Parameter optimization system for car frame die-casting process

Abstract

The invention discloses a frame die-casting process parameter optimization system, and relates to the technical field of vehicle manufacturing. A high-speed image in a frame die-casting process is split into a plurality of image fragments, edge feature extraction is performed on each image fragment, an edge bonding chain of the image fragment is generated, and each edge bonding chain is endowed with a sequence symbol; monitoring whether the image fragments have abnormal conditions in real time by utilizing a lightweight model, and performing separation and image fragment copying processing by utilizing an edge bonding chain aiming at the abnormal conditions; after all the uploaded image fragments are received, a high-speed image is verified and spliced through a sequence symbol and an edge bonding chain, a defect area in the high-speed image is marked through a defect recognition mechanism, and then the optimal die-casting process parameters are determined. The problems that local abnormity is difficult to capture, abnormity processing is not timely and defect identification is not accurate due to overall image analysis in traditional frame die-casting process monitoring are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle manufacturing, in particular to a frame die casting process parameter optimization system. BACKGROUND

[0002] In the production process of the frame die casting, in order to ensure product quality, it is necessary to monitor and optimize the die casting process parameters. The traditional method mainly relies on the monitoring of the overall image or the real-time acquisition of a limited number of key parameters, which is difficult to fully and carefully reflect the complex situation of the die casting process. With the development of image processing technology and intelligent monitoring technology, it is possible to use high-speed images to analyze the die casting process more deeply.

[0003] The traditional process relies on preset injection speed, pressure and mold temperature curve, which is affected by equipment fluctuations and environmental disturbances in actual production, resulting in molding defects such as pores, shrinkage, warping, cold separation, etc. The existing technology mainly collects pressure and temperature signals through sensors, and adjusts the process parameters by using offline modeling or simple open-loop model, which lacks intuitive capture of real-time die casting process and is difficult to efficiently adjust under complex working conditions.

[0004] Therefore, in view of the above problems, there is an urgent need for a frame die casting process parameter optimization system. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a frame die casting process parameter optimization system, which solves the problems of difficult local abnormal capture, untimely abnormal handling, inaccurate defect identification and chaotic production data management in the traditional frame die casting process monitoring.

[0006] To achieve the above object, the present application is implemented by the following technical solutions: a vehicle frame die casting process parameter optimization system, comprising: an image acquisition and blocking module, configured to receive a high-speed image of a vehicle frame die casting process, split the high-speed image into a plurality of image patches according to a preset rule, and each image patch contains a die casting process parameter; a bonding chain generation module, configured to extract edge features of each image patch, generate an edge bonding chain of the image patch, and assign a sequence symbol to each edge bonding chain, and store the sequence symbol, the die casting process parameter of the corresponding image patch, and the edge bonding chain in association; a transmission monitoring and abnormality processing module, configured to deploy a lightweight model at the edge bonding chain for each uploaded image patch, use the lightweight model to monitor whether the image patch has an abnormal condition in real time, and use the edge bonding chain to separate and copy the image patch according to the monitored abnormal condition; an image stitching and defect identification module, configured to verify and stitch the high-speed image through the sequence symbol and the edge bonding chain after receiving all uploaded image patches, form a defect identification mechanism using the edge bonding chain, mark a defect area in the high-speed image using the defect identification mechanism, and determine an optimal die casting process parameter by simulating parameters of the defect area.

[0007] Further, the preset rule specifically includes: dividing the high-speed image into a plurality of adjacent image patches according to a spatial grid, and leaving an overlapping area for adjacent image patches, so that the edge of any image patch contains a die casting process parameter, and each image patch is automatically associated with the image patch acquisition time and the vehicle frame position after division.

[0008] Further, the bonding chain generation module generates a unique sequence symbol for the edge bonding chain of each image patch, and the sequence symbol contains an image patch number, a high-speed image sequence number, and a position index, wherein the image patch number represents the sequence position of the image patch in the high-speed image, the high-speed image sequence number represents the time sequence of the high-speed image in the entire vehicle frame die casting process, and the position index is the physical coordinates of the image patch in the vehicle frame model, used to represent the geometric position of the image patch in the mold.

[0009] Further, the edge bonding chain is composed of an edge feature point sequence and a connection relationship, and sequentially stores the edge coordinates, curvature information, corresponding die casting process parameters, and check codes of the image patch, and the edge bonding chain structure is realized based on a bidirectional connection table.

[0010] Further, the lightweight model is a lightweight neural network based on distance measurement, used to calculate the transmission delay and retrieval check code state between the current image patch and the adjacent image patch in real time, if the transmission delay exceeds a threshold value and the check code does not change, it is judged that the image patch is delayed, and if the check code changes, it is judged that the image patch is invaded.

