System and method for producing glue injection industrial brush by using injection molding machine

Abstract

The invention discloses a system and a method for producing a glue injection industrial brush by using an injection molding machine, and relates to the technical field of industrial brush processing, a multi-parameter coupling index is generated according to acquired working parameters, and the working parameters, the multi-parameter coupling index and a batch identifier in a corresponding time period are bound; the method comprises the following steps: performing multi-angle imaging on brushes, adding mark information sections, performing preprocessing, extracting candidate areas which may have defects, after detecting a plurality of defect areas, constructing a morphological synergy index to quantify the overall defect degree, and obtaining defect data of each brush; after a comprehensive index structure is constructed, an injection molding process feature vector is constructed, kernel mapping and an AWCCA method are utilized to calculate nonlinear correlation and local sensitivity between each working parameter and a form cooperation index, a working parameter interval is formulated, and a parameter modification suggestion is generated; in the production process, parameters are continuously corrected in a small-stride mode, the defect degree of the brush and the stability of the injection molding process are balanced, and more flexible regulation and control are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial brush processing, and specifically to a system and method for producing glue-injected industrial brushes using an injection molding machine. Background Art

[0002] Currently, industrial brush production mainly adopts an injection molding process, where glue or thermoplastic materials are injected into pre-arranged brush filaments to achieve one-time molding. To meet the requirements for durability and consistency of brushes in cleaning, painting, and other special fields, the bonding quality between the glue and the brush filaments is extremely high during the production process. For this reason, modern production lines have introduced multi-sensor data acquisition, machine vision inspection, and data analysis technologies to monitor the temperature, pressure, and flow rate of key parts of the injection molding machine in real time, and at the same time use image processing technology to detect defects in the molded products, realizing the full-process tracking and analysis of process data and product quality.

[0003] In the existing injection molding process, due to the dynamic fluctuations of parameters such as temperature, pressure, and flow rate, the interaction of multiple parameters is difficult to be captured and quantified in real time, resulting in uneven bonding between the glue and the brush filaments, manifested as loose brush filaments or incomplete glue coating. This problem mainly stems from the complex influence of process parameter fluctuations on the flow and curing state of the glue, and it is difficult for the existing technology to effectively associate sensor data with visual inspection results, and thus it is impossible to accurately identify and automatically regulate the key process parameters causing defects, thereby seriously affecting product quality and production efficiency.

[0004] For this reason, the present invention provides a system and method for producing glue-injected industrial brushes using an injection molding machine. Summary of the Invention

[0005] (1) Technical Problems to be Solved

[0006] Aiming at the deficiencies of the existing technology, the present invention provides a system and method for producing glue-injected industrial brushes using an injection molding machine. After multi-angle imaging of the brush and adding marked information, preprocessing is performed, candidate regions that may have defects are extracted. After detecting multiple defect regions, a morphological cooperation index is constructed to quantify the overall defect degree, and defect data of each brush is obtained; after constructing a comprehensive index structure, a feature vector of the injection molding process is constructed, and the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index are calculated using kernel mapping and the AWCCA method, a working parameter interval is formulated and parameter modification suggestions are generated; during the production process, small-step and continuous corrections are made to the parameters to balance the defect degree of the brush and the smoothness of the injection molding process, achieving more flexible regulation; thus solving the technical problems recorded in the background art.

[0007] (2) Technical Solutions

[0008] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0009] A method for producing injection - molded industrial brushes using an injection molding machine, including

[0010] Collecting the working parameters of the injection molding machine by a sensor array arranged at key parts of the injection molding machine, generating a multi - parameter coupling index MCI(t) from the obtained working parameters, and binding the working parameters, the multi - parameter coupling index MCI(t) and the batch identifier within the corresponding time period;

[0011] After performing multi - angle imaging on the brush and adding marker information, pre - processing is carried out, candidate areas where defects may exist are extracted. After detecting multiple defect areas, a morphological synergy index WPMSI is constructed to quantify the overall defect degree, and defect data of each brush is obtained;

[0012] After constructing a comprehensive index structure, an injection - process feature vector X is constructed. Using kernel mapping and the AWCCA method, the non - linear correlation and local sensitivity between each working parameter and the morphological synergy index WPMSI are calculated, and then a working - parameter interval is formulated and a parameter modification suggestion is generated;

[0013] Read the working - parameter modification list in the index structure, superimpose it with the original parameter set to form a new set value, automatically send it to the injection molding machine through the PLC, and online monitor and adaptively correct the working parameters.

[0014] Furthermore, a sensor array is arranged at key parts of the injection molding machine to obtain real - time temperature T(t), pressure P(t) and flow - rate parameter F(t), and a multi - parameter coupling index MCI(t) is generated, where

[0015]

[0016] In the formula: x(τ): includes respectively representing the dynamic change rates of temperature, pressure, and flow at time τ; the weight matrix W is a diagonal matrix, and the main - diagonal elements are respectively used to adjust the relative weights of the influence of temperature, pressure, and flow on the bonding effect; the attenuation kernel function K(t - τ)=e -λ(t-τ) , where λ is a non - negative constant.

[0017] Furthermore, a high - speed industrial camera is used to perform high - precision imaging of the brush from multiple angles or multiple planes to generate original image data I raw (x, y), uniformly adding marker information to all finished - product images of this batch, and performing noise filtering and brightness equalization on the original image I raw (x, y) to generate a pre - processed image I proc (x, y). Based on I proc (x, y), edge enhancement is adopted to extract candidate areas where defects may exist.

