Plug cord injection molding online real-time detection method

Abstract

According to the plug cord injection molding on-line real-time detection method, an industrial camera array is integrated at key nodes of an injection molding machine, technological parameters of mold clamping force and screw rotating speed are synchronously collected, and a comprehensive signal collection system is constructed. By developing a defect recognition algorithm and combining a copper sheet three-dimensional positioning model and a plastic filling dynamic simulation database, micro-offset detection and early warning of bubble formation are realized. An image feature extraction and quantification technology and a three-dimensional reconstruction and micro offset extraction method are utilized to carry out high-precision analysis on the injection molding process. And a parameter adjustment mapping relation is established based on the defect type and severity, an injection molding parameter PID controller is developed, and online correction of the mold locking force and the screw rotating speed is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of injection molding production, and particularly to an on-line real-time detection method for plug wire injection molding. Background Art

[0002] In the field of injection molding production, with the continuous improvement of the manufacturing industry's requirements for product quality and production efficiency, how to precisely control the injection molding process and promptly discover and solve defects in the production process has become a key issue.

[0003] Currently, in the traditional injection molding production process, there are many deficiencies in the detection and control means for product defects. On the one hand, in the defect detection link, most still rely on the method of manual sampling inspection. Manual sampling inspection is not only inefficient and unable to achieve real-time and comprehensive detection of all products on the production line, but also the detection results are easily affected by subjective factors of the inspectors, such as fatigue, experience differences, etc., resulting in some defective products being difficult to be discovered and removed in a timely manner, thus flowing into the market and affecting the overall quality of the products and the reputation of the enterprise.

[0004] On the other hand, even if product defects are discovered, there are also lack of effective automation and precision means in terms of production parameter adjustment. Existing technologies often adjust the injection molding process parameters based on manual experience, which lacks scientific basis and systematicness and is difficult to quickly and accurately find the optimal parameter adjustment scheme for different types and severity levels of defects. For example, when defects such as micro-offset or bubbles appear in the product, the operator may need to try different combinations of parameters such as clamping force and screw rotation speed multiple times to improve the product quality, which not only wastes a lot of time and raw materials, but also reduces the production efficiency and increases the production cost.

[0005] In summary, the existing injection molding production technologies have obvious deficiencies in the real-time and accuracy of defect detection, as well as the automation and precision of production parameter regulation, which seriously restricts the improvement of the quality of injection molded products and the increase of production efficiency, and urgently requires a new technical solution to solve these problems. Summary of the Invention

[0006] To solve the above problems, the present invention proposes the following specific steps for the on-line real-time detection method of plug wire injection molding, which is characterized in that:

[0007] Step 1: Construct a signal acquisition system for the injection molding process, integrate an industrial camera array at key nodes of the injection molding machine, and install it on both sides of the mold opening and closing axis in a bilateral symmetric layout manner to ensure that the copper sheet positioning holes and the plastic melt flow path are at the best shooting angles, and synchronously collect process parameters such as clamping force and screw rotation speed;

[0008] Step 2: Develop a defect recognition algorithm, establish a 3D positioning model of copper sheets and a dynamic simulation database for plastic filling, and achieve micro-offset detection and early warning of bubble formation through feature comparison of real-time image sequences;

[0009] Step 3: Create an adaptive dynamic regulation strategy library, establish a parameter adjustment mapping relationship based on defect types and severity, develop an injection molding parameter PID controller, and achieve online correction of clamping force and screw speed;

[0010] Step 4: Implement closed-loop control verification and iterative optimization, deploy a hardware-in-the-loop test system composed of binocular cameras and actuators on the production line, establish a defect sample database and a parameter adjustment knowledge base, and achieve rapid response to plug-in head injection molding defects.

[0011] The online real-time detection method for plug wire injection molding of the present invention, beneficial effects, the technical effects of the present invention are as follows:

[0012] 1. The defect recognition algorithm developed by the present invention can accurately detect the micro-offset of copper sheets and achieve early warning of bubble formation through fine steps such as image feature extraction and quantization, 3D reconstruction and micro-offset extraction, and offset calculation and defect determination for real-time injection molding image sequences. Compared with traditional methods that rely on manual experience and simple detection means, this algorithm greatly improves the accuracy and reliability of defect detection, reducing the cases of misjudgment and missed judgment.