[0011] Further, the specific analysis of the separation and image slice replication processing using the edge bonding chain is as follows: for image slice delay, according to the order symbol, the position where the image slice determined by the lightweight model to occur delay should be spliced is located, the edge bonding chain of the current image slice is separated from the image slice and goes to the position where the adjacent image slice converges in advance; using the edge coordinates, curvature information and corresponding die casting process parameters of the image slice stored in the edge bonding chain, the built-in image completion model of the edge bonding chain is called to generate a new replication slice at the convergence position, the edge shape of the adjacent image slice is combined for texture synthesis during the replication slice generation process, and the missing die casting process parameters are filled; for image slice intrusion, a new replication slice is generated according to the image slice delay processing mode, while the edge bonding chain of the current image slice is separated from the image slice, the internal resources of the separated image slice are used to perform self-destruction processing; when the new replication slice is generated, the weight of the lightweight model is reinitialized, and the order symbol of all edge bonding chains is updated synchronously.

[0012] Further, the specific analysis of the defect recognition mechanism formed by the edge bonding chain is as follows: after all image slices are spliced through the order symbol and the edge bonding chain, the computing resources occupied by each slice in the edge bonding chain generation and monitoring process are integrated into the defect recognition mechanism according to the resource reorganization strategy, the defect recognition mechanism integration method is as follows: the convolution kernel of the lightweight model and the edge features stored in the edge bonding chain are used as pre-training features, the defect recognition mechanism is constructed by combining the central stored die casting process image database, the defect recognition mechanism has multi-scale convolution layers and attention modules, which are used to extract features from macro framework to micro texture level for the spliced high-speed image, and identify the defect categories caused by process deviation.

[0013] Further, the specific analysis of the parameter simulation of the defect area is as follows: after the defect area is marked by the defect recognition mechanism, according to the defect type and its spatial distribution in the high-speed image, the process parameter simulation mechanism is called for multi-dimensional parameter exploration, the process parameter simulation mechanism is used to establish a digital twin model of the frame die casting process, map the space-time position of the image slice where the defect area is located in the frame die casting process to the physical coordinates, determine the die casting process parameters when the defect occurs by combining the order symbol, filter out the optimal die casting process parameters by monitoring the effects of different die casting process parameters, compare the optimal die casting process parameters with the current die casting process parameters, generate adjustment optimization suggestions and feedback to the die casting equipment control unit.

[0014] The present application has the following beneficial effects: The vehicle frame die casting process parameter optimization system can pay close attention to local details of the vehicle frame die casting process in detail by splitting the high-speed image into multiple image patches and performing edge feature extraction on each patch to generate an edge bonding chain, which helps to more accurately monitor the die casting process parameters and find potential problems, and improves the ability to capture subtle abnormalities compared with monitoring of the overall image; a lightweight model is deployed at the edge bonding chain for each uploaded image patch to realize real-time monitoring of abnormal conditions, and once an abnormality is found, the edge bonding chain can be used for separation and image patch replication processing to timely respond to abnormal conditions and avoid further expansion of the abnormality to affect the die casting quality, thereby improving the stability and reliability of the production process; after splicing the high-speed image, a defect recognition mechanism is formed by using the edge bonding chain to accurately mark the defect area, and by simulating parameters of the defect area, the optimal die casting process parameters can be determined more targetedly, which helps to improve the quality and performance of the vehicle frame die casting, reduce the scrap rate, and reduce the production cost; the sequential symbol, the die casting process parameters of the corresponding image patch, and the edge bonding chain are stored in association, so that the data of the entire die casting process has a clear logical relationship and traceability, facilitating subsequent analysis, query and management of the production data, and providing strong data support for process improvement and production optimization.

[0015] Of course, it is not necessary for any product embodying the present application to achieve all of the above advantages simultaneously. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A vehicle frame die casting process parameter optimization system structure diagram is provided.

[0017] Figure 2 A vehicle frame die casting process parameter optimization system method flow chart is provided.

[0018] Figure 3 A vehicle frame die casting process parameter optimization system logic flow diagram is provided. DETAILED DESCRIPTION

[0019] The vehicle frame die casting process parameter optimization system of the embodiments of the present application realizes fine monitoring through image blocking, real-time abnormal separation processing, accurate defect recognition based on the edge bonding chain, and parameter simulation optimization, thereby effectively improving the vehicle frame die casting quality and production stability.

