A method and device for dynamically adjusting process parameters for plastic mold production

Abstract

The present application relates to the field of production process adjustment, and particularly relates to a process parameter dynamic adjustment method and device for plastic mold production. The method comprises the following steps: performing multi-stage image acquisition on a plastic mold production line to construct a staged image sequence; performing abnormal defect identification on the staged image sequence to mark defect precursor information; performing correlation analysis on the defect precursor information based on the staged image sequence to mark key defect precursors; collecting multi-stage process parameter streams of a real-time production process; performing matching extraction on the multi-stage process parameter streams based on the key defect precursors to obtain defect process parameter streams; performing local process adjustment based on the defect process parameter streams to generate an optimal process adjustment set; and performing normalized mold production operations based on the optimal process adjustment set. The present application improves the product quality, stability and efficiency of plastic mold production by collecting and analyzing multi-stage production data in real time.

Description

Technical Field

[0001] This invention relates to the field of production process adjustment, and in particular to a method and apparatus for dynamically adjusting process parameters in plastic mold production. Background Technology

[0002] In traditional plastic mold manufacturing, process parameters are primarily set using empirical formulas and manual adjustments. For example, parameters such as injection temperature, pressure, cooling time, and screw speed are typically determined based on operator experience or post-production inspection results. While this method can guarantee a basic pass rate to some extent, it often suffers from problems like lag in adjustment, inaccurate parameters, and slow defect response when facing complex process conditions, multiple material applications, and the consistency requirements of batch production. This not only easily leads to quality issues such as dimensional deviations, surface defects, and uneven internal stress, but can also increase raw material waste and reduce production efficiency. With the development of intelligent manufacturing and industrial automation, the amount of data in the production environment is rapidly increasing, and traditional static process parameter control methods are no longer sufficient to meet the needs of dynamic adjustment and real-time optimization. Current process monitoring methods mostly rely on manual inspections, periodic checks, or simple sensor data collection, failing to achieve real-time status perception, defect precursor identification, and automated process adjustment throughout the entire mold production process. This makes it difficult for production lines to respond quickly to sudden process fluctuations and potential defects, affecting production stability and product quality reliability. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a method and apparatus for dynamically adjusting process parameters in plastic mold production, thereby resolving at least one of the aforementioned technical problems.

[0004] To achieve the above objectives, the present invention provides a method for dynamically adjusting process parameters in plastic mold production, comprising the following steps: Step S1: Perform multi-stage image acquisition on the plastic mold production line and construct a staged image sequence; Step S2: Identify abnormal defects in the staged image sequence and mark the precursor information of defects; Step S3: Based on the staged image sequence, perform correlation analysis on the defect precursor information and mark key defect precursors; Step S4: Collect multi-stage process parameter streams of the real-time production process; match and extract the multi-stage process parameter streams based on the precursors of critical defects to obtain the defect process parameter streams; Step S5: Perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; execute normalized mold production operations based on the optimal process adjustment set.

[0005] This specification provides a device for dynamically adjusting process parameters in plastic mold production, used to execute the method for dynamically adjusting process parameters in plastic mold production as described above, including: The image acquisition module is used to acquire images of the plastic mold production line in multiple stages and construct a staged image sequence. The defect identification module is used to identify abnormal defects in staged image sequences and mark defect precursor information. The correlation analysis module is used to perform correlation analysis on defect precursor information based on the staged image sequence and mark key defect precursors; The parameter extraction module is used to collect multi-stage process parameter streams in real-time production; and to match and extract the multi-stage process parameter streams based on key defect precursors to obtain defect process parameter streams. The process adjustment module is used to perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; and to perform normalized mold production operations based on the optimal process adjustment set.

[0006] The beneficial effects of this invention are as follows: By acquiring images of multiple key stages of the production process, each step of mold forming can be clearly recorded, forming a complete visualized sequence of the production process. This staged image sequence provides a rich data foundation for subsequent defect detection, enabling the identification of potential defects at different stages, rather than only discovering problems during final product inspection. Image analysis identifies anomalies or potential defects, marking precursory information before defects fully form, thereby reducing scrap rates. Automatic identification of defect precursors reduces manual inspection of the mold production line, improving production efficiency. By analyzing the relationship between precursory information at each stage and the final defect, it is possible to distinguish which precursors are key factors leading to defects, avoiding ineffective adjustments. Marking key precursors means that subsequent process parameter adjustments can be focused on the most important stages, reducing the number of adjustments and resource waste. Identifying key precursors helps build predictive models, enabling the production line to have early warning capabilities for potential problems, improving overall quality stability. Matching key defect precursors with real-time process parameters at the corresponding stages accurately identifies parameter fluctuations or anomalies causing defects. The extracted defect process parameter stream provides a precise basis for subsequent local process adjustments, avoiding blind adjustments to the entire production line. This establishes a closed-loop mechanism of "defects, parameters, and adjustments," shifting mold manufacturing processes from experience-driven to data-driven, thus enhancing the level of intelligent production. Local optimization of key defect parameters effectively reduces the defect rate and improves the first-pass yield of molded products. The optimal process adjustment set stabilizes the production process, reducing the impact of human experience on product quality. Real-time monitoring, analysis, and adjustment of process parameters enable the production line to respond quickly to process fluctuations, improving flexible manufacturing capabilities. Attached Figure Description

[0007] Fig. 1 This is a schematic diagram of the steps of a method for dynamically adjusting process parameters in plastic mold production according to the present invention; Fig. 2 This is a detailed flowchart illustrating the implementation steps of step S1. Fig. 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0008] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0009] This application provides a method and apparatus for dynamically adjusting process parameters in plastic mold production. The executing entities of the method and apparatus for dynamically adjusting process parameters in plastic mold production include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud data management system.