[0013] 2. The injection molding parameter PID controller developed by the present invention can achieve precise online correction of clamping force and screw speed by introducing proportional, integral, and derivative control. Each PID loop works independently and cooperatively, ensuring precise control of different process parameters, making the injection molding process more stable, and thus effectively improving the consistency and stability of product quality.

[0014] 3. The technical solution of the present invention realizes the full-process automation and intelligence of the injection molding production process from defect detection to parameter regulation, not only improving product quality and reducing the defective rate, but also significantly enhancing the production efficiency and economic benefits of enterprises by reducing raw material waste and shortening the production cycle. Description of the Drawings

[0015] Figure 1 is the flow chart of the present invention;

[0016] Figure 2 is the visualization diagram of the 3D reconstruction of the present invention;

[0017] Figure 3 is the working schematic diagram of the supporting equipment of the present invention Figure 1 ;

[0018] Figure 4 is the working schematic diagram of the supporting equipment of the present inventionFigure 2 . Detailed implementation manners

[0019] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners:

[0020] The present invention collects signals by symmetrically arranging an industrial camera array on both sides of key nodes of an injection molding machine, develops a defect recognition algorithm to warn of defects, creates a regulation strategy library and a PID controller to adjust parameters, and verifies and optimizes through closed-loop control to achieve rapid detection of injection molding defects and precise regulation of parameters, improving product quality and production efficiency. The flow chart of the invention is as Figure 1 shown, and the operation of the supporting equipment is as Figure 3 and 4 shown. The steps of the present invention will be introduced in detail below.

[0021] Step 1: Construct a signal acquisition system for the injection molding process. Integrate an industrial camera array at key nodes of the injection molding machine, and install it on both sides of the mold opening and closing axis in a bilaterally symmetric layout manner to ensure that the copper sheet positioning holes and the plastic melt flow path are at the best shooting angles, and synchronously collect process parameters such as clamping force and screw rotation speed;

[0022] Step 2: Develop a defect recognition algorithm, establish a three-dimensional positioning model of the copper sheet and a dynamic simulation database for plastic filling, and realize micro-offset detection and early warning of bubble formation through feature comparison of real-time image sequences;

[0023] Step 2.1 Image feature extraction and quantization

[0024] To achieve defect recognition and three-dimensional positioning, it is necessary to perform feature extraction and quantization processing on the real-time injection molding image sequence I(x, y) obtained in Step 1. This step aims to extract the target edge and morphological features, and the process includes four sub-steps: image preprocessing, edge detection, contour extraction, and feature quantization.

[0025] Step 2.1.1 Image preprocessing

[0026] To suppress noise interference during the acquisition process, Gaussian filtering is used to smooth the original image I(x, y). x and y respectively represent the horizontal and vertical pixel coordinates of the image, and the Gaussian kernel function expression is as follows:

[0027]

[0028] Among them, G(x, y) is the Gaussian kernel function, and the smoothed image I s (x, y) is obtained after convolution processing:

[0029]

[0030] Among them, σ is the control filtering intensity, k is the kernel radius, and m and n are the relative offsets of each element in the Gaussian kernel.

[0031] Step 2.1.2 Image Gradient Calculation

[0032] Use the Sobel operator to calculate the image gradients G x in the horizontal direction and G y in the vertical direction:

[0033]

[0034] Calculate the gradient magnitude and direction:

[0035]

[0036] Among them, G is the gradient magnitude and θ is the gradient direction angle.

[0037] Step 2.1.3 Non-Maximum Suppression and Double-Threshold Edge Screening

[0038] To obtain a fine edge map, non-maximum suppression needs to be performed on the gradient magnitude image, that is, retain the maximum points in the gradient direction and suppress the expansion of the edge bandwidth.

[0039] Then, perform double-threshold edge connection by setting a high threshold T h and a low threshold T l : If G(x,y)≥T h , mark it as a "strong edge"; if T l ≤G(x,y)<T h , only retain it when connected to a strong edge; discard other points. Finally, obtain a binary edge map E(x,y)∈{0,1}.

[0040] Step 2.1.4 Feature Vector Quantification

[0041] Extract the contour region according to the edge map E(x,y), calculate its shape and position feature vectors for subsequent three-dimensional positioning and defect comparison of the copper sheet. The selected feature dimensions include:

[0042] Contour area A = ∑E(x,y)

[0043] Centroid coordinates

[0044] Principal direction angle

[0045] Among them, μ pq = ∑x p y q E(x,y), where p and q represent the powers of the pixel coordinates in the horizontal and vertical directions respectively.

