cutting impurities.

[0037] S13. Perform preprocessing operations on the filtered image data before labeling. The preprocessing operations include adjusting the image size to a uniform size, converting the image to grayscale or color space, and normalizing the image pixel values ​​so that the image data meets the requirements of subsequent model training. S14. Divide the preprocessed image data into training set, validation set and test set according to a certain ratio to form a dataset for training the YOLO model.

[0038] S2, Labelme marks the predicted bounding box for cutting impurities.

[0039] Specifically, step S2 includes: S21. Start the Labelme annotation tool and import the images from the dataset created in step S1 into the Labelme software interface in sequence. S22. For each image, carefully observe the specific shape, location and distribution of the cutting impurities in the image, and determine different labeling categories based on the type of cutting impurities, including metal fragments, dust clumps and fiber filaments. S23. Using the drawing function in Labelme tool, accurately draw the prediction box along the edge of the cut impurity on the image to ensure that the prediction box completely covers the cut impurity area, while minimizing the wrong selection of non-impurity areas. For cut impurities with irregular shapes, the prediction box should fit its outer contour as closely as possible. S24. Assign a corresponding category label to each drawn prediction box, and accurately fill in the determined cut impurity category name in the Labelme annotation information column to ensure that the category label is consistent with the pre-set category system. S25. Conduct a preliminary check on the labeled images to see if the prediction boxes are drawn accurately and if the category labels are correct. For images that are questionable or inaccurately labeled, go back and observe and label them again to ensure the labeling quality. S26. After all images are labeled, export the data file with predicted bounding box annotations and category labels in the format specified by Labelme, so that it can be used for subsequent training of the YOLO model.

[0040] S3. Train the YOLO model.

[0041] Specifically, step S3 includes: S31. Set the initial training parameters of the YOLO model, including the learning rate, momentum, weight decay coefficient, batch size, and maximum number of iterations. S32. Input the dataset with labeled information obtained after cutting the impurity prediction boxes by Labelme in step S2 into the YOLO model and start the training process according to the set initial training parameters. S33. During training, monitor the model's loss function value, accuracy, recall, and mAP evaluation metrics on the validation set in real time. S34. Based on the monitored changes in evaluation metrics, dynamically adjust the training parameters. For example, when the loss function value stops decreasing or the evaluation metrics fluctuate, reduce the learning rate. S35. Continue training and parameter adjustment until the model's evaluation metrics on the validation set reach the preset optimal threshold, or the number of training iterations reaches the maximum number of iterations.

[0042] S4. Save the optimal YOLO model.

[0043] The following specific example illustrates the process of this invention: 1. Obtain an internal image of the cut impurity containment box 2 and form a dataset (S100).

[0044] 1.1 Image acquisition device parameter settings.

[0045] A high-definition industrial camera with a resolution of at least 1920×1080 is used as the image acquisition device, with a frame rate set to 30 frames per second to ensure that dynamic images of cutting impurities during the cutting machine's operation can be captured. The camera's installation position should ensure a stable relative angle with the cutting impurity containment box 2, so that the acquired image can completely cover the key areas inside the containment box, and the distance between the camera and the containment box should be maintained within 30-50 cm to obtain the best image clarity and viewing angle.

[0046] 1.2 Lighting conditions settings.

[0047] To ensure the quality of the acquired images, a ring of LED light sources is evenly arranged around the impurity collection box 2. The light source color is white, and the light intensity is set to 500-1000 lux. Auxiliary devices such as light shields are used to reduce interference from ambient light, ensuring good contrast in the acquired images and facilitating subsequent image processing and feature extraction.

[0048] 1.3 Data acquisition process control.

[0049] During normal operation of the cutting machine, image data inside the cutting impurity containment box 2 is continuously collected. The collection time is set to 24 hours (which can be adjusted appropriately according to the actual cutting conditions) to ensure that it covers various cutting impurities generated during the operation of the cutting machine at different time periods, thereby enriching the diversity of the dataset.

[0050] 2. Train the YOLO model (S200) using the dataset.

[0051] 2.1 Create a dataset (S1).

[0052] 2.1.1 Collect raw image data (S11).