[0010] Please see Figs. 1 to 3 This invention provides a method for dynamically adjusting process parameters in plastic mold production, comprising the following steps: Step S1: Perform multi-stage image acquisition on the plastic mold production line and construct a staged image sequence; Step S2: Identify abnormal defects in the staged image sequence and mark the precursor information of defects; Step S3: Based on the staged image sequence, perform correlation analysis on the defect precursor information and mark key defect precursors; Step S4: Collect multi-stage process parameter streams of the real-time production process; match and extract the multi-stage process parameter streams based on the precursors of critical defects to obtain the defect process parameter streams; Step S5: Perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; execute normalized mold production operations based on the optimal process adjustment set.

[0011] In the embodiments of the present invention, see Fig. 1 This is a schematic flowchart illustrating the steps of a method for dynamically adjusting process parameters in plastic mold production according to the present invention. In this example, the steps of the method for dynamically adjusting process parameters in plastic mold production include: Step S1: Perform multi-stage image acquisition on the plastic mold production line and construct a staged image sequence; In this embodiment, during the plastic mold production process, multi-stage image acquisition is performed on the production line to record the entire injection molding operation from melt filling, holding pressure, cooling and shaping to demolding. An industrial camera triggers capture along the production cycle, with each stage's image resolution set to 1920×1080 pixels or higher, and a frame rate of 50–100 fps to ensure the capture of dynamic details of the mold. The acquired images include the melt flow front at the initial filling stage, the surface expansion evolution during the holding pressure stage, the edge contraction during the cooling stage, and the structural integrity after demolding. The acquired images undergo filtering and enhancement processing, such as local histogram equalization, edge sharpening, and noise suppression, forming a staged image sequence. Each frame is accompanied by a precise timestamp, synchronized with the production action, achieving a one-to-one correspondence between images and production cycles.

[0012] Step S2: Identify abnormal defects in the staged image sequence and mark the precursor information of defects; In this embodiment, after obtaining the staged image sequence, abnormal defects are identified on the mold surface and structure through image processing and analysis. Local structural features are extracted using edge detection, texture analysis, and gradient change algorithms to identify precursors such as excessively rapid edge shrinkage, surface unevenness, local warping, or abnormal texture. For each image frame, indices are calculated based on local brightness, contrast, and texture uniformity, and compared with normal mold feature thresholds. For example, if the edge gradient is below 10-15 gray units or the local texture variance exceeds a set range, it is determined to be a potential defect precursor. After identification, the abnormal areas are marked on the image to form a defect precursor information mask, and the timestamp and spatial location are recorded simultaneously.

[0013] Step S3: Based on the staged image sequence, perform correlation analysis on the defect precursor information and mark key defect precursors; In this embodiment, after marking defect precursor information, spatiotemporal correlation analysis is performed to distinguish between critical defect precursors and occasional noise. The analysis method correlates the spatial location, morphological characteristics, and evolution trend of defect precursors with the morphological changes of the mold at each production stage, and identifies recurring or continuously expanding abnormal areas through cluster analysis or temporal correlation assessment. Areas with significant edge shrinkage, persistent uneven texture, or rapid local warping are identified as critical defect precursors, and these critical areas are marked in the staged image sequence. During the analysis, the rate of change can be calculated in conjunction with the morphological evolution curve; for example, an edge shrinkage rate exceeding 0.5 mm / s or a decrease in texture uniformity exceeding 15% can be used as criteria for judging critical precursors.

[0014] Step S4: Collect multi-stage process parameter streams of the real-time production process; match and extract the multi-stage process parameter streams based on the precursors of critical defects to obtain the defect process parameter streams; In this embodiment, during the production process, a sensor network deployed at key nodes of the injection molding machine collects real-time process parameters, including melt temperature profiles, injection speed segments, mold temperature distribution, holding pressure decay rate, and cooling time. The collected multi-source data is timestamped and filtered to form a synchronized, smooth, multi-stage process parameter stream. Subsequently, key defect precursor markers are matched with the timestamps of the process parameter stream to extract the process parameter data corresponding to the defect occurrence stage, forming a defect process parameter stream. This stream contains continuous records of all key parameters within the defect stage. During the matching process, time window precision control is used to align the process parameters with the image stage, for example, within ±20 ms, ensuring that the process state of the defect stage is fully captured, achieving an accurate correspondence between process parameters and defect evolution.

[0015] Step S5: Perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; execute normalized mold production operations based on the optimal process adjustment set.

[0016] In this embodiment, based on the analysis results of defect process parameters and molding deviations, local adjustments are made to the injection speed, injection pressure, holding time, mold temperature, and cooling time. The local adjustments are calculated using a multi-objective optimization method to determine the optimal parameter adjustment range. For example, increasing the holding pressure by 5-10%, extending the cooling time by 2-3 seconds, or fine-tuning the injection speed by 2-5% can correct deviations and eliminate early signs of defects. The local adjustment values ​​are integrated with the overall process parameters to generate an optimal process adjustment set, considering constraints such as equipment limits, material properties, and production cycle requirements. After optimization, adjustment commands are issued through the control interface, causing the injection molding machine to operate according to the optimal process during the defect stage, achieving normalized mold production operations. During execution, process parameters and staged images are continuously collected to form a closed-loop feedback, ensuring that the adjustment effect meets expectations, improving molding accuracy, reducing scrap rate, and maintaining production stability, thus achieving dynamic optimization control of plastic mold production.

[0017] In this embodiment, see Fig. 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Start the plastic mold production line to enter the production testing cycle and execute plastic mold production; Multi-stage image acquisition of the production line is performed using an industrial camera to extract multi-stage images of the mold; the multi-stage images of the mold include images of the initial stage of melt filling, surface evolution images of the holding pressure stage, edge shrinkage images of the cooling and shaping stage, and images of the complete structure after demolding. Adaptive enhancement is performed on multi-stage images of the mold, and the filtered image is output. Calculate the timestamp of the filtered image frame by frame; Based on the timestamp, the production stage is identified to obtain a production stage label; The filtered images are fitted with stage labels based on the production stage labels to construct a staged image sequence.

