An automated bearing defect detection system and method

Abstract

The application discloses an automatic bearing defect detection system and method. The system comprises a part supply module, a part conveying module, an illumination and image acquisition module, a visual detection module, a control module and a safety module. The method is as follows: firstly, the part supply module adopts an automatic rotating tray to automatically feed the parts to be detected, and the part conveying module uses a conveying belt and a feeding mechanical arm to convey the bearings to the visual detection area; then the illumination and image acquisition module takes multi-angle photos of the bearings, and sends the bearings into the visual detection module, which uses a detection algorithm and an edge computing device to detect whether the bearing parts have defects, and sends the detection results to the control module; finally, the control module conveys the detected bearings to the waste area and the unloading area respectively; the safety module protects the safety of personnel and equipment during the automatic detection process, and detects and responds to potential dangerous situations in time. The application improves the automation degree, detection efficiency and detection precision of bearing defect detection.

Description

Technical Field

[0001] This invention relates to the field of computer vision inspection technology, and in particular to an automated bearing defect detection system and method. Background Technology

[0002] With technological advancements, image processing and object detection algorithms enable the automatic detection and classification of objects. Currently, computer vision technology is widely used for quality inspection in industrial production. Research on bearing defect detection systems involves using computer vision to capture and analyze bearing images in real time to detect surface defects, dimensional deviations, and other issues. However, modern bearing defect detection systems do not rely on a single technology but combine multiple technologies such as computer vision, deep learning, and sensor technology to build intelligent detection systems. These systems typically feature automation, high efficiency, and high accuracy, meeting the needs of industrial production lines.

[0003] A visual inspection method and apparatus disclosed in application publication number CN113552123A, relating to the field of visual inspection, mainly includes the following steps: acquiring image detection data from at least one image acquisition device remotely, and then generating a detection result using a standard image database and the image detection data. The detection result includes a judgment on whether the image detection data is abnormal, and the corresponding abnormality category in abnormal cases. This method separates image acquisition and processing, improving layout flexibility. Furthermore, at least one image acquisition device can share a standard image database, computing resources, and processing resources, reducing the cost of setting up local image acquisition devices while utilizing better resources and richer data, thereby improving the quality of visual inspection. A visual inspection method and system disclosed in application publication number CN112240888A performs visual inspection on products using a visual inspection system and uses sensors to detect the time and quantity of each product passing by. A synchronous speed measuring device records the initial position information and tracks the movement distance of each product when it passes the sensor. A server controls multiple cameras to simultaneously detect multiple products, acquiring the movement distance of each product in real time. When a product moves to within the camera's shooting range, the server triggers the camera to take a picture of the product. This method solves the problem that single-vision cannot handle complex visual detection content and ensures that the overall detection efficiency is not affected by the camera's shooting speed, thus significantly improving the detection efficiency.

[0004] Existing technologies only use computer vision for product defect detection. The loading and sorting of materials after inspection still require manual operation, which is subject to delays and operational errors caused by human factors. The detection efficiency is low, and there are also problems such as insufficient monitoring data sampling and low detection accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide an automated bearing defect detection system and method with high automation, high detection efficiency, and high detection accuracy.

[0006] The technical solution to achieve the purpose of this invention is: an automated bearing defect detection system, including a parts supply module, a parts conveying module, an illumination and image acquisition module, a vision inspection module, a control module, and a safety module;

[0007] The parts supply module uses an automatic rotating tray to supply parts to be inspected;

[0008] The parts conveying module includes a conveyor belt, a loading robotic arm, a unloading robotic arm, and an operating unit, used to convey the parts to be inspected to the visual inspection area.

[0009] The lighting and image acquisition module includes a supplementary light source and an image acquisition module, which are used to capture the details of the parts to be inspected in order to meet complex visual recognition needs.

[0010] The visual inspection module includes a detection algorithm model and an edge computing device, used to detect whether there are defects in bearing parts;

[0011] The control module includes a main control unit and a control panel, which are used to monitor and control the coordinated operation between various modules, receive and process the detection data from the vision inspection module, and make corresponding control decisions.

[0012] The safety module includes physical protection, an emergency stop button, and safety sensors to protect the safety of personnel and equipment during the testing process and to detect and respond to potential dangerous situations.

[0013] Furthermore, the automatic rotating tray of the parts supply module works with the robotic arm of the parts conveying module to clamp parts. After clamping a part, the automatic rotating tray rotates once to complete the automated feeding process.

[0014] The automatic rotating tray includes guide rods, a parts mounting plate, a stepper motor, and a lifting mechanism arranged from top to bottom. The stepper motor controls the rotation angle of the tray, working in conjunction with the loading robotic arm to complete the loading task. The parts mounting plate is equipped with detachable guide rods to fix the bearing parts and prevent them from slipping during rotation, thus completing the clamping task. The guide rods are adaptable to bearing parts of different sizes, and the number of guide rods is selected according to the actual situation. The lifting mechanism adjusts the height of the tray for contact with the worktable.

