Production process manufacturing execution system based on closed-loop management

Abstract

The invention relates to the technical field of intelligent management, in particular to a closed-loop management-based production process manufacturing execution system, which can find tiny deviation and deterioration trend of a process before a product has physical defects by comparing an injection curve with a standard curve in real time and evaluating the heat balance state of a mold, so that the production process is more accurate. A quality control node is greatly moved forward, and the risk of producing batch defective products is effectively reduced; according to the scheme, through the curve goodness of fit and the comprehensive evaluation index, quantitative management and control of the dynamic form consistency of the whole injection process are achieved; through the heat balance index, quantitative control over the uniformity and stability of the whole-field temperature of the mold is achieved, it is guaranteed that each mold of products are produced under the highly-consistent technological condition from the source, and the consistency of the products in the batches and between the batches is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology, and in particular to a production process manufacturing execution system based on closed-loop management. Background Technology

[0002] In the die-casting manufacturing industry, the stability and consistency of the production process are core factors determining product quality, production efficiency, and cost control. Traditional die-casting process control mainly relies on the experience of operators, through setting and periodically adjusting parameters on the die-casting machine panel (such as injection speed and pressure). However, this method, which depends on human experience, has significant limitations: First, humans cannot perceive and process massive amounts of information in real time, such as millisecond-level key curve data of the injection process and multi-point temperature fields of the mold; second, experience-based adjustments often lag behind the occurrence of quality problems, constituting post-event remediation and making it difficult to achieve pre-event prevention and in-process control.

[0003] Chinese Patent Application No. CN202311833359.2 discloses a cloud-based manufacturing execution system and its manufacturing execution method, including a user interface module, a production planning and scheduling module, an equipment control module, a quality control module, a material management module, a data analysis and reporting module, a security backup module, and a service interface module. This application significantly improves the automation level of the production process through automated production planning, equipment control, quality control, and material management modules, thereby increasing production efficiency and resource utilization, enabling rapid response to production problems, optimizing production plans, reducing downtime, monitoring product quality through sensors, and analyzing production process data through the data analysis and reporting module. This supports traceability and continuous improvement, enhances product quality and process optimization, tracks inventory levels in real time, and ensures timely material supply. Simultaneously, the supply chain coordination module coordinates material procurement and supply chain activities through interaction with suppliers, optimizing supply chain efficiency.

[0004] However, existing technologies still have the following problems: Existing systems can only collect and store isolated data streams such as the injection process, mold temperature, and quality inspection, but cannot automatically and accurately reveal the quantitative causal relationship between fluctuations in specific process parameters and specific quality defects. When problems occur, engineers can only guess the cause based on their experience, which is time-consuming, labor-intensive, and inaccurate. Summary of the Invention

[0005] To address this, the present invention provides a production process execution system based on closed-loop management, which overcomes the problem that existing systems can only collect and store isolated data streams such as injection process, mold temperature, and quality inspection, but cannot automatically and accurately reveal the quantitative causal relationship between fluctuations in specific process parameters and specific quality defects. When problems occur, engineers can only guess the cause based on their experience, which is time-consuming, labor-intensive, and inaccurate.

[0006] To achieve the above objectives, the present invention provides a manufacturing execution system for a production process based on closed-loop management. It includes: The process data sensing and acquisition unit is used to collect real-time operating data of the die casting production line. The operating data includes: injection process curve data from the die casting machine, real-time clamping force data, and temperature data from multiple measuring points of the mold. The production process model library unit contains a pre-stored reference process parameter package that matches different die castings and die casting molds. The reference process parameter package includes standard injection curves, target ranges for mold temperature control, and process window thresholds. The real-time analysis and decision-making central unit is connected to the process data sensing and acquisition unit and the production process model library unit, respectively. It is used to dynamically compare the real-time acquired injection process curve data and real-time clamping force data with the standard injection curve of the corresponding product, calculate the curve matching degree and key parameter deviation, and calculate the comprehensive evaluation index based on the curve matching degree and key parameter deviation. Based on the real-time collected temperature data of the die-casting mold and the target range of mold temperature control, the thermal balance state of the mold is evaluated. Whether to generate a primary process alarm and adjustment instruction based on comprehensive evaluation indicators and mold thermal balance state analysis; The process operation data within the historical time period corresponding to the primary process alarm and adjustment command are correlated with the real-time quality inspection data of the same batch of die castings to establish a mapping relationship between process parameter fluctuations and quality defect types. Based on the mapping relationship, the confidence level of the primary process alarm and adjustment command is verified and the parameters are corrected to generate the final process optimization command. A quality data correlation analysis unit, which is connected to the real-time analysis and decision-making center unit, is used to obtain real-time quality inspection data of the die casting. The instruction closed-loop execution and feedback unit is connected to the process data sensing and acquisition unit and the real-time analysis and decision-making center unit, respectively, to issue the final process optimization instruction to the die-casting machine or mold temperature controller for execution, and to continuously receive data from the process data sensing and acquisition unit.