[0018] In this embodiment, after production preparation is completed, the plastic mold production line is started, allowing the equipment to enter the production testing cycle. During startup, the injection molding machine heats the raw material to a set melting temperature, such as 220–250°C, ensuring the melt has good fluidity and mold-filling properties. After startup, the production line executes injection actions according to a set rhythm, including melt injection, holding pressure, cooling and shaping, and demolding. The time for each stage can be set to 2–4 seconds for injection, 3–5 seconds for holding pressure, and 10–15 seconds for cooling, to adapt to different mold specifications and product structures. During production, stable production of the plastic mold is achieved by controlling key process parameters such as screw speed, injection pressure, and holding pressure. When the production line is running continuously, real-time monitoring of temperature, pressure, and flow sensor data is also required to ensure consistency in each injection process. During injection, an industrial camera is used to acquire multi-stage images of the mold according to the production rhythm. The acquired multi-stage images of the mold include images of the initial melt filling stage, surface evolution images during the holding pressure stage, edge shrinkage images during the cooling and shaping stage, and images of the complete structure after demolding. Industrial cameras are mounted above or to the side of the mold, and capture images synchronously with the production line movements via trigger signals, ensuring clear images of the process status at each stage. The camera resolution can be set to 1920×1080 pixels or higher, with a frame rate of 50–100 fps to capture dynamic changes such as melt filling, pressure holding expansion, and cooling contraction. During acquisition, light source adjustment, shutter control, and automatic exposure compensation ensure uniform image brightness and sufficient contrast, avoiding image feature loss due to shadows or reflections.

[0019] Adaptive enhancement is applied to the acquired multi-stage images of the mold to highlight key structural and process features while suppressing noise and background interference. Adaptive enhancement methods can employ local contrast enhancement, histogram equalization, or multi-scale filtering to dynamically adjust images at different stages. For example, images from the initial stage of melt filling emphasize the flow front contour; images from the surface evolution stage during the holding pressure stage enhance surface protrusions or ripple details; images from the edge contraction stage during cooling and shaping enhance contour lines; and images from the post-demolding stage retain overall geometric features.

[0020] For the enhanced and filtered mold image sequence, timestamps are calculated frame by frame according to the acquisition sequence to accurately mark the production time point corresponding to each frame. Timestamp calculation is obtained by triggering signals synchronized with the injection molding action on the production line, such as generating time stamps at the start of injection, pressure holding, cooling start, and demolding trigger points. The inter-frame time interval can be determined according to the camera sampling frame rate; for example, at 50 fps, the inter-frame time is 20 ms, ensuring that each frame accurately corresponds to the production line operation state. The frame-by-frame timestamp record not only includes the image acquisition time but can also be time-calibrated by combining sensor data such as screw position, injection pressure, or mold temperature, enhancing the spatiotemporal correspondence between images and production process parameters. Using the frame-by-frame timestamps and the production line action cycle, production stage identification is performed on the filtered image sequence. Identification methods can be implemented through time period threshold matching or image feature analysis; for example, the injection period can be identified as the initial stage of mold filling, the pressure holding stage as the surface evolution stage, the cooling period as the edge shrinkage stage, and the demolding period as the structural integrity stage. Auxiliary judgment is made by combining image texture changes, edge detection results, and brightness distribution characteristics to improve the accuracy of stage identification. Through stage recognition, each frame of filtered image obtains a corresponding production stage label, such as "initial mold filling," "holding pressure stage," "cooling and shaping," and "demolding completion." The image sets for each production stage are sorted, aligned, and continuously fitted to ensure the image sequence clearly reflects the process evolution from mold filling to demolding. During the fitting process, edge detection, contour tracking, and curve fitting methods are used to extract stage features, such as the shape of the leading edge of the mold filling, the expansion of the holding pressure surface, the shrinkage edge during cooling, and the final structural contour, and the corresponding stage is marked in the image sequence. The staged image sequence not only visually displays the mold forming process but can also be used for dynamic adjustment of process parameters, such as adjusting injection speed, holding pressure, or cooling time, to achieve real-time production optimization.

[0021] In this embodiment, the specific steps for adaptively enhancing the multi-stage image of the mold and outputting the filtered image are as follows: The multi-stage images of the mold are segmented to obtain multiple local regions; Calculate the brightness and contrast of the multiple local regions; Based on the brightness value and contrast, the difference in brightness distribution between regions is calculated to obtain the difference in brightness distribution between different regions. Based on the brightness distribution difference, perform zoned adaptive brightness adjustment to obtain a brightness-adjusted image; The edge sharpness and detail gradient changes of the brightness-adjusted image are evaluated, local blurring and detail loss are identified, and abnormal locations are marked. Spatial resolution compensation is performed on abnormal locations to construct resolution-enhanced images; Dynamic filtering optimization is performed on the resolution-enhanced image to output a filtered image.

[0022] In this embodiment, each frame of the image is spatially divided into multiple local regions to facilitate independent analysis of local features. The division can employ a grid-based method, dividing the image into several uniform small blocks horizontally and vertically, such as 32×32 pixel or 64×64 pixel grids. The granularity of the division is adjusted according to the mold size and image resolution to ensure that each local region contains sufficient structural information while highlighting local brightness and detail variations. During the division process, adaptive segmentation methods can also be used, incorporating mold geometric features, to distinguish key areas such as gates, cavity edges, or complex structures, thereby ensuring more precise analysis of functionally critical areas. After completing the local region division, the brightness and contrast values ​​of each region are calculated. The brightness value can be obtained by averaging or weighted averaging the pixel grayscale values ​​within the region, reflecting the overall brightness level of that region; the contrast is calculated by the standard deviation or local variance of the pixel grayscale values ​​within the region, characterizing the grayscale variation range and detail levels within the region. During the calculation process, a local weighting algorithm can be used to enhance the sensitivity to edges and fine structures. For example, when calculating the grayscale mean, edge pixels can be given higher weights to more accurately reflect local changes.

