Mold part based on intelligent identification and detection control method thereof

By introducing graph convolution network and multi-camera technology into the mold detection system, combining mobile and clamping components, the problem of mold bottom surface detection is solved, efficient and accurate mold surface defect detection is achieved, and detection efficiency and accuracy are improved.

CN118961712AInactive Publication Date: 2024-11-15YANGZHOU FUTIAN YOUBANG VEHICLE IND CO LTD

Patent Information

Application Number
CN202411023867.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-27
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art is difficult to effectively detect and identify the bottom surface of the mold, resulting in inaccuracy and efficiency of the detection results.

Method used

A mold component based on intelligent identification was designed, using graph convolutional network (GCN) algorithm and multi-camera system, combining movement, rotation and clamping components to realize fully automated detection of the top and bottom surfaces of the mold, through image preprocessing and deep learning models identify defects and generate real-time control signals for feedback.

Benefits of technology

It significantly improves the accuracy and efficiency of mold detection, realizes multi-directional detection of mold surface defects, reduces artificial errors, improves the accuracy and efficiency of detection, and simplifies the operation process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118961712A_ABST
    Figure CN118961712A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of mold parts, and discloses a mold part based on intelligent recognition and a detection control method thereof, the mold part comprises an equipment body, and a moving assembly is arranged in the equipment body. The mold component based on intelligent recognition is provided with the fixing frame, the first motor, the first lead screw, the moving frame, the second motor, the second lead screw, the moving block, the first camera and the second camera, the first camera can move front and back through work of the first motor, and the first camera can move left and right through work of the second motor; defects on the surfaces of the top and the bottom of a mold workpiece are detected and recognized through movement of the first camera and the second camera, the problem that the bottom face of the mold is difficult to detect and recognize is solved, the requirements of people are met, the surface defects of the mold part can be detected and recognized in multiple directions, and the detection efficiency is improved through detection of the multiple cameras. Detection efficiency is high and use by people is convenient.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of mold components, and in particular to a mold component based on intelligent identification and a detection and control method thereof. Background Art

[0002] With the rapid development of modern manufacturing industry, molds, as an important tool for industrial production, have a quality that directly determines the final performance of the product. Traditional mold inspection methods mainly rely on manual labor, which is not only inefficient but also easily affected by subjective judgment, making it difficult to ensure the accuracy of the inspection results.

[0003] With the continuous advancement of technologies such as artificial intelligence and computer vision, mold component detection technology based on intelligent recognition has gradually become a research hotspot. Intelligent recognition devices use image recognition technology to detect and identify surface defects of products. The emergence of this technology has greatly improved the efficiency of production operations, avoided the accuracy of detection results affected by operating conditions, subjective judgment, etc., and achieved better and more accurate surface defect detection and faster identification of product surface defects.

[0004] Product surface defect detection is a type of machine vision technology, which uses computer vision to simulate the functions of human vision, collect and process images from specific objects, calculate, and perform actual detection, control, and application.

[0005] The prior art patent document with publication number CN211318230U provides an intelligent identification device for automobile mold parts that is convenient for multi-angle identification. By setting an arc-shaped rack, a housing, a first motor, a second motor, a gear, etc., the housing can be driven to move along the surface of the rack by the meshing transmission of the gear and the rack, and the rack can be driven to rotate as a whole by the first motor, which greatly increases the adjustment angle of the camera, so that the identification device can detect and identify surface defects of automobile mold parts from multiple angles, and the use effect is good.

[0006] Although the above-mentioned prior art can detect and identify the surface defects of the mold from multiple angles, it is difficult for the device to detect and identify the defects on the bottom surface of the mold during use. It still relies on manual observation to see if there are defects on the bottom surface of the mold. Manual observation is not only inefficient, but also easily affected by subjective judgment, resulting in the accuracy of the detection results being difficult to guarantee. Therefore, we need a mold component based on intelligent identification. Summary of the invention

[0007] The object of the present invention is to provide a mold component based on intelligent identification and a detection and control method thereof, so as to solve the problem of difficulty in detecting and identifying the bottom surface of the mold proposed in the above background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: a mold component based on intelligent identification, comprising a device body, a moving component is arranged inside the device body, a rotating component is arranged on the inner wall of the device body, a clamping component is arranged on the outer wall of the rotating component, and a lifting component is arranged on the outer wall of the device body;

[0009] The moving assembly includes a fixed frame, and a first motor is installed on the outer wall of the fixed frame, the output shaft of the first motor is fixedly connected to the first screw rod through a coupling, and the outer wall of the first screw rod is threadedly connected to the moving frame, a second motor is installed on the inner wall of the moving frame, and the output shaft of the second motor is fixedly connected to the second screw rod through a coupling, a moving block is threadedly connected to the outer wall of the second screw rod, and the outer wall of the moving block is connected to the first camera through bolts, and a second camera is arranged inside the device body;

[0010] The rotating assembly comprises a mounting frame, and a rotating groove is formed on the outer wall of the mounting frame, a mounting groove is formed on the inner wall of the mounting frame, and a ball bearing is arranged on the inner wall of the mounting groove, a fixing seat is detachably connected to the inner wall of the mounting groove, and a fixing bolt is arranged on the outer wall of the fixing seat, a rotating block is rotatably connected to the inner wall of the rotating groove, and a storage seat is fixedly connected to the top of the rotating block, a through groove is formed on the outer wall of the storage seat, and a gear ring is fixedly connected to the outer wall of the storage seat, a fixing plate is fixedly connected to the inner wall of the equipment body, and a reduction motor is installed on the outer wall of the fixing plate, the output shaft of the reduction motor is fixedly connected to the rotating rod through a coupling, and a gear is fixedly connected to the outer wall of the rotating rod;

[0011] The clamping assembly includes a fixed block, and the outer wall of the fixed block is provided with a threaded groove, the inner wall of the threaded groove is threadedly connected with a screw, and one end of the screw is fixedly connected with a fixed bearing, and one end of the fixed bearing is fixedly connected with a clamping block;

[0012] The lifting assembly includes a slide groove, the outer wall of the equipment body is fixedly connected to a mounting block, and an electric telescopic rod is installed on the outer wall of the mounting block, one end of the electric telescopic rod is fixedly connected to a connecting plate, and the bottom of the connecting plate is detachably connected to a connecting frame, a circular groove is opened on the outer wall of the connecting frame, and a long rod is slidably connected to the inner wall of the circular groove, one end of the connecting frame is detachably connected to the mounting plate, and the outer wall of the mounting plate is detachably connected to a third camera.

