Self-adaptive full-automatic spline injection molding system

Abstract

The invention relates to the technical field of intelligent control of injection molding machines, in particular to a self-adaptive full-automatic spline injection molding system. Comprising an injection molding data acquisition module, a spline model generation module, a self-adaptive control module, a finished product quality detection module and a monitoring and early warning module. The data acquisition module acquires multi-dimensional data in the injection molding process in real time, the spline model generation module adopts deep learning to construct a nonlinear spline model, and the adaptive control module adjusts equipment parameters in real time based on the model. The finished product quality detection module analyzes quality features such as sizes and defects through machine vision and generates a detection report. And the monitoring and early warning module monitors the running state of the equipment in real time, identifies fault features and provides adjustment suggestions. The spline injection molding process is comprehensively optimized, and the production efficiency, the finished product quality and the safety and reliability of equipment operation are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent control of injection molding machines, in particular to a self-adaptive fully automatic spline injection molding system. Background Art

[0002] Injection molding technology is a processing method widely used in the production of plastic products, and is widely used in many fields such as automobiles, electronics, medical and consumer products. The injection molding process includes the melting of plastic raw materials, injection into the mold, flow, cooling and demolding. However, with the increasing demand for complex shapes, precise dimensions and high-quality products in modern industry, traditional injection molding systems face many technical bottlenecks in meeting diverse needs, especially in the manufacturing process of high-precision and complex curved parts.

[0003] Spline injection molding is an injection molding technology for the molding of complex curved products. It can dynamically adjust injection parameters during the injection molding process to improve product quality. However, the following problems still exist: the existing spline injection molding system mostly relies on preset process parameters and lacks the ability to dynamically adapt to real-time process fluctuations; the existing technology uses simple linear models or empirical formulas when dealing with complex nonlinear process parameter relationships. This method not only has limited fitting capabilities, but is also easily affected by process fluctuations, resulting in unsatisfactory prediction accuracy and optimization effects; the process adjustment of the existing injection molding system mainly relies on manual experience or simple rules, with slow response speed and insufficient accuracy, making it difficult to cope with real-time changing production environments and complex process requirements, and easily causing product quality instability. Summary of the invention

[0004] To solve the above problems, the present invention provides an adaptive fully automatic spline injection molding system, which solves the problem of how to overcome the influence of nonlinear process relationships by dynamically adjusting process parameters during the spline injection molding process, improve the system's adaptability to real-time process fluctuations and ensure the stability of product quality, thereby improving the spline injection molding production efficiency, finished product quality and the safety and reliability of equipment operation.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] An adaptive fully automatic spline injection molding system, comprising an injection molding data acquisition module, a spline model generation module, an adaptive control module, a finished product quality detection module and a monitoring and early warning module which are sequentially connected in communication;

[0007] The injection molding data acquisition module is used to collect multi-dimensional data of the injection molding process in real time through a sensor network;

[0008] The spline model generation module is used to preprocess the multidimensional data, extract key process parameters, and construct a nonlinear spline model using a spline curve fitting algorithm based on deep learning;

[0009] The adaptive control module is used to generate an optimal process adjustment strategy based on the nonlinear spline model and adopt fuzzy adaptive resonance theory to adjust the key process parameters of the injection molding equipment in real time;

[0010] The finished product quality inspection module is used to perform real-time online inspection of injection molded products through machine vision technology and embedded image processing algorithms, analyze the dimensional accuracy, surface defects, density distribution and internal stress of the finished products, and determine the quality level of the finished products through a prediction model to generate a quality inspection analysis report;

[0011] The monitoring and early warning module is used to use an abnormality detection algorithm to monitor the operating status and process parameters of the injection molding equipment in real time, identify potential fault characteristics under complex nonlinear relationships, generate a monitoring and analysis report, trigger an early warning signal, and provide process adjustment suggestions.

[0012] Furthermore, the multidimensional data includes injection pressure, temperature, flow rate, mold cavity filling status and injection equipment operation status data.

[0013] Furthermore, the operation process of the spline model generation module includes the following steps:

[0014] Dividing the multidimensional data into multiple process stages according to the time series, and extracting the dynamic characteristic parameters of each stage, including key change trends and correlation indicators, in combination with the injection molding process rules;

[0015] Based on the process fluctuation characteristics, an adaptive data window is designed to optimize and analyze the extracted dynamic features window by window to screen out the key process parameters;

[0016] A spline curve fitting algorithm based on deep learning is used to perform nonlinear modeling on the key process parameters, and the parameters of the fitting function are optimized through iterative training to generate a nonlinear spline model.