[0023] A brightness distribution difference matrix is ​​formed by comparing the brightness and contrast differences between each region and its neighboring regions. This matrix reflects the local brightness variations caused by uneven lighting, excessive reflection, or shadows in different areas of the mold surface. Weighting coefficients can be introduced during the calculation process, assigning higher weights to structurally important or key contour areas to enhance sensitivity to brightness anomalies in critical areas. Quantitative analysis of brightness differences between regions identifies areas that are too dark or too bright, providing precise parameters for zoned adaptive brightness adjustment. Adaptive brightness adjustment is performed on each local region to achieve balanced brightness while preserving local details and structural features. Adaptive brightness adjustment can employ local histogram equalization or multi-scale contrast stretching methods, increasing gain in areas with low brightness and compressing areas with high brightness to achieve a more uniform overall brightness distribution. Gain limits can also be set during the adjustment process to prevent over-enhancement that could lead to detail loss or saturation. After processing, a brightness-adjusted image is output, ensuring that key structures on the mold surface are clearly presented in images at each production stage, providing a reliable foundation for subsequent edge sharpness and detail gradient analysis.

[0024] The sharpness of each local region is evaluated by calculating gradient magnitude or detecting edge intensity changes using the Laplacian operator. Regions with gradient magnitudes below a set threshold are identified as areas of local blur or loss of detail and marked as anomalous locations. Simultaneously, edge continuity is detected; interrupted edges or abnormal contour shapes are also marked as anomalous. The threshold can be set according to image resolution, camera acquisition conditions, and mold structure dimensions; for example, the gradient magnitude threshold can be set to 10–15 grayscale units. After marking, anomalous locations are clearly identified in the image.

[0025] For marked abnormal locations, spatial resolution compensation techniques are applied for local enhancement. Local super-resolution reconstruction algorithms, interpolation amplification, or multi-frame information fusion can be used to refine blurred or detail-loss-laden areas at the pixel level, improving the recognizability of local texture and edge information. During the compensation process, guided interpolation is performed by incorporating information from neighboring clear areas to maintain edge coherence and structural integrity. The resulting enhanced image ensures that key features and complex structures of the mold are clearly identifiable at each production stage. Based on this enhanced image, dynamic filtering optimization is performed to balance detail preservation and noise suppression. Dynamic filtering can employ adaptive Gaussian filtering, bilateral filtering, or non-local mean filtering to process images with different brightness, texture, and noise levels. Stronger filtering is applied to flat areas to reduce noise, while lower-intensity filtering is used for edges and textured areas to preserve details. Filtering parameters can be dynamically adjusted based on local image contrast and gradient values; for example, the filter kernel size can change from 3×3 to 7×7 pixels depending on local features. After dynamic filtering optimization, the final filtered image is output, eliminating random noise while preserving key details of the mold surface and edges.

[0026] In this embodiment, see Fig. 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Perform stage-by-stage recognition and instance segmentation on the staged image sequence, and mark the different structural positions of each stage; Visual measurements were performed on the different structural locations to obtain mold morphology parameters at different stages; the morphology parameters included geometric contour, edge change rate, surface texture uniformity, and local shrinkage characteristics. Continuous optical flow tracing of the mold morphology parameters yields the morphology evolution curve; The rate and magnitude of change of each parameter are calculated based on the morphological evolution curve; Based on the rate and magnitude of change, abnormal defects are identified, and precursor information of defects is marked.

[0027] In this embodiment, after obtaining the staged image sequence, each production stage image is identified step-by-step to distinguish different structural parts of the mold, allowing for independent analysis of local features. Step-by-step identification can be initially grouped by combining timestamps and production stage labels. Pixel-level segmentation of different structures in the image is then performed using deep learning instance segmentation methods, such as Mask R-CNN or a segmentation model based on convolutional neural networks. Instance segmentation can identify the melt filling leading edge, the surface expansion area during the holding pressure stage, the edge contraction area during the cooling stage, and the complete structural area after demolding. Each structural location is marked as an independent object, and a corresponding mask is generated. Segmentation accuracy is controlled above 0.85 through cross-validation or the IoU metric to ensure that key structural areas are completely identified. After marking the structural locations, visual measurements are performed on each local area to obtain mold morphological parameters. These parameters include geometric contours, edge change rate, surface texture uniformity, and local contraction features. Geometric contours are extracted using edge detection algorithms, such as the Canny or Sobel operators, to obtain contour curves and calculate contour length, area, and curvature changes. Edge change rate is calculated using the time difference of contour changes across consecutive frames, reflecting the shape evolution speed during the filling and cooling stages. Surface texture uniformity is quantified using local grayscale variance or texture feature extraction algorithms (such as GLCM) to describe the smoothness and uniformity of the mold surface. Local shrinkage features are achieved by measuring edge shrinkage amplitude and thickness changes. During measurement, the dimensional parameters can be calibrated using the camera's spatial resolution (0.1–0.3 mm / pixel) to ensure consistency between physical dimensions and image measurements.