[0013] Preferably, a workpiece is arranged on the top of the storage seat, a lighting lamp is installed inside the equipment body, and an equipment door is hinged on the outer wall of the equipment body.

[0014] Preferably, there are two fixing brackets, and the two fixing brackets are symmetrically arranged in the device body.

[0015] Preferably, the first motor forms a moving structure with the moving frame through a first screw rod, and one end of the first screw rod is threadedly passed through the moving frame.

[0016] Preferably, the second motor forms a moving structure through a second screw rod and a moving block, and one end of the second screw rod is threadedly passed through the moving block.

[0017] Preferably, the mounting frame forms a rotating structure with the rotating block through the rotating groove, and one end of the rotating block extends into the rotating groove for connection.

[0018] Preferably, the fixed plate forms a rotating structure through a reduction motor and a rotating rod, and one end of the reduction motor passes through the fixed plate to be connected to the rotating rod.

[0019] Preferably, the gear teeth of the gear ring mesh with the gear teeth of the gear.

[0020] Preferably, the fixing block forms a threaded structure with the screw rod through the threaded groove, and one end of the screw rod extends into the threaded groove for connection.

[0021] Preferably, the connecting frame forms a sliding structure with the long rod through the circular groove, and one end of the long rod passes through the circular groove.

[0022] By introducing the graph convolutional network (GCN) algorithm, the mold component based on intelligent recognition can significantly improve the accuracy and efficiency of defect detection and realize fully automated detection and feedback control. The specific steps are as follows:

[0023] Step 1: Image data acquisition

[0024] Hardware and sensors and installation:

[0025] Camera: High-resolution industrial camera, the first camera is installed on the outer wall of the moving block, covering the longitudinal area of ​​the mold surface; the second camera is installed inside the equipment body, covering the lateral area of ​​the mold surface;

[0026] Position sensor: Rotary encoder, which measures the displacement of the rotating shaft or linear slider, provides real-time position feedback, and is installed on the motor or slide rail to monitor the movement of the camera; Photoelectric laser displacement sensor, which measures the distance through the laser beam and provides high-precision position information, is used to monitor the precise position of the camera on the slide rail.

[0027] Light sensor: Photoresistor sensor, which changes resistance value according to light intensity, detects ambient light intensity, and is used to adjust the exposure time and gain of the camera; silicon photodiode, which converts light signals into current signals, detects light intensity, and is used to monitor ambient light changes in real time and adjust camera parameters;

[0028] Image preprocessing:

[0029] Image denoising: Perform denoising on the received image data, using Gaussian filtering and mean filtering to remove noise and interference in the image;

[0030] Image enhancement processing: Perform image enhancement processing to adjust the brightness, contrast and sharpness of the image to improve image quality and ensure the accuracy of subsequent defect detection;

[0031] Brightness adjustment: According to the grayscale histogram of the image, the adaptive histogram equalization method is used to improve the contrast of the image;

[0032] Sharpness adjustment: Use the Laplacian operator to enhance the details of the image and highlight the defect features in the image;

[0033] Step 2: Graph Convolutional Network (GCN) and RMSProp Detection Model

[0034] The core operation of the graph convolutional network (GCN) is to convolve the features of the graph nodes. The formula is as follows:

[0035]

[0036] Among them, H (l+1) represents the node feature matrix of the l+1th layer, H (l) represents the node feature matrix of the lth layer, σ is the activation function (ReLU), W (l) is the weight matrix of the lth layer, is the regularized adjacency matrix (the adjacency matrix of the graph plus self-loops), yes The degree matrix of is the inverse square root of the regularized degree matrix;

[0037] Forward propagation derivation of the GCN model:

[0038] Construct the adjacency matrix and degree matrix:

[0039] Adjacency matrix A:

[0040]

[0041] The degree matrix D is a diagonal matrix, where D ii =∑ j A ij , A ij is the element of the adjacency matrix A (indicating the connection relationship between node i and node j), D ii is the diagonal element of the degree matrix D (indicating the degree of node i), ∑ j It means to perform a sum operation on index j;

[0042] Regularized adjacency matrix:

[0043]

[0044]

[0045] in, is the regularized adjacency matrix (representing the adjacency matrix of the graph plus self-loops), A is the original adjacency matrix (representing the connection relationship between nodes in the graph), I is the identity matrix, is the regularized degree matrix The diagonal elements of (representing the degree of node i), j is the neighbor node index, i is the node index, is the regularized adjacency matrix (represents the connection relationship between node i and node j);

[0046] Graph convolution operation:

[0047] For each layer, perform graph convolution operation to convolve the features of each layer of graph nodes;

[0048] Loss calculation and gradient calculation

[0049] Calculate the loss:

[0050] Use the cross entropy loss function L for classification tasks:

[0051]

[0052] Among them, L is the cross entropy loss function, N is the number of samples, C is the number of categories, i is the sample index, c is the category index, and y is the ic is the true label of sample i belonging to category c, is the predicted probability that sample i belongs to category c, log is a logarithmic function;

[0053] Compute the gradient:

[0054] For the loss function L with respect to the weight matrix W (l) Compute the gradient:

[0055]

[0056] in, is the gradient of the weight matrix of the lth layer at the tth iteration, L is the loss function, and W ( l ) is the weight matrix of the lth layer, is the gradient operator;

[0057] RMSProp optimization steps:

[0058] Calculate the current gradient:

[0059]

[0060] Among them, θ t is the parameter value at the tth iteration, g t is the current parameter θ t The gradient under t ) is the loss function, is the gradient operator;

[0061] Compute the exponentially weighted moving average of the squared gradients:

[0062]

[0063] Among them, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, γ is the decay factor, E[g 2 ] t-1 is the exponentially weighted moving average of the squared gradient at the t-1th iteration, g 2 is the square of the current gradient, is the square of the gradient at the tth iteration, g t is the gradient at the tth iteration;

[0064] For each layer's weight matrix W (l) Optimization using RMSProp:

[0065] Update parameters:

[0066]

[0067] Among them, θ t+1 is the parameter value after the t+1th iteration, θ t is the parameter value after the tth iteration, g t is the gradient at the tth iteration, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, g 2 is the square of the gradient, η is the learning rate, and ∈ is a small constant;

[0068] Update the weight matrix:

[0069]

[0070] in, is the weight matrix of the lth layer after the t+1th iteration, is the weight matrix of layer l after the tth iteration, η is the learning rate, is the gradient of the weight matrix of layer l at the tth iteration, E[(g (l) )2 ] t is the exponentially weighted moving average of the squared gradient of the weight matrix of the lth layer at the tth iteration, g (l) is the gradient of the weight matrix of the lth layer, ∈ is a small constant;

[0071] Defect Identification:

[0072] The trained GCN model infers the new image to obtain the defect classification results for each pixel. The recognition results are transmitted to the feedback control unit to adjust the operating parameters of the equipment.