[0015] Furthermore, the operation unit of the parts conveying module controls the loading robotic arm to place the parts at a designated position on the conveyor belt after gripping them, arranging a batch of parts in a 3x4 layout; the operation unit is equipped with a coordinate system to guide the loading robotic arm to accurately place the parts at the designated position on the conveyor belt; the unloading robotic arm, controlled by signals from the operation unit, grips and unloads defective parts; the operation unit controls the speed of the conveyor belt to ensure that the parts are stably transported to the inspection area.

[0016] Furthermore, the supplementary light source of the lighting and image acquisition module adopts a multi-angle light source to reduce shadows and reflections, and the lighting intensity is adjusted according to the material and color of the part to obtain the best image quality; the image acquisition module uses a high-resolution camera to capture the tiny details on the surface of the part, and adopts multi-angle shooting to comprehensively inspect every surface of the part.

[0017] Furthermore, the detection algorithm model of the visual detection module includes a model training unit, a feature extraction unit, a defect detection unit, a feature fusion unit, and a detection head unit;

[0018] The model training unit includes a model selection submodule, a model initialization submodule, a loss function submodule, an optimizer submodule, a batch processing submodule, a regularization submodule, a model evaluation submodule, and a model saving submodule;

[0019] The model selection submodule is used to select a suitable machine learning or deep learning model structure for the problem; the model initialization submodule is used to initialize the model's weights and biases; the loss function submodule is used to optimize the model's loss function, which is a function used to measure the difference between the model's output and the actual labels; the optimizer submodule selects and configures the optimization algorithm used to optimize the model; the batch processing submodule divides the data into mini-batches during training for calculating gradients and updating model parameters; the regularization submodule adds regularization terms to prevent overfitting; the model evaluation submodule is used to evaluate the model's performance on the training set, validation set, or test set; and the model saving submodule saves the trained model weights for later use.

[0020] The feature extraction unit includes a convolutional layer submodule, a pooling layer submodule, a normalization layer submodule, an activation layer submodule, a feature selection submodule, and a feature dimensionality reduction submodule.

[0021] The convolutional layer submodule extracts image features using convolution operations; the pooling layer submodule reduces the spatial dimension of the feature map using pooling operations; the normalization layer submodule normalizes the features; the activation layer submodule applies activation functions to introduce nonlinearity and increase the model's expressive power; the feature selection submodule selects the most representative features for subsequent defect detection; and the feature dimensionality reduction submodule reduces the dimensionality of the features.

[0022] The defect detection unit includes an image preprocessing submodule, a feature extraction submodule, a defect detection submodule, a defect localization submodule, and a visualization submodule;

[0023] The image preprocessing submodule preprocesses the input image, including resizing, cropping, or enhancing image contrast; the feature extraction submodule extracts features from the preprocessed image; the defect detection submodule uses the extracted features to detect defects in the image, including methods using classifiers or regressors; the defect localization submodule locates and marks the positions of detected defects in the image; and the visualization submodule visualizes the detected defect information for user understanding and analysis of the detection results.

[0024] The detection process of the detection algorithm model is as follows:

[0025] (1) The image acquisition module acquires image information of bearing parts and transmits the image information of bearing parts to the image preprocessing submodule in the vision detection module. The image preprocessing submodule performs noise reduction on the acquired image information of bearing parts through filtering algorithm to reduce the error of feature extraction and model training.

[0026] (2) The convolutional layer submodule in the feature extraction unit adopts a lightweight model based on YOLOv8 to perform convolution operations on the preprocessed bearing parts image information; in the backbone network of YOLOv8, the convolutional layer of downsampling operation is replaced with the lightweight adaptive convolutional module LACM. The structure of LACM is divided into two paths: one path obtains an adaptive weight, and the other path uses group convolution to reduce the number of parameters. Finally, it is multiplied with the obtained adaptive weight and then added to obtain the feature parameters with emphasis; in the backbone network of YOLOv8, the Bottleneck layer used by C2f is replaced with the multi-scale convolutional module EMC. The structure of EMC divides the input features into two parts: one part is processed by 3×3 Conv and 5×5 Conv, and the other part retains the original features without operation; finally, the two parts of features are merged together to perform 1×1 Conv; the bearing contour and texture required for bearing parts defect detection are extracted;

[0027] (3) The model selection submodule in the model training unit selects the specified lightweight model based on YOLOv8 improvement, uses preprocessed image data and extracted features to train the deep learning model; finally, YOLOv8m is used as the teacher model to perform knowledge distillation on the lightweight model based on YOLOv8 improvement.

[0028] (4) The model evaluation submodule in the model training unit evaluates the performance of the trained deep learning model. The evaluation metrics include computational accuracy, precision and recall.