[0046] The final output feature vector form is:

[0047] F image = [A, x c , y c , θ main

[0048] Step 2.2 Three-dimensional reconstruction and micro-offset extraction

[0049] In this step, using the image centroid information extracted in Step 2.1, the two-dimensional image data is converted into three-dimensional space coordinates, and compared with the pre-obtained ideal reference position to quantitatively analyze the micro-offset of the copper sheet.

[0050] Step 2.2.1 Camera calibration and projection model establishment

[0051] To achieve an accurate conversion from pixel coordinates to three-dimensional space, first, the Zhang Zhengyou calibration method is used to obtain the camera internal parameter matrix K and the external parameter matrix [R|t], where R is the rotation matrix and t is the translation vector:

[0052]

[0053] Among them, where p = (x, y, 1) T is the point in the image, s is the scale factor, X, Y, Z are the three-dimensional space coordinates, f x and f y are the pixel focal lengths of the camera in the x and y directions, c x and c y are the principal point coordinates. The visualization diagram of the three-dimensional reconstruction is as shown in Figure 2 .

[0054] Step 2.2.2 Obtaining the ideal reference position

[0055] Under the initial state of the mold, the copper sheet image is collected. After the corresponding two-dimensional centroid coordinates are back-projected by binocular vision, the initial three-dimensional coordinates P ref = (X0, Y0, Z0) are calculated.

[0056] Step 2.2.3 Three-dimensional back-projection and real-time reconstruction

[0057] Using the image feature vector F image = [A, x c , y c , θ main extracted in Step 2.1, select the centroid coordinates (x c , y c ) as the main input. According to the camera calibration parameters, a binocular system is used to achieve three-dimensional back-projection:

[0058] ​Let the centroid coordinates in the left and right images be x L and x R , the parallax d = x L - x R , and the depth Z is determined by the following formula:

[0059]

[0060] where B is the baseline distance, f is the physical focal length of the camera, and the three-dimensional coordinates P curr = (X, Y, Z) are obtained by real-time reconstruction through the above formula.

[0061] Step 2.3 Offset calculation and defect determination

[0062] Align the three-dimensional coordinates P curr obtained by real-time reconstruction with the reference position P ref , and calculate the Euclidean distance as the total offset:

[0063]

[0064] where ΔP is the three-dimensional offset of the copper sheet, and X0, Y0, Z0 are the three-dimensional coordinates of the reference position P ref .

[0065] Step 2.3.1 Micro-offset anomaly detection

[0066] Set the micro-offset tolerance δ. When ΔP > δ, it is considered that there is a micro-offset anomaly in the copper sheet.

[0067] Step 2.3.2 Early warning of bubble formation

[0068] Calculate the similarity between the real-time image features and the typical bubble precursor features in the database by using cosine similarity:

[0069]

[0070] where the precursor features F real and F bubble include texture, brightness distribution, and local gradient features. Let the comprehensive judgment function be:

[0071]

[0072] where λ ∈ [0, 1] is the weight coefficient. When D total > ε threshold, the system considers that there is a greater defect risk and issues a warning.

[0073] Step 3: Create an adaptive dynamic regulation strategy library, establish a parameter adjustment mapping relationship based on the defect type and severity, develop an injection molding parameter PID controller, and realize the online correction of the clamping force and screw speed;

[0074] Step 3.1 Establishment of Defect Type - Parameter Mapping Relationship

[0075] According to Steps 2.3.1 and 2.3.2, the system identifies two types of defect indicators: micro - offset ΔP, defect risk function D total ; and establish the following multivariable control mapping:

[0076]

[0077] Step 3.2 PID Controller Design

[0078] To achieve precise online parameter regulation, a PID controller is introduced, and its control formula is:

[0079]

[0080] where e(t) is the control error, that is, the difference between the reference value and the current measured value, and t is the time parameter. For micro - offset, let e(t)= - ΔP, and for bubble characteristics, let e(t)=D target -D total , D target is the reference bubble characteristic; K p , K i , K d are the gain coefficients of proportional, integral, and differential respectively.