[0007] Furthermore, the real-time analysis and decision-making central unit is used to calculate the curve fit degree, including: Key feature segments are extracted from the standard injection curve, including the slow injection segment, the high-speed switching point, and the high-speed injection segment. The real-time collected injection process curves are aligned with the standard injection curve on the time axis using the same feature points as a reference. Calculate the first difference measure between the real-time curve and the standard curve in each of the key feature segments. The first difference measure includes: the root mean square error of the corresponding data points in the key feature segment, the shape similarity coefficient of the curve in the key feature segment, and the absolute deviation between the measured value and the target value of the key parameter in the key feature segment. The first difference measure of each of the key feature segments is weighted and summed to obtain the curve matching degree, wherein the weight of the high-speed injection segment is higher than that of the slow injection segment.

[0008] Furthermore, the real-time analysis and decision-making central unit is used to calculate the deviation of key parameters, including: From the standard injection curve and the associated reference process parameter package, predefined key parameters and their target values ​​are extracted. The key parameters include: the first-speed to second-speed switching position, the peak high-speed injection speed, the pressure build-up point pressure, and the peak pressure boosting pressure. Identify and locate the actual feature points corresponding to predefined key parameters from the preprocessed and calibrated real-time injection process curve data; Calculate the absolute deviation between the actual value and the target value of each of the key parameters; Based on the process window threshold corresponding to the key parameter, the normalized relative deviation is calculated. The normalized relative deviation is the percentage value obtained by dividing the absolute deviation by half of the process window threshold.

[0009] Furthermore, the real-time analysis and decision-making central unit is used to calculate a comprehensive evaluation index based on the curve fit and key parameter deviation, as shown in the following formula: Z=α×S+Σ(βi×|ΔPi|×Ti); Where Z is the comprehensive evaluation index; S is the quantitative value of curve fit; α is the weighting coefficient of fit; ΔPi is the deviation between the measured value and the target value of the i-th key parameter; Ti is the process window threshold corresponding to the i-th key parameter; and βi is the deviation weighting coefficient of the i-th key parameter.

[0010] Furthermore, the real-time analysis and decision-making central unit is used to assess the thermal balance state of the mold, including: Based on the temperature data of each measuring point and its three-dimensional spatial coordinates on the mold, the mold is divided into several logical evaluation areas, and the average temperature, temperature range within the area, and temperature change rate of each logical evaluation area are calculated. Compare the average temperature of each logical evaluation region with the temperature of the corresponding sub-target range in the mold temperature control target range, and calculate the first type of deviation D1; The temperature range within each logical evaluation region is compared with the preset allowable temperature gradient threshold to calculate the second type of deviation D2. The regional temperature change rate of each region during the current production cycle is compared with the baseline change rate under steady-state production conditions to calculate the third type of deviation D3. The formula for the comprehensive heat balance index is: H=w1×f(D1)+w2×g(D2)+w3×h(D3); Where w1, w2, and w3 are the weight coefficients of each deviation component, and w1 + w2 + w3 = 1; f(D1), g(D2), and h(D3) are functions that normalize D1, D2, and D3, respectively. If the comprehensive thermal balance index of any logical evaluation area exceeds the preset thermal balance anomaly threshold, or if the comprehensive thermal balance index of more than a preset number of logical evaluation areas exceeds the warning threshold, then the mold thermal balance status evaluation is determined to be abnormal.

[0011] Furthermore, the real-time analysis and decision-making central unit is used to analyze whether to generate a primary process alarm and adjustment command based on comprehensive evaluation indicators and mold thermal balance status, including: If the comprehensive evaluation index is greater than or equal to the first preset threshold, and the mold thermal balance state is assessed as abnormal, a primary process alarm and adjustment instruction will be generated. If the comprehensive evaluation index is greater than or equal to the first preset threshold, but the mold thermal balance status is assessed as normal, then process prompt information is generated only for the injection process.

[0012] Furthermore, the real-time analysis and decision-making central unit also includes: The process parameter-quality defect mapping library subunit is used to store and manage the mapping relationship model between process parameter fluctuations and quality defect types. The mapping relationship model is established and updated in the following ways: Continuously collect process operation data and corresponding die-casting quality inspection data from historical production cycles; A process fluctuation feature vector is constructed for each injection process, and a defect characterization vector is constructed for the corresponding die casting. Using the process fluctuation feature vector as input and the defect characterization vector as output target, a multi-output regression model is trained or updated periodically or after accumulating a certain amount of data. The weight matrix of this model is the mapping relationship.