[0029] (5) The visualization sub-module in the defect detection unit is electrically connected to the control panel of the control module to visualize the defect information and generate a bearing part defect information report. The bearing part defect information report includes the number of bearing part defects, the location of the defects, and the type of bearing defects.

[0030] Furthermore, the loss function submodule of the visual inspection module uses the composite loss function L. total To simultaneously optimize the bounding box position, size, object confidence, and class probability, the loss function L... total The calculation formula is:

[0031] L total =L coord +L conf +L class

[0032] In the formula, L coord For coordinate loss, the squared difference loss is used to measure the difference between the predicted bounding box and the ground truth bounding box; L conf For confidence loss, MSE or cross-entropy loss is used to evaluate the accuracy of confidence prediction; L class For classification loss, cross-entropy loss is used to evaluate the accuracy of class probability prediction;

[0033] The cross-entropy loss function L is:

[0034]

[0035] In the formula, M is the number of categories, and y o,c p is the observed value. o,c To predict probabilities.

[0036] Furthermore, the control module uses a PLC as the main control unit of the system to control the rotation of the automatic rotating material tray, the operation of the loading robotic arm, the movement of the conveyor belt, and the operation of the unloading robotic arm. It also uses an industrial PC to run a lightweight deep learning model to process the images captured by the image acquisition module, perform defect detection, and feed the results back to the PLC.

[0037] Furthermore, the physical protection uses safety sensors to monitor the environment around the loading and unloading robotic arms in real time, sets safety limits for the loading and unloading robotic arms, and controls the movement range of the loading and unloading robotic arms to not exceed the preset range.

[0038] The emergency stop button is used to stop the movement of the loading and unloading robotic arms in case of an emergency.

[0039] The safety sensor is used to monitor the status parameters of the loading and unloading robotic arms, and to issue an alarm and trigger the emergency stop button in case of abnormal conditions.

[0040] An automated bearing defect detection method based on the aforementioned automated bearing defect detection system includes the following steps:

[0041] Step 1: The parts supply module uses an automatic rotating tray to automatically feed the parts to be inspected.

[0042] Step 2: The parts conveying module uses a conveyor belt and a loading robotic arm to transport the bearings to the vision inspection area;

[0043] Step 3: The lighting and image acquisition module takes multi-angle photos of the bearing and sends them to the visual inspection module;

[0044] Step 4: The visual inspection module uses detection algorithms and edge computing devices to detect whether there are defects in the bearing parts, and sends the detection results to the control module.

[0045] Step 5: The control module sorts the inspected bearings, transporting defective bearings to the scrap area and defect-free bearings to the unloading area.

[0046] Step 6: The safety module protects the safety of personnel and equipment during the automated testing process, and promptly detects and responds to potential hazards.

[0047] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the automated bearing defect detection method.

[0048] Compared with the prior art, the significant advantages of this invention are:

[0049] (1) By automating the feeding, transmission, inspection and unloading processes, the need for manual operation is reduced, and the speed and efficiency of bearing defect detection process are improved;

[0050] (2) The automated system can operate continuously, reducing delays caused by human factors and achieving a highly efficient production process;

[0051] (3) By reducing reliance on manual operations, labor costs can be significantly reduced;

[0052] (4) Using deep learning algorithms to perform 360-degree all-round inspection of bearings can efficiently and accurately identify various defects, such as cracks, scratches or other irregularities, ensuring that only qualified products can enter the next production stage. The inspection is more accurate and consistent, improving the overall quality of the products.

[0053] (5) Using an improved lightweight deep learning target detection algorithm, the number of parameters is greatly reduced while ensuring detection accuracy, making it more suitable for application in industrial real-time detection scenarios. Attached Figure Description

[0054] Figure 1 This is a structural block diagram of an automated bearing defect detection system according to the present invention.

[0055] Figure 2 This is a schematic diagram of the structure of the automated bearing defect detection system provided in an embodiment of the present invention.

[0056] Figure 3 This is a schematic diagram of the structure of the automatic rotating tray in an embodiment of the present invention.

[0057] Figure 4 This is a schematic diagram of the lightweight detection algorithm in an embodiment of the present invention. Detailed Implementation

[0058] Combination Figure 1 This invention provides an automated bearing defect detection system, including a parts supply module, a parts conveying module, an illumination and image acquisition module, a vision inspection module, a control module, and a safety module;

[0059] The parts supply module uses an automatic rotating tray to supply parts to be inspected;

[0060] The parts conveying module includes a conveyor belt, a loading robotic arm, a unloading robotic arm, and an operating unit, used to convey the parts to be inspected to the visual inspection area.

[0061] The lighting and image acquisition module includes a supplementary light source and an image acquisition module, which are used to capture the details of the parts to be inspected in order to meet complex visual recognition needs.