[0081] In this control framework, independent PID loops are constructed for different process parameters. Each PID loop works independently, and the final output is synthesized for overall process parameter adjustment:

[0082] Clamping force control loop, using the detected micro - offset e ΔP (t) to generate a control signal u F (t) acting on the torque value of the clamping force torque motor to dynamically adjust the clamping force F lock ;

[0083] Screw speed control loop, based on the bubble risk index to calculate the error e D (t)=D target -D total (t), output the control signal u w (t) acting on the speed value of the screw speed motor to correct the screw speed w screw .

[0084] Step 4: Implement closed - loop control verification and iterative optimization. Deploy a hardware - in - the - loop test system composed of a binocular camera and an actuator on the production line, establish a defect sample database and a parameter adjustment knowledge base, and achieve rapid response to plug - in head injection molding defects.

[0085] The above are only the preferred embodiments of the present invention, and are not any other form of limitation to the present invention. Any modification or equivalent change made according to the technical essence of the present invention still belongs to the scope of protection required by the present invention.

Claims

1. The on-line real-time detection method for the injection molding of plug wires is characterized by the following specific steps: Step 1: Construct a signal acquisition system for the injection molding process. Integrate an industrial camera array at the key nodes of the injection molding machine, and install it on both sides of the mold opening and closing axis in a bilateral symmetric layout manner to ensure that the copper sheet positioning holes and the plastic melt flow path are at the best shooting angles, and synchronously collect the process parameters of the clamping force and the screw speed. Step 2: Develop a defect recognition algorithm, establish a three-dimensional positioning model of the copper sheet and a dynamic simulation database for plastic filling, and realize the micro-offset detection and early warning of bubble formation through the feature comparison of real-time image sequences. Step 3: Create an adaptive dynamic regulation strategy library, establish a parameter adjustment mapping relationship based on the defect type and severity, develop a PID controller for injection molding parameters, and realize the on-line correction of the clamping force and the screw speed. Step 4: Implement closed-loop control verification and iterative optimization. Deploy a hardware-in-the-loop test system composed of binocular cameras and actuators on the production line, establish a defect sample database and a parameter adjustment knowledge base, and realize the rapid response to the injection molding defects of the plug head.