[0053] Image data of the cutting machine in operation was collected under different cutting scenarios. During the acquisition process, images were taken from at least 10 different perspectives (e.g., top-down, bottom-up, side-view, etc.), covering at least 5 different lighting conditions (e.g., natural bright light, low-light, direct light, diffused light, etc.) and 3 different cutting conditions (e.g., high-speed cutting, low-speed high-precision cutting, irregular shape cutting, etc.). A total of no less than 10,000 raw images were collected to provide a rich data foundation for subsequent model training.

[0054] 2.1.2 Filtering raw image data (S12).

[0055] Develop screening rules to remove images that are blurry (based on image sharpness evaluation metrics, such as variance assessment of the Laplacian operator, with a variance below 10 defined as a blurry image), have poor image quality (based on evaluation of image contrast, brightness, etc., with a contrast below 20 or a brightness range exceeding 10% of the 0-255 range defined as a low-quality image), and do not contain effective information from cropping impurities (screened through preliminary manual inspection). Retain effective images of good quality, ensuring that the number of screened images is no less than 80% of the original images, i.e., no less than 8000 images.

[0056] 2.1.3 Preprocessing operation (S13).

[0057] Image resizing: The filtered images are uniformly resized to 416×416 pixels, which is a commonly used and suitable data input size in YOLO model training, and helps the model learn image features.

[0058] Color space conversion and normalization: The image will be converted to grayscale (if the original image is a color image, this operation can be performed depending on the actual situation; grayscale images can reduce the amount of computation). At the same time, the pixel values ​​of the image will be normalized to a range between 0 and 1 to improve the stability and convergence speed of model training.

[0059] 2.1.4 Dataset partitioning (S14).

[0060] The preprocessed image data is divided into a training set, a validation set, and a test set in a ratio of 7:2:1, with approximately 5600 images in the training set, approximately 1600 images in the validation set, and approximately 800 images in the test set, to form a dataset for training the YOLO model.

[0061] 2.2 Label me marks the predicted box for cutting impurities (S2).

[0062] 2.2.1 Start the annotation tool and import data (S21).

[0063] Launch the Labelme annotation tool and import all the images from the dataset created in step S1 into the Labelme software interface in sequence to ensure that each image can be loaded and processed correctly.

[0064] 2.2.2 Determine the labeling category (S22).

[0065] For each image, the specific morphology, location, and distribution of the cut impurities were carefully observed. Based on the type of cut impurities, they were clearly divided into the following three annotation categories: Metal scrap: It is usually in the form of irregular small particles or polyhedrons, with a distinct metallic luster.

[0066] Dust clumps: These are generally loose but sticky granular clumps with irregular shapes and relatively low density.

[0067] Fiber filaments: long and thin with a certain degree of flexibility, appearing in strip or filament shape.

[0068] 2.2.3 Draw the prediction box (S23).

[0069] Using the drawing function in the Labelme tool, accurately draw the prediction box along the edge of the cut impurity. For regularly shaped cut impurities, the maximum distance between the edge of the prediction box and the edge of the impurity should not exceed 2 pixels; for irregularly shaped cut impurities, the prediction box should fit its outer contour as closely as possible, and the similarity between the covered area and the actual impurity area should reach more than 90%. At the same time, ensure that the prediction box completely covers the cut impurity area, and that the number of pixels that are mistakenly selected in non-impurity areas does not exceed 5% of the total number of pixels in the selected area.

[0070] 2.2.4 Assign category labels (S24).

[0071] Assign a corresponding category label to each drawn prediction box, and accurately fill in the determined cutting impurity category name in the Labelme annotation information field to ensure that the label is consistent with the pre-set "metal scrap", "dust clumps" and "fiber filaments" category system to avoid label confusion or errors.

[0072] 2.2.5 Inspection and Correction (S25).

[0073] All labeled images undergo a 100% comprehensive review, employing a two-person cross-checking method to ensure accurate drawing of prediction boxes and correct category labels. Images with questionable or inaccurate labeling (e.g., inconsistent results between two reviewers, prediction boxes covering less than 90% of impurities, or misselected boxes exceeding 5%) are returned for re-observation and labeling until all image labeling quality meets the standards.

[0074] 2.2.6 Export data file (S26).

[0075] After all images are labeled, export the data file containing the predicted bounding box annotations and category labels in the JSON format specified by Labelme, and save it to a dedicated folder for later use in training the YOLO model.

[0076] 2.3 Training the YOLO model (S3).

[0077] 2.3.1 Set initial training parameters (S31).