[0062] The visual inspection module includes a detection algorithm model and an edge computing device, used to detect whether there are defects in bearing parts;

[0063] The control module includes a main control unit and a control panel, which are used to monitor and control the coordinated operation between various modules, receive and process the detection data from the vision inspection module, and make corresponding control decisions.

[0064] The safety module includes physical protection, an emergency stop button, and safety sensors to protect the safety of personnel and equipment during the testing process and to detect and respond to potential dangerous situations.

[0065] As a specific example, the automatic rotating tray of the parts supply module works with the robotic arm of the parts conveying module to clamp parts. After clamping a part, the automatic rotating tray rotates once to complete the automated feeding process.

[0066] The automatic rotating tray includes guide rods, a parts mounting plate, a stepper motor, and a lifting mechanism arranged from top to bottom. The stepper motor controls the rotation angle of the tray, working in conjunction with the loading robotic arm to complete the loading task. The parts mounting plate is equipped with detachable guide rods to fix the bearing parts and prevent them from slipping during rotation, thus completing the clamping task. The guide rods are adaptable to bearing parts of different sizes, and the number of guide rods is selected according to the actual situation. The lifting mechanism adjusts the height of the tray for contact with the worktable.

[0067] As a specific example, the operation unit of the parts conveying module controls the loading robotic arm to place the parts at a designated position on the conveyor belt after gripping them, arranging a batch of parts in a 3x4 layout; the operation unit is equipped with a coordinate system to guide the loading robotic arm to accurately place the parts at the designated position on the conveyor belt; the unloading robotic arm, controlled by signals from the operation unit, grips and unloads defective parts; the operation unit controls the speed of the conveyor belt to ensure that the parts are stably transported to the inspection area.

[0068] As a specific example, the lighting and image acquisition module uses a multi-angle light source to reduce shadows and reflections, and the lighting intensity is adjusted according to the material and color of the part to obtain the best image quality; the image acquisition module uses a high-resolution camera to capture tiny details on the surface of the part, and uses multi-angle shooting to comprehensively inspect every surface of the part.

[0069] As a specific example, the detection algorithm model of the visual detection module includes a model training unit, a feature extraction unit, a defect detection unit, a feature fusion unit, and a detection head unit;

[0070] The model training unit includes a model selection submodule, a model initialization submodule, a loss function submodule, an optimizer submodule, a batch processing submodule, a regularization submodule, a model evaluation submodule, and a model saving submodule;

[0071] The model selection submodule is used to select a suitable machine learning or deep learning model structure for the problem; the model initialization submodule is used to initialize the model's weights and biases; the loss function submodule is used to optimize the model's loss function, which is a function used to measure the difference between the model's output and the actual labels; the optimizer submodule selects and configures the optimization algorithm used to optimize the model, such as stochastic gradient descent (SGD) or Adam; the batch processing submodule divides the data into mini-batches during training to calculate gradients and update model parameters; the regularization submodule adds regularization terms to prevent the model from overfitting; the model evaluation submodule is used to evaluate the model's performance on the training set, validation set, or test set; and the model saving submodule saves the trained model weights for later use.

[0072] The feature extraction unit includes a convolutional layer submodule, a pooling layer submodule, a normalization layer submodule, an activation layer submodule, a feature selection submodule, and a feature dimensionality reduction submodule.

[0073] The convolutional layer submodule extracts image features using convolution operations; the pooling layer submodule reduces the spatial dimension of the feature map and lowers computational complexity through pooling operations; the normalization layer submodule normalizes the features, improving the model's stability and convergence speed; the activation layer submodule applies activation functions to introduce nonlinearity, increasing the model's expressive power; the feature selection submodule selects the most representative features for subsequent defect detection; and the feature dimensionality reduction submodule reduces the feature dimensionality, lowering computational costs and improving the model's generalization ability.

[0074] The defect detection unit includes an image preprocessing submodule, a feature extraction submodule, a defect detection submodule, a defect localization submodule, and a visualization submodule;

[0075] The image preprocessing submodule preprocesses the input image, including resizing, cropping, or enhancing image contrast; the feature extraction submodule extracts features from the preprocessed image; the defect detection submodule uses the extracted features to detect defects in the image, including methods using classifiers or regressors; the defect localization submodule locates and marks the positions of detected defects in the image; and the visualization submodule visualizes the detected defect information for user understanding and analysis of the detection results.

[0076] The detection process of the detection algorithm model is as follows:

[0077] (1) The image acquisition module acquires image information of bearing parts and transmits the image information of bearing parts to the image preprocessing submodule in the vision detection module. The image preprocessing submodule performs noise reduction on the acquired image information of bearing parts through filtering algorithm to reduce the error of feature extraction and model training.