[0073] Step 3: Feedback Control

[0074] Receive recognition results:

[0075] Input data: Output of the defect identification unit, including the type, location and severity of the defect;

[0076] Operation: The feedback control unit receives these recognition results and stores them in an internal data structure;

[0077] Generate control signals:

[0078] Input data: Identify the result data, including the location and severity of defects;

[0079] Operation: The feedback control unit generates corresponding control signals according to the location and type of the defect, and decides whether to adjust the speed and force of the servo motor according to the severity of the defect;

[0080] A control signal indicating the adjustment required for the first motor, controlling the first motor to move forward and backward, so as to adjust the position of the first camera so that the first camera can better observe the defective part;

[0081] A control signal indicates the adjustment that needs to be made to the second motor, and controls the second motor to move left and right, so as to adjust the position of the second camera so that the second camera can better observe the defective part;

[0082] Adjust the electric push rod: control signal, indicate the adjustment that the electric push rod needs to make, control the electric push rod to push and pull action to ensure smooth demoulding of the finished product during the molding process.

[0083] Adjust the camera position and angle: control signals to indicate the camera needs to be adjusted, and control the adjustment of the camera position and angle to obtain clearer images for further analysis;

[0084] Adjust the speed and strength of the servo motor: control signals to indicate the adjustments that need to be made to the servo motor, adjust the speed and strength of the servo motor to ensure a smooth and efficient molding process, and dynamically adjust the parameters of the molding process according to the location and type of defects detected to optimize the quality of the finished product;

[0085] Step 4: Generate reports and store data

[0086] Collect test data: After the test is completed, collect all test result data from the defect identification unit, summarize all defect information, and prepare to generate a report;

[0087] Generate test report: Summarize the test data and create a new test report document, recording the detailed information of each defect in the report;

[0088] Save the test report: format the test report and save it to a local storage device or cloud storage, making sure the file name and path are easy to identify and access, using standard naming conventions;

[0089] Storing test data: Aggregated test data, storing the test data in a database within the device, generating a unique record identifier for each test;

[0090] Database management: existing records and new test data in the database, regular database backup to prevent data loss, database indexing and query functions to facilitate subsequent data analysis and report retrieval, ensure data security and privacy, and control access rights to data.

[0091] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0092] First, the present invention is provided with a fixed frame, a first motor, a first screw rod, a movable frame, a second motor, a second screw rod, a movable block, a first camera and a second camera. The first motor can be operated to realize the forward and backward movement of the first camera, and the second motor can be operated to realize the left and right movement of the first camera. The first camera and the second camera are moved to detect and identify the defects on the top and bottom surfaces of the mold workpiece, which solves the problem of difficulty in detecting and identifying the bottom surface of the mold and meets people's needs. The device can detect and identify the surface defects of the mold parts from multiple directions. Through multiple camera detection, the detection efficiency is high and it is convenient for people to use.

[0093] Second, the present invention is provided with a mounting frame, a rotating groove, a mounting groove, a ball, a fixed seat, a fixing bolt, a rotating block, a storage seat, a through groove, a gear ring, a fixing plate, a reduction motor, a rotating rod and a gear. Through the operation of the reduction motor, the storage seat can drive the clamped and fixed mold workpiece to rotate, which is convenient for the third camera to detect and identify defects on the side of the mold workpiece. The structure is simple and easy to use.

[0094] Third, the present invention is provided with a fixed block, a threaded groove, a screw, a fixed bearing and a clamping block. By rotating the screw, the clamping block can be moved to clamp and fix the mold workpiece, making the mold workpiece more stable when rotating and preventing the mold workpiece from being thrown out.

[0095] Fourth, the present invention is provided with a slide groove, a mounting block, an electric telescopic rod, a connecting plate, a connecting frame, a circular groove, a long rod, a mounting plate and a third camera. The third camera can be raised and lowered by the operation of the electric telescopic rod, which is convenient for adjustment according to the height of the mold workpiece, so that the third camera can better detect and identify defects on the side of the mold workpiece.

[0096] Fifth, the present invention provides a mold component based on intelligent identification, which realizes efficient and accurate detection of mold surface defects by integrating multiple cameras and deep learning algorithms. The system uses graph convolutional networks (GCN) to process and analyze the collected image data, automatically identifies various defects such as cracks and depressions, and significantly improves the accuracy and efficiency of detection. Through the RMSProp optimization method, the rapid convergence and stability of the model during the training process are ensured. The detection results are fed back to the control unit to generate real-time control signals, dynamically adjust the camera position, angle, and speed and strength of the servo motor to ensure the smooth and efficient molding process. After the detection is completed, the system automatically generates a test report and stores the data in the equipment database for subsequent analysis and quality tracking. Compared with traditional manual detection, the present invention significantly reduces human errors, improves detection accuracy and efficiency, reduces the workload of operators, and has extremely high application value and market prospects. By innovatively combining advanced machine learning algorithms with industrial detection technology, the present invention has achieved a breakthrough in intelligent and automated detection in the field of mold manufacturing, and promoted the advancement of industry technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic diagram of the main structure of the present invention;

[0098] Figure 2 It is a schematic diagram of the structure of the fixed frame and the movable frame of the present invention;

[0099] Figure 3 It is a schematic diagram of the structure of the fixing frame and the first motor of the present invention;

[0100] Figure 4 For the present invention Figure 2 The enlarged structural diagram at A in the middle;

[0101] Figure 5 It is a schematic diagram of the structure of the mounting frame and the fixing seat of the present invention;

[0102] Figure 6This is a schematic diagram of the gear ring and gear structure of the present invention;

[0103] Figure 7 This is a schematic diagram of the structure of the connecting frame and the long rod of the present invention;

[0104] Figure 8 This is a flow chart of the deep learning algorithm for mold components based on intelligent identification of the present invention.

