Aluminum profile extrusion die liquid nitrogen cooling control method and system

Abstract

This application relates to the field of aluminum profile extrusion technology, specifically disclosing a method and system for controlling liquid nitrogen cooling of aluminum profile extrusion dies. The method includes the following steps: continuously collecting multi-source temperature data and operating status data of the extrusion production line within a preset historical period; establishing a mapping relationship between operating status and temperature response, and predicting the operating status change sequence; generating die temperature prediction curves and profile temperature prediction curves based on the operating status change sequence and the mapping relationship; calculating the deviations from the corresponding preset temperature control targets, generating die cooling demand curves and profile cooling demand curves, and adjusting the injection strategy of the liquid nitrogen injection device to perform cooling intervention in advance. The technical solution of this application can calculate the cooling demand curve matching future heat load changes in advance and drive the liquid nitrogen injection device to perform forward-looking flow and mode adjustments, overcoming the response lag problem caused by thermal inertia in traditional control.

Description

Technical Field

[0001] This application relates to the field of aluminum profile extrusion technology, and in particular to a liquid nitrogen cooling control method and system for aluminum profile extrusion dies. Background Technology

[0002] In aluminum extrusion production, precise control of the die operating temperature and the profile exit temperature is crucial for ensuring product quality, extending die life, and improving production efficiency. Liquid nitrogen injection technology, with its strong cooling capacity and inert protection (anti-oxidation), has become the mainstream temperature control method in the industry. Currently, this technology generally adopts a closed-loop control strategy based on single-point temperature feedback. This involves installing an infrared thermometer at the extruder exit to monitor the profile temperature, and a PID controller adjusting the liquid nitrogen valve opening in real time based on the deviation between the measured value and the target value to change the flow rate in an attempt to maintain temperature stability.

[0003] However, the actual extrusion process is full of periodic operating disturbances, which seriously challenge the traditional control mode mentioned above. The main disturbances include: 1) Periodic aluminum rod replacement: During the rod replacement interval, the die temperature drops sharply due to the temporary loss of heat input from the aluminum rod, and then rises rapidly after the new rod enters, forming violent fluctuations; 2) Online fixed-length cutting: The cutting action causes instantaneous changes in the extrusion speed, which changes the frictional heat generation and disturbs the temperature; 3) Planned speed adjustment: Different profile specifications require switching extrusion speeds, which directly changes the system's thermal load.

[0004] Faced with these inherent disturbances, traditional single-point temperature PID feedback control exposes a series of fundamental defects. First, the system response is severely lagging. From regulating the valve to liquid nitrogen vaporization, mold heat transfer, and ultimately affecting the temperature measurement point, thermal inertia is extremely high. By the time the controller detects a temperature deviation and begins adjustment, the actual temperature has already deviated further due to changes in operating conditions, causing the control action to always "catch up" with the error, easily leading to continuous oscillations at switching points such as bar changing. Second, information perception is incomplete. Monitoring only the profile outlet temperature cannot detect changes in the aluminum bar feed temperature in advance, nor can it reflect the true temperature distribution inside the mold, lacking foresight regarding thermal disturbances originating from these links. Most importantly, its control logic is purely passive feedback. The controller only responds after a temperature change occurs, completely unable to predict future events such as "about to change bars" or "about to cut," and therefore has no possibility of adjusting the cooling strategy in advance before the switching of operating conditions.

[0005] The aforementioned lag and passivity in control directly leads to the difficulty in suppressing temperature fluctuations during the production process. This not only affects the stability of profile dimensions, surface quality (such as oxidation and scratches), and uniformity of mechanical properties, but may also shorten the service life of the mold due to cyclic thermal stress, while also causing waste of liquid nitrogen. Summary of the Invention

[0006] This application proposes a liquid nitrogen cooling control method and system for aluminum profile extrusion dies, aiming to solve the technical problem that existing aluminum profile module temperature control schemes are difficult to suppress temperature fluctuations during the production process.