[0017] Furthermore, the formula of the nonlinear spline model is as follows:

[0018]

[0019] Where y(t) represents the predicted value of the key process parameters at time t; φ i (t) represents the spline basis function; ω i represents the weight of the spline basis function; ψ j (t) represents the dynamic characteristic function of the injection molding process; exp(-β j ·ψj (t)) represents the nonlinear response function; α j Indicates the contribution strength of each feature function to the model output; β j represents the time attenuation coefficient of the dynamic response; γ·sin(ωt) represents the potential periodic fluctuations in the injection molding process; γ represents the intensity of the influence of periodic fluctuations on process parameters; ω represents the period length of process parameter fluctuations; δ·t p represents the time trend term; n represents the number of spline basis functions; m represents the number of dynamic characteristic functions.

[0020] Furthermore, the operation process of the adaptive control module includes the following steps:

[0021] Based on the nonlinear spline model, the dynamic multi-dimensional data transmitted by the injection molding data acquisition module is received in real time, and the current process state is dynamically analyzed by using a fuzzy reasoning algorithm in combination with historical process data to identify the degree of deviation of key parameters;

[0022] Using fuzzy adaptive resonance theory, the deviation of key parameters is evaluated in real time, a multi-dimensional deviation matrix is ​​generated, and real-time adjustment constraints are established according to the process optimization goals;

[0023] Based on the multi-dimensional deviation matrix and constraints, the process adjustment strategy is calculated through a dynamic optimization algorithm to generate the best process adjustment plan;

[0024] The optimal process adjustment plan is output, the operating parameters of the equipment are adjusted in real time, and the adjustment results are verified by closed-loop feedback.

[0025] Furthermore, the formula of the fuzzy adaptive resonance theory is as follows:

[0026]

[0027] Where A(t) represents the optimal process adjustment solution at time t; W i represents the weight of the i-th key process parameter; P i (t) represents the deviation value of the i-th key process parameter; K i represents the nonlinear suppression factor of the i-th parameter; Q i (t) represents the dynamic relationship between different parameters in the injection molding process; D i (t) represents the dynamic changes of real-time acquisition in multidimensional data; R i (t) represents the influence of random fluctuations in the correction parameters; T i Indicates the response delay of the injection molding equipment to the key process parameters; M represents the global dynamic adjustment coefficient; represents the rate of change of the i-th key process parameter; m represents the total number of key process parameters.

[0028] Furthermore, the operation process of the finished product quality inspection module includes the following steps:

[0029] The appearance characteristics and internal structure data of finished products are collected through multi-angle industrial cameras, laser scanning sensors and stress detection equipment, and multi-source data are integrated;

[0030] Divide the finished product into regions, identify key inspection areas, and use regional specialization algorithms to perform detailed analysis on key areas to generate finished product feature parameters;

[0031] Based on the embedded deep learning algorithm, the characteristic parameters of the finished product are gradually detected to identify defects including dimensional deviation, surface scratches, uneven color and internal bubbles; combining historical quality data and process parameters to establish a dynamic defect feature library for real-time optimization of defect recognition accuracy;

[0032] Combined with machine learning algorithms, the test results are comprehensively analyzed, the quality scores of the finished products in multiple dimensions are calculated, the finished products are graded, and a visual quality report is generated.

[0033] Furthermore, the machine learning algorithm includes a decision tree or a support vector machine algorithm to achieve intelligent evaluation of the quality level of the finished product.

[0034] Furthermore, the monitoring and early warning module specifically adopts an abnormal feature association analysis method based on a knowledge graph, and realizes multi-source early warning signal recognition and intelligent fault root cause tracing under complex nonlinear working conditions by establishing a dynamic association network between process parameters, equipment status and failure modes.

[0035] The beneficial effects of the present invention are:

[0036] In the present invention, the injection molding data acquisition module collects multi-dimensional data in real time through a sensor network, covers the key process parameters in the injection molding process, provides comprehensive and accurate basic data, and lays a solid foundation for subsequent modeling and control optimization. The spline model generation module adopts a spline curve fitting algorithm based on deep learning, which can accurately extract the complex relationship between nonlinear process parameters in the injection molding process, effectively solve the nonlinear problem that is difficult to fit by traditional methods, and ensure the accuracy and robustness of the model. Based on the spline model, the adaptive control module generates the best process adjustment strategy through fuzzy adaptive resonance theory, which can optimize and adjust the key process parameters in real time, effectively reduce process errors, and improve the consistency and quality of injection molded products. The finished product quality detection module combines machine vision technology and embedded image processing algorithms, and can comprehensively detect the dimensional accuracy, surface defects, density distribution and internal stress of injection molded products. At the same time, it evaluates the quality level through a prediction model to provide data support for subsequent process improvements. The monitoring and early warning module adopts an abnormal detection algorithm to monitor the operating status and key process parameters of the injection molding equipment in real time, can identify potential fault characteristics under complex nonlinear relationships, and timely generate early warning signals and adjustment suggestions to reduce equipment failure rates and improve system stability and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a module schematic diagram of a self-adaptive fully automatic spline injection molding system of the present invention.