2. The on-line real-time detection method for plug wire injection molding according to claim 1, characterized in that: The development of the defect recognition algorithm in Step 2 can be expressed as: Step 2.1 Image feature extraction and quantization To realize defect recognition and three-dimensional positioning, it is necessary to perform feature extraction and quantization processing on the real-time injection molding image sequence I(x, y) obtained in Step 1; this step aims to extract the target edge and morphological features, and the process includes four sub-steps: image preprocessing, edge detection, contour extraction and feature quantization. Step 2.1.1 Image preprocessing To suppress the noise interference in the acquisition process, the original image I(x, y) is smoothed by Gaussian filtering. x and y respectively represent the horizontal and vertical pixel coordinates of the image, and the Gaussian kernel function expression is as follows: where \(G(x,y)\) is the Gaussian kernel function, and the smoothed image \(I\) is obtained after convolution processing s (x,y): where σ is the control filter strength, k is the kernel radius, and m and n are the relative offsets of each element in the Gaussian kernel. Step 2.1.2 Image gradient calculation Calculate the gradients of the image in the horizontal direction \(G\) and the vertical direction \(G\) using the Sobel operator: x in the vertical direction \(G\) y : Calculate the gradient amplitude and direction: where G is the gradient amplitude and θ is the gradient direction angle. Step 2.1.3 Non-maximum suppression and double-threshold edge screening To obtain a fine edge map, non-maximum suppression needs to be performed on the gradient magnitude image, that is, the maximum points are retained in the gradient direction to suppress the expansion of the edge bandwidth; then by setting a high threshold T h and a low threshold T l for double-threshold edge connection: if G(x,y)≥T h , it is marked as a "strong edge"; if T l ≤G(x,y)<T h , it is only retained when connected to a strong edge; other points are discarded; finally, a binary edge map E(x,y)∈{0,1} is obtained; Step 2.1.4 Feature vector quantization According to the edge map E(x, y), extract the contour area and calculate its shape and position feature vectors for subsequent three-dimensional positioning of the copper sheet and defect comparison. The selected feature dimensions include: Contour area A = ∑E(x, y) Centroid coordinates Main direction angle where μ pq = ∑x p y q E(x, y), p and q respectively represent the powers of pixel coordinates in the horizontal and vertical directions; The final output feature vector form is: F image = [A, x c , y c , θ main ​ Step 2.2 Three-dimensional reconstruction and micro-offset extraction In this step, using the centroid information of the image extracted in Step 2.1, convert the two-dimensional image data into three-dimensional space coordinates, and compare it with the pre-obtained ideal reference position to quantitatively analyze the micro-offset of the copper sheet. Step 2.2.1 Camera calibration and projection model establishment To achieve an accurate conversion from pixel coordinates to three-dimensional space, first use the Zhang-Zhengyou calibration method to obtain the camera internal parameter matrix K and the external parameter matrix [R|t]. R is the rotation matrix and t is the translation vector: Among them, where \(p=(x,y,1)\) T is a point in the image, \(s\) is the scale factor, \(X\), \(Y\), \(Z\) are three-dimensional space coordinates, \(f\) x and \(f\) y are the pixel focal lengths of the camera in the \(x\) and \(y\) directions, \(c\) x and \(c\) y are the principal point coordinates; Step 2.2.2 Obtain the ideal reference position Collect the copper sheet image in the initial state of the mold. After the corresponding two-dimensional centroid coordinates are back-projected by binocular vision, calculate the initial three-dimensional coordinates P ref =(X0, Y0, Z0); Step 2.2.3 Three-dimensional back-projection and real-time reconstruction Using the image feature vector F extracted in step 2.1 image = [A, x c , y c , θ main , select the centroid coordinates (x c , y c ) as the main input; according to the camera calibration parameters, use the binocular system to achieve three-dimensional back-projection: Let the centroid coordinates in the left and right images be x L and x R , the disparity d = x L - x R , and the depth Z is determined by the following formula: Among them, B is the baseline distance, f is the physical focal length of the camera, and the three-dimensional coordinates P curr =(X, Y, Z) obtained by real-time reconstruction through the above formula; Step 2.3 Offset calculation and defect determination Align the three-dimensional coordinates P obtained by real-time reconstruction curr with the reference position P ref and calculate the Euclidean distance as the total offset: Among them, ΔP is the three-dimensional offset of the copper sheet, and X0, Y0, and Z0 are the three-dimensional coordinates of the reference position P ref ; Step 2.3.1 Micro-offset anomaly detection Set the micro-offset tolerance δ. When ΔP > δ, it is considered that there is a micro-offset anomaly in the copper sheet. Step 2.3.2 Early Warning of Bubble Formation The similarity between the real-time image features and the typical bubble precursor features in the database is calculated by using cosine similarity: Among them, the precursor feature F real and F bubble include texture, brightness distribution, and local gradient features. Let the comprehensive decision function be: Among them, λ ∈ [0, 1] is the weight coefficient; when D total > ε threshold, the system considers that there is a greater risk of defects and issues a warning.

3. The in-line real-time detection method for plug wire injection molding according to claim 1, characterized in that: The online correction of the clamping force and the screw speed in Step 3 can be expressed as follows: Step 3.1 Establishment of Defect Type-Parameter Mapping Relationship According to steps 2.3.1 and 2.3.2, the system identifies two types of defect metrics: the micro-offset ΔP and the defect risk function D total ; and establishes the following multivariate control mapping: Step 3.2 PID Controller Design To achieve precise online parameter regulation, a PID controller is introduced, and its control formula is: Among them, e(t) is the control error, that is, the difference between the reference value and the current measured value, and t is the time parameter; for micro-offset, let e(t) = -ΔP, and for bubble characteristics, let e(t) = D target -D total , D target is the reference bubble characteristic; K p , K i , K d are the gain coefficients of proportional, integral and differential respectively; In this control framework, independent PID loops are constructed for different process parameters. Each PID loop works independently, and the final output is synthesized for the adjustment of the overall process parameters: The clamping force control circuit uses the detected micro-offset e ΔP (t) to generate a control signal u F (t) acts on the torque value of the clamping force torque motor to dynamically adjust the clamping force F lock ; Screw rotation speed control loop, calculating the error e based on the bubble risk index D (t) = D target -D total (t), outputting the control signal u w (t) acts on the screw rotation speed motor speed value to correct the screw rotation speed w screw .

Images