[0105] Among them: 1. Equipment body; 2. Moving assembly; 201. Fixed frame; 202. First motor; 203. First screw rod; 204. Moving frame; 205. Second motor; 206. Second screw rod; 207. Moving block; 208. First camera; 209. Second camera; 3. Rotating assembly; 301. Mounting frame; 302. Rotating slot; 303. Mounting slot; 304. Ball bearing; 305. Fixed seat; 306. Fixing bolt; 307. Rotating block; 308. Storage seat; 309. Through slot; 310 , gear ring; 311, fixed plate; 312, reduction motor; 313, rotating rod; 314, gear; 4, clamping assembly; 401, fixed block; 402, threaded groove; 403, screw; 404, fixed bearing; 405, clamping block; 5, lifting assembly; 501, slide groove; 502, mounting block; 503, electric telescopic rod; 504, connecting plate; 505, connecting frame; 506, round groove; 507, long rod; 508, mounting plate; 509, third camera; 6, workpiece; 7, lighting; 8, equipment door. DETAILED DESCRIPTION

[0106] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0107] See also Figure 1-7 , the present invention provides a technical solution:

[0108] Embodiment 1

[0109] A mold component based on intelligent identification includes a device body 1, a moving component 2 is arranged inside the device body 1, a rotating component 3 is arranged on the inner wall of the device body 1, a clamping component 4 is arranged on the outer wall of the rotating component 3, and a lifting component 5 is arranged on the outer wall of the device body 1; the moving component 2 includes a fixed frame 201, and a first motor 202 is installed on the outer wall of the fixed frame 201, the output shaft of the first motor 202 is fixedly connected to a first screw rod 203 through a coupling, and the outer wall of the first screw rod 203 is threadedly connected to a moving frame 204, and a second motor 205 is installed on the inner wall of the moving frame 204, and the output shaft of the second motor 205 is fixedly connected to the first screw rod 203 through a coupling. The device body 1 is fixedly connected with a second screw rod 206, the outer wall of the second screw rod 206 is threadedly connected with a moving block 207, and the outer wall of the moving block 207 is connected with a first camera 208 by bolts, and a second camera 209 is arranged inside the device body 1; the rotating component 3 includes a mounting frame 301, and the outer wall of the mounting frame 301 is provided with a rotating groove 302, the inner wall of the mounting frame 301 is provided with a mounting groove 303, and the inner wall of the mounting groove 303 is provided with a ball 304, the inner wall of the mounting groove 303 is detachably connected with a fixing seat 305, and the outer wall of the fixing seat 305 is provided with a fixing bolt 306, the inner wall of the rotating groove 302 is rotatably connected with a rotating block 307, and The top of the rotating block 307 is fixedly connected with a storage seat 308, the outer wall of the storage seat 308 is provided with a through groove 309, and the outer wall of the storage seat 308 is fixedly connected with a gear ring 310, the inner wall of the equipment body 1 is fixedly connected with a fixing plate 311, and the outer wall of the fixing plate 311 is installed with a reduction motor 312, the output shaft of the reduction motor 312 is fixedly connected with a rotating rod 313 through a coupling, and the outer wall of the rotating rod 313 is fixedly connected with a gear 314; the clamping assembly 4 includes a fixing block 401, and the outer wall of the fixing block 401 is provided with a threaded groove 402, the inner wall of the threaded groove 402 is threadedly connected with a screw 403, and one end of the screw 403 is fixedly connected It is connected to a fixed bearing 404, one end of which is fixedly connected to a clamping block 405; the lifting component 5 includes a slide groove 501, the outer wall of the equipment body 1 is fixedly connected to a mounting block 502, and the outer wall of the mounting block 502 is installed with an electric telescopic rod 503, one end of the electric telescopic rod 503 is fixedly connected to a connecting plate 504, and the bottom of the connecting plate 504 is detachably connected to a connecting frame 505, the outer wall of the connecting frame 505 is provided with a circular groove 506, and the inner wall of the circular groove 506 is slidably connected to a long rod 507, one end of the connecting frame 505 is detachably connected to a mounting plate 508, and the outer wall of the mounting plate 508 is detachably connected to a third camera 509.

[0110] Through the above technical solution, the first camera 208 can be moved forward and backward by the operation of the first motor 202, and can be moved left and right by the operation of the second motor 205. The defects on the top and bottom surfaces of the mold workpiece 6 can be detected and identified by moving the first camera 208 and the second camera 209, which solves the problem of difficulty in detecting and identifying the bottom surface of the mold and meets people's needs; the reduction motor 312 can drive the holder 308 to rotate the clamped mold workpiece 6, so that the third camera 509 can detect the defects on the side of the mold workpiece 6. The detection and identification is carried out with a simple structure and is easy to use; by rotating the screw 403, the clamping block 405 can be moved to clamp and fix the mold workpiece 6, so that the mold workpiece 6 is more stable when rotating and the mold workpiece 6 is prevented from being thrown out; through the operation of the electric telescopic rod 503, the third camera 509 can be raised and lowered, which is convenient for adjustment according to the height of the mold workpiece 6, so that the third camera 509 can better detect and identify defects on the side of the mold workpiece 6. The device can detect and identify surface defects of mold components from multiple directions. Through detection by multiple cameras, the detection efficiency is high and it is easy for people to use.

[0111] Specifically, a workpiece 6 is disposed on the top of the storage seat 308 , a lighting lamp 7 is installed inside the equipment body 1 , and an equipment door 8 is hinged on the outer wall of the equipment body 1 .

[0112] Through the above technical scheme, by providing a lighting lamp 7, good lighting can be achieved, which is convenient for the photo detection and identification work of multiple cameras. Multiple cameras are connected to the external identification device host through connecting lines, which is the existing technology. The host detects, identifies and analyzes the images taken by the cameras. The device is controlled by a controller in the electric control cabinet. A control panel is provided on one side of the equipment body 1, and an observation window is provided on the equipment door 8 for people to observe. The mounting frame 301 is fixed on the inner wall of the equipment body 1, and the fixing block 401 is fixed on the top of the storage seat 308. The slide groove 501 is opened on one side of the equipment body 1 to facilitate the passage of the connecting frame 505.

[0113] Specifically, there are two fixing frames 201 , and the two fixing frames 201 are symmetrically arranged in the device body 1 .

[0114] Through the above technical solution, the fixing frame 201 is fixed on the inner wall of the device body 1. Through the upper and lower settings of the fixing frame 201, the upper moving mechanism facilitates the movement of the second camera 209, and the lower moving mechanism facilitates the movement of the first camera 208.

[0115] Specifically, the first motor 202 forms a moving structure with the moving frame 204 through the first screw rod 203 , and one end of the first screw rod 203 is threadedly penetrated through the moving frame 204 .

[0116] Through the above technical solution, the connection effect between the first motor 202, the first screw rod 203 and the movable frame 204 is strengthened. A bearing is provided at one end of the first screw rod 203. The first screw rod 203 rotates more stably by relying on the bearing. A screw groove is opened on the movable frame 204, and the thread in the screw groove matches the thread on the first screw rod 203.

