Intelligent detection control method for injection molding defects of cooperative robot

Abstract

The invention belongs to the technical field of intelligent control, and particularly relates to a cooperative robot injection molding defect intelligent detection control method which comprises the following steps: acquiring surface point cloud data of an injection molding part through a three-dimensional vision sensor; acquiring injection molding process parameters in real time, wherein the injection molding process parameters comprise mold temperature, mold locking force and melt flow velocity; a depth support vector data description model is adopted to calculate an abnormal score, and when the abnormal score exceeds a threshold value, defect classification is triggered; classifying the defects through the meta-learning prototype network model and outputting a process adjustment suggestion; the adjustment suggestion is fed back to an injection molding machine control system to complete closed-loop control; according to the scheme, a grading processing mechanism from bimodal acquisition and defect classification from anomaly detection is realized, a closed-loop process adjustment system is formed, namely, a closed loop from detection to classification, suggestion and execution is formed, the technical problems of defect identification lagging and process adjustment isolation in a traditional detection method are solved, and the detection efficiency is improved. And intelligent tracing adjustment control of early-stage processing parameters is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and particularly to an intelligent detection and control method for injection molding defects of a collaborative robot. Background Art

[0002] With the rapid development of the automotive manufacturing industry, the quality requirements for automotive plastic injection parts are getting higher and higher. As an important part of automobiles, the quality of automotive plastic injection parts directly affects the performance and safety of automobiles. Therefore, defect detection of automotive plastic injection parts is an important link to ensure the quality of automobiles, but the existing defect detection systems for automotive plastic injection parts cannot meet the growing needs.

[0003] The existing defect detection systems for automotive plastic injection parts have the following problems: 1) Traditional visual inspection can only identify defects but cannot trace the source; 2) Die debugging depends on manual experience, and each die modification causes a production stop of 3 - 5 hours on average.

[0004] The existing solutions have problems such as the disconnection between detection and process parameters, and the repeated investment in multi - process equipment. Therefore, it is necessary to propose a new solution to solve these problems. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent detection and control method for injection molding defects of a collaborative robot to solve the problems raised in the background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: An intelligent detection and control method for injection molding defects of a collaborative robot, comprising: Collecting the surface point cloud data of the injection molded part through a three - dimensional vision sensor; Obtaining the injection molding process parameters in real - time, including mold temperature, clamping force, and melt flow rate; Calculating the anomaly score using a deep support vector data description model, and triggering defect classification when it exceeds the threshold; Classifying the defects through a meta - learning prototype network model and outputting process adjustment suggestions; Feeding back the adjustment suggestions to the injection molding machine control system to complete the closed - loop control.

[0007] For the intelligent detection and control method for injection molding defects of a collaborative robot described in the present invention, wherein, the vector data description model is Deep SVDD, and its training process includes: Constructing a three - dimensional convolutional neural network to process the point cloud data; Mapping the process parameters into a conditional vector through a fully - connected layer; Dynamically adjusting the center of the hypersphere: , Among them, C0 is a multi-dimensional initial center vector, W is a multi-dimensional learnable weight matrix, and P t is a multi-dimensional process parameter vector.

[0008] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the dimensions of the multi-dimensional initial center vector, the multi-dimensional learnable weight matrix, and the multi-dimensional process parameter vector are all 32 dimensions, and the process parameters are preprocessed by Min-Max normalization.

[0009] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the meta-learning prototype network model is ProtoNet, and the defect classification process of the ProtoNet includes: Calculating the class prototype center: ; Calculating the classification probability based on the Euclidean distance: Among them, S K represents the set of samples of the k-th class, is the set cardinality, is the feature extraction function.

[0010] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the dynamic update of the class prototype center adopts: , where A is an attenuation coefficient of 0.7 - 0.9.

[0011] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the generation of the process adjustment suggestion includes Establishing a defect type - process parameter mapping library; When it is identified as a flash defect, an adjustment instruction of "increasing the clamping pressure by 10 - 15%" is automatically generated.

[0012] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the construction of the mapping library includes Recording the mold temperature deviation and pressure fluctuation coefficient corresponding to the historical defect samples; Screening the key influencing parameters through the Pearson correlation coefficient.

[0013] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the sampling accuracy of the three-dimensional vision sensor ≤ 0.05 mm, and the time synchronization error with the process parameter acquisition device < 10 ms.

[0014] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, among them, the intelligent detection and control method of injection molding defects of the collaborative robot further includes: Establishing a defect traceability log in the MES system; Trigger a secondary alarm for continuously occurring similar defects.

