An electric control rear shell housing multi-point die casting control system and method

Abstract

The present application relates to the technical field of die casting, and particularly relates to an electric control rear shell shell multi-point die casting control system and method, which adopts homomorphic encryption and secure multi-party computation to complete die casting state reasoning under the condition that edge nodes are not decrypted, and takes into account data security and real-time performance; constructs a die casting process data cube based on a directed acyclic graph, dynamically associates different batches according to process similarity and evolution path, and improves process control continuity; automatically extracts deep correlations between process parameters, filling modes, defect types and failure rates through periodic data mining, dynamically updates homomorphic die casting models and alarm rules, so that the models have self-evolution ability; integrates historical fault libraries and real-time working conditions into decision-making, optimizes the speed, pressure and exhaust timing of multi-point pouring in cooperation with the prediction of failure rate as a constraint, and generates control instructions with the lowest comprehensive failure risk under the premise of ensuring the forming quality.

Description

Technical Field

[0001] This invention relates to the field of die casting technology, and in particular to an electronically controlled multi-point die casting control system and method for rear shell. Background Technology

[0002] As a critical protective and load-bearing structural component for control units in new energy vehicles, smart grids, and industrial automation equipment, the molding quality of the electronic control system's rear housing directly affects its heat dissipation efficiency, electromagnetic shielding performance, and long-term operational reliability. These housings typically feature thin walls, multiple cavities, dense heat dissipation fins, and complex mounting clips, demanding extremely high dimensional accuracy, internal density, and airtightness. Die casting, with its high efficiency, near-net-shape forming capability, and excellent mechanical properties, has become the mainstream manufacturing process for electronic control system rear housings.

[0003] However, in the existing technology, there is a lack of dynamic reasoning based on the structured correlation between different injection batches based on process similarity, and it is difficult to achieve synergistic optimization control between molding quality and component failure risk. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-point die-casting system and method for an electronically controlled rear shell, which solves the technical problems in the prior art where there is a lack of dynamic reasoning based on the structured association between different injection batches based on process similarity, and it is difficult to achieve collaborative optimization control between molding quality and component failure risk.

[0005] To achieve the above objectives, the present invention provides a multi-point die-casting control method for an electronically controlled rear shell, comprising: After the die casting process is started, multi-source process parameters are collected through a multi-source sensor network and transmitted to the edge computing node after being encrypted by homomorphic encryption or secure multi-party computation protocol. Without decryption, the edge computing node directly calls the homomorphic die casting state model to perform forward reasoning on the ciphertext and outputs state ciphertext to characterize the filling mode, exhaust efficiency and temperature field uniformity. Each injection stroke is encapsulated as a data block. Based on the directed acyclic graph structure, the data blocks of different batches are associated according to process similarity and evolution path to form a traceable die casting process data cube, which stores multi-source process parameters, status ciphertext, operation logs and quality tags. Data mining is performed periodically or triggered on the data cube to extract deep correlation rules between process parameters and filling patterns, defect types and component failure rates, and the internal weights of the homomorphic die casting state model and the alarm rules of the process decision model are updated based on the mining results. When executing the process decision model, the historical fault database of components and real-time operating condition data are accessed to dynamically predict the failure rate of each molded component in the current production cycle. The predicted failure rate is embedded as a constraint in the decision-making process of multi-point injection speed, pressure and venting timing. Control instructions that minimize the overall failure risk under the premise of ensuring molding quality are generated and sent to the die casting machine for execution.

[0006] The edge computing node receives the ciphertext of process parameters generated by random masking or polynomial encoding; loads a pre-deployed homomorphic die casting state model containing only addition and multiplication operations; inputs the ciphertext into the model, performs forward propagation calculation layer by layer in the ciphertext domain, and directly outputs the encrypted filling front position, air entrapment risk index and temperature gradient deviation value as the state ciphertext.