[0038] Figure 2 It is a flowchart of the operation process of the spline model generation module provided by an embodiment of the present invention.

[0039] Figure 3 It is a flowchart of the operation process of the adaptive control module provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] See also Figure 1-3 As shown, the present invention relates to an adaptive fully automatic spline injection molding system.

[0041] Example

[0042] An adaptive fully automatic spline injection molding system, comprising an injection molding data acquisition module, a spline model generation module, an adaptive control module, a finished product quality detection module and a monitoring and early warning module which are sequentially connected in communication;

[0043] The injection molding data acquisition module is used to collect multi-dimensional data of the injection molding process in real time through a sensor network; the multi-dimensional data includes injection molding pressure, temperature, flow rate, mold cavity filling status and injection molding equipment operation status data.

[0044] It should be noted that a high-precision pressure sensor is installed inside the injection molding machine to monitor the real-time injection pressure changes in the mold cavity.

[0045] Temperature sensors are distributed on the nozzle, mold wall and melt flow channel to monitor temperature changes in key areas.

[0046] Ultrasonic flow rate sensors are installed at the barrel and cavity flow channel to monitor the melt flow rate in real time.

[0047] Optical sensors and displacement sensors are combined to monitor the cavity filling status and dynamic changes of the cavity.

[0048] Vibration sensors and current sensors are installed on the main moving parts and power supply system of the injection molding machine respectively to monitor the stability of the equipment's operating status.

[0049] Using high-speed data acquisition cards and wireless transmission modules, the data collected by the sensors are transmitted to the central processing system at a frequency of milliseconds. Distributed collection and centralized storage are realized to ensure the synchronization and consistency of multi-source heterogeneous data.

[0050] The spline model generation module is used to preprocess the multidimensional data, extract key process parameters, and construct a nonlinear spline model using a spline curve fitting algorithm based on deep learning;

[0051] The operation process of the spline model generation module includes the following steps:

[0052] Dividing the multidimensional data into multiple process stages according to the time series, and extracting the dynamic characteristic parameters of each stage, including key change trends and correlation indicators, in combination with the injection molding process rules;

[0053] Specifically, the process stage division rules: Based on the injection molding process rules, the entire process is divided into the following stages:

[0054] Injection stage: the melt enters the mold cavity and the dynamic changes of pressure and flow rate are monitored;

[0055] Holding stage: the pressure is maintained before the melt solidifies, with a focus on the cavity pressure curve;

[0056] Cooling stage: the melt is cooled and solidified, and the mold temperature and cooling time are extracted;

[0057] Demolding stage: When the product is taken out, monitor the mold opening and closing status and vibration characteristics.

[0058] Use moving average and first-order difference methods to calculate the trend of parameter changes (such as injection pressure rise rate, cavity filling speed changes). Extract statistical characteristics such as standard deviation and maximum-minimum value of process parameters to reflect stability and fluctuation range. Identify the inflection point of pressure or temperature curve and mark the start and end time of process stage. Use correlation coefficient to analyze the correlation between multidimensional parameters (such as the linear relationship between pressure and flow rate); use mutual information method to detect nonlinear correlation and explore hidden correlation of key parameters.

[0059] Based on the process fluctuation characteristics, an adaptive data window is designed to optimize and analyze the extracted dynamic features window by window to screen out the key process parameters;

[0060] Specifically, according to the dynamic characteristics of different process stages, variable windows are designed to perform local analysis of data and optimize the feature extraction effect. For the rapidly changing injection stage, a small time window (such as 50 milliseconds) is used to capture high-frequency dynamic features; for the more stable pressure holding and cooling stages, a long time window (such as 500 milliseconds) is used to extract global trends. Automatic window optimization is achieved through sliding windows and dynamic adjustment mechanisms (such as adaptive adjustment based on signal change rate). Dynamic features (such as local fluctuations, trend change rate) are calculated in each window; the differences between the window and the global features are compared to identify window-specific key change points.

[0061] Calculate the feature importance score of each parameter based on a decision tree or random forest model; filter out key parameters with importance scores above a preset threshold, such as the pressure rise rate during the injection phase or the mold temperature change during the cooling phase. Use principal component analysis (PCA) to reduce the dimensionality of high-dimensional features and retain the main features; remove redundant features (such as features that are highly correlated with other parameters) to ensure that the model input is concise and efficient.

[0062] A spline curve fitting algorithm based on deep learning is used to perform nonlinear modeling on the key process parameters, and the parameters of the fitting function are optimized through iterative training to generate a nonlinear spline model.