[0013] Furthermore, the real-time analysis and decision-making central unit is used to call the current mapping relationship model in the process parameter-quality defect mapping library subunit after generating the primary process alarm and adjustment instructions, and to perform confidence verification and parameter correction on the instructions.

[0014] Furthermore, the real-time analysis and decision-making central unit is also used to trigger a parameter correction process, including: Based on the overall confidence assessment, different correction strategies are dynamically selected by comparing it with a preset set of confidence interval thresholds. When the overall confidence level assessment falls within the middle confidence range, a local iterative correction strategy is selected; When the overall confidence assessment falls into the low confidence range, a global optimization correction strategy is selected.

[0015] Furthermore, the primary process alarm and adjustment instructions include the anomaly type, anomaly severity, and recommended adjustment parameters.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: by comparing the injection curve with the standard curve in real time and evaluating the thermal balance of the mold, this invention can detect minor deviations and deterioration trends in the process before physical defects appear in the product, thus significantly advancing the quality control point and effectively reducing the risk of batch defects. This solution achieves quantitative control over the dynamic morphological consistency of the entire injection process through curve matching and comprehensive evaluation indicators; and through the thermal balance index, it achieves quantitative control over the temperature uniformity and stability of the entire mold, fundamentally ensuring that each batch of products is produced under highly consistent process conditions, greatly improving the consistency of products within and between batches.

[0017] Furthermore, the process parameter fluctuation-quality defect mapping relationship established in this invention transforms the experience and intuition of senior process engineers into a calculable and verifiable digital model. When quality problems occur, the system can quickly analyze and point out the most likely root cause of process fluctuations, such as the decrease in the curve matching degree of the high-speed injection section leading to an increase in the probability of air entrainment defects. This changes the inefficient mode of relying on manual item-by-item inspection, greatly shortens the problem analysis time, and improves the accuracy of diagnosis.

[0018] Furthermore, by pre-extracting three key feature segments with clear physical meaning—the slow injection segment, the high-speed switching point, and the high-speed injection segment—this method focuses the evaluation of the curve on the process phase that plays a decisive role in product quality. This greatly enhances the direct correlation between the evaluation results and the final product performance (such as internal density, surface quality, and contour clarity), and significantly improves the engineering guidance value of the evaluation.

[0019] Furthermore, the normalized relative deviation data in this invention is a high-quality feature with standardized and uniform scale. It is beneficial to use relative deviation as an input feature when establishing the mapping relationship between process parameter fluctuations and quality defect types. The model can more clearly learn the quantitative relationship between the degree to which different parameters deviate from their normal range and the quality results. In the parameter correction process, whether it is local iteration or global optimization, the objective function or constraint condition composed of relative deviation can ensure that the algorithm searches within a reasonable process space, making the optimization results more in line with engineering practice.

[0020] Furthermore, this invention creatively integrates the morphological consistency index with the positional deviation index of multiple key parameters through weighted fusion using weighting coefficients and process window thresholds to generate a single comprehensive evaluation index. This condenses the complex process state into a numerical value that can be directly used for comparison and decision-making, greatly simplifying the subsequent judgment logic for whether to issue an alarm. It only requires comparing the comprehensive evaluation index with a preset threshold. The construction of the comprehensive evaluation index and the mold thermal balance index marks the evolution of die-casting process monitoring from scattered and one-sided parameter monitoring to systematic, multi-dimensional, and quantifiable state assessment. It provides high-fidelity, high-information-density core perception data for the entire closed-loop intelligent optimization system, which is a key prerequisite for achieving predictive maintenance and precise process optimization.

[0021] Furthermore, this invention systematically constructs a process fluctuation feature vector (including curve fit, multi-parameter relative deviation, and multi-region thermal deviation) and a defect characterization vector (defect type and severity encoded by multiple labels), and trains a multi-output regression model. This successfully encodes complex, nonlinear causal relationships into an executable mathematical matrix, enabling the model to instantly obtain a quantitative answer to the question of what kind of fluctuation leads to what kind of defect.

[0022] Furthermore, in this invention, when the system has a high degree of confidence in its own diagnosis (high overall confidence), it quickly executes the suggestions of the primary alarm, ensuring decision-making and execution efficiency in clear and typical problem scenarios. When the system has doubts about the diagnosis (low overall confidence), it does not act rashly, but instead initiates a more complex and cautious parameter correction process. This reflects that the system acknowledges the limitations of the model and rules, and that when faced with atypical, complex, or newly emerging abnormal patterns, it needs to call more advanced optimization algorithms for exploration. The confidence verification and diversion mechanism of this invention elevates the entire system from open-loop automated execution to an intelligent closed loop with self-evaluation and self-adjustment capabilities. It not only ensures the safety valve for correct single decisions, but also serves as the engine driving the entire system to continuously learn and evolve in practice. Attached Figure Description

[0023] Figure 1 This is a structural block diagram of the manufacturing execution system for production processes based on closed-loop management according to the present invention; Figure 2 A flowchart for analyzing whether to generate primary process alarms and adjustment instructions. Detailed Implementation

[0024] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0025] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0026] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] Please see Figure 1 As shown, it is a structural block diagram of the production process manufacturing execution system based on closed-loop management of the present invention.