[0117] Specifically, the second motor 205 forms a moving structure through the second screw rod 206 and the moving block 207 , and one end of the second screw rod 206 is threadedly penetrated through the moving block 207 .

[0118] Through the above technical scheme, the connection effect between the second motor 205 and the second screw rod 206 and the moving block 207 is strengthened. A bearing is arranged at one end of the second screw rod 206. The second screw rod 206 rotates more stably by relying on the bearing. A screw groove is provided on the moving block 207. The thread in the screw groove matches the thread on the second screw rod 206. A groove is also provided on the moving block 207. A limit rod is slidably provided on the inner wall of the groove. When the moving block 207 moves, it moves more stably on the limit rod by relying on the groove.

[0119] Specifically, the mounting frame 301 forms a rotating structure with the rotating block 307 through the rotating groove 302, and one end of the rotating block 307 extends into the rotating groove 302 for connection.

[0120] Through the above technical solution, the connection effect between the mounting frame 301 and the rotating groove 302 and the rotating block 307 is strengthened. When the storage seat 308 rotates, the storage seat 308 drives the rotating block 307 to rotate more stably in the rotating groove 302. The rotating block 307 supports the storage seat 308. When the rotating block 307 rotates, the outer wall friction drives the ball 304 to roll, so that the rotating block 307 rotates more smoothly. There are multiple balls 304, which are arranged in the mounting groove 303 and limited by the fixed seat 305.

[0121] Specifically, the fixed plate 311 forms a rotating structure with the rotating rod 313 through the reduction motor 312 , and one end of the reduction motor 312 passes through the fixed plate 311 to be connected with the rotating rod 313 .

[0122] Through the above technical solution, the connection effect between the fixed plate 311 and the reduction motor 312 and the rotating rod 313 is strengthened. When the reduction motor 312 is working, it can well drive the rotating rod 313 to rotate. A bearing is provided at one end of the rotating rod 313, and the rotating rod 313 rotates more stably relying on the bearing.

[0123] Specifically, the gear teeth of the gear ring 310 mesh with the gear teeth of the gear 314 .

[0124] Through the above technical solution, when the gear 314 rotates, the gear 314 can well drive the gear ring 310 to rotate, and the gear ring 310 rotates to drive the storage seat 308 to rotate.

[0125] Specifically, the fixing block 401 forms a threaded structure with the screw rod 403 through the threaded groove 402, and one end of the screw rod 403 extends into the threaded groove 402 for connection.

[0126] Through the above technical scheme, the connection effect between the fixed block 401 and the threaded groove 402 and the screw 403 is strengthened, the threads in the threaded groove 402 match the threads on the screw 403, and a handle is provided at one end of the screw 403 to facilitate people to better rotate the screw 403. The outer wall of the clamping block 405 is provided with a rubber anti-slip pad, which can clamp the mold workpiece 6 well.

[0127] Specifically, the connecting frame 505 forms a sliding structure with the long rod 507 through the circular groove 506 , and one end of the long rod 507 passes through the circular groove 506 .

[0128] Through the above technical solution, the connection effect between the connecting frame 505, the circular groove 506 and the long rod 507 is strengthened. When the connecting frame 505 is lifted or lowered, the connecting frame 505 moves more stably on the long rod 507 relying on the circular groove 506, and the connecting frame 505 moves more stably in the slide groove 501.

[0129] When in use, firstly, the device is connected to an external power supply, the external power supply supplies power to the device, the illuminating lamp 7 provides lighting, the operator places the mold workpiece 6 on the storage seat 308, and rotates the screw rod 403 to move the clamping block 405 to clamp and fix the mold workpiece 6, and the first motor 202 works to drive the first screw rod 203 to rotate, so that the moving frame 204 moves, the moving frame 204 drives the moving block 207 to move, and the moving block 207 drives the first camera 208 to move, and the second motor 205 works to drive the second screw rod 206 to rotate, so that the moving block 207 moves, and the moving block 207 drives the first camera 208 to move, and by moving the first camera 208 and the second camera 209 forward, backward, left and right, the surface of the top and bottom of the mold workpiece 6 can be inspected and identified for defects, and the reduction motor 312 works to drive the rotating rod 313 to rotate, the rotating rod 313 drives the gear 314 to rotate, the gear 314 drives the gear ring 310 to rotate, and the gear ring 310 drives the storage seat 308 to rotate, and the storage seat 308 is rotated. The seat 308 drives the rotating block 307 to rotate more stably in the rotating groove 302. The rotation of the rotating block 307 drives the ball and 304 to roll, so that the rotating block 307 rotates more smoothly. The third camera 509 detects and identifies the side of the continuously rotating mold workpiece 6, and drives the connecting plate 504 to move through the electric telescopic rod 503. The connecting plate 504 drives the connecting frame 505 to move. The connecting frame 505 relies on the circular groove 506 to move on the long rod 507 more stably. The connecting frame 505 drives the mounting plate 508 to move, and the mounting plate 508 drives the third camera 509 to move. Through the lifting and moving of the third camera 509, the side of the mold workpiece 6 can be well detected and identified, which is convenient for detecting and identifying mold workpieces 6 at different heights, so that the device can detect and identify surface defects of mold components from multiple directions. Through multiple camera detection, the detection efficiency is high and it is convenient for people to use. This completes all the work. The content not described in detail in this manual belongs to the existing technology known to professional and technical personnel in this field.

[0130] like Figure 8 , by introducing the graph convolutional network (GCN) algorithm, the mold component based on intelligent recognition can significantly improve the accuracy and efficiency of defect detection and realize fully automated detection and feedback control. The specific steps are as follows:

[0131] Step 1: Image data acquisition

[0132] Hardware and sensors and installation:

[0133] Camera: High-resolution industrial camera, the first camera is installed on the outer wall of the moving block, covering the longitudinal area of ​​the mold surface; the second camera is installed inside the equipment body, covering the lateral area of ​​the mold surface;

[0134] Position sensor: Rotary encoder, which measures the displacement of the rotating shaft or linear slider, provides real-time position feedback, and is installed on the motor or slide rail to monitor the movement of the camera; Photoelectric laser displacement sensor, which measures the distance through the laser beam and provides high-precision position information, is used to monitor the precise position of the camera on the slide rail.