[0063] Specifically, a nonlinear spline model is constructed using a deep learning framework to describe the dynamic relationship between key parameters with high precision. A hybrid model based on a deep neural network is used to combine B-spline and radial basis function (RBF) to construct a spline fitting function; the network input is the key process parameter, and the output is the node position of the spline control point and the fitting curve. The loss function of the spline fitting error is defined, including the fitting error of the curve (such as mean square error) and smoothness constraints (such as the smoothness of the first-order and second-order derivatives). The loss weight is dynamically adjusted during the training process to balance accuracy and curve smoothness.

[0064] Divide the selected key parameters into training sets, validation sets, and test sets to ensure that typical situations in all process stages are covered. Improve the robustness of the model through data enhancement (such as adding noise and random sampling). Use adaptive learning rate optimization algorithms (such as Adam) to dynamically adjust the learning rate to accelerate convergence; perform accelerated training in a GP U environment to process high-dimensional complex data sets. Add L2 regularization terms to avoid model overfitting; terminate training early when the validation set error stops decreasing to prevent overtraining.

[0065] The spline model is output in the form of a nonlinear function, which contains the following information: the coordinates of the control points; the equation of each spline curve (such as the B-spline basis function); the association weights between parameters and the results of sensitivity analysis. The model can be directly input into the adaptive control module for real-time optimization of injection molding process parameters.

[0066] Furthermore, the formula of the nonlinear spline model is as follows:

[0067]

[0068] Where y(t) represents the predicted value of the key process parameters at time t; φ i (t) represents the spline basis function; w i represents the weight of the spline basis function; ψ j (t) represents the dynamic characteristic function of the injection molding process; exp(-β j ·ψ j (t)) represents a nonlinear response function, which is used to describe the attenuation characteristics of the above process data over time; α j Indicates the contribution strength of each feature function to the model output; β j represents the time attenuation coefficient of the dynamic response; γ·sin(ωt) represents the potential periodic fluctuations in the injection molding process; γ represents the intensity of the influence of periodic fluctuations on process parameters; ω represents the period length of process parameter fluctuations; δ·t p It represents the time trend term, which is used to describe the long-term trend (such as the overall change of the operating status of the injection molding equipment); n represents the number of spline basis functions; m represents the number of dynamic characteristic functions.

[0069] The adaptive control module is used to generate an optimal process adjustment strategy based on the nonlinear spline model and adopt fuzzy adaptive resonance theory to adjust the key process parameters of the injection molding equipment in real time;

[0070] The operation process of the adaptive control module includes the following steps:

[0071] Based on the nonlinear spline model, the dynamic multi-dimensional data transmitted by the injection molding data acquisition module is received in real time, and the current process state is dynamically analyzed by using a fuzzy reasoning algorithm in combination with historical process data to identify the degree of deviation of key parameters;

[0072] Specifically, during the operation of the injection molding equipment, the adaptive control module combines historical process data (such as process parameter database or historical optimization cases) and uses a nonlinear spline model to pre-process the data, including denoising, normalization, and data mapping, to ensure the integrity and accuracy of the data. A fuzzy reasoning algorithm is used to dynamically analyze the current process status, such as classifying the cavity pressure fluctuation range as "normal", "high", or "low", to identify the degree of deviation of each key process parameter and its potential impact on the quality of the finished product.

[0073] Using fuzzy adaptive resonance theory, the deviation of key parameters is evaluated in real time, a multi-dimensional deviation matrix is ​​generated, and real-time adjustment constraints are established according to the process optimization goals;

[0074] Specifically, based on the fuzzy reasoning results, the fuzzy adaptive resonance theory is used to evaluate the deviation of key parameters in real time. For example, when the cavity pressure fluctuation exceeds the set range, its deviation amount, deviation direction and correlation with other parameters are evaluated.

[0075] Generate a multidimensional deviation matrix, the rows of which represent key parameters (such as cavity pressure, temperature, etc.), the columns represent time series, and the matrix elements are deviation values.

[0076] Combined with process optimization goals (such as reducing product warpage and improving dimensional accuracy), define constraints for real-time adjustment, including but not limited to:

[0077] Adjustment range: Ensure that the adjusted parameters do not exceed the safe operating range of the equipment;

[0078] Adjustment priority: give priority to adjusting the parameters that have the greatest impact on product quality;

[0079] Time response: Ensure that the adjustment action is completed within the specified time.

[0080] Based on the multi-dimensional deviation matrix and constraints, the process adjustment strategy is calculated through a dynamic optimization algorithm to generate the best process adjustment plan;

[0081] Specifically, the main goal of process adjustment is to eliminate the deviation of key parameters and return them to the target range, while ensuring the stability of equipment operation and production efficiency. During the adjustment process, avoid process fluctuations or equipment failures caused by rapid and large adjustments. For example: if the cavity pressure deviation is large, the adjustment needs to be carried out step by step (such as adjusting 5MPa each time) to avoid excessive adjustments at one time that may cause pressure instability.