[0078] (2) The convolutional layer submodule in the feature extraction unit adopts a lightweight model based on YOLOv8 to perform convolution operations on the preprocessed bearing parts image information; in the YOLOv8 backbone network, the convolutional layer of downsampling operation is replaced with the lightweight adaptive convolutional module LACM. The structure of LACM is divided into two paths: one path obtains an adaptive weight, and the other path uses group convolution to reduce the number of parameters. Finally, it is multiplied with the obtained adaptive weight and then added to obtain the feature parameters with emphasis; this reduces the number of parameters while making the feature extraction more accurate; in the YOLOv8 backbone network, the Bottleneck layer used by C2f is replaced with the multi-scale convolutional module EMC. The structure of EMC divides the input features into two parts: one part is processed by 3×3 Conv and 5×5 Conv, and the other part retains the original features without operation; finally, the two parts of features are merged together to perform 1×1 Conv, which achieves the purpose of reducing parameters and improving feature representation; the bearing contour and texture required for bearing parts defect detection are extracted;

[0079] (3) The model selection submodule in the model training unit selects the specified lightweight model based on YOLOv8 improvement, uses preprocessed image data and extracted features to train the deep learning model; finally, YOLOv8m is used as the teacher model to perform knowledge distillation on the lightweight model based on YOLOv8 improvement to improve its feature extraction and fusion capabilities and improve detection accuracy.

[0080] (4) The model evaluation submodule in the model training unit evaluates the performance of the trained deep learning model. The evaluation metrics include computational accuracy, precision and recall.

[0081] (5) The visualization sub-module in the defect detection unit is electrically connected to the control panel of the control module to visualize the defect information and generate a bearing part defect information report. The bearing part defect information report includes the number of bearing part defects, the location of the defects, and the type of bearing defects.

[0082] As a concrete example, the loss function submodule of the visual inspection module uses the composite loss function L. total To simultaneously optimize the bounding box position, size, object confidence, and class probability, the loss function L... total The calculation formula is:

[0083] L total =L coord +L conf +L class

[0084] In the formula, L coord For coordinate loss, the squared difference loss is used to measure the difference between the predicted bounding box and the ground truth bounding box; Lconf For confidence loss, MSE or cross-entropy loss is used to evaluate the accuracy of confidence prediction; L class For classification loss, cross-entropy loss is used to evaluate the accuracy of class probability prediction;

[0085] The cross-entropy loss function L is:

[0086]

[0087] In the formula, M is the number of categories, and y o,c p is the observed value. o,c To predict probabilities.

[0088] As a specific example, the control module uses a PLC as the main control unit of the system to control the rotation of the automatic rotating tray, the operation of the loading robotic arm, the movement of the conveyor belt, and the operation of the unloading robotic arm. It also uses an industrial PC to run a lightweight deep learning model to process images captured by the image acquisition module, perform defect detection, and feed the results back to the PLC.

[0089] As a specific example, the physical protection uses safety sensors to monitor the environment around the loading and unloading robotic arms in real time, sets safety limits for the loading and unloading robotic arms, and controls the movement range of the loading and unloading robotic arms to not exceed the preset range.

[0090] The emergency stop button is used to stop the movement of the loading and unloading robotic arms in case of an emergency.

[0091] The safety sensor is used to monitor the status parameters of the loading and unloading robotic arms, and to issue an alarm and trigger the emergency stop button in case of abnormal conditions.

[0092] The present invention also provides an automated bearing defect detection method based on the aforementioned automated bearing defect detection system, comprising the following steps:

[0093] Step 1: The parts supply module uses an automatic rotating tray to automatically feed the parts to be inspected.

[0094] Step 2: The parts conveying module uses a conveyor belt and a loading robotic arm to transport the bearings to the vision inspection area;

[0095] Step 3: The lighting and image acquisition module takes multi-angle photos of the bearing and sends them to the visual inspection module;

[0096] Step 4: The visual inspection module uses detection algorithms and edge computing devices to detect whether there are defects in the bearing parts, and sends the detection results to the control module.

[0097] Step 5: The control module sorts the inspected bearings, transporting defective bearings to the scrap area and defect-free bearings to the unloading area.

[0098] Step 6: The safety module protects the safety of personnel and equipment during the automated testing process, and promptly detects and responds to potential hazards.

[0099] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the automated bearing defect detection method.

[0100] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0101] Example

[0102] This embodiment provides a mechanical structure diagram of an automated bearing defect detection system, such as... Figure 2 As shown, the mechanical structure of the automated bearing defect detection system includes a loading robot arm 1, an automated production line 2, a visual tray 3, a control panel 4, a bottom camera device 5, a gantry support 6, a top camera 7, a bottom camera 8, a side camera 9, and a loading robot arm 10, wherein:

[0103] The unloading robot arm 1 is responsible for sorting and clamping defective bearing parts. After completion, it sends a signal to the control module to carry out the next batch of inspection work.

[0104] The automated production line 2 is responsible for transporting parts. It is controlled by the control module and works with two robotic arms and a vision inspection module to complete the entire inspection system.