[0015] For the intelligent detection and control method of injection molding defects of the collaborative robot described in the present invention, wherein the secondary alarm includes: When the same mold continuously produces more than 3 repeated defects, a maintenance prompt is automatically pushed.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This solution realizes a hierarchical processing mechanism from dual-modal acquisition and anomaly detection to defect classification, forming a closed-loop process adjustment system, that is, forming a closed-loop from detection to classification, suggestion, and execution, solving the technical pain points of lagging defect recognition and isolated process adjustment in traditional detection methods, realizing intelligent source adjustment control of early processing parameters, avoiding the disadvantages that the parameter adjustment of mold equipment in the traditional production process needs to be stopped for manual detection before it can be completed, and significantly improving production efficiency. Brief Description of the Drawings

[0017] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of the method steps of the present invention. Detailed Embodiments

[0019] The terms "first", "second", "third", and "fourth" in the description and claims of the present invention and the drawings thereof are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0020] Referring to "embodiments" herein means that a specific feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0021] "A plurality of" means two or more. "And / or" describes the relationship between associated objects and indicates that there can be three relationships. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0022] Moreover, terms indicating directions such as "upper, lower, left, right, upper end, lower end, longitudinal" etc. are all referenced based on the posture position of the device or equipment described in this solution during normal use.

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0024] This embodiment discloses a collaborative robot injection molding defect intelligent detection and control method as shown in Figure 1 and includes: Step a: Collect the surface point cloud data of the injection molded part through a three-dimensional vision sensor; Step b: Obtain the injection molding process parameters in real time, including mold temperature, clamping force, and melt flow rate; Step c: Use a deep support vector data description model to calculate the anomaly score, and trigger defect classification when it exceeds the threshold; Step d: Classify the defects through a meta-learning prototype network model and output process adjustment suggestions; Step e: Feed back the adjustment suggestions to the injection molding machine control system to complete the closed-loop control.

[0025] This solution realizes a hierarchical processing mechanism from dual-modal acquisition and anomaly detection to defect classification, forming a closed-loop process adjustment system, that is, forming a closed-loop from detection to classification, suggestion, and execution, solving the technical pain points of lagging defect identification and isolated process adjustment in traditional detection methods, realizing intelligent retroactive adjustment control of early processing parameters, avoiding the drawbacks that parameter adjustment of mold equipment in the traditional production process requires manual detection after shutdown, and significantly improving production efficiency.

[0026] In this embodiment, the vector data description model is Deep SVDD, and its training process includes: Construct a three-dimensional convolutional neural network to process the point cloud data; Map the process parameters to a conditional vector through a fully connected layer; Dynamically adjust the center of the hypersphere: , Among them, C0 is the multidimensional initial center vector, specifically 256 dimensions; W is the multidimensional learnable weight matrix, specifically the dimension is ;P t It is a multidimensional process parameter vector, specifically six dimensions (temperature / pressure / speed, etc., normalized values). By dynamically correcting the center of the hypersphere through process parameters, the false alarm problem of traditional single-function Deep SVDD in variable-condition injection molding production is solved. By dynamically correcting the core parameters of the detection model through process parameters (mold temperature / clamping force, etc.), the anomaly score calculation is adaptively adjusted according to production conditions, reducing the measured false detection rate (compared with the fixed-center model).

[0027] In this embodiment, the dimensions of the multidimensional initial center vector, the multidimensional learnable weight matrix, and the multidimensional process parameter vector are all 32 dimensions, and the process parameters are preprocessed by Min-Max normalization; The data preprocessing process is as follows: (1) Eliminate outliers exceeding ±3σ, where σ is the pressure fluctuation coefficient; (2) Calculate normalization based on feature columns to eliminate dimensional interference and improve the convergence speed of the weight matrix W; Limit the encoding dimension and preprocessing method of process parameters to ensure that high-dimensional features fully represent the process status (32 dimensions balance information volume and computational efficiency).

[0028] In this embodiment, the meta-learning prototype network model is ProtoNet, and the defect classification process of ProtoNet includes: Computing Prototype Center: ; Calculate the classification probability based on Euclidean distance: ; Among them, S K Represents the k-th class sample set, which is the minimum sample size of a single class and the sample size is greater than 50; is the number of samples (cardinality) of the K-th class support set, which must satisfy ≥50 to ensure statistical validity; is a feature extraction function used to output the input sample X (d-dimensional feature vector, d is a positive integer); For the feature extraction function of the support set sample X i The output result is, represents the i-th sample of the K-th class, The output of is a d-dimensional feature vector (typical dimension d = 128–4096); C K is the prototype center vector of the Kth class, with the same dimension as are the same, and the value range of i is: i ∈ {1, 2,..., |S K |} (support set sample index); d(x, y) represents the Euclidean distance function, and -d(x, y) is the negative Euclidean distance metric. The larger this value is, the higher the similarity between sample x and the class center C K is; K' is the summation index, traversing all categories (K' ∈ {1, 2,..., N}), and the value range is also K ∈ {1, 2,..., N}; Construct prototypes through the mean value of the support set samples of the meta-learning prototype network model to solve the problem of scarce samples in new defect categories.