[0135] Light sensor: Photoresistor sensor, which changes resistance value according to light intensity, detects ambient light intensity, and is used to adjust the exposure time and gain of the camera; silicon photodiode, which converts light signals into current signals, detects light intensity, and is used to monitor ambient light changes in real time and adjust camera parameters;

[0136] Image preprocessing:

[0137] Image denoising: Perform denoising on the received image data, using Gaussian filtering and mean filtering to remove noise and interference in the image;

[0138] Image enhancement processing: Perform image enhancement processing to adjust the brightness, contrast and sharpness of the image to improve image quality and ensure the accuracy of subsequent defect detection;

[0139] Brightness adjustment: According to the grayscale histogram of the image, the adaptive histogram equalization method is used to improve the contrast of the image;

[0140] Sharpness adjustment: Use the Laplacian operator to enhance the details of the image and highlight the defect features in the image;

[0141] Step 2: Graph Convolutional Network (GCN) and RMSProp Detection Model

[0142] The core operation of the graph convolutional network (GCN) is to convolve the features of the graph nodes. The formula is as follows:

[0143]

[0144] Among them, H (l+1) represents the node feature matrix of the l+1th layer, H (l) represents the node feature matrix of the lth layer, σ is the activation function (ReLU), W (l) is the weight matrix of the lth layer, is the regularized adjacency matrix (the adjacency matrix of the graph plus self-loops), yes The degree matrix of is the inverse square root of the regularized degree matrix;

[0145]

[0146]

[0147] Forward propagation derivation of the GCN model:

[0148] Construct the adjacency matrix and degree matrix:

[0149] Adjacency matrix A:

[0150]

[0151] The degree matrix D is a diagonal matrix, where D ii =∑ j A ij , A ij is the element of the adjacency matrix A (indicating the connection relationship between node i and node j), D ii is the diagonal element of the degree matrix D (indicating the degree of node i), ∑ j It means to perform a sum operation on index j;

[0152] Regularized adjacency matrix:

[0153]

[0154]

[0155] in, is the regularized adjacency matrix (representing the adjacency matrix of the graph plus self-loops), A is the original adjacency matrix (representing the connection relationship between nodes in the graph), I is the identity matrix, is the regularized degree matrix The diagonal elements of (representing the degree of node i), j is the neighbor node index, i is the node index, is the regularized adjacency matrix (represents the connection relationship between node i and node j);

[0156] Graph convolution operation:

[0157] For each layer, perform graph convolution operation to convolve the features of each layer of graph nodes;

[0158] Loss calculation and gradient calculation

[0159] Calculate the loss:

[0160] Use the cross entropy loss function L for classification tasks:

[0161]

[0162] Among them, L is the cross entropy loss function, N is the number of samples, C is the number of categories, i is the sample index, c is the category index, and y is the ic is the true label of sample i belonging to category c, is the predicted probability that sample i belongs to category c, log is a logarithmic function;

[0163] Compute the gradient:

[0164] For the loss function L with respect to the weight matrix W (l) Compute the gradient:

[0165]

[0166] in, is the gradient of the weight matrix of the lth layer at the tth iteration, L is the loss function, and W (l) is the weight matrix of the lth layer, is the gradient operator;

[0167] RMSProp optimization steps:

[0168] Let γ = 0.9, small constant ∈ = 10 -8 , the initial value E[g 2 ]0=0

[0169]

[0170] Then the square of the gradient is:

[0171] Calculate the current gradient:

[0172]

[0173] Among them, θ t is the parameter value at the tth iteration, g t is the current parameter θ t The gradient under t ) is the loss function, is the gradient operator;

[0174] Compute the exponentially weighted moving average of the squared gradients:

[0175]

[0176] Among them, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, γ is the decay factor, E[g 2 ] t-1 is the exponentially weighted moving average of the squared gradient at the t-1th iteration, g 2 is the square of the current gradient, is the square of the gradient at the tth iteration, g t is the gradient at the tth iteration;

[0177]

[0178] For each layer's weight matrix W (l) Optimization using RMSProp:

[0179] Update parameters:

[0180]

[0181] Among them, θ t+1 is the parameter value after the t+1th iteration, θ t is the parameter value after the tth iteration, g t is the gradient at the tth iteration, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, g 2 is the square of the gradient, η = 0.01 learning rate, ∈ is a small constant;

[0182] Update the weight matrix:

[0183]

[0184]

[0185] in, is the weight matrix of the lth layer after the t+1th iteration, is the weight matrix of layer l after the tth iteration, η is the learning rate, is the gradient of the weight matrix of layer l at the tth iteration, E[(g (l) ) 2 ] t is the exponentially weighted moving average of the squared gradient of the weight matrix of the lth layer at the tth iteration, g (l) is the gradient of the weight matrix of the lth layer, ∈ is a small constant;

[0186] Defect Identification:

[0187] The trained GCN model infers the new image to obtain the defect classification results for each pixel. The recognition results are transmitted to the feedback control unit to adjust the operating parameters of the equipment.

[0188] Step 3: Feedback Control

[0189] Receive recognition results:

[0190] Input data: Output of the defect identification unit, including the type, location and severity of the defect;

[0191] Operation: The feedback control unit receives these recognition results and stores them in an internal data structure;

[0192] Generate control signals:

[0193] Input data: Identify the result data, including the location and severity of defects;

[0194] Operation: The feedback control unit generates corresponding control signals according to the location and type of the defect, and decides whether to adjust the speed and force of the servo motor according to the severity of the defect;

[0195] A control signal indicating the adjustment required for the first motor, controlling the first motor to move forward and backward, so as to adjust the position of the first camera so that the first camera can better observe the defective part;

[0196] A control signal indicates the adjustment that needs to be made to the second motor, and controls the second motor to move left and right, so as to adjust the position of the second camera so that the second camera can better observe the defective part;

[0197] Adjust the electric push rod: control signal, indicate the adjustment that the electric push rod needs to make, control the electric push rod to push and pull action to ensure smooth demoulding of the finished product during the molding process.

[0198] Adjust the camera position and angle: control signals to indicate the camera needs to be adjusted, and control the adjustment of the camera position and angle to obtain clearer images for further analysis;

[0199] Adjust the speed and strength of the servo motor: control signals to indicate the adjustments that need to be made to the servo motor, adjust the speed and strength of the servo motor to ensure a smooth and efficient molding process, and dynamically adjust the parameters of the molding process according to the location and type of defects detected to optimize the quality of the finished product;

[0200] Step 4: Generate reports and store data

[0201] Collect test data: After the test is completed, collect all test result data from the defect identification unit, summarize all defect information, and prepare to generate a report;

[0202] Generate test report: Summarize the test data and create a new test report document, recording the detailed information of each defect in the report;

[0203] Save the test report: format the test report and save it to a local storage device or cloud storage, making sure the file name and path are easy to identify and access, using standard naming conventions;

[0204] Storing test data: Aggregated test data, storing the test data in a database within the device, generating a unique record identifier for each test;

[0205] Database management: existing records and new test data in the database, regular database backup to prevent data loss, database indexing and query functions to facilitate subsequent data analysis and report retrieval, ensure data security and privacy, and control access rights to data.