[0028] The manufacturing execution system based on closed-loop management provided in this embodiment includes: The process data sensing and acquisition unit is used to collect real-time operating data of the die casting production line. The operating data includes: injection process curve data from the die casting machine, real-time clamping force data, and temperature data from multiple measuring points of the mold. The production process model library unit contains a pre-stored reference process parameter package that matches different die castings and die casting molds. The reference process parameter package includes standard injection curves, target ranges for mold temperature control, and process window thresholds. The real-time analysis and decision-making central unit is connected to the process data sensing and acquisition unit and the production process model library unit, respectively. It is used to dynamically compare the real-time acquired injection process curve data and real-time clamping force data with the standard injection curve of the corresponding product, calculate the curve matching degree and key parameter deviation, and calculate the comprehensive evaluation index based on the curve matching degree and key parameter deviation. Based on the real-time collected temperature data of the die-casting mold and the target range of mold temperature control, the thermal balance state of the mold is evaluated. Whether to generate a primary process alarm and adjustment instruction based on comprehensive evaluation indicators and mold thermal balance state analysis; The process operation data within the historical time period corresponding to the primary process alarm and adjustment command are correlated with the real-time quality inspection data of the same batch of die castings to establish a mapping relationship between process parameter fluctuations and quality defect types. Based on the mapping relationship, the confidence level of the primary process alarm and adjustment command is verified and the parameters are corrected to generate the final process optimization command. A quality data correlation analysis unit, which is connected to the real-time analysis and decision-making center unit, is used to obtain real-time quality inspection data of the die casting. The instruction closed-loop execution and feedback unit is connected to the process data sensing and acquisition unit and the real-time analysis and decision-making center unit, respectively, to issue the final process optimization instruction to the die-casting machine or mold temperature controller for execution, and to continuously receive data from the process data sensing and acquisition unit.

[0029] Specifically, in this embodiment, the timeline of the data closed-loop flow is as follows: real-time data acquisition and status assessment; generation of a primary alarm when conditions are met; simultaneously, acquisition of process data and quality data (as a validation set) of recently produced workpieces with quality inspection results; invocation of the mapping model to predict quality defects based on the abnormal characteristics of the primary alarm, comparison with actual quality inspection results, and calculation of confidence level; determination of whether to directly execute instructions or enter the parameter correction process based on high / low confidence level; execution of optimization instructions, and storage of new result data in the database for future updates to the mapping model.

[0030] This invention, by comparing the injection curve with the standard curve in real time and evaluating the thermal balance of the mold, can detect minute deviations and deterioration trends in the process before physical defects appear in the product. This significantly advances the quality control point and effectively reduces the risk of batch defects. This solution achieves quantitative control over the dynamic morphological consistency of the entire injection process through curve matching and comprehensive evaluation indicators. Through the thermal balance index, it achieves quantitative control over the temperature uniformity and stability of the entire mold, ensuring that each batch of products is produced under highly consistent process conditions from the source, greatly improving the consistency of products within and between batches.

[0031] The process parameter fluctuation-quality defect mapping relationship established in this invention transforms the experience and intuition of senior process engineers into a calculable and verifiable digital model. When quality problems occur, the system can quickly analyze and point out the most likely root cause of process fluctuations, such as the decrease in the curve matching degree of the high-speed injection section leading to an increase in the probability of air entrainment defects. This changes the inefficient mode of relying on manual item-by-item inspection, greatly shortens the problem analysis time, and improves the accuracy of diagnosis.

[0032] Specifically, the real-time analysis and decision-making central unit is used to calculate the curve fit, including: Key feature segments are extracted from the standard injection curve, including the slow injection segment, the high-speed switching point, and the high-speed injection segment. The real-time collected injection process curves are aligned with the standard injection curve on the time axis using the same feature points as a reference. Calculate the first difference measure between the real-time curve and the standard curve in each of the key feature segments. The first difference measure includes: the root mean square error of the corresponding data points in the key feature segment, the shape similarity coefficient of the curve in the key feature segment, and the absolute deviation between the measured value and the target value of the key parameter in the key feature segment. The first difference measure of each of the key feature segments is weighted and summed to obtain the curve matching degree, wherein the weight of the high-speed injection segment is higher than that of the slow injection segment.