[0082] Extract the parameter with the largest current deviation from the deviation matrix, and calculate the target parameter to be adjusted first in combination with the constraints. For example, if the cavity pressure deviation is the largest and exceeds the allowable range, adjust the cavity pressure first to ensure that it returns to the target value range. If multiple parameter deviations are out of range at the same time, formulate a joint adjustment strategy based on the correlation between parameters (such as pressure and temperature coupling).

[0083] When calculating the optimal adjustment value, avoid adjusting multiple parameters at the same time to reduce the risk of process instability caused by the adjustment. For example: first adjust the cavity pressure to restore it to the normal range; then gradually adjust secondary parameters such as temperature and injection speed. Verify the feasibility of each adjustment plan to ensure that it does not violate the equipment operation safety constraints. For example: before adjusting the cavity pressure, confirm the response capacity of the hydraulic system to avoid exceeding the hydraulic limit.

[0084] The optimal process adjustment plan is output, the operating parameters of the equipment are adjusted in real time, and the adjustment results are verified by closed-loop feedback.

[0085] Specifically, the adaptive control module sends the generated optimal process adjustment plan to the equipment control module in the form of a digital signal, and the adjustment is performed through the automation system. For example, the screw motion parameters are adjusted through the servo controller; the mold temperature is accurately adjusted through the temperature control system.

[0086] After the adjustment is completed, the equipment operating parameters and product quality indicators (such as molding size, weight, etc.) are monitored in real time.

[0087] Perform closed-loop feedback verification on the adjustment results: If the adjustment results meet the process goals, the current parameters will be saved in the historical process database; if the adjustment results do not meet expectations, analyze the reasons and further optimize the adjustment strategy.

[0088] During the whole process, the adaptive control module runs continuously to form a closed-loop control system that dynamically adjusts, verifies and optimizes.

[0089] Furthermore, the formula of the fuzzy adaptive resonance theory is as follows:

[0090]

[0091] Where A(t) represents the optimal process adjustment solution at time t; W i represents the weight of the i-th key process parameter; P i (t) represents the deviation value of the i-th key process parameter; K i Represents the nonlinear suppression factor of the i-th parameter, which is used to reduce the excessive impact of certain high-fluctuation parameters (such as injection pressure or flow rate) on the overall strategy and avoid too frequent equipment adjustments; Q i(t) represents the dynamic relationship between different parameters in the injection molding process, such as the coupling between the mold cavity filling state and the flow rate or the effect of the injection temperature on the internal stress distribution; D i (t) represents the dynamic changes of real-time acquisition in multidimensional data; R i (t) represents the influence of random fluctuations in the correction parameters; T i Indicates the response delay of the injection molding equipment to the key process parameters; M represents the global dynamic adjustment coefficient; represents the rate of change of the i-th key process parameter; m represents the total number of key process parameters.

[0092] The finished product quality inspection module is used to perform real-time online inspection of injection molded products through machine vision technology and embedded image processing algorithms, analyze the dimensional accuracy, surface defects, density distribution and internal stress of the finished products, and determine the quality level of the finished products through a prediction model to generate a quality inspection analysis report;

[0093] The operation process of the finished product quality inspection module includes the following steps:

[0094] The appearance characteristics and internal structure data of finished products are collected through multi-angle industrial cameras, laser scanning sensors and stress detection equipment, and multi-source data are integrated;

[0095] Specifically, multi-angle industrial cameras (covering multiple viewing angles of the finished product), laser scanning sensors (for measuring dimensions and surface profiles), and stress detection equipment (such as infrared stress testers or ultrasonic sensors) are installed on the injection molding production line to ensure that all equipment is seamlessly connected to the quality inspection module.

[0096] Industrial cameras: Real-time capture of the finished product's appearance and acquisition of multi-angle surface feature data. For example, the top view captures the finished product's plane dimensional features; the side view captures the thickness and edge integrity.

[0097] Laser scanning sensor: Generates high-precision 3D point cloud data of finished products through non-contact laser scanning for dimensional accuracy and density distribution analysis.

[0098] Stress detection equipment: Use infrared or ultrasonic technology to detect the internal stress distribution of the finished product and record abnormal areas.

[0099] Divide the finished product into regions, identify key inspection areas, and use regional specialization algorithms to perform detailed analysis on key areas to generate finished product feature parameters;

[0100] Specifically, the surface and interior of the finished product are divided into regions based on the 3D point cloud data and image features of the finished product. For example, the finished product is divided into edge regions (such as interfaces and fastener parts) and central regions (such as planes and interiors) using a region segmentation algorithm. Critical areas that may have quality risks are marked, such as joints, curved areas, or parts with thickness changes.