[0028] After obtaining the mold morphology parameters for consecutive stages, the evolution of each structural position of the mold in the time series is tracked using optical flow tracing methods. Optical flow methods can employ dense optical flow (such as the Farnebäck algorithm) or sparse optical flow based on feature points (such as the Lucas-Kanade method) to track the displacement and deformation of edge points, contour points, or texture keypoints in consecutive image frames. By tracking each structural position in consecutive frames, a morphological evolution curve over time is obtained, including the movement of geometric contours, edge changes, and the trend of changes in local thickness or shrinkage over time. The curve can be sampled according to the time step, with the temporal resolution consistent with the image acquisition frame rate; for example, a frame rate of 50 fps corresponds to sampling once every 20 ms. After obtaining the mold morphology evolution curve, numerical analysis is performed on each structural parameter to calculate its rate of change and amplitude. The rate of change can be obtained by dividing the parameter difference between consecutive sampling points by the time interval, reflecting the instantaneous speed of morphological change; the amplitude of change is quantified by the difference between the maximum and minimum values ​​of the curve, describing the overall morphological shift or shrinkage. The rate and magnitude of change are calculated for edge variations, contour shifts, texture uniformity, and local shrinkage features, forming a multi-dimensional index matrix. To improve accuracy, the curves can be smoothed, such as by using moving averages or low-pass filtering to remove noise fluctuations.

[0029] Anomaly analysis is performed on the structural parameters of each mold using the calculated rate and magnitude of change. By setting thresholds or employing statistical analysis methods, such as the mean ± 2σ or a deviation range based on historical process data, it is determined whether abnormal morphological changes exist at each structural location. Anomalies include excessively rapid edge movement, excessive local shrinkage, uneven texture, or contour deformation; these can be marked as precursors to defects. The marking process can generate masks or annotate key points in the staged image to record the location and severity of potential warping, shrinkage, or cracking.

[0030] In this embodiment, step S3 includes the following steps: Collect multi-stage process parameter streams during real-time production; Three-dimensional structural analysis was performed on the complete image of the demolded structure to obtain the three-dimensional structural parameters; Surface morphology reconstruction is performed based on 3D structural parameters to construct a 3D point cloud model; Load the CAD standard shape model; perform point cloud registration on the 3D point cloud model according to the CAD standard shape model, and calculate the forming shape deviation value; Based on the defect precursor information, correlation analysis is performed on the molding shape deviation value to mark the key defect precursors.

[0031] In this embodiment, key process parameters of the production line are collected in real time during the plastic mold production process to form a multi-stage process parameter stream. The collected parameters include injection pressure, holding pressure, screw speed, melt temperature, mold temperature, cooling time, and demolding force. Data acquisition is triggered synchronously according to the production cycle, with the sampling frequency typically set at 100 Hz or higher to ensure the capture of subtle changes in each stage of injection, holding pressure, cooling, and demolding. During the acquisition process, each process parameter can be digitized from sensor signals to form a time-series data stream, with each parameter bound to a corresponding timestamp to ensure precise matching between process status and production time. After mold demolding, a complete image of the demolded structure is acquired, and its three-dimensional structure is analyzed to obtain the geometric features of mold forming. Three-dimensional analysis can be achieved through multi-view stereo reconstruction or depth information obtained from structured light / depth cameras, converting two-dimensional image information into spatial coordinates. During the analysis, the mold contour, edges, and key protrusions and concavities are extracted, and the spatial position, curvature, surface normal vector, and thickness distribution of each surface point are calculated to form a complete set of three-dimensional structural parameters. To improve analytical accuracy, the image can be filtered for noise reduction, edge contrast enhanced, and local structure sharpened, achieving sub-pixel accuracy in feature point detection. The analytically obtained 3D structural parameters not only reflect the mold's geometric dimensions but also include surface morphology features.

[0032] Based on the 3D structural parameters, the surface topography of the mold is reconstructed to generate a corresponding 3D point cloud model. During the reconstruction process, each spatial point obtained from the analysis is used as a point cloud node, and surface continuity is preserved between adjacent nodes to form a high-density point cloud model. The point cloud distribution can be optimized using voxel mesh or surface fitting algorithms to match the point cloud density with the complex structural features of the mold. For example, the spacing between edge contour points is controlled within 0.1–0.5 mm to ensure that minute surface changes are captured. The point cloud model simultaneously records spatial coordinates and surface normal vectors for subsequent registration and deviation analysis. By constructing the 3D point cloud model, the surface topography and geometry of the mold after demolding can be fully expressed. After obtaining the 3D point cloud model of the mold, a pre-designed CAD standard shape model is loaded as a reference. A point cloud registration algorithm aligns the actual mold point cloud with the CAD model in spatial coordinates to achieve precise matching. The registration method can employ the Iterative Closest Point (ICP) algorithm or its improved version, iteratively minimizing the Euclidean distance error between point clouds based on the initial coarse registration, ensuring that the deviation calculation accuracy reaches 0.1–0.3 mm. After registration is completed, each point cloud point is matched with the corresponding CAD model surface point, and the forming shape deviation value is calculated, including local height difference, surface curvature difference and edge geometric error.

[0033] After obtaining the molding shape deviation values, they are correlated with defect precursor information for analysis. Defect precursor information may include indicators such as surface blistering, localized shrinkage, warping, cracks, or thickness anomalies. The analysis method establishes a correspondence between deviation thresholds and defect risks; for example, if an edge deviation exceeds 0.5 mm or a curvature anomaly exceeds a set threshold, it is determined that there may be precursors to warping or shrinkage. Spatial clustering or anomaly detection is performed on the molding deviation data to identify key areas, and these areas are marked on the 3D point cloud model as defect precursor locations.