[0018] Furthermore, multi-scale feature extraction is performed on the candidate regions to obtain a set of multi-dimensional feature vectors Φ k , where k is the index of the candidate region; for each candidate region ROI k , its feature vector Φ k is input into the trained defect determination model. If it is determined as a defect region, a defect record can be generated; after detecting multiple defect regions, the overall defect degree is quantified through the following morphological cooperation index WPMSI.

[0019] Furthermore, the sampling records of the same batch and the brush defect information are extracted, and after matching, a comprehensive index structure is constructed; the forming period [t start , t end of the brush is retrieved, and WPMSI is associated with T(t), P(t), F(t), MCI(t) in the interval [t start , t end ; for each aligned product, the sequences of T(t), P(t), F(t), MCI(t) in the time interval [t start , t end are selected to construct the injection molding process feature vector X. At the same time, the corresponding defect index WPMSI is regarded as the output scalar Y and introduced into the kernel function κ(·,·) to map X to a high-dimensional feature space.

[0020] Furthermore, adaptive weighted canonical correlation analysis is introduced to quantify the non-linear correlation degree between X and Y. Among them, the correlation degree R c is the optimal correlation measure between the two in the kernel space, and is defined as follows:

[0021]

[0022] where: i represents the i-th product, X i is the injection molding process feature vector corresponding to the i-th product, Y i is the defect index WPMSI of the product, κ(·,·) is the kernel function, ω i is the adaptive weighting coefficient, and u and v are the direction vectors found in the kernel space respectively.

[0023] Furthermore, based on the correlation degree R c obtained from the analysis of each batch and the local sensitivity, the parameter range that has the greatest impact on the defect is marked. If MCI(t) continuously exceeds the corresponding threshold within a certain period of time and at the same time corresponds to a WPMSI defect index higher than expected, the corresponding high-fluctuation interval is taken as the key control object;

[0024] Based on the selected key parameters and historical data, an artificial neural network (ANN) is used to establish a defect index model between WPMSI and process parameters. Taking the defect index model as the objective function, solve for the parameter combination that minimizes WPMSI, output specific parameter suggestions, and obtain recommended injection parameters.

[0025] Furthermore, extract the recommended injection parameters output for the current batch, including the intervals for maintaining or restricting T(t), P(t), and F(t), as well as the corresponding injection speed adjustment plan. Compare them with the original default settings of the injection molding machine and generate a parameter modification list. Use a programmable logic controller or a host computer system to perform initial calibration on the execution components such as the heating zones, injection pressure units, and flow valves of the injection molding machine to generate a new set of initial adjustment parameters.

[0026] Furthermore, re-acquire T(t), P(t), F(t), and MCI(t), and introduce a dynamic tracking function to perform higher-frequency comparison of the key fluctuations during injection molding. If the following equation holds at any moment τ:

[0027] |MCI(τ) - MCI ref | > Ω MC

[0028] Then it is determined that there is an over-limit phenomenon of process fluctuations at this time and adaptive correction is required. Among them, MCI ref is the reference coupling index benchmark, and Ω MC is the threshold of the allowable fluctuation interval. When process fluctuation over-limit or abnormal defect trend is detected, the adaptive control equation is automatically triggered for secondary correction.

[0029] A system for producing injection-molded industrial brushes using an injection molding machine, including

[0030] A data acquisition unit that collects the working parameters of the injection molding machine by means of a sensor array arranged at key parts of the injection molding machine, generates a multi-parameter coupling index MCI(t) from the obtained working parameters, and binds the working parameters, multi-parameter coupling index MCI(t) within the corresponding time period with the batch identifier;

[0031] A defect detection unit that performs multi-angle imaging on the brush, adds marked information, performs preprocessing, extracts candidate areas that may have defects, and after detecting multiple defect areas, constructs a morphological cooperation index WPMSI to quantify the overall defect degree and obtain the defect data of each brush;

[0032] A recommended output unit that constructs an integrated index structure and then constructs an injection process feature vector X, calculates the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index WPMSI using kernel mapping and the AWCCA method, and then formulates the working parameter interval and generates parameter modification suggestions;

[0033] The parameter correction unit reads the working parameter modification list in the index structure, superimposes it with the original parameter set to form a new set value, automatically issues it to the injection molding machine through the PLC, and online monitors and adaptively corrects the working parameters.

[0034] (III) Beneficial effects

[0035] The present invention provides a system and method for producing injection-molded industrial brushes with an injection molding machine, having the following beneficial effects:

[0036] 1. Using continuous sampling and high-order operations to generate the multi-parameter coupling index MCI(t) to quantify the combined fluctuations of parameters such as temperature, pressure, and flow rate, providing an intuitive indicator for process state evaluation; when MCI(t) is large, it indicates that there are significant fluctuations in the process, which may cause unstable bonding between the brush filaments and the glue, providing an early warning basis for subsequent defect detection and process control.

[0037] 2. Adopting multi-scale feature extraction to effectively capture the subtle changes in the contact area between the brush filaments and the glue, enhancing the detection ability for defects such as local glue leakage and loose brush filaments; using the morphological synergy index WPMSI to quantitatively evaluate each candidate area, combining the local wave packet energy and morphological complexity to achieve accurate quantification of the defect severity; by finding the optimal projection in the kernel space, complex non-linear mapping relationships can be captured while introducing adaptive weighting to assign higher weights to key products or products with large fluctuations, improving the sensitivity to severe defects.