[0206] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A mold component based on intelligent recognition, comprising a device body (1), characterized in that: A moving assembly (2) is arranged inside the equipment body (1), a rotating assembly (3) is arranged on the inner wall of the equipment body (1), a clamping assembly (4) is arranged on the outer wall of the rotating assembly (3), and a lifting assembly (5) is arranged on the outer wall of the equipment body (1); The moving assembly (2) comprises a fixed frame (201), and a first motor (202) is installed on the outer wall of the fixed frame (201), the output shaft of the first motor (202) is fixedly connected to a first screw rod (203) via a coupling, and the outer wall of the first screw rod (203) is threadedly connected to a moving frame (204), a second motor (205) is installed on the inner wall of the moving frame (204), and the output shaft of the second motor (205) is fixedly connected to a second screw rod (206) via a coupling, the outer wall of the second screw rod (206) is threadedly connected to a moving block (207), and the outer wall of the moving block (207) is bolted to a first camera (208), and a second camera (209) is arranged inside the device body (1); The rotating assembly (3) comprises a mounting frame (301), wherein the outer wall of the mounting frame (301) is provided with a rotating groove (302), the inner wall of the mounting frame (301) is provided with a mounting groove (303), and the inner wall of the mounting groove (303) is provided with a ball (304), the inner wall of the mounting groove (303) is detachably connected to a fixing seat (305), and the outer wall of the fixing seat (305) is provided with a fixing bolt (306), the inner wall of the rotating groove (302) is rotatably connected to a rotating block (307), and the rotating block (3 07) is fixedly connected to the top of a storage seat (308), a through slot (309) is formed on the outer wall of the storage seat (308), and a gear ring (310) is fixedly connected to the outer wall of the storage seat (308), a fixed plate (311) is fixedly connected to the inner wall of the equipment body (1), and a reduction motor (312) is installed on the outer wall of the fixed plate (311), the output shaft of the reduction motor (312) is fixedly connected to a rotating rod (313) through a coupling, and a gear (314) is fixedly connected to the outer wall of the rotating rod (313); The clamping assembly (4) comprises a fixed block (401), and the outer wall of the fixed block (401) is provided with a threaded groove (402), the inner wall of the threaded groove (402) is threadedly connected with a screw rod (403), one end of the screw rod (403) is fixedly connected with a fixed bearing (404), and one end of the fixed bearing (404) is fixedly connected with a clamping block (405); The lifting assembly (5) comprises a slide groove (501), the outer wall of the device body (1) is fixedly connected with a mounting block (502), and the outer wall of the mounting block (502) is installed with an electric telescopic rod (503), one end of the electric telescopic rod (503) is fixedly connected with a connecting plate (504), and the bottom of the connecting plate (504) is detachably connected with a connecting frame (505), the outer wall of the connecting frame (505) is provided with a circular groove (506), and the inner wall of the circular groove (506) is slidably connected with a long rod (507), one end of the connecting frame (505) is detachably connected with a mounting plate (508), and the outer wall of the mounting plate (508) is detachably connected with a third camera (509).

2. A mold component based on intelligent identification according to claim 1, characterized in that: A workpiece (6) is arranged on the top of the storage seat (308), a lighting lamp (7) is installed inside the equipment body (1), and an equipment door (8) is hinged on the outer wall of the equipment body (1).

3. A mold component based on intelligent identification according to claim 1, characterized in that: The number of the fixing frames (201) is two, and the two fixing frames (201) are symmetrically arranged in the device body (1).

4. The mold component based on intelligent identification according to claim 1, characterized in that: The first motor (202) forms a moving structure through the first screw rod (203) and the moving frame (204), and one end of the first screw rod (203) is threadedly inserted through the moving frame (204).

5. The mold component based on intelligent identification according to claim 1, characterized in that: The second motor (205) forms a moving structure through a second screw rod (206) and a moving block (207), and one end of the second screw rod (206) is threadedly inserted through the moving block (207).

6. The mold component based on intelligent identification according to claim 1, characterized in that: The mounting frame (301) forms a rotating structure with the rotating block (307) through the rotating groove (302), and one end of the rotating block (307) extends into the rotating groove (302) for connection.

7. The mold component based on intelligent identification according to claim 1, characterized in that: The fixed plate (311) forms a rotating structure through a reduction motor (312) and a rotating rod (313), and one end of the reduction motor (312) passes through the fixed plate (311) and is connected to the rotating rod (313).

8. The mold component based on intelligent identification according to claim 1, characterized in that: The gear teeth of the gear ring (310) mesh with the gear teeth of the gear (314).

9. The mold component based on intelligent identification according to claim 1, characterized in that: The fixing block (401) forms a threaded structure with the screw rod (403) through the threaded groove (402), and one end of the screw rod (403) extends into the threaded groove (402) for connection.

10. The mold component based on intelligent identification according to claim 1, characterized in that: The connecting frame (505) forms a sliding structure with the long rod (507) through the circular groove (506), and one end of the long rod (507) passes through the circular groove (506).