[0020] The general idea of the embodiments of the present application is as follows: The video of the die casting process is collected by a high-speed industrial camera, each frame of image is split into multiple image patches according to a preset grid by an edge computing gateway, and features are extracted at the edges of the patches to generate a bonding chain. The bonding chain carries process parameters and sequence symbols, which are used to detect missing or tampered fragments during uploading. Once an anomaly occurs, the saved edge information is used to generate a duplicate fragment and merge it with the adjacent patch in advance to maintain transmission integrity. After all the patches are uploaded, the image is spliced according to the sequence symbols and edge bonding chain, and the resources occupied by the bonding chain are reorganized into a defect identification mechanism for defect detection on the spliced image. The optimal die casting parameters are determined through process parameter simulation and fed back to the die casting machine actuators to achieve closed-loop control.

[0021] Please refer to Figure 1 、 Figure 2 、 Figure 3 , the embodiment of the present application provides a technical scheme: a vehicle frame die casting process parameter optimization system, comprising: an image acquisition and blocking module, configured to receive high-speed images of a vehicle frame die casting process, and split the high-speed images into multiple image patches according to a preset rule, each image patch containing die casting process parameters; a bonding chain generation module, configured to extract edge features of each image patch to generate an edge bonding chain of the image patch, and assign a sequence symbol to each edge bonding chain, and store the sequence symbol, the die casting process parameters of the corresponding image patch and the edge bonding chain in association; a transmission monitoring and abnormality processing module, configured to deploy a lightweight model at the edge bonding chain for each uploaded image patch, use the lightweight model to monitor whether the image patch has an abnormal condition in real time, and use the edge bonding chain to separate and duplicate the image patch according to the monitored abnormal condition; an image splicing and defect identification module, configured to verify and splice the high-speed images through the sequence symbol and the edge bonding chain after receiving all the uploaded image patches, form a defect identification mechanism using the edge bonding chain, mark the defect area in the high-speed image using the defect identification mechanism, and determine the optimal die casting process parameters through parameter simulation on the defect area.

[0022] Specifically, the die casting process parameters refer to key process variables that directly affect the filling, solidification and quality of the castings during the vehicle frame die casting production process. These parameters are recorded in real time and embedded in the high-speed image patches as the core basis for defect analysis and parameter optimization. The embedding methods include but are not limited to: adding a structured data field containing process parameters in the file header or metadata area of the image patch; or encoding the process parameters as invisible watermarks and embedding them in the frequency domain of the patch image. The die casting process parameters specifically include the following parameters: injection-related parameters: injection speed, injection pressure; temperature-related parameters: mold temperature, pouring temperature; vacuum and exhaust parameters: vacuum degree, exhaust time; time-related parameters: holding time, cooling time, opening time; auxiliary process parameters: release agent parameters, pressure chamber fullness.

[0023] The die casting process parameters are collected in real time by sensors and associated with the corresponding area of the high-speed image. For example, a certain section of the injection speed corresponds to the image of the metal liquid filling a certain frame position. The image fragments are embedded by the image acquisition and blocking module, and then bound with the fragment edge features through the adhesion chain. Finally, the mapping of the defect area and the corresponding process parameters is realized during defect identification, providing specific adjustment objects for parameter simulation optimization.

[0024] Specifically, the establishment method of the association relationship is to set a timestamp for each image acquisition time or frame, and synchronously bind all the die casting process parameter data collected by the sensor at that time. When the image is blocked, the process parameter data at the same time is embedded into the image fragment of the corresponding physical coordinate area according to the timestamp and the preset coordinate mapping table. For example, the injection speed value at a certain time is associated and stored with the fragment corresponding to the metal liquid front position in the image at that time.

[0025] The image acquisition and blocking module is deployed on a high-speed industrial camera and an edge computing gateway. The high-speed industrial camera is fixed in the working area of the die casting machine and connected with the gateway through a high-speed interface to collect high-speed images of the die casting process in real time. Due to the fast transient change in the die casting process, this module uses high frame rate exposure and high-speed transmission technology to ensure the capture of the dynamic of the mold filling and cavity filling.