[0038] 3. Using kernel mapping and adaptive weighted canonical correlation analysis to combine multi-dimensional process parameters into a feature vector, revealing the non-linear relationships between temperature, pressure, flow rate, MCI(t), and the defect index WPMSI; determining the parameter interval that has the greatest impact on defects based on correlation analysis and local sensitivity data; when MCI(t) continuously exceeds the predetermined threshold during a certain period and is accompanied by a high WPMSI, this interval is marked as the key control object; the output control suggestions can be automatically executed by the parameter optimization control module and also provide a scientific basis for manual intervention, thereby reducing the defect rate and improving production consistency.

[0039] 4. Through the adaptive control equation, the parameters can be corrected in small steps and continuously during the production process, avoiding process instability caused by a large-scale adjustment at one time. The loss function C can combine indicators such as WPMSI and MCI(t) to balance the defect degree of the brush and the smoothness of the injection molding process, achieving more flexible control.

[0040] 5. Apply the regulatory suggestions to the production site through the parameter modification and distribution mechanism, and fine-tune the process fluctuations through real-time monitoring and the adaptive control equation to form a closed-loop feedback system. Finally, through multiple rounds of iteration, the continuous reduction of the brush defect rate and the significant improvement of process stability are achieved, enhancing the product quality consistency and production efficiency. Description of the Drawings

[0041] Figure 1 It is a schematic flow chart of the method for producing the injection-molded industrial brush of the present invention;

[0042] Figure 2 It is a schematic structural diagram of the system for producing the injection-molded industrial brush of the present invention. Detailed Embodiments

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0044] Please refer to Figure 1 , the present invention provides a method for producing an injection-molded industrial brush using an injection molding machine, including,

[0045] Step 1. Collect the working parameters of the injection molding machine by the sensor array arranged at the key parts of the injection molding machine, generate the multi-parameter coupling index MCI(t) from the obtained working parameters, and bind the working parameters, the multi-parameter coupling index MCI(t) and the batch identifier in the corresponding time period;

[0046] Step 101. At the key parts such as the injection port, the mold cavity and the injection channel of the injection molding machine, arrange temperature sensors, pressure sensors and flow sensors respectively to construct a sensor array for obtaining the real-time temperature T(t), pressure P(t) and flow parameters F(t), where t represents the time series, and establish a unified PROCESS_DATA data structure for all the collected temperature, pressure and flow data, and the fields include {timestamp, T(t), P(t), F(t), batch identifier};

[0047] During use, deploy temperature, pressure and flow sensors at the injection port, the mold cavity and the runner inlet of the injection molding machine to ensure obtaining real-time data at the key parts and covering the entire process flow; Batch differentiation: Identify and segment the data of each production cycle through the batch identifier (Batch), which helps with traceability, comparison and fault analysis.

[0048] Step 102: During the operation of the injection molding machine, the sensor array continuously collects the real-time values of temperature T(t), pressure P(t), and flow parameter F(t) at a predefined sampling frequency; and generates a multi-parameter coupling index MCI(t) to measure the multi-parameter interaction intensity from the starting time 0 to the current time t, which is defined as follows: Let K(t - τ) = E -λ(t-τ) ; where: T(τ), P(τ), and F(τ) represent the temperature, pressure, and flow rate at time τ respectively;

[0049] represents the instantaneous rate of change of the above parameters with respect to time, and α T , α P , α F are the weight factors for the three parameters of temperature, pressure, and flow rate. x(τ) is a three-dimensional vector that records the instantaneous rate of change at time τ; W is a diagonal matrix corresponding to the relative weights of the three parameters in the coupling analysis. K(t - τ) = e -λ(t-τ) represents assigning different decay weights to the changes at different times, and λ is the decay coefficient.

[0050] Therefore, the form of the multi-parameter coupling index MCI(t) is as follows:

[0051]

[0052] Here,

[0053] Parameter meaning: x(τ): contains respectively represent the dynamic change rates of temperature, pressure, and flow rate at time τ, which are used to capture the fluctuation amplitude and frequency of key parameters during the injection molding process; the weight matrix W, which is a diagonal matrix, and the main diagonal elements are respectively are used to adjust the relative proportions of the effects of temperature, pressure, and flow rate on the bonding effect, so as to adapt to the requirements of different material glues or brush wire characteristics; the decay kernel function K(t - τ) = e -λ(t-τ) , and λ is a non-negative constant.

[0054] When in use, the multi-parameter coupling index MCI(t) is generated by continuous sampling and high-order operations to quantify the joint fluctuations of parameters such as temperature, pressure, and flow rate, providing an intuitive indicator for process state evaluation; when MCI(t) is relatively large, it indicates that there are significant fluctuations in the process, which may cause unstable bonding between the brush wire and the glue, providing a warning basis for subsequent defect detection and process regulation.