[0007] In a first aspect, this application provides a method for controlling liquid nitrogen cooling of aluminum profile extrusion dies, which uses a liquid nitrogen injection device to coordinately control the temperature of the extrusion die and the profile temperature. The method includes the following steps: S1, continuously collect multi-source temperature data and working status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum bar feeding temperature; S2, Based on the multi-source temperature data and working status data within the preset historical period, establish a mapping relationship between working status and temperature response, and predict the sequence of working status changes that the extrusion production line will undergo in the future preset period. S3, Based on the working state change sequence and the mapping relationship between the working state and temperature response, generate the mold temperature prediction curve and the profile temperature prediction curve for the future preset time period. S4. Based on the mold temperature prediction curve and the profile temperature prediction curve, calculate the deviation from the corresponding preset temperature control target, and generate the mold cooling demand curve and the profile cooling demand curve within the preset future time period. S5. Based on the mold cooling demand curve and the profile cooling demand curve, adjust the injection strategy of the liquid nitrogen injection device to perform cooling intervention in advance.

[0008] As an improvement to the above solution, the working status data includes: the extrusion stage based on the control system signal recognition of the extrusion production line; the preset historical cycle covers multiple complete aluminum rod replacement operation cycles and online cutting operation cycles.

[0009] As an improvement to the above solution, S2 includes: S21. Based on the multi-source temperature data and working status data within the preset historical period, the data is classified according to the identified extrusion stages, and the corresponding multi-source temperature data for each extrusion stage is statistically analyzed to generate typical temperature responses of the mold and profile under various extrusion stages, so as to establish a mapping relationship between working status and temperature response. S22, Analyze the working status data within the preset historical period, identify the periodic patterns of operations such as replacing aluminum rods and online cutting, and generate a sequence of working status changes within the preset future time period based on the periodic patterns.

[0010] As an improvement to the above solution, S4 includes: S41, the future preset time period is divided into multiple consecutive time intervals of equal length; S42, For each time interval, calculate the average temperature or integral temperature of the mold temperature prediction curve and the profile temperature prediction curve within the interval, and compare it with the corresponding mold target temperature and profile target temperature to obtain the average temperature deviation of the mold and the average temperature deviation of the profile for the corresponding time interval. S43, based on the average temperature deviation of the mold and the average temperature deviation of the profile, obtain the initial cooling demand of the mold and the initial cooling demand of the profile for the corresponding time interval from the preset temperature deviation-cooling demand mapping relationship; S44, Based on the working status data of the time interval, determine the mold cooling coefficient and the profile cooling coefficient; S45, Multiply the initial cooling demand of the mold and the initial cooling demand of the profile by the corresponding cooling coefficients of the mold and the profile respectively to obtain the final cooling demand of the mold and the final cooling demand of the profile for the corresponding time interval. S46, connect the final cooling demand of the mold and the final cooling demand of the profile calculated in each time interval along the time axis to generate the mold cooling demand curve and the profile cooling demand curve respectively.

[0011] As an improvement to the above scheme, before S4, the following is also included: The temperature of the extrusion die and the temperature of the profile are acquired in real time. During the predetermined future time period, the temperature of the extrusion die and the temperature of the profile are compared with the predicted temperature curves of the die and the profile to obtain a comprehensive temperature deviation. When the overall temperature deviation continues to exceed the preset deviation threshold, the mold cooling coefficient and profile cooling coefficient in the subsequent production stages are corrected based on the overall temperature deviation.

[0012] As an improvement to the above solution, the step of comparing the temperature of the extrusion die and the temperature of the profile with the predicted temperature curves of the die and the profile respectively during the predetermined future time period to obtain the comprehensive temperature deviation includes: During the predetermined future time period, the average temperature deviation between the extrusion die temperature and the die temperature prediction curve, and the average temperature deviation between the profile temperature and the profile temperature prediction curve are calculated to obtain the die temperature deviation and the profile temperature deviation. Based on the temperature deviation of the mold and the temperature deviation of the profile, the comprehensive temperature deviation is calculated using the weighted root mean square error method.

[0013] As an improvement to the above solution, the mold cooling coefficient is corrected using a first formula, and the profile cooling coefficient is corrected using a second formula. The first formula is: C1' = C1 × (1 + a × T1); The second formula is: C2' = C2 × (1 + b × T2); Wherein, C1 and C2 are the mold cooling coefficient and profile cooling coefficient before correction, respectively; C1' and C2' are the mold cooling coefficient and profile cooling coefficient after correction, respectively; T1 and T2 are the mold temperature deviation and profile temperature deviation, respectively; and a and b are the preset mold cooling coefficient correction sensitivity and profile cooling coefficient correction sensitivity, respectively.