[0033] Specifically, in this embodiment, the shape similarity coefficient is calculated using a dynamic time warping algorithm, which is used to nonlinearly stretch or compress the real-time curve and the standard curve on the time axis to find the optimal shape matching path, and the cumulative distance or warped path length is used as the basis for calculating the shape similarity coefficient. Before calculating the curve matching degree, the real-time collected injection process curve data is processed by moving average filtering or Kalman filtering to eliminate noise interference, and according to the equipment identification of the die casting machine, the corresponding sensor error correction coefficient is called to calibrate the filtered data.

[0034] In this invention, by pre-extracting three key feature segments with clear physical meaning—the slow injection segment, the high-speed switching point, and the high-speed injection segment—this method focuses the evaluation of the curve on the process phase that plays a decisive role in product quality. This greatly enhances the direct correlation between the evaluation results and the final product performance (such as internal density, surface quality, and contour clarity), and significantly improves the engineering guidance value of the evaluation.

[0035] Specifically, the real-time analysis and decision-making central unit is used to calculate the deviation of key parameters, including: From the standard injection curve and the associated reference process parameter package, predefined key parameters and their target values ​​are extracted. The key parameters include: the first-speed to second-speed switching position, the peak high-speed injection speed, the pressure build-up point pressure, and the peak pressure boosting pressure. Identify and locate the actual feature points corresponding to predefined key parameters from the preprocessed and calibrated real-time injection process curve data; Calculate the absolute deviation between the actual value and the target value of each of the key parameters; Based on the process window threshold corresponding to the key parameter, the normalized relative deviation is calculated. The normalized relative deviation is the percentage value obtained by dividing the absolute deviation by half of the process window threshold.

[0036] Specifically, in this embodiment, identifying and locating actual feature points includes: for the speed transition from one speed to two speed, locating the displacement point where the speed undergoes a step change by analyzing the first derivative of the real-time injection displacement curve; for the high-speed injection speed peak, finding the global maximum value point of the speed curve within the identified high-speed injection stage; for the pressure build-up point, locating the inflection point on the pressure curve after the injection ends where the pressure changes from a rapid increase to a stable maintenance; and for the peak pressure boost, finding the maximum value point of the pressure curve within a set time window after the pressure build-up point.

[0037] In this invention, the normalized relative deviation data is a high-quality feature with standardized and uniform scale. It is beneficial to use relative deviation as an input feature when establishing the mapping relationship between process parameter fluctuations and quality defect types. The model can learn more clearly the quantitative relationship between the degree to which different parameters deviate from their normal range and the quality results. In the parameter correction process, whether it is local iteration or global optimization, the objective function or constraint condition composed of relative deviation can ensure that the algorithm searches within a reasonable process space, making the optimization results more in line with engineering practice.

[0038] Specifically, the real-time analysis and decision-making central unit is used to calculate a comprehensive evaluation index based on curve fit and key parameter deviation, as shown in the following formula: Z=α×S+Σ(βi×|ΔPi|×Ti); Where Z is the comprehensive evaluation index; S is the quantitative value of curve fit; α is the weighting coefficient of fit; ΔPi is the deviation between the measured value and the target value of the i-th key parameter; Ti is the process window threshold corresponding to the i-th key parameter; and βi is the deviation weighting coefficient of the i-th key parameter.

[0039] Specifically, in this embodiment, the lower the comprehensive evaluation index, the better the process consistency, and the higher the curve fit, the better the fit.

[0040] Specifically, the real-time analysis and decision-making central unit is used to assess the thermal balance state of the mold, including: Based on the temperature data of each measuring point and its three-dimensional spatial coordinates on the mold, the mold is divided into several logical evaluation areas, and the average temperature, temperature range within the area, and temperature change rate of each logical evaluation area are calculated. Compare the average temperature of each logical evaluation region with the temperature of the corresponding sub-target range in the mold temperature control target range, and calculate the first type of deviation D1; The temperature range within each logical evaluation region is compared with the preset allowable temperature gradient threshold to calculate the second type of deviation D2. The regional temperature change rate of each region during the current production cycle is compared with the baseline change rate under steady-state production conditions to calculate the third type of deviation D3. The formula for the comprehensive thermal balance index is as follows: H=w1×f(D1)+w2×g(D2)+w3×h(D3); Where w1, w2, and w3 are the weight coefficients of each deviation component, and w1 + w2 + w3 = 1; f(D1), g(D2), and h(D3) are functions that normalize D1, D2, and D3, respectively. If the comprehensive thermal balance index of any logical evaluation area exceeds the preset thermal balance anomaly threshold, or if the comprehensive thermal balance index of more than a preset number of logical evaluation areas exceeds the warning threshold, then the mold thermal balance status evaluation is determined to be abnormal.