[0055] Step 103: Write the obtained multi-parameter coupling index MCI(t) into the PROCESS_DATA data structure simultaneously to form a record {timestamp T(t), P(t), F(t), MCI(t), batch identifier}; after each production batch ends, bind all the records of {timestamp T(t), P(t), F(t), MCI(t)} within the corresponding time period with the batch identifier, and write the batch data into the PROCESS_DATA database;

[0056] If continuous production spans multiple batches, perform segmented processing according to the batch identifier. The records of each batch are stored independently, forming a multi-batch data set in the same database. When it is necessary to trace the forming process of a specific brush product, the temperature, pressure flow, and MCI(t) records corresponding to it can be quickly retrieved according to the batch identifier, and associated with the defect detection results generated in the subsequent steps;

[0057] It should be noted that: by integrating the weighted norm of the parameter change rate over the time interval [0, t] and combining with a kernel function for attenuation, a coupling strength index that comprehensively reflects the coupling degree of temperature, pressure, and flow during the injection molding process is obtained. The larger the MCI(t) value, the higher the degree of fluctuation exists within this time period, which may lead to an increase in the risk of instability in the bonding of brush filaments and glue, and is provided for key evaluation in subsequent processes (such as machine vision defect detection).

[0058] When in use, structure and store all the collected parameter data and the calculated MCI(t) according to batches (Batch) to form a complete historical data record; the unified data format ensures the matching and joint analysis with machine vision detection data in the subsequent process, and realizes the precise comparison between process data and finished product quality.

[0059] Step Two: After performing multi-angle imaging on the brush and adding marker information, perform preprocessing, extract candidate regions that may have defects. After detecting multiple defect regions, construct a morphological synergy index WPMSI to quantify the overall defect degree and obtain the defect data of each brush;

[0060] Step 201: After the brush product is demolded, use a high-speed industrial camera to perform high-precision multi-angle or multi-plane imaging on the brush to generate original image data I raw (x, y). To ensure the precise matching of the image data with the corresponding batch of PROCESS_DATA, uniformly add marker information such as batch identifier (BatchID) and product serial number (ProductID) to all the finished product images of this batch, and temporarily store the image data in the VISUAL_DATA structure. Its field examples are: {timestamp, BatchID, ProductID, I raw(x,y)}; By attaching BatchID and ProductID to each image, the subsequent defect analysis results can be seamlessly compared with the temperature T(t), pressure P(t), flow F(t) and other data of the same batch. Multi-view imaging can also improve the observation accuracy of the brush wire position and glue wrapping status, avoiding blind spots caused by a single angle;

[0061] When in use, a high-speed industrial camera is used for multi-angle shooting to ensure that high-quality original images of key parts of the brush are obtained. Each finished product is also attached with a batch (Batch) and product number (Item) to ensure accurate association with PROCESS_DATA data, laying the foundation for subsequent quantitative defect analysis.

[0062] Step 202: original image I raw (x, y) performs noise filtering and brightness equalization to generate the preprocessed image I proc (x,y), in I proc Based on (x, y), edge enhancement and morphological operations are used to extract candidate regions (ROIs) where defects may exist;

[0063] In order to accurately characterize the defects between the bristle and the glue (such as loose bristles or incomplete glue wrapping), high-order methods such as discrete wavelet transform or adjustable scale convolution kernel are introduced to extract multi-scale features of the candidate region (ROI) and obtain a set of multi-dimensional feature vectors Φ k , where k is the index of the candidate area; through image preprocessing, the interference factors in imaging are eliminated and the recognizability of the glue-wrapped boundary and the brush wire contour is enhanced. Compared with traditional threshold segmentation, multi-scale feature extraction can better capture the subtle bonding differences between glue and brush wire, and is especially suitable for detecting local glue leakage or gap areas.

[0064] Step 203: for each candidate region ROI k , its feature vector Φ k Input the trained defect determination model (such as a classifier based on machine learning or deep learning). If it is determined to be a defect area, a defect record can be generated, for example, {BatchID, ProductID, k, DefectType}; after multiple defect areas are detected, the overall defect degree is quantified by the following morphological synergy index WPMSI, where:

[0065] Let Ω k represents the coordinate domain corresponding to the kth defect area in the image plane, Φ k is the multi-dimensional feature vector extracted from the defect area (for example, obtained by wavelet transform, morphological operation or other high-order feature extraction methods), W(Φ k ,ψ s,θ), which represents projecting the feature vector Φ k onto a set of wavelet bases ψ s,Ω to obtain the wave packet coefficient vector;

[0066] In addition, a shape metric function is introduced to characterize the shape complexity or contour irregularity of the defect region Ω k and can be defined according to the specific requirements of industrial brush defects. For example:

[0067] Perimeter - area ratio: Convexity - concavity coefficient: Based on the difference between Ω k and its minimum bounding rectangle (or minimum convex hull) to measure the contour complexity; Fractal dimension: Quantify the morphological richness of the defect by calculating the filling characteristics of the region at multiple scales;

[0068] Define the morphological cooperation index WPMSI as follows:

[0069]

[0070] where: m is the total number of current brush defect regions, ||W(Φ k , ψ s,θ )||2 represents calculating the L2 norm of the wave packet coefficients of the k - th defect region to measure the energy accumulation of features under different wavelet bases, γ is a non - linear amplification factor for the energy of wave packet coefficients, used to emphasize higher - intensity defect patterns, embed the shape complexity of the defect region into the energy measurement in exponential form, μ is the shape weight coefficient; β is the coefficient for quadratic non - linear amplification of the overall integral result, which can further enhance the discrimination of severe defects, dA represents the spatial coordinate integration of the defect region Ω k ;