[0101] Classify the divided areas and determine the areas that need to be inspected, for example: whether there is dimensional deviation at the interface of the finished product; whether there are scratches or uneven colors on the smooth surface; whether there are bubbles or stress concentration areas in the internal structure.

[0102] For key areas, we use regional specialization algorithms to conduct detailed analysis, for example:

[0103] Size detection algorithm: In the edge area, measure key parameters such as thickness, width, and gap;

[0104] Surface defect detection algorithm: detects tiny scratches, dents or color changes in smooth areas;

[0105] Stress concentration detection algorithm: In the internal stress distribution data, mark the area where the stress exceeds the set threshold.

[0106] Generate a multidimensional feature list containing regional characteristic parameters as the basis for the next step of analysis.

[0107] Based on the embedded deep learning algorithm, the characteristic parameters of the finished product are gradually detected to identify defects including dimensional deviation, surface scratches, uneven color and internal bubbles; combining historical quality data and process parameters to establish a dynamic defect feature library for real-time optimization of defect recognition accuracy;

[0108] Specifically, the embedded deep learning model is trained to identify various defects in finished products, including: using point cloud data to compare the target model to detect oversized or undersized size deviations; analyzing image features through a convolutional neural network (CNN) to identify the length, depth, and distribution of scratches; identifying color difference areas through color histogram analysis; and identifying the location and volume of bubbles based on abnormal stress or density distribution.

[0109] The detected defects and their characteristic parameters (such as location, size, shape) are recorded in the dynamic defect feature library. Combined with historical quality inspection data and process parameters (such as injection temperature and holding time), a correlation model between defects and process conditions is established. By dynamically updating the feature library and continuously training the deep learning model, the ability to identify small or new defects is improved.

[0110] Combined with machine learning algorithms, comprehensive analysis of the test results is performed to calculate the quality scores of the finished products in multiple dimensions, the finished products are graded, and a visual quality report is generated; the machine learning algorithm includes a decision tree or a support vector machine algorithm to realize intelligent evaluation of the quality grade of the finished product.

[0111] Specifically, machine learning algorithms are used to conduct comprehensive analysis of the detected multi-dimensional defect data, such as:

[0112] Decision tree algorithm: Step by step, analyze the impact of each defect on the quality of the finished product. For example, a scratch depth greater than 0.2 mm is considered a serious defect, affecting the grade of the finished product; a size deviation less than 1% is considered a minor defect.

[0113] Support Vector Machine Algorithm (SVM): Multidimensional parameters are input into the SVM model, and the quality scores of finished products are classified according to the sample training results.

[0114] Generate a quality inspection analysis report containing the following information:

[0115] Inspection overview: finished product inspection time, inspection equipment, finished product number;

[0116] Defect distribution: Multi-dimensional visualization charts showing defect location, type, and degree (such as scratch location distribution heat map);

[0117] Quality grade: clearly mark the comprehensive quality grade of the finished product;

[0118] Optimization suggestions: Provide targeted process improvement suggestions based on the test results, such as increasing the holding time or lowering the injection temperature.

[0119] The monitoring and early warning module is used to use an abnormality detection algorithm to monitor the operating status and process parameters of the injection molding equipment in real time, identify potential fault characteristics under complex nonlinear relationships, and generate a monitoring and analysis report, trigger an early warning signal, and provide process adjustment suggestions; the monitoring and early warning module specifically adopts an abnormal feature association analysis method based on a knowledge graph, and by establishing a dynamic association network between process parameters, equipment status and fault modes, it realizes multi-source early warning signal identification and intelligent fault root cause tracing under complex nonlinear working conditions.

[0120] It should be noted that the abnormal characteristics are graded according to their severity:

[0121] Level 1 abnormality: A slight deviation from the normal range (such as a small temperature fluctuation) will generate a prompt message.

[0122] Secondary abnormality: A significant deviation from the normal range (such as a rapid drop in injection pressure) generates an early warning and recommends shutdown for inspection.

[0123] Level 3 abnormality: When key parameters exceed the safety range, an alarm is immediately triggered and the emergency shutdown procedure is initiated.

[0124] The node design for knowledge graph construction is as follows:

[0125] Process parameter nodes: injection pressure, temperature, flow rate, filling time, etc.

[0126] Equipment status nodes: vibration characteristics, motor current, voltage fluctuations, and mechanical component status.

[0127] Failure mode nodes: such as mold blockage, insufficient lubrication, equipment aging, sensor failure, etc.