[0041] Specifically, in this embodiment, the method of dividing the mold into multiple logical evaluation regions includes: pre-defining division rules based on the physical structure and thermal conductivity characteristics of the die-casting mold; the division rules are: dividing the cavity surface, gating system, hot spots and insert areas into independent logical evaluation regions, and dividing the mold frame portion away from the cavity into one or more background regions.

[0042] Specifically, in this embodiment, the formula for calculating the first type of deviation D1 is: D1=|Tavg-Ttarget| / (0.5×ΔTrange); where, Tavg is the average temperature of the region, Ttarget is the median of the corresponding sub-target range, and ΔTrange is the width of the sub-target range.

[0043] The formula for calculating the second type of deviation D2 is: D2=max(0,(Tmax-Tmin-Gthreshold) / Gthreshold) Where Tmax and Tmin are the extreme temperatures in the region, respectively, and Gthreshold is the preset allowable temperature gradient threshold.

[0044] This invention creatively integrates the morphological consistency index with the positional deviation index of multiple key parameters through weighted fusion using weighting coefficients and process window thresholds to generate a single comprehensive evaluation index. This condenses the complex process state into a numerical value that can be directly used for comparison and decision-making, greatly simplifying the subsequent judgment logic for whether to issue an alarm. It only requires comparing the comprehensive evaluation index with a preset threshold. The construction of the comprehensive evaluation index and the mold thermal balance index marks the evolution of die casting process monitoring from scattered and one-sided parameter monitoring to systematic, multi-dimensional, and quantifiable state assessment. It provides high-fidelity, high-information-density core perception data for the entire closed-loop intelligent optimization system, which is a key prerequisite for achieving predictive maintenance and precise process optimization.

[0045] Please see Figure 2 As shown, it is a flowchart for determining whether to generate a primary process alarm and adjustment command.

[0046] Specifically, the real-time analysis and decision-making central unit is used to analyze whether to generate a primary process alarm and adjustment command based on comprehensive evaluation indicators and mold thermal balance status, including: If the comprehensive evaluation index is greater than or equal to the first preset threshold, and the mold thermal balance state is assessed as abnormal, a primary process alarm and adjustment instruction will be generated. If the comprehensive evaluation index is greater than or equal to the first preset threshold, but the mold thermal balance status is assessed as normal, then process prompt information is generated only for the injection process.

[0047] Specifically, in this embodiment, the process prompt information is for operator reference and does not trigger subsequent confidence verification and parameter correction processes.

[0048] Specifically, the real-time analysis and decision-making central unit also includes: The process parameter-quality defect mapping library subunit is used to store and manage the mapping relationship model between process parameter fluctuations and quality defect types. The mapping relationship model is established and updated in the following ways: Continuously collect process operation data and corresponding die-casting quality inspection data from historical production cycles; A process fluctuation feature vector is constructed for each injection process, and a defect characterization vector is constructed for the corresponding die casting. Using the process fluctuation feature vector as input and the defect characterization vector as output target, a multi-output regression model is trained or updated periodically or after accumulating a certain amount of data. The weight matrix of this model is the mapping relationship.

[0049] Specifically, in this embodiment, the defect characterization vector adopts a multi-label encoding form, where each dimension represents the presence or absence and severity level of a quality defect type; the dimensions of the process fluctuation feature vector include: the curve fit S, the normalized relative deviation of each key parameter, and the temperature field anomaly feature value composed of the thermal balance deviations D1 and D2 of each logical evaluation area of ​​the mold. This invention systematically constructs a process fluctuation feature vector (including curve fit, multi-parameter relative deviation, and multi-region thermal deviation) and a defect characterization vector (defect type and severity encoded by multiple labels), and trains a multi-output regression model. This successfully encodes complex, nonlinear causal relationships into an executable mathematical matrix, enabling the model to instantly obtain a quantitative answer to the question of what kind of fluctuation leads to what kind of defect.

[0050] Specifically, the real-time analysis and decision-making central unit is used to call the current mapping relationship model in the process parameter-quality defect mapping library subunit after generating primary process alarms and adjustment instructions, and to perform confidence verification and parameter correction on the instructions.

[0051] Specifically, in this embodiment, the parameter correction process includes: with the goal of minimizing the predicted defect type probability vector, using gradient descent or linear programming, under the constraints of the mapping relationship of the multi-output regression model, to solve in reverse the optimal process parameter adjustment direction and magnitude that reduces the predicted defect probability, thereby correcting the recommended adjustment parameters in the primary alarm and generating the final process optimization instruction.