[0071] W(Φ k , ψ s,θ ) characterizes the high - frequency or fine - structure information of the defect region, as well as the texture features of the contact surface between the brush filaments and the glue. If the energy value of this vector is large, it indicates that there are abnormal fluctuations in this region at multiple scales or multi - direction frequency bands, usually corresponding to more severe or more quantify the defect shape complexity. If the defect contour is irregular or diffuse, then the value will increase accordingly, indicating that this defect is more difficult to handle or more likely to cause functional damage (such as a large range of loose brush filaments);

[0072] Furthermore, μ controls the influence intensity of the morphological complexity in the exponential function, which can give higher weights to the defective areas with complex morphology (such as large-area glue leakage or uneven wrapping). γ and β are non-linear amplification coefficients used to highlight serious defects during the accumulation process. By adjusting these two parameters, the system can maintain flexibility when statistically analyzing a large number of small defects or a small number of large defects. Using the exponential form multiplies the influence of the morphological complexity on the energy value. The more complex the morphology, the larger the exponential product, which also indicates a potentially greater negative impact of the defect on the product performance.

[0073] The defect data information of each detected brush contains {BatchID, ProductID, WPMSI, DefectMap}, as well as the type, coordinate range, and severity level of each defective area, etc., which are uniformly written into the VISUAL_DATA structure to form a set of visual inspection results corresponding to the PROCESS_DATA.

[0074] During use, through preprocessing means such as noise filtering, brightness equalization, and edge enhancement, the image clarity is improved and environmental interference is reduced. Precise positioning of candidate areas: Multiscale feature extraction (such as discrete wavelet transform or adjustable-scale convolution kernels) is used to effectively capture the subtle changes in the contact area between the brush filaments and the glue, enhancing the detection ability for defects such as local glue leakage and loose brush filaments. The morphological cooperation index WPMSI is used to quantitatively evaluate each candidate area, combining the local wave packet energy and morphological complexity to achieve accurate quantification of the defect severity. The calculated WPMSI is associated with Batch and Item and stored in VISUAL_DATA to ensure effective comparison with the process parameters in PROCESS_DATA subsequently, supporting accurate defect cause analysis.

[0075] Step 3: After constructing the comprehensive index structure, construct the injection molding process feature vector X, and use the kernel mapping and AWCCA methods to calculate the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index WPMSI, and then formulate the working parameter range and generate parameter modification suggestions.

[0076] Step 301: Extract the PROCESS_DATA under the same batch identifier BatchID, which contains the sampling records of T(t), P(t), F(t), and MCI(t) over time. At the same time, obtain the brush defect information under the corresponding BatchID from VISUAL_DATA, such as WPMSI, defect area mapping, etc.

[0077] Match the two according to BatchID and ProductID, and construct a comprehensive index structure ANALYSIS_DATA, which is {BatchID, ProductID, T(t), P(t), F(t), MCI(t), WPMSI, …};

[0078] To match the defect index WPMSI of a specific brush with its corresponding forming process, it is necessary to retrieve the forming period [t start , t end of this brush in PROCESS_DATA, and associate WPMSI with T(t), P(t), F(t), MCI(t) in the interval [t start , t end , so as to conduct targeted analysis subsequently to ensure that the analysis results are based on accurate time sequence correspondence;

[0079] When constructing VISUAL_DATA, it should be ensured that all records are indexed using the same batch identifier and product serial number, so that all parameters and defect information of a specific brush can be quickly retrieved in subsequent steps. The accuracy of the time window mapping directly affects the subsequent causal association inference between temperature, pressure, flow rate fluctuations and defect generation;

[0080] When used, it can ensure that the forming data of each brush finished product corresponds one-to-one with the visual inspection results, providing a data basis for establishing the causal relationship between defects and process parameters in the future.

[0081] Step 302: For each aligned product, select the sequences of T(t), P(t), F(t), MCI(t) in the time interval [t start , t end , construct the injection molding process feature vector X, and at the same time regard the corresponding defect index WPMSI as the output scalar Y and introduce the kernel function κ(·,·) to map X to a high-dimensional feature space to capture the potential non-linear relationship between the parameters and the defect index;

[0082] Introduce adaptive weighted canonical correlation analysis to quantify the non-linear correlation degree between X and Y. Among them, the correlation degree R c is the optimal correlation measure between the two in the kernel space. The higher the value, the more significant the correlation between the injection molding process characteristics and the defect index. It is defined as follows:

[0083]

[0084] Among them: i represents the i-th product, X i is the injection molding process feature vector corresponding to the i-th product, and Y iis the defect index WPMSI of the product, κ(·,·) is the kernel function (such as RBF kernel or polynomial kernel) for realizing nonlinear mapping, ω i is the adaptive weighting coefficient, which can be set according to the importance of the product in production or the fluctuation of MCI(t). u and v are the direction vectors found in the kernel space respectively, which are used to maximize the projection correlation between X and Y;

[0085] Write the R c value and the results of the local sensitivity analysis of each parameter (between T(t), P(t), F(t), MCI(t) and WPMSI) into ANALYSIS_DATA to form: {BatchID, ProductID, R c , PartialSensitivity,...}; PartialSensitivity can be further broken down into the influence weights on temperature, pressure, flow rate and coupling index MCI(t), which are used to identify which process parameters are strongly correlated with defects;

[0086] By finding the optimal projection in the kernel space, complex nonlinear mapping relationships can be captured while introducing adaptive weighting, assigning higher weights to critical products or products with large fluctuations, improving the sensitivity to serious defects. By writing the result data into ANALYSIS_DATA uniformly, it can be directly docked with the subsequent steps, thus avoiding parameter name conflicts or data inconsistencies.