[0078] Set the initial training parameters for the YOLO model to ensure the stability and efficiency of the training process. Specific parameter settings are as follows: Learning rate: The initial learning rate is set to 0.001, which is a commonly used range for initial learning rates in deep learning training. This helps the model converge quickly to the vicinity of the optimal solution.

[0079] Momentum: The momentum parameter is set to 0.9. Momentum allows the model to accumulate an exponentially weighted average of previous gradients during training, which helps to accelerate convergence and avoid getting trapped in local optima.

[0080] Weight decay coefficient: The weight decay coefficient is set to 0.0005 to prevent model overfitting. By regularizing the model weights, the complexity of the model is limited.

[0081] Batch size: The batch size is set to 64. This parameter determines the number of samples input to the model each time it is trained. A suitable batch size can achieve a balance between computational efficiency and training effect.

[0082] Maximum number of iterations: Set the maximum number of iterations to 500 to ensure that the model has enough training rounds to converge to the optimal state.

[0083] 2.3.2 Start training (S32).

[0084] The dataset with labeled information obtained after cutting the impurity prediction boxes by Labelme in step S2 is input into the YOLO model. The training process begins according to the initial training parameters set in step S31, and the training loss value and validation set index are recorded for each iteration.

[0085] 2.3.3 Real-time monitoring and evaluation indicators (S33).

[0086] During training, the model's performance on the validation set is monitored in real time, including but not limited to: Loss function value: measures the difference between the model's prediction and the true label. The lower the loss function value, the better the model fits.

[0087] Accuracy: The proportion of correctly predicted samples out of the total number of predicted samples, used to evaluate the overall predictive accuracy of the model.

[0088] Recall: The proportion of samples that are actually positive that are correctly predicted as positive, reflecting the model's ability to capture positive samples.

[0089] mAP (mean Average Precision): This indicator comprehensively considers the average precision of the model under different IoU (Intersection over Union) thresholds and is an important metric for evaluating the overall performance of object detection models.

[0090] 2.3.4 Dynamically adjust training parameters (S34).

[0091] Based on the monitored changes in evaluation metrics, training parameters are dynamically adjusted to ensure the effectiveness and stability of model training. The specific adjustment strategy is as follows: When the loss function value no longer decreases in 10 consecutive iterations (i.e., the decrease is less than 0.001) or the evaluation metrics (such as accuracy, recall, mAP) fluctuate by more than 5%, the learning rate is reduced by 10% to finely adjust the model parameters and avoid missing the optimal solution.

[0092] If, after adjusting the learning rate, the loss function value does not decrease significantly (the decrease is less than 0.001) after 20 consecutive iterations and the evaluation metric is not significantly optimized, then the weight decay coefficient should be appropriately increased (by 0.0001 each time) to further prevent the model from overfitting.

[0093] 2.3.5 Continue training and adjusting until convergence (S35).

[0094] Continue training and parameter tuning until the model's evaluation metrics on the validation set reach the preset optimal threshold, as follows: The loss function value is less than 0.01; Accuracy rate no less than 90%; The recall rate shall not be less than 85%; mAP is not less than 0.8.

[0095] Alternatively, training can continue until the maximum number of iterations (500) is reached. The model trained when either the optimal evaluation metric or the maximum number of iterations is reached is considered the final trained YOLO model.

[0096] 2.4 Save the optimal YOLO model (S4).

[0097] Once the optimal evaluation metric or the maximum number of iterations has been reached, the trained YOLO model is saved in a standard model format (such as .pt) in a dedicated model storage folder for later retrieval and loading in practical applications.

[0098] 3. Control the operating status of the dust removal device 300 based on the judgment result (S300).

[0099] 3.1 Acquire new images and input them into the model for judgment (S200 and subsequent).

[0100] In actual operation, the internal images of the cutting impurity containment box (2) are continuously acquired. After preprocessing (size adjustment, normalization, etc.) in the same way as the training model in step S200, the images are input into the trained YOLO model. The model is then used to determine whether the anchor frame used to indicate the cutting impurities can be identified.

[0101] 3.2 Dust removal device 300 operation control (S300 specific control).

[0102] 3.2.1 When the anchor frame can be identified (S300 - Identification status).

[0103] If the anchor frame used to indicate cutting impurities can be identified, it indicates that cutting impurities exist in the cutting impurity holding box 2, and the dust removal device 300 is then activated. Specific control parameters are as follows: Hair dryer 8: Set the hair dryer speed to 1500-2000 rpm and the air pressure to 800-1000 Pa to ensure sufficient suction to pick up impurities.