[0052] In this invention, when the system has a high degree of confidence in its own diagnosis (high overall confidence), it quickly executes the suggestions of the primary alarm, ensuring decision-making and execution efficiency in clear and typical problem scenarios. When the system has doubts about the diagnosis (low overall confidence), it does not act rashly, but instead initiates a more complex and cautious parameter correction process. This reflects that the system acknowledges the limitations of the model and rules, and that when faced with atypical, complex, or newly emerging abnormal patterns, it needs to call more advanced optimization algorithms for exploration. The confidence verification and diversion mechanism of this invention elevates the entire system from open-loop automated execution to an intelligent closed loop with self-evaluation and self-adjustment capabilities. It not only ensures the safety valve for correct single decisions, but also serves as the engine driving the entire system to continuously learn and evolve in practice.

[0053] Specifically, the real-time analysis and decision-making central unit is also used to trigger a parameter correction process, including: Based on the overall confidence assessment, different correction strategies are dynamically selected by comparing it with a preset set of confidence interval thresholds. When the overall confidence level assessment falls within the middle confidence range, a local iterative correction strategy is selected; When the overall confidence assessment falls into the low confidence range, a global optimization correction strategy is selected.

[0054] Specifically, in this embodiment, the local iterative correction strategy is as follows: based on the adjustment parameters recommended by the primary alarm, combined with the gradient information of the multi-output regression model, a small-range, multi-round iterative search is performed to find a suboptimal solution that minimizes the deviation from the original recommended parameters while satisfying the quality improvement goal; the global optimization correction strategy is as follows: ignoring the recommended parameters of the primary alarm, the multi-output regression model is regarded as a black-box objective function, and within the preset feasible domain of all process parameters, a heuristic optimization algorithm is used to directly search for the process parameter combination that makes the overall predicted defect probability vector optimal.

[0055] Specifically, the primary process alarm and adjustment instructions include the anomaly type, anomaly severity, and recommended adjustment parameters.

[0056] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A manufacturing execution system for production processes based on closed-loop management, characterized in that, include: The process data sensing and acquisition unit is used to collect real-time operating data of the die casting production line. The operating data includes: injection process curve data from the die casting machine, real-time clamping force data, and temperature data from multiple measuring points of the mold. The production process model library unit contains a pre-stored reference process parameter package that matches different die castings and die casting molds. The reference process parameter package includes standard injection curves, target ranges for mold temperature control, and process window thresholds. The real-time analysis and decision-making central unit is connected to the process data sensing and acquisition unit and the production process model library unit, respectively. It is used to dynamically compare the real-time acquired injection process curve data and real-time clamping force data with the standard injection curve of the corresponding product, calculate the curve matching degree and key parameter deviation, and calculate the comprehensive evaluation index based on the curve matching degree and key parameter deviation. Based on the real-time collected temperature data of the die-casting mold and the target range of mold temperature control, the thermal balance state of the mold is evaluated. Whether to generate a primary process alarm and adjustment instruction based on comprehensive evaluation indicators and mold thermal balance state analysis; The process operation data within the historical time period corresponding to the primary process alarm and adjustment command are correlated with the real-time quality inspection data of the same batch of die castings to establish a mapping relationship between process parameter fluctuations and quality defect types. Based on the mapping relationship, the confidence level of the primary process alarm and adjustment command is verified and the parameters are corrected to generate the final process optimization command. A quality data correlation analysis unit, which is connected to the real-time analysis and decision-making center unit, is used to obtain real-time quality inspection data of the die casting. The instruction closed-loop execution and feedback unit is connected to the process data sensing and acquisition unit and the real-time analysis and decision-making center unit, respectively, to issue the final process optimization instruction to the die-casting machine or mold temperature controller for execution, and to continuously receive data from the process data sensing and acquisition unit.

2. The manufacturing execution system based on closed-loop management according to claim 1, characterized in that, The real-time analysis and decision-making central unit is used to calculate the curve fit degree, including: Key feature segments are extracted from the standard injection curve, including the slow injection segment, the high-speed switching point, and the high-speed injection segment. The real-time collected injection process curves are aligned with the standard injection curve on the time axis using the same feature points as a reference. Calculate the first difference measure between the real-time curve and the standard curve in each of the key feature segments. The first difference measure includes: the root mean square error of the corresponding data points in the key feature segment, the shape similarity coefficient of the curve in the key feature segment, and the absolute deviation between the measured value and the target value of the key parameter in the key feature segment. The first difference measure of each of the key feature segments is weighted and summed to obtain the curve matching degree, wherein the weight of the high-speed injection segment is higher than that of the slow injection segment.