[0026] In this embodiment, the collected high-speed images are blocked on the edge computing gateway according to the preset rules. The preset rules divide the high-speed image into a plurality of adjacent image fragments based on a spatial grid. The grid size is adaptively adjusted according to the field of view of the die casting equipment, the spatial scale of the process parameter change, and the camera resolution. The minimum acceptable fragment size is determined by calculating the motion speed and temperature gradient of the process area. The adaptive adjustment algorithm of the grid size is based on the following rules: first, determine the upper limit of the number of image fragments according to the field of view of the die casting equipment. Second, determine the lower limit of the physical size of each fragment according to the camera resolution and the spatial scale of the process parameter change to ensure that key process changes can be captured. Finally, dynamically adjust the fragment size in combination with the calculated average motion speed of the process area. The faster the motion speed, the larger the fragment size, but it must be ensured that it is not less than the aforementioned lower limit of the physical size. In order to ensure that adjacent fragments can be seamlessly spliced, an overlapping area is set at the boundary of the grid, and the overlapping width is determined according to the camera resolution and the process parameter accuracy requirement.

[0027] Each image fragment is automatically associated with the acquisition timestamp of the fragment, the physical coordinates in the frame mold coordinate system, and the die casting process parameters after splitting. The process parameters are derived from the injection speed, pressure curve, mold temperature, and vacuum degree collected by the PLC. By embedding these parameters in the fragment, it is ensured that the complete process state can still be inferred based on the remaining fragments when any fragment is lost.

[0028] The image acquisition and blocking module realizes fine blocking of the high-speed die casting process, binds each slice with process parameters, provides sufficient slice redundancy to ensure transmission reliability, and lays a foundation for subsequent edge bonding chain generation and parameter reconstruction.

[0029] Specifically, the edge bonding chain generation module is deployed on the acceleration chip of the edge computing gateway, and is used to perform edge feature extraction on each image slice and generate an edge bonding chain for the slice. An edge detection algorithm based on gradient and texture is adopted, and the pixel gradient of the slice edge is analyzed in combination with the flow characteristics of the molten metal in the die casting process to extract an edge feature point sequence.

[0030] In the edge feature extraction process in the present embodiment, edge curvature information and die casting process parameters corresponding to the edge are recorded, an edge feature vector is constructed, and a sequence symbol is assigned to each edge bonding chain according to a preset sequence rule. The sequence symbol is a unique identifier composed of an image slice number, a high-speed image sequence number, and a physical coordinate in the vehicle frame model. The image slice number represents the relative sequence position of the slice in the high-speed image; the high-speed image sequence number represents the time sequence of the high-speed image in the video; and the physical coordinate represents the geometric position of the slice in the mold.

[0031] The edge bonding chain structure is implemented by using a bidirectional connection table, and the nodes store the edge feature point sequence and the die casting process parameters. The edge bonding chain is connected by pointers at both ends to form a ring topology. Each chain also generates a check code, which is calculated by the edge features and the process parameters through a hash function, and is used to detect whether the slice has been tampered with or damaged during transmission.

[0032] The edge bonding chain not only carries edge information for subsequent splicing, but also can save process parameters, texture information, and texture weights of the corresponding image slice. When an image slice is lost, a new duplicate slice is generated in the overlapping area of adjacent image slices using the information in the edge bonding chain, ensuring that each image slice contains complete process parameters.

[0033] Through the edge bonding chain, each image slice is closely bound with edge features and process parameters. The design of the sequence symbol not only marks the position of the slice in space and time, but also ensures that the system can quickly locate the delayed segment and reconstruct it, improving data integrity and security and providing a reliable basis for subsequent splicing and defect identification.

[0034] Specifically, the transmission monitoring and abnormality handling module is deployed in the real-time operating system of the edge computing gateway. A lightweight model is deployed at the edge bonding chain of each uploaded image slice, which runs in an independent thread to monitor the transmission status of each slice. The lightweight model uses deep network compression and pruning techniques to reduce the number of parameters while maintaining recognition performance, enabling efficient operation on resource-limited gateway devices.

[0035] The input of the lightweight model includes the transmission delay of the current slice and adjacent slices, the change of the check code, and the difference in the sequence symbol. By calculating the transmission distance between slices, i.e., the difference in sending time and the number of network hops, and combining the check code comparison, it can be determined in real time whether there is an abnormal condition in the slice. When the transmission delay exceeds the set threshold and the check code does not change, it is determined that the slice delay exceeds the expected value, i.e., the slice can be considered lost; when the check code changes, it is determined that the slice may be invaded or transmission error occurs. To accurately distinguish, the transmission delay is combined for judgment: if the transmission delay is greater than the threshold and the check code changes, it is determined that the slice is highly likely to be invaded; if the transmission delay does not exceed the threshold but the check code changes, it is determined as transmission error or data damage.

[0036] Once an abnormality is detected, the slice separation and replication processing is performed using the bonding chain. For a lost slice, the lightweight model instructs the chain to separate from the slice and locate the position that should be spliced through the sequence symbol. The chain then goes to the adjacent slice located and converges in advance. During the convergence process, the lightweight model calls the image completion model inside the chain to generate a new replicated slice using the saved edge features, curvature information, and process parameters, with the edge shape of the adjacent slice as a constraint.