[0007] Specifically, data mining tasks are executed periodically or triggered in a manner that includes: An improved association rule mining algorithm was adopted, with nodes in the multidimensional storage network as transaction items, and support and confidence thresholds were set to mine strong association rules between high-speed filling speed range, specific exhaust groove state and air entrainment defects. A time-series-based causal analysis algorithm was used to analyze the time-delay causal chain between different opening sequences of multi-point ingates and the local wall thickness deviation, residual stress distribution, and final warping deformation of the rear shell. The discovered strong correlation rules and time-delay causal chains are used to automatically adjust the internal weight parameters of the homomorphic die-casting state model, or as the basis for generating process alarm rules.

[0008] The construction of a directed acyclic graph structure includes: A single injection stroke is defined as a data block containing a block header and a block body. The block header stores the hash value, timestamp, and batch identifier of the previous block, while the block body stores the desensitized process parameters, status ciphertext, and operation logs. Based on the evolution trajectory of process parameters and the similarity of molding quality between different batches, directed edges and lateral similarity edges are established to form a nonlinear multidimensional storage network. Based on this, a data cube is constructed with batch, time, parameter type, and quality label as dimensions, which supports any node to trace back to the process and quality information of all previous related nodes.

[0009] Specifically, the dynamic prediction of the failure rate of each molded component and its integration into decision-making include: Temperature cycle curves, pressure peaks, and thermal stress-strain data of the mold under different cumulative injection cycles are extracted from the historical fault database. Combined with real-time working condition data, the probability of thermal fatigue cracks, slider wear rate, and ejector pin fracture risk index are predicted respectively. The predicted failure rates are normalized and used as constraints for multi-objective optimization. Together with the preset threshold for molding quality, they form the optimization objective function. The multi-point injection opening sequence, segmented injection speed curve, and exhaust valve opening angle that minimize the overall failure risk are obtained and executed via fieldbus.

[0010] Before starting die casting, the die casting machine's mold clamping mechanism, injection mechanism, hydraulic system, cooling system, and electrical control system are checked for status. Once they are in normal working condition, the die casting process starts. If any subsystem is abnormal, an alarm message is generated and startup is prohibited.

[0011] The multi-source sensor network includes a temperature sensor array and a pressure sensor array installed in the mold cavity, a displacement sensor and a speed sensor installed at the injection punch, and a pressure sensor and a flow sensor installed in the hydraulic system of the die-casting machine. Each sensor collects data synchronously according to a preset sampling frequency, and the sampling frequency is not less than twice the frequency of change of key parameters during the die-casting process.

[0012] The present invention also provides a multi-point die-casting molding control system for an electronically controlled rear shell, used to execute the multi-point die-casting molding control method for an electronically controlled rear shell as described above. It includes a data acquisition module, an encrypted transmission module, a encrypted reasoning module, a data storage module, a data mining module, and a decision control module; The data acquisition module is configured on the multi-source sensor network of the die-casting machine and mold for real-time acquisition of multi-source process parameters. The encrypted transmission module is used to encrypt multi-source process parameters using homomorphic encryption or secure multi-party computation protocol before transmitting them to the edge computing node; The ciphertext reasoning module is deployed on edge computing nodes and is used to call the homomorphic die-casting state model to perform forward reasoning on the ciphertext without decryption and output the state ciphertext. The data storage module is used to encapsulate each injection stroke into a data block, and associate different batches of data blocks based on a directed acyclic graph structure to form a die casting process data cube. The data mining module is used to perform data mining on the data cube periodically or triggered by events, extract deep association rules, and update model weights and alarm rules. The decision control module is used to access the historical fault database of components and real-time operating condition data, dynamically predict the failure rate, embed multi-objective optimization decision-making, and generate control commands to be sent to the die-casting machine actuator.

[0013] This invention discloses a multi-point die-casting system and method for an electrically controlled rear shell. Through homomorphic encryption and a secure multi-party computation protocol, it directly completes die-casting state reasoning without decryption at edge nodes, avoiding leakage risks and balancing data security and real-time performance. A die-casting process data cube is constructed based on a directed acyclic graph structure, dynamically associating different batches according to process similarity and evolution paths. This overcomes the limitations of traditional isolated batch management and significantly improves the continuity of process control. Furthermore, periodic data mining automatically extracts deep correlation rules between process parameters and filling patterns, defect types, and component failure rates, dynamically updating the homomorphic die-casting state model and alarm rules accordingly. This enables the model to self-evolve and continuously adapt to actual production conditions. The historical component failure database and real-time operating data are integrated into process decision-making. Using predicted failure rates as constraints, the speed, pressure, and venting timing of multi-point injection are collaboratively optimized, achieving the generation of control commands with the lowest overall failure risk while ensuring molding quality. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0015] Figure 1 This is a flowchart of the multi-point die-casting molding control method for the electronically controlled rear shell of the present invention. Detailed Implementation