[0034] In this embodiment, the specific steps for collecting the multi-stage process parameter flow of the real-time production process are as follows: A sensor network is deployed at key nodes of the injection molding machine to collect real-time injection molding process parameters, including melt temperature profile, injection speed segment values, mold temperature distribution, holding pressure decay rate, and cooling time. Perform multi-source data time-series alignment on real-time injection molding process parameters and extract time-synchronized process parameters; Calculate the self-noise of the timing synchronization process parameters to generate self-noise data; Noise filtering is performed based on self-noise data to obtain noise filtering process parameters; Calculate the timestamps of the noise filtering process parameters, perform time-series matching on the timestamps of the filtered image, and generate a multi-stage process parameter stream.

[0035] In this embodiment, a sensor network is deployed at key nodes of the injection molding machine, such as the barrel, screw, mold cavity, cooling pipes, and holding pressure system, to collect key process parameters in real time during the injection molding process. Melt temperature is measured along different positions of the screw using high-precision thermocouples or infrared temperature sensors, with a sampling frequency set to 100 Hz to capture rapid fluctuations in melt temperature. Injection speed is calculated using a screw position sensor and a speed sensing module, achieving an accuracy of ±1 mm / s. Mold temperature is recorded by arranging multiple thermocouples or thermistors along the cavity surface, recording the temperature distribution in each key area. Holding pressure is monitored by a pressure sensor, and a pressure decay curve is collected. Cooling time is obtained using a liquid cooling circuit temperature sensor and a timestamp. First, the sensor data is interpolated or resampled to unify data from different sampling frequencies onto a unified time axis, for example, to a 100 Hz time resolution. Then, synchronization is performed based on the timestamps, combining parameters such as melt temperature, injection speed, mold temperature, holding pressure, and cooling time corresponding to the same time point to form a complete time-series vector, constituting the time-synchronized process parameters. During the multi-source data alignment process, interpolation or smoothing can be performed to handle abnormal delays or missing data, ensuring the integrity of parameters at each time point.

[0036] After obtaining the timing synchronization process parameters, self-noise is calculated for each parameter to quantify random noise and interference fluctuations in sensor measurements. Self-noise calculation can employ moving standard deviation or sliding variance methods based on a small window. For example, local standard deviations can be calculated using 5–10 consecutive time points to obtain noise estimates for each time point. For temperature, pressure, and velocity signals, self-noise values ​​reflect the amplitude of sensor fluctuations in static or steady-state conditions and can be used to identify occasional abnormal peaks. By generating self-noise data for each parameter, filtering can remove measurement noise while preserving the true signal changes, improving the signal-to-noise ratio of the process parameters. Noise filtering is performed on the timing synchronization process parameters to remove random fluctuations and spike noise. Adaptive filtering or low-pass filtering can be used. For example, the filtering intensity is adjusted according to the local self-noise amplitude, with stronger filtering at time points with large noise amplitudes and weaker filtering in stable regions to ensure that signal details are not overly smoothed. For temperature and injection speed curves, exponentially weighted moving averages or Kalman filtering methods can be used, with filtering parameters automatically adjusted according to the noise amplitude.

[0037] After obtaining the noise-filtering process parameters, their timestamps are calculated and calibrated, and then precisely matched with the timestamps of the filtered images. Through matching, each frame of the staged image is aligned with the injection molding process parameters at the corresponding time point, forming a multi-stage process parameter stream. Linear interpolation or nearest-neighbor alignment methods can be used during the matching process to associate image frames with the process parameters closest to the timestamp. The final multi-stage process parameter stream contains a sequence of synchronized data for each production stage, including images and parameters such as melt temperature, injection speed, mold temperature, holding pressure, and cooling time.

[0038] In this embodiment, the specific steps of step S4 are as follows: Defect attribution analysis is performed based on the early signs of critical defects to identify defect factors; Based on defect factors, trace the production stage and mark the stage caused by the defect; Based on the stage caused by the defect, the multi-stage process parameter flow is matched and extracted to obtain the defect process parameter flow.

[0039] In this embodiment, after marking defect precursors, defect attribution analysis is performed on each key precursor to identify the root causes of potential defects. The analysis method combines molding morphology deviation data, morphology evolution curves, and local structural anomalies, establishing a correspondence between precursors and potential defects through statistical, association rule, or machine learning methods. For example, for areas with excessive edge shrinkage or uneven texture, the molding deviation curve can be used to analyze the time point of occurrence and the affected structural location, thereby determining whether it may be caused by excessive injection speed, insufficient holding pressure, uneven cooling, or abnormal melt temperature. During the attribution process, a multi-parameter weighted method can be used to quantify the degree of influence of different process parameters on defects, generating a set of defect factors, including information such as influence type, severity, and spatial location. After obtaining the defect factors, a production stage tracing analysis is performed to determine the stage at which the defect occurred or formed during production. Using staged image sequences and multi-stage process parameter flows, the defect area is mapped to each stage of melt filling, holding pressure, cooling and shaping, or demolding. By matching timestamps and spatial locations, the specific stage at which the defect most likely occurred is determined. For example, edge shrinkage or warping typically occurs during the cooling and setting stage, while surface blistering may originate from the initial stage of melt filling or insufficient holding pressure. During the tracing process, temporal correlation analysis can be used to compare the defect location with the morphological evolution curves and process parameter changes within each stage, identifying stages highly correlated with abnormal changes.

[0040] After identifying the stage that caused the defect, the corresponding time period is matched and extracted with the multi-stage process parameter stream to generate a defect process parameter stream. The matching method uses timestamp alignment to associate the noise-filtering process parameters of the defect occurrence stage with staged images and morphological evolution curves, forming a complete parameter sequence, including key parameters such as melt temperature, injection speed, mold temperature, holding pressure, and cooling time. During extraction, parameter fluctuation amplitudes, abnormal peaks, and trend changes can be marked to form a process parameter stream containing defect feature information for subsequent analysis and optimization. The defect process parameter stream can quantify the process state at the time of defect formation, enabling dynamic optimization control based on defect precursors. This ensures that the production process can be precisely adjusted for key defect factors in subsequent mold injection, reducing the probability of potential defects.