[0037] For a slice detected as being invaded, the processing method is similar, except that when the chain separates from the slice, it will retain part of the resources to perform self-destruction and blurring to prevent sensitive information leakage. After separation and replication are completed, a new bonding chain and lightweight model are generated and synchronized with the sequence symbol of other slices.

[0038] The lightweight model structure includes an input layer, a number of quantized convolution layers, and an attention mechanism layer for extracting transmission state features; the model size is reduced through pruning and quantization; and the output is given after the activation function to give the abnormality classification result. The lightweight model interacts with the transmission protocol stack of the edge gateway through the socket interface to obtain transmission statistics and sends abnormal events to the upper layer to trigger the chain processing logic.

[0039] Image completion model is a generative model that reconstructs the missing region according to the surrounding content. Its essence is to use the texture, color and structure information of the existing region to infer the pixels of the occluded or damaged region through a deep neural network. Image completion model is specifically considered as a conditional generation task: the input is an image with missing or occluded regions, and the output is a complete image with the missing region filled. In the training stage, the image completion model learns the statistical characteristics and structural rules of natural images, so that it can generate details that are coordinated with the original image in the inference stage.

[0040] The image completion model is constructed in an encoder-decoder structure or a generative adversarial network. The specific methods include: encoding the image into multi-scale features through convolutional layers to capture the context information around the missing region; the decoder part decodes these features into pixel values to realize the reconstruction of the missing region; in order to better handle large area missing, the model often integrates self-attention mechanism or residual block, allowing the network to focus on key texture and edge clues in a global range; in training, the complete picture is randomly occluded to construct training samples, and the model is optimized through pixel reconstruction loss, perceptual loss and adversarial loss, so that it can learn to generate natural and coherent content in the missing region.

[0041] The image completion model is embedded in the adhesion chain to generate replacement fragments for missing or invaded image fragments. Its functions mainly include: according to the edge features and curvature information of adjacent fragments, texture synthesis and edge extension are performed on the missing region to ensure that the spliced picture is visually seamless. Using the die casting process parameters saved in the chain, these parameters are filled into the newly generated fragments, so that the generated fragments not only look like the original fragments, but also contain complete process data. When an exception occurs during transmission, the missing fragments are quickly copied to maintain the integrity of the data chain and improve the robustness of the system, avoiding the influence of the loss of a single fragment on the restoration of the entire process parameters. Figure 1 By deploying a lightweight model on the edge device, real-time transmission monitoring and abnormality judgment are realized, which not only solves the problem of difficult deployment of traditional deep models in embedded environments, but also guarantees the consistency and security of image fragments during transmission. After the loss or invasion of fragments, they can be immediately copied and restored, improving the robustness of the system.

[0042]

[0043] ​Specifically, when all image patches are uploaded to the central server through the edge computing gateway, the image stitching and defect identification module begins to work. This module is deployed on the cloud server and edge gateway combined platform. According to the sequence symbol and edge bonding chain verification of the uploaded patch integrity, the edge features and curvature information in the chain are used to seamlessly stitch each patch into a complete high-speed image sequence according to the spatial and temporal order determined by the sequence symbol. In the stitching process, the overlapping area is fused to eliminate gaps and visual discontinuity through weighted average and texture matching.

[0044] In this embodiment, after image stitching is completed, the computing resources occupied by each patch in the bonding chain generation and monitoring process are integrated to form a defect identification mechanism. The resource integration strategy includes using the convolution kernel and attention weight in the lightweight model as pre-training features to train a deep defect identification network together with the edge features saved by the edge bonding chain. This network has multi-scale convolution layers, spatial attention and channel attention modules, which can extract features from macro contours to micro textures at multiple levels, and identify various defect categories such as incomplete filling, air hole shrinkage, etc.

[0045] The defect identification mechanism is in communication link with the central storage die casting process image database. A large number of annotated reference process images in the database are called to build a contrast learning task to improve the recognition accuracy. By comparing the stitched image with the reference image, the spatial position of the defect area, the defect type, and the time when the defect is formed are identified.

[0046] After identifying the defect area, the sequence symbol of the defect segment is mapped to the process parameters, and the process parameter simulation mechanism is called for parameter optimization. If necessary, the edge computing gateway can be revisited to adjust the compensation strategy in real time, and the optimal parameters are fed back to the die casting machine PLC control logic to form a closed-loop control.