[0014] As an improvement to the above solution, the liquid nitrogen injection device includes a first mechanism for injecting nitrogen gas into the extrusion die, and a second mechanism for injecting nitrogen gas into the aluminum profile leaving the extrusion die. S5 includes: Based on the mold cooling demand curve, adjust the injection flow rate and injection method of the first mechanism in each time interval; Based on the profile cooling demand curve, adjust the injection flow rate and injection method of the second mechanism in each time interval; The nitrogen injection methods include: average injection mode, pulse injection mode, and gradient injection mode.

[0015] As an improvement to the above solution, the liquid nitrogen cooling control method for aluminum profile extrusion dies in this application further includes the following steps: Obtain the surface quality test results of the extruded aluminum profile, including the oxidation index and the tensile index; If the oxidation index continues to exceed a preset first preset threshold for a first preset number of consecutive production cycles, a downward adjustment to the cooling coefficient of the profile will be triggered. If the tear index continues to exceed the preset second preset threshold for a second preset number of consecutive production cycles, an upward adjustment of the mold cooling coefficient will be triggered.

[0016] Secondly, this application provides a liquid nitrogen cooling control system for aluminum profile extrusion dies, which coordinates the temperature of the extrusion die and the profile temperature through a liquid nitrogen injection device. The system includes: The data acquisition module is used to continuously collect multi-source temperature data and operating status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum rod feeding temperature; The status prediction module is used to establish a mapping relationship between working status and temperature response based on multi-source temperature data and working status data within the preset historical period, and to predict the sequence of working status changes that the extrusion production line will undergo in the future preset period. The temperature prediction module is used to generate mold temperature prediction curves and profile temperature prediction curves within the preset future time period based on the working state change sequence and the mapping relationship between the working state and temperature response. The demand generation module is used to calculate the deviation from the corresponding preset temperature control target based on the mold temperature prediction curve and the profile temperature prediction curve, and generate the mold cooling demand curve and the profile cooling demand curve in the preset future time period. The strategy adjustment module is used to adjust the injection strategy of the liquid nitrogen injection device according to the mold cooling demand curve and the profile cooling demand curve, so as to perform cooling intervention in advance.

[0017] The technical solution according to the embodiments of this application has at least the following beneficial effects: The technical solution of this application continuously collects multi-source temperature and operating condition historical data, establishes and uses the operating condition-temperature mapping relationship to predict future production event sequences, and then generates temperature prediction curves for molds and profiles. Based on this prediction, the cooling demand curve matching future heat load changes is calculated in advance, and the liquid nitrogen injection device is driven to perform forward-looking flow and mode adjustments, thereby transforming the control logic from passive temperature deviation feedback to active operating condition prediction feedforward, overcoming the response lag problem caused by thermal inertia in traditional control.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0020] Figure 1 This is a schematic flowchart illustrating a liquid nitrogen cooling control method for aluminum profile extrusion dies provided in an embodiment of this application.

[0021] Figure 2 This is a flowchart illustrating S2 in an embodiment of this application.

[0022] Figure 3 This is a flowchart illustrating step S4 in an embodiment of this application.

[0023] Figure 4This is a schematic diagram of the architecture of a liquid nitrogen cooling control system for an aluminum profile extrusion die, provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical methods, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] It should be noted that the meaning of "multiple" (or "more than") in the description of the embodiments of this application refers to two or more, and "greater than," "less than," "exceeding," etc. are understood to exclude the number itself, while "above," "below," "within," etc. are understood to include the number itself. If "first," "second," etc. are used in the description, they are only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0026] In aluminum profile extrusion production, although liquid nitrogen cooling is the mainstream temperature control method, the commonly used single-point temperature PID feedback control mode cannot predict and respond to disturbances in advance when faced with inherent periodic operating conditions such as aluminum rod replacement, online cutting, and speed adjustment. This is due to the severe system thermal inertia lag, limited single-point information perception, and completely passive feedback logic. As a result, the control action always lags behind the temperature change, causing the mold and profile temperature to oscillate continuously and uncontrollably during the production process, which in turn damages product quality, mold life, and leads to resource waste.

[0027] In this regard, such as Figure 1 As shown, this application provides a method for controlling liquid nitrogen cooling of aluminum profile extrusion dies. The method uses a liquid nitrogen injection device to coordinate the control of the extrusion die temperature and the profile temperature. The method includes the following steps: S1, continuously collect multi-source temperature data and working status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum bar feeding temperature; S2, Based on the multi-source temperature data and working status data within the preset historical period, establish a mapping relationship between working status and temperature response, and predict the sequence of working status changes that the extrusion production line will undergo in the future preset period. S3, Based on the working state change sequence and the mapping relationship between the working state and temperature response, generate the mold temperature prediction curve and the profile temperature prediction curve for the future preset time period. S4. Based on the mold temperature prediction curve and the profile temperature prediction curve, calculate the deviation from the corresponding preset temperature control target, and generate the mold cooling demand curve and the profile cooling demand curve within the preset future time period. S5. Based on the mold cooling demand curve and the profile cooling demand curve, adjust the injection strategy of the liquid nitrogen injection device to perform cooling intervention in advance.