[0105] The number and size of the visual tray 3 are set according to the specific number and size of the parts. It is made of polycarbonate (PC) material with high transparency. PC material not only has high transparency, but also has good impact resistance and temperature resistance, making it suitable for use in different environments. Its high transparency also allows the camera below the detection area to capture images.

[0106] Control Panel 4 is the interface through which users interact with the part defect detection system, allowing operators to set detection parameters, start and stop the detection process, monitor system status, and view alarms and notifications.

[0107] Bottom camera device 5 supports the bottom camera and includes a smart spotlight that adjusts its brightness according to indoor lighting.

[0108] The gantry bracket 6 is a support component used to support the camera and control panel, etc.

[0109] The top-mounted camera 7 is a key hardware component responsible for capturing image data of parts on the production line. These images are used to identify any potential defects through visual algorithms. The choice of camera is crucial to ensuring the accuracy and efficiency of the inspection system; the camera needs to be compatible with the defect detection algorithm so that the model can accurately analyze the captured images, and it needs to include intelligent spotlights that adjust brightness according to indoor lighting conditions, taking images of the parts from above.

[0110] The bottom camera 8 is an upward-facing component, featuring intelligent spotlights that adjust brightness according to indoor lighting conditions.

[0111] There are two side cameras (9 in total), responsible for shooting from the left and right perspectives, and they include intelligent spotlights that adjust brightness according to indoor lighting.

[0112] The loading robot arm 10 is responsible for picking up parts from the material tray and placing them neatly on the visible tray of the conveyor belt in a certain arrangement. After a batch of parts is clamped, the conveyor belt starts to operate.

[0113] like Figure 3 The diagram shows the mechanical structure of the automatic rotating tray, including guide rods 11, a parts mounting plate 12, a stepper motor 13, and a lifting mechanism 14. The stepper motor 13 precisely controls the motor's stepping, allowing for precise control of the tray's rotation angle. This is suitable for scenarios requiring precise position control, working in conjunction with a robotic arm to complete the loading task. The upper parts mounting plate 12 is equipped with three detachable guide rods 11 to secure bearing parts and prevent slippage during rotation, ensuring precise clamping. The three guide rods 11 accommodate bearing parts of different sizes, and the number can be selected based on actual needs. The lower lifting mechanism 14 raises the tray from one height to another, facilitating connection with other equipment or workbenches and ensuring smooth transitions between different processing stages. Using the lifting mechanism 14 reduces the need for manual handling of parts or products, thereby reducing labor intensity and improving production safety and efficiency.

[0114] Figure 4 The detection algorithm structure of the visual inspection module includes a model training module, a feature extraction module, a defect detection module, a feature fusion module, and a detection head module;

[0115] The model training module includes a model selection module, a model initialization module, a loss function module, an optimizer module, a batch processing module, a regularization module, a model evaluation module, and a model saving module;

[0116] The feature extraction module includes a convolutional layer module, a pooling layer module, a normalization layer module, an activation layer module, a feature selection module, and a feature dimensionality reduction module;

[0117] The defect detection module includes an image preprocessing module, a feature extraction module, a defect detection module, a defect localization module, and a visualization module;

[0118] The detection algorithm flow of the visual inspection module is as follows:

[0119] (1) The image information of bearing parts acquired by the high-definition camera device in the image acquisition module is stored in the storage module. The image of the bearing parts is transmitted to the image preprocessing module in the defect detection module. The image denoising module in the image preprocessing module performs denoising processing on the acquired bearing image through a filtering algorithm to reduce the error of feature extraction and model training.

[0120] (2) The convolutional layer module in the feature extraction module uses a convolutional neural network deep learning model to perform convolution operations on the preprocessed bearing part image information to extract the bearing contour and texture required for bearing part defect detection.

[0121] (3) The model selection module in the model training module selects the specified convolutional neural network model, and uses preprocessed image data and extracted features to train a high-performance deep learning model.

[0122] (4) The model evaluation module in the model training module evaluates the performance of the trained deep learning model. The evaluation metrics include computational accuracy, precision and recall.

[0123] (5) The visualization module in the defect detection module is electrically connected to the external control panel device to visualize the defect information and generate a bearing part defect information report. The bearing part defect information report includes the number of bearing part defects, the location of the defects, and the type of bearing defects.

[0124] Furthermore, the deep learning model is a lightweight improvement based on the YOLOv8 algorithm. It uses a self-developed LACM module to complete the downsampling operation in the backbone network and designs adaptive weight selection and grouped convolution, which reduces the model parameters while improving accuracy.

[0125] Using a self-developed EMC module as the bottleneck layer in the backbone network, the bottleneck structure was reconstructed using the idea of ​​grouped convolution, which further improved the detection accuracy and reduced the number of model parameters.

[0126] Knowledge distillation was performed on the improved model, and the accuracy of detection was further improved by using the teacher model.