[0047] By using the computing resources reorganized by the bonding chain to build a defect identification mechanism, additional hardware consumption is avoided; multi-scale convolution and attention mechanism improve the accuracy and robustness of defect detection; at the same time, linkage with the process database realizes high-precision comparison, so that defect identification and parameter optimization form a closed loop, improving production quality.

[0048] Specifically, the process parameter simulation mechanism is deployed on the collaborative computing platform of the cloud server and the die casting machine controller, which is used for simulation analysis of the die casting process parameters according to the defect identification results. The process parameter simulation mechanism constructs a digital twin model of the die casting process of the vehicle frame, and maps the spatial position and timestamp of the defect area in the image to the specific part and process stage in the physical casting.

[0049] According to the mapping result, key process variables such as injection speed, pressure gradient, mold temperature, vacuum degree, alloy composition, etc. are selected to establish the parameter-quality response relationship in the physical model. By calling the simulation module, the influence of different parameter combinations on the defect index is predicted, and the process parameter with the greatest influence on the defect is selected through sensitivity analysis. The defect index includes porosity, surface roughness, and size deviation.

[0050] Based on the simulation results, a set of optimal die casting process parameters is selected from the multi-dimensional parameter space. This parameter combination can minimize the defect index and meet the production rhythm. The optimal parameters are compared with the current production parameters to generate adjustment suggestions, which are sent to the die casting machine PLC control unit through the edge gateway. The PLC adjusts the injection speed, pressure, and temperature by controlling actuators such as servo valves, hydraulic systems, heaters, and vacuum pumps, thereby optimizing the die casting process in real time. The pressure and temperature data collected by the sensors are fed back to the simulation model, forming a closed loop for updating.

[0051] Using digital twin models and sensitivity analysis, key process parameters that cause defects can be quickly located and optimization schemes can be generated. The closed loop formed by the PLC and sensors enables real-time adjustments and precise control, improving product quality and reducing waste.

[0052] Specifically, in the transmission monitoring module, abnormal conditions are divided into two categories: fragment loss and fragment intrusion.

[0053] Fragment loss judgment: The lightweight model continuously calculates the transmission time difference and network hop count of adjacent fragments. When the transmission delay of a fragment is greater than the threshold value, and the corresponding check code does not change in the network, it is considered that the fragment is accidentally lost during transmission. Fragment loss may be caused by network congestion, buffer overflow, or link failure.

[0054] Fragment intrusion judgment: If the transmission delay is greater than the threshold value, and the check code of the fragment changes, i.e., it is inconsistent with the check code calculated when it was generated, it is considered that the fragment may have been tampered with or invaded during transmission.

[0055] For fragment loss processing, the replication process is immediately started: the edge bonding chain is separated from the fragment and locates the target splicing position through the sequential symbol. The edge bonding chain goes to the positioning location and meets the adjacent chain in advance; using the edge coordinates, curvature information, and process parameters saved in the edge bonding chain, a new fragment is reconstructed through the built-in image completion model. The completion model is based on a convolutional neural network and combines the texture features of adjacent fragments to generate the missing part and supplement the lost process parameters. After replication, a new bonding chain and lightweight model are generated to replace the lost fragment.

[0056] For the processing of the invaded fragments, the missing fragments are recovered through the replication process. After the chain is separated from the invaded fragment, part of the resources are used to perform the blurring process, smearing or scrambling the pixels of the invaded fragment to protect the process secrets, and then destroying the fragment. The new replicated fragment is uploaded together with other fragments to ensure the integrity of the image.

[0057] After the completion of the abnormal processing of the fragments, all the bonded chains are reordered according to the new order symbol and synchronized to other edge nodes through control messages, ensuring consistent splicing in the future.

[0058] By accurately distinguishing between the different abnormal conditions of loss and invasion and adopting targeted processing strategies, both the integrity and continuity of data transmission are ensured, and security is strengthened; the fragment replication model enhances the fault tolerance capability, enabling the complex die casting process to run stably under network fluctuations.

[0059] Specifically, the defect identification mechanism is an intelligent analysis module established by integrating the computing resources occupied by the edge bonding chains after image splicing is completed. The convolution kernel and attention weight of the lightweight model deployed during the generation and transmission monitoring of the edge bonding chain, the edge features saved in the chain, and the intermediate features of the temporarily used image completion model are unified into a multi-task training library.

[0060] When establishing the defect identification mechanism, these pre-trained features are normalized and weight fused to form initial model parameters. Combined with the reference samples in the die casting process image database, the identification network is subjected to transfer learning. The identification network adopts a pyramid multi-scale convolution structure, gradually increases the receptive field from top to bottom, focuses on the defect area combined with the spatial attention module, and filters out the feature channels related to the defect through the channel attention module, improving the recognition sensitivity of the model.