[0128] Edge design: Establish dynamic association rules between nodes through historical data and domain knowledge. For example:

[0129] Cavity pressure fluctuations → temperature increase → longer filling time → finished product defects (node ​​path).

[0130] Embed the knowledge graph and use the graph neural network model to extract key features and multi-parameter nonlinear relationships. Update node status in real-time data streams, dynamically adjust association weights, and identify potential complex failure modes. Based on the knowledge graph and real-time monitoring data, match possible failure paths and locate the root cause. For example: if cavity pressure fluctuations and filling time anomalies occur at the same time, combined with graph analysis, it may be caused by increased melt flow resistance.

[0131] Based on real-time monitoring data and anomaly detection models, multi-level warnings are triggered:

[0132] Prompt level: Prompts the operator with suggestions for process adjustments, such as optimizing the injection temperature or reducing the injection speed.

[0133] Alarm level: When key equipment parameters reach dangerous thresholds, an alarm is triggered immediately and shutdown inspection is recommended.

[0134] Based on the knowledge graph and anomaly types, combined with historical adjustment records and optimization rules, specific suggestions are generated, such as:

[0135] Abnormal cavity pressure → It is recommended to check the melt temperature and flow channel patency;

[0136] The equipment vibrates too much → Check the wear of mechanical parts and add lubrication.

[0137] Manual Adjustment: Multiple adjustment options are provided through the display interface for the operator to choose from.

[0138] Automatic Adjustment: Adjustment suggestions are sent directly to the control module to optimize equipment parameters (such as reducing screw speed or increasing holding time).

[0139] In summary, the present invention realizes high-frequency and precise collection of multi-dimensional parameters through the injection molding data acquisition module, covering key indicators such as pressure, temperature, flow rate, cavity filling status and equipment operation status. Combining high-precision sensors with high-speed data transmission technology, the synchronization and consistency of data are guaranteed, providing a reliable foundation for subsequent process optimization and model construction. Based on the spline curve fitting algorithm of deep learning, nonlinear modeling is performed on the multi-dimensional data of the injection molding process. By extracting dynamic features and key process parameters, a high-precision spline model is generated, which accurately describes the dynamic changes and correlation characteristics of process data, and improves the description ability of complex process.

[0140] The present invention uses fuzzy adaptive resonance theory to analyze parameter deviations in the injection molding process in real time, dynamically generate the best process adjustment strategy, and realize real-time optimization of key parameters through closed-loop control. It effectively reduces product defects, improves process stability and finished product quality, and enhances the adaptability and robustness of the system. Combined with multi-angle industrial cameras, laser scanning sensors and stress detection equipment, the appearance characteristics, dimensional accuracy and internal stress of the finished product are comprehensively detected. Through embedded deep learning algorithms and dynamic defect feature libraries, not only the accuracy of defect identification is improved, but also the dynamic correlation analysis of defects and process conditions can be realized, providing guidance for further process optimization.

[0141] The abnormal feature association analysis method based on the knowledge graph in the present invention constructs a dynamic network between process parameters, equipment status and failure mode, and can identify potential failure characteristics under complex nonlinear relationships. The fault prediction and response capabilities of the system are improved through the hierarchical early warning mechanism and intelligent root cause tracing. Efficient data management is achieved through the collaborative design of distributed acquisition and centralized storage; the dynamic optimization capability of the adaptive control module reduces human intervention and debugging time, and improves production efficiency. At the same time, accurate real-time monitoring and automatic adjustment strategies effectively reduce equipment wear and energy consumption, and reduce operating costs.

[0142] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An adaptive fully automatic spline injection molding system, characterized in that: It includes an injection molding data acquisition module, a spline model generation module, an adaptive control module, a finished product quality detection module and a monitoring and early warning module which are sequentially connected in communication; The injection molding data acquisition module is used to collect multi-dimensional data of the injection molding process in real time through a sensor network; The spline model generation module is used to preprocess the multidimensional data, extract key process parameters, and construct a nonlinear spline model using a spline curve fitting algorithm based on deep learning; The adaptive control module is used to generate an optimal process adjustment strategy based on the nonlinear spline model and adopt fuzzy adaptive resonance theory to adjust the key process parameters of the injection molding equipment in real time; The finished product quality inspection module is used to perform real-time online inspection of injection molded products through machine vision technology and embedded image processing algorithms, analyze the dimensional accuracy, surface defects, density distribution and internal stress of the finished products, and determine the quality level of the finished products through a prediction model to generate a quality inspection analysis report; The monitoring and early warning module is used to use an abnormality detection algorithm to monitor the operating status and process parameters of the injection molding equipment in real time, identify potential fault characteristics under complex nonlinear relationships, generate a monitoring and analysis report, trigger an early warning signal, and provide process adjustment suggestions.