3. The manufacturing execution system based on closed-loop management according to claim 1, characterized in that, The real-time analysis and decision-making central unit is used to calculate the deviation of key parameters, including: From the standard injection curve and the associated reference process parameter package, predefined key parameters and their target values ​​are extracted. The key parameters include: the first-speed to second-speed switching position, the peak high-speed injection speed, the pressure build-up point pressure, and the peak pressure boosting pressure. Identify and locate the actual feature points corresponding to predefined key parameters from the preprocessed and calibrated real-time injection process curve data; Calculate the absolute deviation between the actual value and the target value of each of the key parameters; Based on the process window threshold corresponding to the key parameter, the normalized relative deviation is calculated. The normalized relative deviation is the percentage value obtained by dividing the absolute deviation by half of the process window threshold.

4. The manufacturing execution system based on closed-loop management according to claim 3, characterized in that, The real-time analysis and decision-making central unit is used to calculate a comprehensive evaluation index based on curve fit and key parameter deviation, as shown in the following formula: Z=α×S+Σ(βi×|ΔPi|×Ti); Where Z is the comprehensive evaluation index; S is the quantitative value of curve fit; α is the weighting coefficient of fit; ΔPi is the deviation between the measured value and the target value of the i-th key parameter; Ti is the process window threshold corresponding to the i-th key parameter; and βi is the deviation weighting coefficient of the i-th key parameter.

5. The manufacturing execution system based on closed-loop management according to claim 1, characterized in that, The real-time analysis and decision-making central unit is used to assess the thermal balance state of the mold, including: Based on the temperature data of each measuring point and its three-dimensional spatial coordinates on the mold, the mold is divided into several logical evaluation areas, and the average temperature, temperature range within the area, and temperature change rate of each logical evaluation area are calculated. Compare the average temperature of each logical evaluation region with the temperature of the corresponding sub-target range in the mold temperature control target range, and calculate the first type of deviation D1; The temperature range within each logical evaluation region is compared with the preset allowable temperature gradient threshold to calculate the second type of deviation D2. The regional temperature change rate of each region during the current production cycle is compared with the baseline change rate under steady-state production conditions to calculate the third type of deviation D3. The formula for the comprehensive heat balance index is: H=w1×f(D1)+w2×g(D2)+w3×h(D3); Where w1, w2, and w3 are the weight coefficients of each deviation component, and w1 + w2 + w3 = 1; f(D1), g(D2), and h(D3) are functions that normalize D1, D2, and D3, respectively. If the comprehensive thermal balance index of any logical evaluation area exceeds the preset thermal balance anomaly threshold, or if the comprehensive thermal balance index of more than a preset number of logical evaluation areas exceeds the warning threshold, then the mold thermal balance status evaluation is determined to be abnormal.

6. The manufacturing execution system based on closed-loop management according to claim 5, characterized in that, The real-time analysis and decision-making central unit is used to analyze whether to generate primary process alarms and adjustment instructions based on comprehensive evaluation indicators and mold thermal balance status, including: If the comprehensive evaluation index is greater than or equal to the first preset threshold, and the mold thermal balance state is assessed as abnormal, a primary process alarm and adjustment instruction will be generated. If the comprehensive evaluation index is greater than or equal to the first preset threshold, but the mold thermal balance status is assessed as normal, then process prompt information is generated only for the injection process.

7. The manufacturing execution system based on closed-loop management according to claim 6, characterized in that, The real-time analysis and decision-making central unit also includes: The process parameter-quality defect mapping library subunit is used to store and manage the mapping relationship model between process parameter fluctuations and quality defect types. The mapping relationship model is established and updated in the following ways: Continuously collect process operation data and corresponding die-casting quality inspection data from historical production cycles; A process fluctuation feature vector is constructed for each injection process, and a defect characterization vector is constructed for the corresponding die casting. Using the process fluctuation feature vector as input and the defect characterization vector as output target, a multi-output regression model is trained or updated periodically or after accumulating a certain amount of data. The weight matrix of this model is the mapping relationship.

8. The manufacturing execution system based on closed-loop management according to claim 7, characterized in that, The real-time analysis and decision-making central unit is used to call the current mapping relationship model in the process parameter-quality defect mapping library subunit after generating primary process alarms and adjustment instructions, and to perform confidence verification and parameter correction on the instructions.

9. The manufacturing execution system based on closed-loop management according to claim 8, characterized in that, The real-time analysis and decision-making central unit is also used to trigger parameter correction processes, including: Based on the overall confidence assessment, different correction strategies are dynamically selected by comparing it with a preset set of confidence interval thresholds. When the overall confidence level assessment falls within the middle confidence range, a local iterative correction strategy is selected; When the overall confidence assessment falls into the low confidence range, a global optimization correction strategy is selected.

10. The manufacturing execution system based on closed-loop management according to claim 1, characterized in that, The primary process alarm and adjustment instructions include the anomaly type, anomaly severity, and recommended adjustment parameters.

Images