[0061] The defect identification mechanism extracts local features on the spliced image through a sliding window, matches the features with the reference image features, and marks the specific location and category of the defect area; at the same time, it calculates the size, shape, and duration of the defect area in the time sequence to provide necessary information for subsequent parameter simulation. The identified defect area is linked to the process parameters in the order symbol mapping manner, thereby forming a three-dimensional correlation index of time, space, and parameters.

[0062] The completed defect identification mechanism serves as a cloud service and interacts with the edge gateway, PLC, and database through standardized interfaces. The defect identification mechanism realizes efficient construction through resource integration and transfer learning, and enhances the focusing ability on defect features with the help of the attention mechanism; the training strategy improves the recognition performance of the system on new defect types, significantly improving the die casting process quality control level.

[0063] Specifically, the parameter simulation method of the defect area is based on the digital twin model and the sensitivity analysis. For each labeled defect area, the process parameters at the time when the area is generated are retrieved through sequential symbol retrieval to construct a parameter vector. Then, partial differential equations are established in the physical simulation model to describe the molten metal flow, heat transfer and solidification process, and the finite element or finite volume discrete solution is obtained to get the influence of different parameter combinations on the defect index.

[0064] A global optimization algorithm is used to search for the optimal solution in a multi-dimensional parameter space, considering the influence of injection speed curve, pressure gradient, alloy temperature and mold surface temperature on defect formation. At the same time, the simulation model is corrected online combined with the real-time data collected by the sensor of the die casting machine, to improve the prediction accuracy.

[0065] To avoid time-consuming large-scale calculation, the simulation model is combined with the machine learning agent model to quickly evaluate the pros and cons of parameter combinations, and dynamic switching is performed between numerical simulation and agent model. The final set of optimal process parameters will be sent to the die casting machine PLC through the control protocol, and the PLC will control the servo valve to adjust the injection speed, the heater to adjust the mold temperature, and the vacuum pump to adjust the vacuum degree in the mold cavity, etc., to realize the targeted optimization of the defect area.

[0066] The parameter simulation method using digital twin simulation combined with agent model and global optimization algorithm can search for the nearly optimal process parameter combination in a short time; combined with real-time sensor data to improve model accuracy, thereby effectively reducing the defect rate and improving the consistency and reliability of die casting products.

[0067] In summary, the present application has at least the following effects: When the image is blocked, it is ensured that each piece contains process parameters, avoiding parameter fragmentation, reducing the cost of resampling, and adapting to continuous production; the adhesive chain combines edge features and sequential symbols to improve the splicing accuracy and realize rapid tracing of defects and parameters; the edge deployment lightweight model monitors transmission abnormalities in real time, generates a duplicate segment when lost, and isolates and destroys when invaded, ensuring data continuity and security; the defect identification reuses the adhesive chain resources, accurately associates abnormal parameters, and outputs the optimal scheme through simulation, without the need for trial and error. The whole forms a closed loop, adapts to the high safety requirements of the vehicle frame, improves the pass rate, reduces the cost and production interruption risk.

[0068] Those skilled in the art will appreciate that embodiments of the application can be provided as methods. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer-usable program code embodied in the medium.

[0069] The application is described with reference to the Figures according to which methods of the application are schematically illustrated. It should be understood that each flow of the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart block or blocks. Figure 1 These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks.

[0070] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks.

[0071] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks.

[0072] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those of skill in the art once they have the benefit of the present disclosure. Therefore, the appended claims are intended to cover all such variations and modifications as falling within the scope of the application.

[0073] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A system for optimizing process parameters in vehicle frame die casting, characterized in that, include: The image acquisition and segmentation module is used to receive high-speed images of the chassis die-casting process and split the high-speed images into multiple image segments according to preset rules. Each image segment contains die-casting process parameters. The adhesive chain generation module is used to extract edge features for each image segment, generate the edge adhesive chain for that image segment, assign a sequence number to each edge adhesive chain, and associate and store the sequence number, the die-casting process parameters of the corresponding image segment, and the edge adhesive chain. The transmission monitoring and anomaly handling module is used to deploy a lightweight model at the edge bonding chain for each uploaded image segment. The lightweight model is used to monitor whether there are any anomalies in the image segment in real time. For the detected anomalies, the edge bonding chain is used to separate and copy the image segments. The image stitching and defect recognition module is used to verify and stitch high-speed images by using sequence symbols and edge bonding chains after receiving all uploaded image segments. It uses the edge bonding chains to form a defect recognition mechanism, marks defect areas in the high-speed images, and determines the optimal die-casting process parameters by simulating the parameters of the defect areas.