2. The adaptive fully automatic spline injection molding system according to claim 1, characterized in that: The multi-dimensional data includes injection pressure, temperature, flow rate, mold cavity filling status and injection equipment operation status data.

3. The adaptive fully automatic spline injection molding system according to claim 1, characterized in that: The operation process of the spline model generation module includes the following steps: Dividing the multidimensional data into multiple process stages according to the time series, and extracting the dynamic characteristic parameters of each stage, including key change trends and correlation indicators, in combination with the injection molding process rules; Based on the process fluctuation characteristics, an adaptive data window is designed to optimize and analyze the extracted dynamic features window by window to screen out the key process parameters; A spline curve fitting algorithm based on deep learning is used to perform nonlinear modeling on the key process parameters, and the parameters of the fitting function are optimized through iterative training to generate a nonlinear spline model.

4. The adaptive fully automatic spline injection molding system according to claim 3, characterized in that: The formula of the nonlinear spline model is as follows: Where y(t) represents the predicted value of the key process parameters at time t; φ i (t) represents the spline basis function; w i represents the weight of the spline basis function; ψ j (t) represents the dynamic characteristic function of the injection molding process; exp(-β j ·ψ j (t)) represents the nonlinear response function; α j Indicates the contribution strength of each feature function to the model output; β j represents the time attenuation coefficient of the dynamic response; γ·sin(ωt) represents the potential periodic fluctuations in the injection molding process; γ represents the intensity of the influence of periodic fluctuations on process parameters; ω represents the period length of process parameter fluctuations; δ·t p represents the time trend term; n represents the number of spline basis functions; m represents the number of dynamic characteristic functions.

5. The adaptive fully automatic spline injection molding system according to claim 1, characterized in that: The operation process of the adaptive control module includes the following steps: Based on the nonlinear spline model, the dynamic multi-dimensional data transmitted by the injection molding data acquisition module is received in real time, and the current process state is dynamically analyzed by using a fuzzy reasoning algorithm in combination with historical process data to identify the degree of deviation of key parameters; Using fuzzy adaptive resonance theory, the deviation of key parameters is evaluated in real time, a multi-dimensional deviation matrix is ​​generated, and real-time adjustment constraints are established according to the process optimization goals; Based on the multi-dimensional deviation matrix and constraints, the process adjustment strategy is calculated through a dynamic optimization algorithm to generate the best process adjustment plan; The optimal process adjustment plan is output, the operating parameters of the equipment are adjusted in real time, and the adjustment results are verified by closed-loop feedback.

6. The adaptive fully automatic spline injection molding system according to claim 5, characterized in that: The formula of the fuzzy adaptive resonance theory is as follows: Where A(t) represents the optimal process adjustment solution at time t; W i represents the weight of the i-th key process parameter; P i (t) represents the deviation value of the i-th key process parameter; K i represents the nonlinear suppression factor of the i-th parameter; Q i (t) represents the dynamic relationship between different parameters in the injection molding process; D i (t) represents the dynamic changes of real-time acquisition in multidimensional data; R i (t) represents the influence of random fluctuations in the correction parameters; T i Indicates the response delay of the injection molding equipment to the key process parameters; M represents the global dynamic adjustment coefficient; represents the rate of change of the i-th key process parameter; m represents the total number of key process parameters.

7. The adaptive fully automatic spline injection molding system according to claim 1, characterized in that: The operation process of the finished product quality inspection module includes the following steps: The appearance characteristics and internal structure data of finished products are collected through multi-angle industrial cameras, laser scanning sensors and stress detection equipment, and multi-source data are integrated; Divide the finished product into regions, identify key inspection areas, and use regional specialization algorithms to perform detailed analysis on key areas to generate finished product feature parameters; Based on the embedded deep learning algorithm, the characteristic parameters of the finished product are gradually detected to identify defects including dimensional deviation, surface scratches, uneven color and internal bubbles; combining historical quality data and process parameters to establish a dynamic defect feature library for real-time optimization of defect recognition accuracy; Combined with machine learning algorithms, the test results are comprehensively analyzed, the quality scores of the finished products in multiple dimensions are calculated, the finished products are graded, and a visual quality report is generated.

8. The adaptive fully automatic spline injection molding system according to claim 7, characterized in that: The machine learning algorithm includes a decision tree or a support vector machine algorithm to realize intelligent evaluation of the quality level of the finished product.

9. The adaptive fully automatic spline injection molding system according to claim 1, characterized in that: The monitoring and early warning module specifically adopts an abnormal feature association analysis method based on a knowledge graph, and realizes multi-source early warning signal recognition and intelligent fault root cause tracing under complex nonlinear working conditions by establishing a dynamic association network between process parameters, equipment status and failure modes.

Images