[0041] In this embodiment, the specific steps of step S5 are as follows: Calculate the stage duration length of the defect process parameter flow to obtain an adjustable time window; Based on the molding shape deviation value, the process parameters are locally adjusted to obtain the local adjustment value; The optimal process adjustment set is generated by adjusting the multi-objective constraints of the entire process based on the local adjustment values. The optimal process adjustment set includes injection speed, injection pressure, holding time, mold temperature and cooling time.

[0042] Based on the optimal process adjustment set and adjustable time window, production control is adjusted and control commands are output. Based on control commands, routine mold production operations are carried out.

[0043] In this embodiment, a time analysis is performed on the defect occurrence stage to calculate its duration. The start and end points of each process parameter are calibrated using timestamps to obtain the complete defect occurrence time interval. For example, abnormal edge shrinkage during the cooling stage may last 8–12 seconds, while abnormal mold filling during the initial injection stage may last 2–4 seconds. This duration interval is defined as the adjustable time window, used to limit the time range within which targeted process adjustments can be made during production. During the calculation, morphological evolution curves and staged image sequences can be combined to ensure that the time window accurately corresponds to the generation and evolution of the defect. The adjustable time window not only reflects the temporal characteristics of the defect occurrence stage but also provides a constraint boundary for local process adjustments, ensuring that the adjustment only affects the defect-related stage, avoiding unnecessary interference with normal stages, and improving the accuracy and efficiency of adjustments. After determining the adjustable time window, the defect process parameter flow is locally analyzed and adjusted using the molding morphology deviation value. The goal of local process adjustment is to correct the mold morphology that deviates from the design by modifying key process parameters. For example, if the edge shrinkage is too large, the corresponding holding pressure and holding time may need to be increased by 5-10%. If the surface texture is uneven, the injection speed segment values ​​may need to be reduced by 2-5% to reduce bubbles caused by turbulence. Local adjustment values ​​can be calculated using multi-objective optimization methods, combined with deviation values ​​and sensitivity analysis of process parameters, to ensure that the adjustment range can eliminate precursors of defects without introducing new anomalies. During the adjustment process, upper and lower limits are set for each parameter, such as injection speed 50-120 mm / s, holding pressure 40-80 MPa, mold temperature ±5°C, and cooling time ±2 seconds, to ensure that the adjustment is feasible and meets the requirements of equipment safety and process stability.

[0044] After obtaining the local adjustment values, they are integrated with the overall process parameters for multi-objective constraint optimization to generate the optimal process adjustment set. The optimization process considers multiple objectives simultaneously, including reducing molding shape deviations, lowering the probability of defects, maintaining mold dimensional accuracy, and ensuring stable production cycle time. Multi-objective optimization algorithms, such as weighted least squares, genetic algorithms, or gradient descent methods, are used to jointly optimize parameters such as injection speed, injection pressure, holding time, mold temperature, and cooling time. Constraints during optimization include equipment performance limits, material properties, and production cycle time requirements; for example, the injection pressure must not exceed 80 MPa, and the mold temperature fluctuation must not exceed ±5°C. After generating the optimal process adjustment set, it is combined with an adjustable time window to form a production control strategy targeting the defect stage. The control strategy includes real-time adjustment commands for the injection speed, injection pressure, holding time, mold temperature, and cooling time of the injection molding machine. Each command corresponds to a precise execution time within the adjustable time window. For example, adjusting the mold temperature gradient or extending the holding time 3 seconds before the cooling stage, and fine-tuning the speed during the injection stage to optimize melt flow. Control commands are issued through the CNC interface or the injection molding machine's PLC module to ensure that the equipment performs operations according to optimized parameters.

[0045] After the control command is issued, the injection molding machine executes routine production operations according to the command, achieving real-time optimization and adjustment during defect stages. During execution, process parameters such as injection speed, injection pressure, holding time, mold temperature, and cooling time dynamically change within an adjustable time window based on the optimal adjustment set, while ensuring the production cycle time and mold integrity in subsequent stages. Process parameters and staged images can continue to be collected during production, forming a closed-loop feedback loop. The actual execution effect is compared with the optimal target, and fine-tuning is made if necessary. Through continuous control and optimization, key defects in the mold production process are effectively suppressed, molding accuracy is improved, scrap rate is reduced, and a stable production rhythm is maintained. This combination of routine production and dynamic adjustment ensures high-quality output and process stability of the plastic mold throughout the entire production cycle.

[0046] In this embodiment, a device for dynamically adjusting process parameters for plastic mold production is provided, used to execute the method for dynamically adjusting process parameters for plastic mold production as described above, including: The image acquisition module is used to acquire images of the plastic mold production line in multiple stages and construct a staged image sequence. The defect identification module is used to identify abnormal defects in staged image sequences and mark defect precursor information. The correlation analysis module is used to perform correlation analysis on defect precursor information based on the staged image sequence and mark key defect precursors; The parameter extraction module is used to collect multi-stage process parameter streams in real-time production; and to match and extract the multi-stage process parameter streams based on key defect precursors to obtain defect process parameter streams. The process adjustment module is used to perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; and to perform normalized mold production operations based on the optimal process adjustment set.

[0047] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0048] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for dynamically adjusting process parameters in plastic mold production, characterized in that, Includes the following steps: Step S1: Perform multi-stage image acquisition on the plastic mold production line and construct a staged image sequence; Step S2: Identify abnormal defects in the staged image sequence and mark the precursor information of defects; Step S3: Based on the staged image sequence, perform correlation analysis on the defect precursor information and mark key defect precursors; Step S4: Collect multi-stage process parameter streams from the real-time production process; Based on the key defect precursors, the multi-stage process parameter stream is matched and extracted to obtain the defect process parameter stream; Step S5: Perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; execute normalized mold production operations based on the optimal process adjustment set.