[0127] In summary, this lightweight network based on the YOLOv8 deep learning model improves detection accuracy while reducing the number of model parameters, making it easy to use in edge devices for real-time detection tasks. The detection results can be visualized for staff, making it more intuitive, reducing manual intervention, and increasing detection efficiency, thereby enabling automated bearing defect detection.

[0128] It should be emphasized that the foregoing examples are merely illustrative of the technical content of the present invention and are not intended to limit it. Even though the present invention has been specifically described herein through preferred examples, those skilled in the art should recognize that adjustments or equivalent substitutions can be made to the technical solutions of the present invention. These changes should not deviate from the core concept and scope of the present invention, and all such adjustments and substitutions should be considered within the scope of protection of the present invention.

Claims

1. An automated bearing defect detection system, characterized in that, It includes a parts supply module, a parts conveying module, a lighting and image acquisition module, a vision inspection module, a control module, and a safety module; The parts supply module uses an automatic rotating tray to supply parts to be inspected; The parts conveying module includes a conveyor belt, a loading robotic arm, a unloading robotic arm, and an operating unit, used to convey the parts to be inspected to the visual inspection area. The lighting and image acquisition module includes a supplementary light source and an image acquisition module, which are used to capture the details of the parts to be inspected in order to meet complex visual recognition needs. The visual inspection module includes a detection algorithm model and an edge computing device, used to detect whether there are defects in bearing parts; The control module includes a main control unit and a control panel, which are used to monitor and control the coordinated operation between various modules, receive and process the detection data from the vision inspection module, and make corresponding control decisions. The safety module includes physical protection, an emergency stop button, and safety sensors to protect the safety of personnel and equipment during the detection process and to detect and respond to potential dangerous situations. The detection algorithm model of the visual detection module includes a model training unit, a feature extraction unit, a defect detection unit, a feature fusion unit, and a detection head unit; The model training unit includes a model selection submodule, a model initialization submodule, a loss function submodule, an optimizer submodule, a batch processing submodule, a regularization submodule, a model evaluation submodule, and a model saving submodule; The feature extraction unit includes a convolutional layer submodule, a pooling layer submodule, a normalization layer submodule, an activation layer submodule, a feature selection submodule, and a feature dimensionality reduction submodule. The defect detection unit includes an image preprocessing submodule, a feature extraction submodule, a defect detection submodule, a defect localization submodule, and a visualization submodule; The detection process of the detection algorithm model is as follows: (1) The image acquisition module acquires the image information of the bearing parts and transmits the image information of the bearing parts to the image preprocessing submodule in the vision detection module. The image preprocessing submodule performs noise reduction on the acquired image information of the bearing parts through a filtering algorithm to reduce the error in feature extraction and model training. (2) The convolutional layer submodule in the feature extraction unit adopts a lightweight model based on YOLOv8 to perform convolution operations on the preprocessed bearing parts image information; in the backbone network of YOLOv8, the convolutional layer of downsampling operation is replaced with the lightweight adaptive convolutional module LACM. The structure of LACM is divided into two paths: one path obtains an adaptive weight, and the other path uses group convolution to reduce the number of parameters. Finally, it is multiplied with the obtained adaptive weight and then added to obtain the feature parameters with emphasis; in the backbone network of YOLOv8, the Bottleneck layer used by C2f is replaced with the multi-scale convolutional module EMC. The structure of EMC divides the input features into two parts: one part goes through 3×3 Conv and 5×5 Conv, and the other part retains the original features without operation; finally, the two parts of features are merged together to perform 1×1 Conv; the bearing contour and texture required for bearing parts defect detection are extracted. (3) The model selection submodule in the model training unit selects the specified lightweight model based on YOLOv8 improvement, uses preprocessed image data and extracted features to train the deep learning model; finally, YOLOv8m is used as the teacher model to perform knowledge distillation on the lightweight model based on YOLOv8 improvement. (4) The model evaluation submodule in the model training unit evaluates the performance of the trained deep learning model. The evaluation metrics include computational accuracy, precision and recall. (5) The visualization sub-module in the defect detection unit is electrically connected to the control panel of the control module to visualize the defect information and generate a bearing part defect information report. The bearing part defect information report includes the number of bearing part defects, the location of the defects, and the type of bearing defects.

2. The automated bearing defect detection system according to claim 1, characterized in that, The automatic rotating tray of the parts supply module works with the robotic arm of the parts conveying module to clamp parts. After clamping a part, the automatic rotating tray rotates once to complete the automated feeding process. The automatic rotating tray includes guide rods, a parts mounting plate, a stepper motor, and a lifting mechanism arranged from top to bottom. The stepper motor controls the rotation angle of the tray, working in conjunction with the loading robotic arm to complete the loading task. The parts mounting plate is equipped with detachable guide rods to fix the bearing parts and prevent them from slipping during rotation, thus completing the clamping task. The guide rods are adaptable to bearing parts of different sizes, and the number of guide rods is selected according to the actual situation. The lifting mechanism adjusts the height of the tray for contact with the worktable.