2. The vehicle frame die-casting process parameter optimization system according to claim 1, characterized in that, The preset rules specifically include: dividing the high-speed image into multiple adjacent image segments according to the spatial grid, and leaving an overlapping area for adjacent image segments, so that the edge of any image segment contains the die-casting process parameters, and automatically associating the acquisition time and chassis position of each image segment after division.

3. The vehicle frame die-casting process parameter optimization system according to claim 1, characterized in that, The adhesive chain generation module generates a unique sequence symbol for the edge adhesive chain of each image segment. The sequence symbol includes the image segment number, the high-speed image sequence number, and the position index. The image segment number indicates the sequential position of the image segment in the high-speed image, the high-speed image sequence number indicates the temporal order of the high-speed image in the entire frame die casting process, and the position index is the physical coordinate of the image segment in the frame model, which is used to indicate the geometric position of the image segment in the mold.

4. The vehicle frame die-casting process parameter optimization system according to claim 3, characterized in that, The edge bonding chain consists of an edge feature point sequence and connection relationships, which sequentially stores the edge coordinates, curvature information, corresponding die-casting process parameters, and check codes of the image segment. The edge bonding chain structure is implemented based on a bidirectional connection table.

5. The vehicle frame die-casting process parameter optimization system according to claim 4, characterized in that, The lightweight model is a lightweight neural network based on distance metric, used to calculate the transmission delay and retrieval checksum status between the current image slice and its neighboring image slices in real time. If the transmission delay exceeds the threshold and the checksum does not change, it is determined to be an image slice delay. If the checksum changes, it is determined that the image slice has been compromised.

6. The vehicle frame die-casting process parameter optimization system according to claim 5, characterized in that, The specific analysis of the separation and image segmentation copying process using edge adhesive chains is as follows: To address image segmentation delay, the lightweight model locates the position where image segments identified as having delays should be stitched together based on the sequence number. It then instructs the edge bonding chain of the current image segment to separate from the image segment and move to the located position to merge with its adjacent image segments in advance. Using the edge coordinates, curvature information, and corresponding die-casting process parameters of the image segments stored in the edge bonding chain, the built-in image completion model of the edge bonding chain is called to generate a new copy segment at the merging point. During the generation of the copy segment, texture synthesis is performed by combining the edge shapes of adjacent image segments, and missing die-casting process parameters are filled in. In response to the intrusion of image segments, a new copy segment is generated according to the image segment delay processing method. While the edge glue chain of the current image segment is separated from the image segment, the internal resources of the separated image segment are used to perform self-destruction processing. When a new copy fragment is generated, the weights of the lightweight model are reinitialized, and the sequence number of all edge glue chains is updated synchronously.

7. The vehicle frame die-casting process parameter optimization system according to claim 1, characterized in that, The specific analysis of the defect recognition mechanism using edge bonding chains is as follows: After all image segments are stitched together by verifying the sequence number and edge bonding chains, the computing resources occupied by each segment during the generation and monitoring of edge bonding chains are integrated into a defect recognition mechanism according to the resource reorganization strategy. The integration method of the defect recognition mechanism is as follows: the convolution kernel of the lightweight model and the edge features stored in the edge bonding chains are used as pre-trained features, and the defect recognition mechanism is constructed by combining the centrally stored die-casting process image database. The defect recognition mechanism has multi-scale convolutional layers and attention modules, which are used to extract features from the macro framework to the micro texture level for the stitched high-speed image step by step, and identify the defect category caused by process deviation.

8. The vehicle frame die-casting process parameter optimization system according to claim 1, characterized in that, The specific analysis of parameter simulation for the defect area is as follows: After the defect area is marked using the defect identification mechanism, the process parameter simulation mechanism is invoked to explore multi-dimensional parameters based on the defect type and its spatial distribution in the high-speed image. The process parameter simulation mechanism is used to establish a digital twin model of the chassis die-casting process, mapping the spatiotemporal position of the image slice where the defect area is located to physical coordinates in the chassis die-casting process, and determining the die-casting process parameters when the defect occurs by combining the sequence number. The optimal die-casting process parameters are selected by monitoring the effect of simulating different die-casting process parameters, and the optimal die-casting process parameters are compared with the current die-casting process parameters to generate adjustment and optimization suggestions and feed them back to the die-casting equipment control unit.

Images