2. The method for dynamically adjusting process parameters for plastic mold production according to claim 1, characterized in that, The specific steps of step S1 are as follows: Start the plastic mold production line to enter the production testing cycle and execute plastic mold production; Multi-stage image acquisition of the production line is performed using an industrial camera to extract multi-stage images of the mold; the multi-stage images of the mold include images of the initial stage of melt filling, surface evolution images of the holding pressure stage, edge shrinkage images of the cooling and shaping stage, and images of the complete structure after demolding. Adaptive enhancement is performed on multi-stage images of the mold, and the filtered image is output. Calculate the timestamp of the filtered image frame by frame; Based on the timestamp, the production stage is identified to obtain a production stage label; The filtered images are fitted with stage labels based on the production stage labels to construct a staged image sequence.

3. The method for dynamically adjusting process parameters for plastic mold production according to claim 2, characterized in that, The specific steps for adaptively enhancing the multi-stage image of the mold and outputting the filtered image are as follows: The multi-stage images of the mold are segmented to obtain multiple local regions; Calculate the brightness and contrast of the multiple local regions; Based on the brightness value and contrast, the difference in brightness distribution between regions is calculated to obtain the difference in brightness distribution between different regions. Based on the brightness distribution difference, perform zoned adaptive brightness adjustment to obtain a brightness-adjusted image; The edge sharpness and detail gradient changes of the brightness-adjusted image are evaluated, local blurring and detail loss are identified, and abnormal locations are marked. Spatial resolution compensation is performed on abnormal locations to construct resolution-enhanced images; Dynamic filtering optimization is performed on the resolution-enhanced image to output a filtered image.

4. The method for dynamically adjusting process parameters for plastic mold production according to claim 1, characterized in that, The specific steps of step S2 are as follows: Perform stage-by-stage recognition and instance segmentation on the staged image sequence, and mark the different structural positions of each stage; Visual measurements were performed on the different structural locations to obtain mold morphology parameters at different stages; the morphology parameters included geometric contour, edge change rate, surface texture uniformity, and local shrinkage characteristics. Continuous optical flow tracing of the mold morphology parameters yields the morphology evolution curve; The rate and magnitude of change of each parameter are calculated based on the morphological evolution curve; Based on the rate and magnitude of change, abnormal defects are identified, and precursor information of defects is marked.

5. The method according to claim 1, characterized in that, The specific steps of step S3 are as follows: Collect multi-stage process parameter streams during real-time production; Three-dimensional structural analysis was performed on the complete image of the demolded structure to obtain the three-dimensional structural parameters; Surface morphology reconstruction is performed based on 3D structural parameters to construct a 3D point cloud model; Load the CAD standard shape model; perform point cloud registration on the 3D point cloud model according to the CAD standard shape model, and calculate the forming shape deviation value; Based on the defect precursor information, correlation analysis is performed on the molding shape deviation value to mark the key defect precursors.

6. The method for dynamically adjusting process parameters for plastic mold production according to claim 5, characterized in that, The specific steps for collecting the multi-stage process parameter flow of the real-time production process are as follows: A sensor network is deployed at key nodes of the injection molding machine to collect real-time injection molding process parameters, including melt temperature profile, injection speed segment values, mold temperature distribution, holding pressure decay rate, and cooling time. Perform multi-source data time-series alignment on real-time injection molding process parameters and extract time-synchronized process parameters; Calculate the self-noise of the timing synchronization process parameters to generate self-noise data; Noise filtering is performed based on self-noise data to obtain noise filtering process parameters; Calculate the timestamps of the noise filtering process parameters, perform time-series matching on the timestamps of the filtered image, and generate a multi-stage process parameter stream.

7. The method for dynamically adjusting process parameters for plastic mold production according to claim 1, characterized in that, The specific steps of step S4 are as follows: Defect attribution analysis is performed based on the early signs of critical defects to identify defect factors; Based on defect factors, trace the production stage and mark the stage caused by the defect; Based on the stage caused by the defect, the multi-stage process parameter flow is matched and extracted to obtain the defect process parameter flow.

8. The method for dynamically adjusting process parameters for plastic mold production according to claim 1, characterized in that, The specific steps of step S5 are as follows: Calculate the stage duration length of the defect process parameter flow to obtain an adjustable time window; Based on the molding shape deviation value, the process parameters are locally adjusted to obtain the local adjustment value; Based on local adjustment values, perform multi-objective constraint adjustments on the entire process flow to generate the optimal process adjustment set; Based on the optimal process adjustment set and adjustable time window, production control is adjusted and control commands are output. Based on control commands, routine mold production operations are carried out.

9. The method for dynamically adjusting process parameters for plastic mold production according to claim 8, characterized in that, The optimal process adjustment set includes injection speed, injection pressure, holding time, mold temperature, and cooling time.

10. A device for dynamically adjusting process parameters in plastic mold production, characterized in that, The method for dynamically adjusting process parameters for plastic mold production as described in claim 1 includes: The image acquisition module is used to acquire images of the plastic mold production line in multiple stages and construct a staged image sequence. The defect identification module is used to identify abnormal defects in staged image sequences and mark defect precursor information. The correlation analysis module is used to perform correlation analysis on defect precursor information based on the staged image sequence and mark key defect precursors; The parameter extraction module is used to collect multi-stage process parameter streams in real-time production; and to match and extract the multi-stage process parameter streams based on key defect precursors to obtain defect process parameter streams. The process adjustment module is used to perform local process adjustment based on the defect process parameter flow to generate an optimal process adjustment set; and to perform normalized mold production operations based on the optimal process adjustment set.

Images