[0028] This method is primarily applied to aluminum profile hot extrusion production lines equipped with liquid nitrogen jet cooling systems. The "extrusion production line" includes an aluminum rod heating furnace, an extrusion press (including an extrusion cylinder and die), a discharge conveyor, and an online cutting device. The "multi-source temperature data" refers to temperature information from different physical locations on the production line, reflecting different thermal states: the extrusion die temperature is typically obtained by embedding high-temperature thermocouples near the die pad or working belt, directly reflecting the thermal load of the die body; the profile temperature is obtained by installing a non-contact infrared thermometer at the extrusion press outlet, reflecting the final product's exit thermal state; the aluminum rod feed temperature is obtained by installing thermocouples near the extrusion cylinder feed inlet, and is a key initial condition affecting heat input. The "operating status data" refers to signals describing the extrusion process and operational events, such as digital signals read from the extrusion press's programmable logic controller (PLC) used to identify different stages such as "normal extrusion," "aluminum rod replacement," and "online cutting." The “preset historical period” is not a fixed time length, but a data collection period that must cover multiple complete production cycles containing the aforementioned periodic disturbances. For example, it must include at least dozens of complete “aluminum rod replacement-stable extrusion-cutting” cycles to ensure that the extracted patterns are statistically significant.

[0029] The acquisition of multi-source temperature data and operational status data provides comprehensive, synchronous, and reliable input information for the subsequent modeling, prediction, and decision-making processes. Multi-source temperature data spatially covers the entire process chain from heat input and heat exchange to heat output, enabling the perception of the global thermal state; multi-source operational status data temporally identifies production processes and events. Continuously collecting and storing these two types of historically aligned data, especially covering multiple complete production cycles, is crucial for extracting statistically significant operational condition-temperature response patterns from history, providing a robust experience base for intelligent prediction. For example, standard industrial temperature sensors (thermocouples, infrared thermometers) can be deployed at relevant workstations, and analog sensor signals can be converted into digital quantities at a fixed sampling period (e.g., 100 milliseconds to 1 second) using data acquisition cards or distributed I / O modules. These digital signals, along with operational status signals read from PLC communication interfaces (e.g., Ethernet), are then stored in an industrial database or time-series database.

[0030] Based on the historical data accumulated in the previous step, a quantitative or qualitative mapping relationship between "working state" and "temperature response" is established through statistical analysis to understand the temperature change patterns triggered by different working conditions (such as bar changing and cutting). Based on the analysis of historical working state sequences, the inherent periodic patterns of production operations (such as average bar changing interval and cutting rhythm) are identified, and based on this, the types of working condition events that will occur in the future and their time series are predicted. This step extends the knowledge of control from "current state" to "future expectations" and is a prerequisite for feedforward. Establishing the mapping relationship can be achieved using classification and regression methods from data mining. For example, the historical dataset can be segmented according to working state labels (such as "steady-state extrusion section_A speed", "bar changing transition section", and "cutting disturbance section"), and the statistical characteristics (such as mean, rise / fall slope, and peak value) of the temperature sequence in each data segment can be calculated to form a "typical temperature response template" for that type of working condition, which is then stored in the database. Working condition prediction can be achieved using a simple prediction model based on state machines and time series analysis. For example, a counter can be maintained to record the number of extruded aluminum bars. When the number of bars approaches the historical average per batch, it can be predicted that the bar replacement stage is about to begin. Alternatively, the approximate time of the next cut can be predicted by analyzing the intervals of the cutting signals.

[0031] Next, using the sequence of changes in operating states, combined with the established mapping relationship between operating states and temperature response, continuous curves are inferred and synthesized to show how the mold and profile temperatures will change over time in the future. These two predicted curves represent key forward-looking indicators of mold heat load and profile exit quality, providing clear quantitative targets for formulating cooling strategies. For example, a curve generation method based on template splicing can be used. The control system, based on the future operating condition sequence predicted by S2, sequentially retrieves the corresponding "typical temperature response templates" from the mapping relationship library. Then, these template curves are aligned and spliced ​​on the time axis according to the predicted start and end times. This generates two predicted temperature curves, representing the future temperature trends of the mold and profile, respectively.