[0087] When in use, adopt kernel mapping and adaptive weighted canonical correlation analysis (AWCCA) to combine multi-dimensional process parameters into feature vectors, revealing the nonlinear relationships between temperature, pressure, flow rate, MCI(t) and the defect index WPMSI; calculate the local sensitivity of each parameter to defects and quantify their comprehensive correlation degree, providing a basis for identifying key influencing factors and abnormal intervals.

[0088] Step 303, based on the correlation degree R c and local sensitivity obtained from the analysis of each batch (or a certain time window), mark the parameter ranges that have the greatest impact on defects (for example, temperature T within a certain interval, pressure P near a certain critical value). If MCI(t) continuously exceeds the corresponding threshold within a certain period of time and at the same time corresponds to a WPMSI defect index higher than expected, mark the corresponding high-fluctuation interval as the key control object;

[0089] Based on the selected key parameters and historical data, a defect index model between WPMSI and process parameters is established using an artificial neural network ANN; taking the defect index model as the objective function, solve for the parameter combination that minimizes WPMSI, where T(t), P(t), F(t) need to satisfy the physical constraints of the process (such as upper and lower limits of temperature, pressure, etc.); at this time, output specific parameter suggestions such as keeping the temperature curve within [T min ,T max and the peak pressure not exceeding a certain critical value, etc., to obtain recommended injection parameters;

[0090] Write the suggestions into the suggestion field of the index structure ANALYSIS_DATA. By comparing the significant correlation degrees between different parameter intervals and the defect index, the optimal or near-optimal combination for reducing the defect rate can be sought in the multi-dimensional parameter space. The output regulation suggestions can be automatically executed in the subsequent steps or provided as a reference for manual intervention decision-making.

[0091] According to the correlation analysis and local sensitivity data, determine the parameter interval that has the greatest impact on defects (for example, the temperature is within a certain range, near the pressure critical value); at the same time, when MCI(t) continuously exceeds the predetermined threshold during a certain period and is accompanied by a high WPMSI, this interval is marked as the key regulation object; the output regulation suggestions can not only be automatically executed by the parameter optimization control module but also provide a scientific basis for manual intervention, thereby reducing the defect rate and improving production consistency.

[0092] Step 4: Read the working parameter modification list in the index structure, superimpose it with the original parameter set to form a new set value, and automatically send it to the injection molding machine through the PLC, and online monitor and adaptively correct the working parameters;

[0093] Step 401: Extract the recommended injection parameters output for the current batch, including the intervals that may need to be maintained or restricted for T(t), P(t), F(t), and the corresponding injection speed adjustment scheme, compare it with the original default settings of the injection molding machine, and generate a parameter modification list. The format can be defined as: {BatchID, ΔT, ΔP, ΔF, SpeedAdj,...}, where ΔT, ΔP, ΔF are the adjustment amounts of temperature, pressure, and flow rate relative to the original set values respectively, and SpeedAdj is the correction coefficient of the injection speed, etc.;

[0094] Use a programmable logic controller or a host computer system to perform initial calibration on the execution components such as the heating zone, injection pressure unit, and flow valve of the injection molding machine, generate a new set of initial adjustment parameters, register and associate them with the corresponding BatchID for subsequent comparison and traceability. Keep a record of the differences between the old set values and the new set values to facilitate subsequent verification of the adjustment effect and continuous optimization.

[0095] Read the parameter modification list from ANALYSIS_DATA, compare it with the existing original injection molding parameter set, calculate the new target setting value, and use the PLC or the host computer system to send the new setting value to the temperature control unit, pressure unit, flow valve, and injection speed adjustment mechanism of the injection molding machine to achieve the initial process adjustment; write the successfully effective adjustment value into the data management device in the format of the initial adjustment parameter set and associate it with the corresponding Batch to provide a basis for subsequent comparison and traceability.

[0096] Step 402. After the start of a new round of production, repeat the sensor acquisition in Step 1: Obtain T(t), P(t), F(t), and MCI(t) in real time to form new process data. Introduce a dynamic tracking function to perform a higher-frequency comparison of the key fluctuations during injection molding. If the following equation holds at any moment τ:

[0097] |MCI(τ) - MCI ref | > Ω MC

[0098] Then it is determined that there is a phenomenon of process fluctuation exceeding the limit at this time and adaptive correction is required; where MCI ref is the reference coupling index benchmark and Ω MC is the allowable fluctuation interval threshold;

[0099] When it is detected that the process fluctuation exceeds the limit or the defect trend is abnormal, the following adaptive control equation is automatically triggered for secondary correction of ΔT, ΔP, ΔF, etc. Among them,

[0100] where: ΔT old , ΔP old , ΔF old is the parameter adjustment amount during the current execution, and ΔT new , ΔP new , ΔF new is the new adjusted amount after correction, η T , η P , η F is the step coefficient of each parameter, which determines the speed of the correction action; represents a certain comprehensive loss function the partial derivatives of temperature, pressure, and flow rate, which are used to quantify the comprehensive influence on the brush defect index (such as WPMSI) and the injection molding process fluctuation (MCI(t));