11. The method for detecting and controlling mold components based on intelligent identification according to claim 1, characterized in that: Through the intelligent identification of mold parts by the graph convolutional network GCN algorithm, the accuracy and efficiency of defect detection are improved, and fully automated detection and feedback control are achieved. The specific steps are as follows: Step 1: Image data acquisition Camera: High-resolution industrial camera, the first camera is installed on the outer wall of the moving block, covering the longitudinal area of ​​the mold surface; the second camera is installed inside the equipment body, covering the lateral area of ​​the mold surface; Position sensor: Rotary encoder, which measures the displacement of a rotating shaft or linear slider, provides real-time position feedback, and is mounted on a motor or slide rail to monitor the movement of the camera; Photoelectric laser displacement sensor, which measures distance through a laser beam and provides high-precision position information, is used to monitor the precise position of the camera on the slide rail; Light sensor: Photoresistor sensor, which changes resistance value according to light intensity, detects ambient light intensity, and is used to adjust the exposure time and gain of the camera; silicon photodiode, which converts light signals into current signals, detects light intensity, and is used to monitor ambient light changes in real time and adjust camera parameters; Image preprocessing: Image denoising: Perform denoising on the received image data, using Gaussian filtering and mean filtering to remove noise and interference in the image; Image enhancement processing: Perform image enhancement processing to adjust the brightness, contrast and sharpness of the image to improve image quality and ensure the accuracy of subsequent defect detection; Brightness adjustment: According to the grayscale histogram of the image, the adaptive histogram equalization method is used to improve the contrast of the image; Sharpness adjustment: Use the Laplacian operator to enhance the details of the image and highlight the defect features in the image; Step 2: Graph Convolutional Network GCN and RMSProp Detection Model The core operation of the graph convolutional network (GCN) is to convolve the features of the graph nodes. The formula is as follows: Among them, H (l+1) represents the node feature matrix of the l+1th layer, H (l) represents the node feature matrix of the lth layer, σ is the activation function (ReLU), W (l) is the weight matrix of the lth layer, is the regularized adjacency matrix, which represents the adjacency matrix of the graph plus self-loops, yes The degree matrix of is the inverse square root of the regularized degree matrix; Forward propagation of the GCN model: Construct the adjacency matrix and degree matrix: Adjacency matrix A: The degree matrix D is a diagonal matrix, where D ii =∑ j A ij , A ij is an element of the adjacency matrix A, representing the connection relationship between node i and node j, D ii is the diagonal element of the degree matrix D, indicating the degree of node i, ∑ j It means to perform a sum operation on index j; Regularized adjacency matrix: in, is the regularized adjacency matrix (representing the adjacency matrix of the graph plus self-loops), A is the original adjacency matrix, representing the connection relationship between nodes in the graph, I is the unit matrix, is the regularized degree matrix The diagonal elements of represent the degree of node i, j is the neighbor node index, i is the node index, is the regularized adjacency matrix Represents the connection relationship between node i and node j; Graph convolution operation: For each layer, perform graph convolution operation to convolve the features of each layer of graph nodes; Loss calculation and gradient calculation Calculate the loss: Use the cross entropy loss function L for classification tasks: Among them, L is the cross entropy loss function, N is the number of samples, C is the number of categories, i is the sample index, c is the category index, and y is the ic is the true label of sample i belonging to category c, is the predicted probability that sample i belongs to category c, log is a logarithmic function; Compute the gradient: For the loss function L with respect to the weight matrix W (l) Compute the gradient: in, is the gradient of the weight matrix of the lth layer at the tth iteration, L is the loss function, and W (l) is the weight matrix of the lth layer, is the gradient operator; RMSProp optimization steps: Calculate the current gradient: Among them, θ t is the parameter value at the tth iteration, g t is the current parameter θ t The gradient under t ) is the loss function, is the gradient operator; Compute the exponentially weighted moving average of the squared gradients: Among them, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, y is the decay factor, E[g 2 ] t-1 is the exponentially weighted moving average of the squared gradient at the t-1th iteration, g 2 is the square of the current gradient, is the square of the gradient at the tth iteration, g t is the gradient at the tth iteration; For each layer's weight matrix W (l) Optimization using RMSProp: Update parameters: Among them, θ t+1 is the parameter value after the t+1th iteration, θ t is the parameter value after the tth iteration, g t is the gradient at the tth iteration, E[g 2 ] t is the exponentially weighted moving average of the squared gradient at the tth iteration, g 2 is the square of the gradient, η is the learning rate, and ∈ is a small constant; Update the weight matrix: in, is the weight matrix of layer l after iteration t+1, is the weight matrix of layer l after the tth iteration, η is the learning rate, is the gradient of the weight matrix of layer l at the tth iteration, E[(g (l) ) 2 ] t is the exponentially weighted moving average of the squared gradient of the weight matrix of the lth layer at the tth iteration, g (l) is the gradient of the weight matrix of the lth layer, ∈ is a small constant; Defect Identification: The trained GCN model infers the new image to obtain the defect classification results for each pixel. The recognition results are transmitted to the feedback control unit to adjust the operating parameters of the equipment. Step 3: Feedback Control Receive recognition results: Input data: Output of the defect identification unit, including the type, location and severity of the defect; Operation: The feedback control unit receives these recognition results and stores them in an internal data structure; Generate control signals: Input data: Identify the result data, including the location and severity of defects; Operation: The feedback control unit generates corresponding control signals according to the location and type of the defect, and decides whether to adjust the speed and force of the servo motor according to the severity of the defect; A control signal indicating the adjustment required for the first motor, controlling the first motor to move forward and backward, so as to adjust the position of the first camera so that the first camera can better observe the defective part; A control signal indicates the adjustment that needs to be made to the second motor, and controls the second motor to move left and right, so as to adjust the position of the second camera so that the second camera can better observe the defective part; Adjust the electric push rod: control signal, indicate the adjustment required for the electric push rod, control the electric push rod to push and pull, so as to ensure the smooth demoulding of the finished product during the molding process; Adjust the camera position and angle: control signals to indicate the camera needs to be adjusted, and control the adjustment of the camera position and angle to obtain clearer images for further analysis; Adjust the speed and strength of the servo motor: control signals to indicate the adjustments that need to be made to the servo motor, adjust the speed and strength of the servo motor to ensure a smooth and efficient molding process, and dynamically adjust the parameters of the molding process according to the location and type of defects detected to optimize the quality of the finished product; Step 4: Generate reports and store data Collect test data: After the test is completed, collect all test result data from the defect identification unit, summarize all defect information, and prepare to generate a report; Generate test report: Summarize the test data and create a new test report document, recording the detailed information of each defect in the report; Save the test report: format the test report and save it to a local storage device or cloud storage, making sure the file name and path are easy to identify and access, using standard naming conventions; Storing test data: Aggregated test data, storing the test data in a database within the device, generating a unique record identifier for each test; Database management: existing records and new test data in the database, regular database backup to prevent data loss, database indexing and query functions to facilitate subsequent data analysis and report retrieval, ensure data security and privacy, and control access rights to data.

Citation Information

Patent Citations

  • Steel bridge corrosion detection robot and corrosion detection method

    CN117067226A

  • Metal mold online detection method and equipment based on computer vision

    CN117974665A

  • Visual inspection device for appearance of hardware plastic product

    CN217058686U

  • Sheet metal part bending and punching device

    CN220717438U

  • Die steel surface smoothness detection device

    CN220893243U

Cited By

  • Testing device for liquid crystal screen production

    CN119827525A

  • A testing device for liquid crystal screen production

    CN119827525B

  • Stamping die detection device

    CN120901117A

  • Rotation measurement device, measurement method thereof and rotation measurement equipment

    CN121275773A