[0016] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0017] Please refer to Figure 1 This invention provides a multi-point die-casting control method for an electronically controlled rear shell, comprising: S1. After starting die casting, multi-source process parameters are collected through a multi-source sensor network and transmitted to the edge computing node after being encrypted by homomorphic encryption or secure multi-party computation protocol. In this specific embodiment, the multi-source sensor network includes a temperature sensor array and a pressure sensor array installed in the mold cavity, a displacement sensor and a speed sensor installed at the injection punch, and a pressure sensor and a flow sensor installed in the hydraulic system of the die-casting machine; each sensor synchronously collects data according to a preset sampling frequency, and the sampling frequency is not less than twice the frequency of change of key parameters during the die-casting process.

[0018] S2. Without decryption, the edge computing node directly calls the homomorphic die-casting state model to perform forward reasoning on the ciphertext and outputs the state ciphertext used to characterize the filling mode, exhaust efficiency and temperature field uniformity. In this specific implementation, the edge computing node receives the ciphertext of process parameters generated by random masking or polynomial encoding; loads a pre-deployed homomorphic die casting state model containing only addition and multiplication operations; inputs the ciphertext into the model, performs forward propagation calculation layer by layer in the ciphertext domain, and directly outputs the encrypted filling front position, air entrapment risk index, and temperature gradient deviation value as the state ciphertext.

[0019] S3. Each injection stroke is encapsulated as a data block. Based on the directed acyclic graph structure, the data blocks of different batches are associated according to process similarity and evolution path to form a traceable die casting process data cube, which stores multi-source process parameters, status ciphertext, operation logs and quality labels. In this specific implementation, the construction of the directed acyclic graph structure includes: A single injection stroke is defined as a data block containing a block header and a block body. The block header stores the hash value, timestamp, and batch identifier of the previous block, while the block body stores the desensitized process parameters, status ciphertext, and operation logs. Based on the evolution trajectory of process parameters and the similarity of molding quality between different batches, directed edges and lateral similarity edges are established to form a nonlinear multidimensional storage network. Based on this, a data cube is constructed with batch, time, parameter type, and quality label as dimensions, which supports any node to trace back to the process and quality information of all previous related nodes.

[0020] S4. Perform data mining on the data cube periodically or triggered to extract deep correlation rules between process parameters and filling patterns, defect types and component failure rates, and update the internal weights of the homomorphic die casting state model and the alarm rules of the process decision model based on the mining results. In this specific implementation, data mining tasks are executed periodically or triggered periodically, specifically including: An improved association rule mining algorithm was adopted, with nodes in the multidimensional storage network as transaction items, and support and confidence thresholds were set to mine strong association rules between high-speed filling speed range, specific exhaust groove state and air entrainment defects. A time-series-based causal analysis algorithm was used to analyze the time-delay causal chain between different opening sequences of multi-point ingates and the local wall thickness deviation, residual stress distribution, and final warping deformation of the rear shell. The discovered strong correlation rules and time-delay causal chains are used to automatically adjust the internal weight parameters of the homomorphic die-casting state model, or as the basis for generating process alarm rules.

[0021] S5. When executing the process decision model, the historical fault database of components and real-time operating condition data are accessed to dynamically predict the failure rate of each molding component in the current production cycle. The predicted failure rate is embedded as a constraint in the decision-making process of multi-point injection speed, pressure and venting timing. Control instructions that minimize the overall failure risk under the premise of ensuring molding quality are generated and sent to the die casting machine for execution.