[0101] Update the actual execution parameter record of the current batch at any moment τ during the production process with the new parameter value after correction to form {BatchID, τ, T exec , P exec , F exec, {MCI(τ)}, so that subsequent defect detection and associated data analysis can trace back to the corresponding true process state according to the specific correction moment;

[0102] Through the adaptive control equation, parameters can be corrected in small steps and continuously during the production process, avoiding process instability caused by a large-scale adjustment at one time. In the loss function C, indicators such as WPMSI and MCI(t) can be combined to balance the degree of brush defects and the smoothness of the injection molding process, realizing a more flexible regulation;

[0103] After the new settings take effect, the sensors continue to collect T(t), P(t), F(t), MCI(t) in real time and write them into PROCESS_DATA to form a continuous process data stream; using the dynamic tracking function to monitor, if MCI(t) exceeds the preset threshold at any moment, the adaptive control equation will be automatically triggered to make small-step corrections to ΔT, ΔP, ΔF, ensuring that the production process is always within the target parameter range; the corrected parameters and corresponding timestamps are recorded in the data management device, providing detailed data for real-time feedback and subsequent re-analysis.

[0104] Step 403: After completing this round of production, repeatedly obtain new defect detection indicators (such as WPMSI), associate the new detection indicators with T(t), P(t), F(t), MCI(t) corresponding to this round of production and all corrected parameters, and write them into ANALYSIS_DATA;

[0105] If the detection result still indicates a high proportion of uneven brush adhesion, further call Step Three to output optimization suggestions again based on the new collected data and defect indicators, and then return to this step (Step Four). Through multiple iterations, a relative balance and optimal region are achieved between the fluctuations in the injection molding process and the finished product defect rate, realizing a stable improvement in the adhesion quality of the brush filaments and glue; this process actually forms a closed loop of data collection - defect detection - correlation analysis - parameter optimization - re-collection and re-analysis. When the system is verified through multiple iterations and the defect rate is significantly reduced and stabilized, the best injection molding parameters for this round can be solidified stage by stage for subsequent production reference.

[0106] After the production cycle ends, machine vision inspection is carried out again to obtain the latest defect index WPMSI, and these inspection results are associated and written into ANALYSIS_DATA with the process parameter data currently being executed; if the defect rate is still high, return to step 303 to perform correlation analysis again based on the latest data, output a new round of parameter modification suggestions, and achieve multi-round closed-loop optimization until the defect rate steadily decreases and reaches the expected goal. Apply the regulation suggestions to the production site through the parameter modification distribution mechanism, and fine-tune the process fluctuations through real-time monitoring and adaptive control equations to form a closed-loop feedback system; finally, through multiple rounds of iteration, the continuous reduction of the brush defect rate and the significant improvement of process stability are achieved, enhancing product quality consistency and production efficiency.

[0107] Please refer to Figure 2 , the present invention provides a system for producing injection-molded industrial brushes using an injection molding machine, including,

[0108] A data acquisition unit, which collects the working parameters of the injection molding machine by means of a sensor array arranged at key parts of the injection molding machine, generates a multi-parameter coupling index MCI(t) from the obtained working parameters, and binds the working parameters, the multi-parameter coupling index MCI(t) within the corresponding time period with the batch identifier;

[0109] A defect detection unit, which performs multi-angle imaging on the brush, adds marker information and then performs preprocessing, extracts candidate areas that may have defects, and after detecting multiple defect areas, constructs a morphological cooperation index WPMSI to quantify the overall defect degree and obtain the defect data of each brush;

[0110] A suggestion output unit, which constructs an integrated index structure and then constructs an injection molding process feature vector X, calculates the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index WPMSI by using kernel mapping and AWCCA methods, and then formulates a working parameter range and generates parameter modification suggestions;

[0111] A parameter correction unit, which reads the working parameter modification list in the index structure, superimposes it with the original parameter set to form a new set value, and automatically issues it to the injection molding machine through the PLC, and online monitors and adaptively corrects the working parameters.

[0112] Those of ordinary skill in the art can realize that, combining the units and algorithm steps of each example described in the embodiments disclosed herein, can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0113] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0114] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only for some logical function divisions. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application and should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for producing injection-molded industrial brushes with an injection molding machine, characterized in that: including, collecting the working parameters of the injection molding machine by a sensor array arranged at key parts of the injection molding machine, generating a multi-parameter coupling index MCI(t) from the obtained working parameters, and binding the working parameters, the multi-parameter coupling index MCI(t) and the batch identifier within the corresponding time period; performing multi-angle imaging on the brush, adding marker information and then performing preprocessing, extracting candidate areas where defects may exist, detecting multiple defect areas, constructing a morphological cooperation index WPMSI to quantify the overall defect degree, and obtaining the defect data of each brush; constructing an integrated index structure and then constructing an injection process feature vector X, calculating the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index WPMSI by using kernel mapping and AWCCA methods, and then formulating a working parameter range and generating parameter modification suggestions; reading the working parameter modification list in the index structure, superimposing it with the original parameter set to form a new set value, automatically sending it to the injection molding machine through the PLC, and online monitoring and adaptively correcting the working parameters.

2. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 1, characterized in that: arranging a sensor array at key parts of the injection molding machine for obtaining real-time temperature T(t), pressure P(t) and flow rate parameter F(t), and generating a multi-parameter coupling index MCI(t), wherein, where: x(τ): includes respectively representing the dynamic change rates of temperature, pressure, and flow at time τ; the weight matrix W is a diagonal matrix, and the main diagonal elements are respectively used to adjust the relative proportions of the effects of temperature, pressure, and flow on the bonding effect; the attenuation kernel function K(t - τ) = e -λ(t-τ) , where λ is a non-negative constant.

3. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 2, characterized in that: Use a high-speed industrial camera to perform high-precision imaging of the brush from multiple angles or multiple planes to generate the original image data I raw (x,y), uniformly add marking information to all finished product images of this batch, and perform noise filtering and brightness equalization on the original image I raw (x,y) to generate the preprocessed image I proc (x,y), based on I proc (x,y), adopt edge enhancement to extract candidate regions with defect risks 4. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 3, characterized in that: Perform multi-scale feature extraction on the candidate regions to obtain a set of multi-dimensional feature vectors Φ k , where k is the index of the candidate region; for each candidate region ROI k , its feature vector Φ k is input into the trained defect determination model. If it is determined to be a defect region, a defect record can be generated. After detecting multiple defect regions, the overall defect degree is quantified by the following morphological cooperation index WPMSI.

5. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 4, characterized in that: Extract the sampling records and brush defect information of the same batch, match them, and construct a comprehensive index structure; retrieve the brush forming period [t start , t end , and associate WPMSI with T(t), P(t), F(t), MCI(t) in the interval [s tart , t end ; for each aligned product, select the sequences of T(t), P(t), F(t), MCI(t) in the time interval [t start , t end , construct the injection molding process feature vector X, and at the same time regard the corresponding defect index WPMSI as the output scalar Y and introduce the kernel function κ(·,·) to map X to a high-dimensional feature space.

6. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 5, characterized in that: Adaptive weighted canonical correlation analysis is introduced to quantify the non-linear correlation degree between X and Y, where the correlation degree R c is the optimal correlation measure between the two in the kernel space and is defined as follows: where: i represents the i-th product, X i is the injection molding process feature vector corresponding to the i-th product, Y i is the defect index WPMSI of the product, k(·,·) is the kernel function, ω i is the adaptive weighting coefficient, and u and v are the direction vectors found in the kernel space respectively.

7. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 6, characterized in that: Based on the relevance R obtained from the analysis of each batch c and local sensitivity, the parameter range that has the greatest impact on defects will be marked. If MCI(t) continuously exceeds the corresponding threshold within a certain period of time and simultaneously corresponds to a WPMSI defect index higher than expected, the corresponding high-fluctuation interval will be regarded as the key regulation object; based on the selected key parameters and historical data, establishing a defect index model between WPMSI and process parameters by using an artificial neural network ANN; taking the defect index model as the objective function, solving the parameter combination that minimizes WPMSI, outputting specific parameter suggestions, and obtaining recommended injection parameters.

8. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 7, characterized in that: extracting the recommended injection parameters output for the current batch, including the intervals for maintaining or restricting T(t), P(t), F(t), and the corresponding injection speed adjustment scheme, comparing them with the original default settings of the injection molding machine, and generating a parameter modification list; using a programmable logic controller or a host computer system to perform initial calibration on the execution components such as the heating zone, injection pressure unit, and flow valve of the injection molding machine, and generating a new set of initial adjustment parameters.

9. The method for producing a glue-injected industrial brush using an injection molding machine according to claim 8, characterized in that: re-obtaining T(t), P(t), F(t) and MCI(t), introducing a dynamic tracking function to perform a higher-frequency comparison on the key fluctuations during injection molding. If the following formula holds at any moment τ: |MCI(τ)-MCI ref |>Ω MC It is then determined that there is a phenomenon of excessive process fluctuation at this time, and adaptive correction is required; where MCI ref is the reference coupling index benchmark, and Ω MC is the threshold of the allowable fluctuation range; when excessive process fluctuation or abnormal defect trend is detected, the adaptive control equation is automatically triggered and secondary correction is performed.

10. A system for producing injection-molded industrial brushes using an injection molding machine, characterized in that: including, The data acquisition unit collects the working parameters of the injection molding machine through a sensor array arranged at key parts of the injection molding machine, generates a multi-parameter coupling index MCI(t) from the obtained working parameters, and binds the working parameters, the multi-parameter coupling index MCI(t) within the corresponding time period, and the batch identifier; The defect detection unit performs multi-angle imaging on the brush, adds marked information and then performs preprocessing, extracts candidate regions that may have defects, constructs a morphological cooperation index WPMSI to quantify the overall defect degree after detecting multiple defect regions, and obtains the defect data of each brush; The suggestion output unit constructs an overall index structure and then constructs the injection process feature vector X, calculates the non-linear correlation and local sensitivity between each working parameter and the morphological cooperation index WPMSI by using the kernel mapping and AWCCA methods, and then formulates the working parameter interval and generates parameter modification suggestions; The parameter correction unit reads the working parameter modification list in the index structure, superimposes it with the original parameter set to form a new set value, automatically sends it to the injection molding machine through the PLC, and online monitors and adaptively corrects the working parameters.

Images