[0029] In this embodiment, the dynamic update of the class prototype center adopts: ; Among them, A is an attenuation coefficient of 0.7 - 0.9, and specifically, 0.85 is the best; C K’ is the updated prototype center vector of the Kth class, and C K is the prototype center vector of the Kth class before update; is the feature vector of the new sample X new (the dimension is the same as that of the class prototype center C K ); X new is the 4096-dimensional feature of the new sample.

[0030] Among them, C K and P(Y = K, X) and C K’ In these three models, K represents the class label index (classindex), referring to the target class number being calculated or updated. Specifically, in the clustering task, K identifies the Kth cluster, and in the classification task, K identifies the Kth class, and the value range is K ∈ {1, 2,..., N} (N is the total number of classes).

[0031] In this embodiment, the generation of the process adjustment suggestion includes: Establish a defect type - process parameter mapping library; When it is identified as a flash defect, an adjustment instruction of "increase the clamping pressure by 10 - 15%" is automatically generated.

[0032] In this embodiment, the construction of the mapping library includes Recording the mold temperature deviation and pressure fluctuation coefficient corresponding to historical defect samples; Screening key influencing parameters through the Pearson correlation coefficient.

[0033] In this embodiment, the sampling accuracy of the 3D vision sensor is ≤0.05 mm, and the time synchronization error with the process parameter acquisition device is <10 ms. The micron-level point cloud accuracy can capture subtle defects such as sink marks and flash, while the millisecond-level synchronization can ensure the spatio-temporal consistency of process parameters and point clouds, significantly reducing the data alignment error rate.

[0034] In this embodiment, the intelligent detection and control method for injection molding defects of the collaborative robot further includes: Establishing a defect traceability log in the MES system; Triggering a secondary alarm for continuously occurring similar defects; Specifically, the secondary alarm includes: When the same mold continuously produces more than 3 repeated defects, automatically push a maintenance prompt; By recording the defect-process-mold association data in the log, the traceability analysis time can be greatly shortened, and the secondary alarm triggered by 3 consecutive similar defects (such as the maintenance of the cooling water channel) can prevent batch quality accidents, significantly reducing the equipment downtime rate.

[0035] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. An intelligent detection and control method for injection molding defects of a collaborative robot, characterized in that, Including: Collecting the surface point cloud data of the injection molded part through a 3D vision sensor; Obtaining the injection molding process parameters in real time, including mold temperature, clamping force, and melt flow rate; Calculating the anomaly score using a deep support vector data description model, and triggering defect classification when the threshold is exceeded; Classifying the defects through a meta-learning prototype network model and outputting process adjustment suggestions; Feeding back the adjustment suggestions to the injection molding machine control system to complete closed-loop control.

2. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 1, characterized in that The vector data description model is Deep SVDD, and its training process includes: Constructing a 3D convolutional neural network to process the point cloud data; Mapping the process parameters to a conditional vector through a fully connected layer; Dynamic adjustment of the hypersphere center: , Among them, C0 is a multi-dimensional initial center vector, W is a multi-dimensional learnable weight matrix, and P t is a multi-dimensional process parameter vector.

3. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 2, characterized in that The dimensions of the multi-dimensional initial center vector, the multi-dimensional learnable weight matrix, and the multi-dimensional process parameter vector are all 32-dimensional, and the process parameters are preprocessed by Min-Max normalization.

4. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 1, wherein, The meta-learning prototype network model is ProtoNet, and the defect classification process of the ProtoNet includes: Computing class prototype center: ; Calculate the classification probability based on the Euclidean distance: where S K represents the set of samples of the k-th class, is the set cardinality, is the feature extraction function.

5. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 4, wherein The dynamic update of the class prototype center adopts: , where A is an attenuation coefficient of 0.7 - 0.

9.

6. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 1, characterized in that, The generation of the process adjustment suggestions includes Establishing a defect type - process parameter mapping library; When flash defect is identified, automatically generating an adjustment instruction of "increase the clamping pressure by 10 - 15%".

7. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 6, characterized in that The construction of the mapping library includes Recording the mold temperature deviation and pressure fluctuation coefficient corresponding to the historical defect samples; Screening the key influencing parameters through the Pearson correlation coefficient.

8. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 1, wherein The sampling accuracy of the 3D vision sensor is ≤0.05 mm, and the time synchronization error with the process parameter acquisition device is <10 ms.

9. The intelligent detection and control method for injection molding defects of a collaborative robot according to any one of claims 1-8, characterized in that, The intelligent detection and control method for injection molding defects of the collaborative robot also includes: Establishing a defect traceability log in the MES system; Triggering a secondary alarm for continuously occurring similar defects.

10. The intelligent detection and control method for injection molding defects of a collaborative robot according to claim 9, characterized in that The secondary alarm includes: When the same mold generates more than 3 repeated defects continuously, automatically pushing a maintenance prompt.

Images