[0104] Exhaust fan 13: Set the exhaust fan speed to 2000-2500 rpm and the maximum air volume to 1000-1500 cubic meters per hour to efficiently discharge the cutting impurities in the holding box to the outside of the holding box.

[0105] 3.2.2 When the anchor frame cannot be identified (S300 - Unidentifiable situation).

[0106] If the anchor frame used to indicate cutting impurities cannot be identified, it indicates that there may be no cutting impurities in the cutting impurity holding box 2 temporarily, or the amount of impurities is very small. At this time, the dust removal device 300 is controlled to stop operating to avoid unnecessary energy consumption.

[0107] It should be noted that this invention, by acquiring images of the interior of the cutting impurity containment box and using a YOLO model for precise identification, can accurately determine the actual situation of the cutting impurities. When impurities are detected, the dust removal device is activated promptly, avoiding the ineffective operation of traditional timed dust removal methods when impurities are few. This effectively improves dust removal efficiency, ensuring the cutting machine is always in a good working environment, reducing the impact of impurity accumulation on equipment performance, and thus improving the cutting machine's efficiency and cutting quality. Based on the judgment results of the YOLO model, this invention intelligently controls the operation of the dust removal device; that is, it activates the blower and exhaust fan for dust removal when the cutting impurity anchor frame is detected, and stops operation when it cannot be detected. This intelligent control method avoids unnecessary energy consumption, significantly reducing energy costs compared to traditional fixed-mode dust removal systems, and aligns with the development trend of energy conservation and environmental protection.

[0108] In a preferred embodiment, this application also provides an electronic device, the electronic device comprising: The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the intelligent dust removal method for the cutting machine. The computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the method of the present invention.

[0109] This invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0110] It will be understood by those skilled in the art that the method steps of the present invention can be performed by a computer program instructing related hardware, such as a computer device or processor, which may be stored in a non-transitory computer-readable storage medium. When the computer program is executed, the steps of the present invention are performed. Depending on the context, any references herein to memory, storage, database, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0111] It is understood that this invention, by acquiring images of the interior of the cutting impurity containment box and using the YOLO model for precise identification, can accurately determine the actual situation of the cutting impurities. When impurities are identified, the dust removal device is activated promptly, avoiding the ineffective operation of traditional timed dust removal methods when impurities are few. This effectively improves dust removal efficiency, ensuring the cutting machine is always in a good working environment, reducing the impact of impurity accumulation on equipment performance, and thus improving the cutting machine's working efficiency and cutting quality. Based on the judgment results of the YOLO model, this invention intelligently controls the operation of the dust removal device; that is, it activates the blower and exhaust fan for dust removal when the cutting impurity anchor frame is identified, and stops operation when it cannot be identified. This intelligent control method avoids unnecessary energy consumption, significantly reducing energy costs compared to traditional fixed-mode dust removal systems, and conforms to the development trend of energy conservation and environmental protection.

[0112] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0113] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A cutting machine intelligent dust removal system, characterized in that, The system includes: Cutting impurity holding box (2), the cutting impurity holding box (2) is set at the bottom of the cutting machine and is used to hold the cutting impurities of the cutting machine; The acquisition module (100) includes a camera (17) for acquiring internal images of the cut impurity containment box (2) to form a dataset; The control module (200) is used to train the YOLO model using the dataset. After the YOLO model is trained, it acquires the internal image of the cutting impurity holding box (2) and inputs it into the trained YOLO model. The trained YOLO model is used to determine whether the anchor frame used to indicate the cutting impurities can be identified, and the operation status of the dust removal device (300) is controlled according to the judgment result. The dust removal device (300) includes a fan unit connected to the cutting impurity holding box (2), the fan unit being used to discharge the cutting impurities in the cutting impurity holding box (2) to the outside of the cutting impurity holding box (2) during operation; The process of training a YOLO model using a dataset includes: S1. Create a dataset; S2, Labelme marks the predicted bounding box for cutting impurities; S3. Train the YOLO model; S4. Save the optimal YOLO model; Step S2 specifically includes: S21. Start the Labelme annotation tool and import the images from the dataset created in step S1 into the Labelme software interface in sequence. S22. For each image, carefully observe the specific shape, location and distribution of the cutting impurities in the image, and determine different labeling categories based on the type of cutting impurities, including metal fragments, dust clumps and fiber filaments. S23. Using the drawing function in Labelme tool, accurately draw the prediction box along the edge of the cut impurity on the image to ensure that the prediction box completely covers the cut impurity area, while minimizing the selection of non-impurity areas. For cut impurities with irregular shapes, the prediction box should fit its outer contour as closely as possible. S24. Assign a corresponding category label to each drawn prediction box, and accurately fill in the determined cut impurity category name in the Labelme annotation information column to ensure that the category label is consistent with the pre-set category system. S25. Conduct a preliminary check on the labeled images to see if the prediction boxes are drawn accurately and if the category labels are correct. For images that are questionable or inaccurately labeled, go back and observe and label them again to ensure the labeling quality. S26. After all images are labeled, export the data file with predicted bounding box annotations and category labels in the format specified by Labelme, so that it can be used for subsequent training of the YOLO model.