3. The automated bearing defect detection system according to claim 1, characterized in that, The operation unit of the parts conveying module controls the loading robotic arm to place the parts at designated positions on the conveyor belt after gripping them, arranging a batch of parts in a 3x4 layout. The operation unit is equipped with a coordinate system to guide the loading robotic arm to accurately place the parts at designated positions on the conveyor belt. The unloading robotic arm, controlled by signals from the operation unit, grips and unloads defective parts. The operation unit controls the speed of the conveyor belt to ensure that the parts are stably transported to the inspection area.

4. The automated bearing defect detection system according to claim 1, characterized in that, The lighting and image acquisition module uses a multi-angle light source to reduce shadows and reflections. The lighting intensity is adjusted according to the material and color of the part to obtain the best image quality. The image acquisition module uses a high-resolution camera to capture tiny details on the surface of the part and takes pictures from multiple angles to comprehensively inspect every surface of the part.

5. The automated bearing defect detection system according to claim 1, characterized in that, The model selection submodule is used to select a suitable machine learning or deep learning model structure for the problem; the model initialization submodule is used to initialize the model's weights and biases; the loss function submodule is used to optimize the model's loss function, which is a function used to measure the difference between the model's output and the actual label; the optimizer submodule selects and configures the optimization algorithm used to optimize the model. During training, the batch processing submodule divides the data into mini-batches for calculating gradients and updating model parameters; the regularization submodule adds regularization terms to prevent overfitting; the model evaluation submodule evaluates the model's performance on the training, validation, or test sets; and the model saving submodule saves the trained model weights for later use. The convolutional layer submodule extracts image features using convolution operations; the pooling layer submodule reduces the spatial dimension of the feature map using pooling operations; the normalization layer submodule normalizes the features; the activation layer submodule applies activation functions to introduce nonlinearity and increase the model's expressive power; the feature selection submodule selects the most representative features for subsequent defect detection; and the feature dimensionality reduction submodule reduces the dimensionality of the features. The image preprocessing submodule preprocesses the input image, including resizing, cropping, or enhancing image contrast; the feature extraction submodule extracts features from the preprocessed image. The defect detection submodule uses extracted features to detect defects in images, including methods using classifiers or regressors. The defect location submodule locates and marks the positions of detected defects in the image; the visualization submodule visualizes the detected defect information for users to understand and analyze the detection results.

6. The automated bearing defect detection system according to claim 1, characterized in that, The loss function submodule of the visual inspection module uses a composite loss function. To simultaneously optimize the bounding box position, size, object confidence, and class probability, the loss function... The calculation formula is: In the formula, For coordinate loss, the squared difference loss is used to measure the difference between the predicted bounding box and the true bounding box; For confidence loss, MSE or cross-entropy loss is used to evaluate the accuracy of confidence prediction; For classification loss, cross-entropy loss is used to evaluate the accuracy of class probability prediction; The cross-entropy loss function L is: In the formula, For the number of categories, For observation values, To predict probabilities.

7. The automated bearing defect detection system according to claim 1, characterized in that, The control module uses a PLC as the main control unit of the system to control the rotation of the automatic rotating material tray, the operation of the loading robotic arm, the movement of the conveyor belt, and the operation of the unloading robotic arm. It also uses an industrial PC to run a lightweight deep learning model to process images captured by the image acquisition module, perform defect detection, and feed the results back to the PLC.

8. The automated bearing defect detection system according to claim 1, characterized in that, The physical protection uses safety sensors to monitor the environment around the loading and unloading robotic arms in real time, sets safety limits for the loading and unloading robotic arms, and controls the movement range of the loading and unloading robotic arms to not exceed the preset range. The emergency stop button is used to stop the movement of the loading and unloading robotic arms in case of an emergency. The safety sensor is used to monitor the status parameters of the loading and unloading robotic arms, and to issue an alarm and trigger the emergency stop button in case of abnormal conditions.

9. An automated bearing defect detection method based on the automated bearing defect detection system according to any one of claims 1 to 8, characterized in that, Includes the following steps: Step 1: The parts supply module uses an automatic rotating tray to automatically feed the parts to be inspected. Step 2: The parts conveying module uses a conveyor belt and a loading robotic arm to transport the bearings to the vision inspection area; Step 3: The lighting and image acquisition module takes multi-angle photos of the bearing and sends them to the visual inspection module; Step 4: The visual inspection module uses detection algorithms and edge computing devices to detect whether there are defects in the bearing parts, and sends the detection results to the control module. Step 5: The control module sorts the inspected bearings, transporting defective bearings to the scrap area and defect-free bearings to the unloading area. Step 6: The safety module protects the safety of personnel and equipment during the automated testing process, and promptly detects and responds to potential hazards.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the automated bearing defect detection method as described in claim 9.

Images