[0022] In this specific implementation, the failure rate of each molded component is dynamically predicted and embedded into the decision-making process, specifically including: Temperature cycle curves, pressure peaks, and thermal stress-strain data of the mold under different cumulative injection cycles are extracted from the historical fault database. Combined with real-time working condition data, the probability of thermal fatigue cracks, slider wear rate, and ejector pin fracture risk index are predicted respectively. The predicted failure rates are normalized and used as constraints for multi-objective optimization. Together with the preset threshold for molding quality, they form the optimization objective function. The solution is used to obtain the multi-point injection opening sequence, segmented injection speed curve, and exhaust valve opening angle that minimize the overall failure risk. The results are then sent and executed via fieldbus.

[0023] Before starting die casting, the status of the die casting machine's mold clamping mechanism, injection mechanism, hydraulic system, cooling system, and electrical control system is checked. Once they are in normal working condition, the die casting process starts. If any subsystem is abnormal, an alarm message is generated and startup is prohibited.

[0024] The update of the homomorphic die-casting state model adopts an incremental learning approach, which only uses newly added production data to fine-tune the model weight parameters. While maintaining the original generalization ability of the model, it absorbs new process rules. After the updated model is verified by the encrypted domain validation set to ensure that the accuracy index is not lower than the preset threshold, it is automatically deployed to each edge computing node to replace the old version model.

[0025] When the control commands output by the process decision model exceed the physical limits of the die-casting machine's actuator, an alarm signal is generated and the machine automatically switches to a safety protection mode. The safety protection mode includes, but is not limited to, reducing the injection speed to a preset safety value, stopping injection and opening the mold closing mechanism, or triggering an emergency shutdown.

[0026] After die casting, the outer shell of the electronic control rear shell is inspected for appearance defects, measured for dimensional accuracy, and tested for air tightness. The test results are stored as quality tags in the data cube and associated with the corresponding injection stroke data blocks for monitoring signals in subsequent data mining tasks.

[0027] Working Principle: Through homomorphic encryption and secure multi-party computation protocols, die-casting state inference is completed directly without decryption at edge nodes, avoiding leakage risks and balancing data security and real-time performance. A die-casting process data cube is constructed based on a directed acyclic graph structure, dynamically associating different batches according to process similarity and evolution paths. This overcomes the limitations of traditional isolated batch management, significantly improving the continuity of process control. Furthermore, periodic data mining automatically extracts deep correlation rules between process parameters and filling patterns, defect types, and component failure rates, dynamically updating the homomorphic die-casting state model and alarm rules accordingly. This enables the model to self-evolve and continuously adapt to actual production conditions. Integrating historical component failure databases and real-time operating condition data into process decision-making, the predicted failure rate serves as a constraint to collaboratively optimize the speed, pressure, and venting timing of multi-point injection, achieving the generation of control commands with the lowest overall failure risk while ensuring molding quality.

[0028] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of this application still fall within the scope of this application.

Claims

1. A method for controlling the multi-point die-casting of an electronically controlled rear shell, characterized in that, include: After the die casting process is started, multi-source process parameters are collected through a multi-source sensor network and transmitted to the edge computing node after being encrypted by homomorphic encryption or secure multi-party computation protocol. Without decryption, the edge computing node directly calls the homomorphic die casting state model to perform forward reasoning on the ciphertext and outputs state ciphertext to characterize the filling mode, exhaust efficiency and temperature field uniformity. Each injection stroke is encapsulated as a data block. Based on the directed acyclic graph structure, the data blocks of different batches are associated according to process similarity and evolution path to form a traceable die casting process data cube, which stores multi-source process parameters, status ciphertext, operation logs and quality tags. Data mining is performed periodically or triggered on the data cube to extract deep correlation rules between process parameters and filling patterns, defect types and component failure rates, and the internal weights of the homomorphic die casting state model and the alarm rules of the process decision model are updated based on the mining results. When executing the process decision model, the historical fault database of components and real-time operating condition data are accessed to dynamically predict the failure rate of each molded component in the current production cycle. The predicted failure rate is embedded as a constraint in the decision-making process of multi-point injection speed, pressure and venting timing. Control instructions that minimize the overall failure risk under the premise of ensuring molding quality are generated and sent to the die casting machine for execution.

2. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 1, characterized in that, Edge computing nodes receive ciphertext of process parameters generated by random masking or polynomial encoding; load a pre-deployed homomorphic die-casting state model containing only addition and multiplication operations; input the ciphertext into the model, perform forward propagation calculation layer by layer in the ciphertext domain, and directly output the encrypted filling front position, air entrapment risk index, and temperature gradient deviation value as state ciphertext.

3. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 2, characterized in that, Execute data mining tasks periodically or on a trigger-based basis, specifically including: An improved association rule mining algorithm was adopted, with nodes in the multidimensional storage network as transaction items, and support and confidence thresholds were set to mine strong association rules between high-speed filling speed range, specific exhaust groove state and air entrainment defects. A time-series-based causal analysis algorithm was used to analyze the time-delay causal chain between different opening sequences of multi-point ingates and the local wall thickness deviation, residual stress distribution, and final warping deformation of the rear shell. The discovered strong correlation rules and time-delay causal chains are used to automatically adjust the internal weight parameters of the homomorphic die-casting state model, or as the basis for generating process alarm rules.

4. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 3, characterized in that, The construction of a directed acyclic graph structure includes: A single injection stroke is defined as a data block containing a block header and a block body. The block header stores the hash value, timestamp, and batch identifier of the previous block, while the block body stores the desensitized process parameters, status ciphertext, and operation logs. Based on the evolution trajectory of process parameters and the similarity of molding quality between different batches, directed edges and lateral similarity edges are established to form a nonlinear multidimensional storage network. Based on this, a data cube is constructed with batch, time, parameter type, and quality label as dimensions, which supports any node to trace back to the process and quality information of all previous related nodes.

5. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 4, characterized in that, Dynamically predict the failure rate of each molded component and embed it into the decision-making process, specifically including: Temperature cycle curves, pressure peaks, and thermal stress-strain data of the mold under different cumulative injection cycles are extracted from the historical fault database. Combined with real-time working condition data, the probability of thermal fatigue cracks, slider wear rate, and ejector pin fracture risk index are predicted respectively. The predicted failure rates are normalized and used as constraints for multi-objective optimization. Together with the preset threshold for molding quality, they form the optimization objective function. The multi-point injection opening sequence, segmented injection speed curve, and exhaust valve opening angle that minimize the overall failure risk are obtained and executed via fieldbus.

6. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 5, characterized in that, Before starting die casting, the status of the die casting machine's mold clamping mechanism, injection mechanism, hydraulic system, cooling system, and electrical control system is checked. Once they are in normal working condition, the die casting process starts. If any subsystem is abnormal, an alarm message is generated and startup is prohibited.

7. The multi-point die-casting molding control method for the electronically controlled rear shell as described in claim 6, characterized in that, The multi-source sensor network includes temperature and pressure sensor arrays installed in the mold cavity, displacement and speed sensors installed at the injection punch, and pressure and flow sensors installed in the hydraulic system of the die-casting machine. Each sensor collects data synchronously according to a preset sampling frequency, which is no less than twice the frequency of change of key parameters during the die-casting process.

8. A multi-point die-casting molding control system for an electrically controlled rear shell, used to execute the multi-point die-casting molding control method for an electrically controlled rear shell as described in claim 7, characterized in that, It includes a data acquisition module, an encrypted transmission module, a encrypted reasoning module, a data storage module, a data mining module, and a decision control module; The data acquisition module is configured on the multi-source sensor network of the die-casting machine and mold for real-time acquisition of multi-source process parameters. The encrypted transmission module is used to encrypt multi-source process parameters using homomorphic encryption or secure multi-party computation protocol before transmitting them to the edge computing node; The ciphertext reasoning module is deployed on edge computing nodes and is used to call the homomorphic die-casting state model to perform forward reasoning on the ciphertext without decryption and output the state ciphertext. The data storage module is used to encapsulate each injection stroke into a data block, and associate different batches of data blocks based on a directed acyclic graph structure to form a die casting process data cube. The data mining module is used to perform data mining on the data cube periodically or triggered by events, extract deep association rules, and update model weights and alarm rules. The decision control module is used to access the historical fault database of components and real-time operating condition data, dynamically predict the failure rate, embed multi-objective optimization decision-making, and generate control commands to be sent to the die-casting machine actuator.

Images