2. The cutting machine intelligent dust removal system according to claim 1, characterized in that, The blower unit includes a blower (8) and an exhaust fan (13). The cutting impurity holding box (2) is provided with a blower port (4) and an exhaust port (5). The blower port (4) is connected to the blower (8) through a pipe, and the exhaust port (5) is connected to the exhaust fan (13) through a pipe.

3. A cutting machine intelligent dust removal method, characterized in that, The method applicable to the intelligent dust removal system for cutting machines according to claim 2 includes: S100. Obtain an internal image of the cutting impurity containment box (2) to form a dataset; S200. Use the dataset to train the YOLO model. After the YOLO model is trained, obtain the internal image of the cut impurity containment box (2) and input it into the trained YOLO model. Use the trained YOLO model to determine whether the anchor box used to indicate the cut impurities can be identified. S300: Control the operating status of the dust removal device (300) based on the judgment result.

4. The intelligent dust removal method for cutting machines according to claim 3, characterized in that, The step of controlling the operating status of the cutting machine based on the judgment result includes: If the anchor frame used to indicate the cutting of impurities can be identified, control the dust removal device (300) to operate; If the anchor frame used to indicate the cutting of impurities cannot be identified, the dust removal device (300) is controlled to stop operating.

5. The intelligent dust removal method for cutting machines according to claim 4, characterized in that, The step of controlling the operating status of the cutting machine based on the judgment result specifically includes: If the anchor frame used to indicate the cutting impurities can be identified, control the operation of the blower (8) and the exhaust fan (13); If the anchor frame used to indicate cutting impurities cannot be identified, control the blower (8) and exhaust fan (13) to stop operating.

6. The intelligent dust removal method for a cutting machine according to claim 3, characterized in that, Step S1 specifically includes: S11. Collect raw image data containing cutting impurities during the operation of the cutting machine. The raw image data comes from multiple cutting machine working scenarios with different perspectives, different lighting conditions, and different cutting conditions. S12. Filter the collected raw image data to remove blurry, poor image quality, and images that do not contain effective information about cutting impurities. S13. Perform preprocessing operations on the filtered image data before labeling. The preprocessing operations include adjusting the image size to a uniform size, converting the image to grayscale or color space, and normalizing the image pixel values ​​so that the image data meets the requirements of subsequent model training. S14. Divide the preprocessed image data into training set, validation set and test set according to a certain ratio to form a dataset for training the YOLO model.

7. The intelligent dust removal method for a cutting machine according to claim 3, characterized in that, Step S3 specifically includes: S31. Set the initial training parameters of the YOLO model, including the learning rate, momentum, weight decay coefficient, batch size, and maximum number of iterations. S32. Input the dataset with labeled information obtained after cutting the impurity prediction boxes by Labelme in step S2 into the YOLO model and start the training process according to the set initial training parameters. S33. During training, monitor the model's loss function value, accuracy, recall, and mAP evaluation metrics on the validation set in real time. S34. Based on the monitored changes in evaluation metrics, dynamically adjust the training parameters. For example, when the loss function value stops decreasing or the evaluation metrics fluctuate, reduce the learning rate. S35. Continue training and parameter adjustment until the model's evaluation metrics on the validation set reach the preset optimal threshold, or the number of training iterations reaches the maximum number of iterations.

8. An electronic device, characterized in that, include: Memory; The processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the intelligent dust removal method for a cutting machine according to any one of claims 3 to 7.

Images