[0032] S4 is for formulating control commands. It compares the "temperature prediction curve" generated by S3 with the preset "temperature control target," calculates the cooling amount required to eliminate the prediction deviation at each future moment, and outputs it as two flow demand curves that change over time, thus quantifying the temperature control problem into a flow control problem. For example, calculations can be based on a simplified heat balance model. The deviation between the predicted temperature and the target temperature (ΔT) is multiplied by an empirical coefficient (K) that integrates the system's heat capacity and liquid nitrogen heat exchange efficiency to obtain the instantaneous cooling demand (Q = K × ΔT). This operation is performed for each calculation point to form the cooling demand curve. The empirical coefficient K can be obtained through process calibration. Specifically, when the extrusion production line is in a stable operating state, the average liquid nitrogen injection flow rate Q_s required to maintain temperature stability and the steady-state deviation ΔT_s between the measured temperature of the die or profile and the preset target temperature are recorded simultaneously. Then, the simplified heat balance model is used to back-calculate K = Q_s / ΔT_s. To ensure the representativeness and practicality of this coefficient, this calibration process should be repeated under various typical production conditions.

[0033] Based on the forward-looking "cooling demand curve" generated by S4, S5 directly and proactively controls the liquid nitrogen injection device (such as regulating valves), ensuring that the cooling action is performed before the predicted temperature fluctuations occur. This achieves feedforward control and overcomes the lag caused by system thermal inertia. For example, a real-time controller (such as a PLC) uses the cooling demand curve as a setpoint and outputs an analog signal (such as 4-20mA) to the electro-hydraulic proportional valve on the liquid nitrogen pipeline, precisely controlling its opening so that the actual injection flow rate tracks the demand curve. Alternatively, in situations requiring rapid switching, pulse width modulation (PWM) can be used to control high-speed switching valves for early intervention.

[0034] The core principle of this method is prediction and feedforward control based on multi-source historical data. By continuously collecting multi-source temperature and operating status data from the extrusion production line, it provides an information foundation covering the entire process chain for subsequent steps. Step S2 performs two key functions based on this data: first, establishing a mapping relationship between "operating status" and "temperature response" to understand specific temperature change patterns caused by different production events; second, analyzing the periodic patterns of historical operating status to predict the sequence of operating status changes that will occur within a preset time period. Step S3 uses the predicted sequence of operating status changes, invokes the established mapping relationship, and infers and synthesizes the corresponding mold temperature prediction curve and profile temperature prediction curve, thereby achieving a quantitative expectation of future temperature changes. Step S4 compares the two generated temperature prediction curves with the preset temperature control target, calculates the cooling amount required to eliminate prediction deviations at various future times, and generates cooling demand curves for the mold and profile, thus transforming the control target into a clear execution instruction. Finally, step S5 directly adjusts the injection strategy of the liquid nitrogen injection device based on the generated forward-looking cooling demand curve, performing cooling intervention before the predicted temperature fluctuations actually occur.

[0035] In summary, the technical solution of this application continuously collects multi-source temperature and operating condition historical data, establishes and utilizes the operating condition-temperature mapping relationship to predict future production event sequences, and then generates temperature prediction curves for molds and profiles. Based on this prediction, the cooling demand curve matching future heat load changes is calculated in advance, and the liquid nitrogen injection device is driven to perform forward-looking flow and mode adjustments. This transforms the control logic from passive temperature deviation feedback to active operating condition prediction feedforward, overcoming the response lag problem caused by thermal inertia in traditional control, and achieving high-precision and high-stability collaborative control of mold and profile temperatures under periodic disturbances.

[0036] In some embodiments of this application, the working status data preferably includes: the extrusion stage based on the control system signal recognition of the extrusion production line; the preset historical cycle covers multiple complete aluminum rod replacement operation cycles and online cutting operation cycles.

[0037] The extrusion stages, such as "normal extrusion," "changing aluminum rods," "online cutting," or "shutdown," describe the discrete states of the production process on the timeline. By stipulating that identification must be based on "control system signals of the extrusion production line" (typically referring to the output of a programmable logic controller (PLC) or a distributed control system (DCS), the accuracy, real-time nature, and unambiguity of stage determination are ensured. This avoids errors and delays that might arise from relying on manual operator recording or inferences based on indirect parameters (such as zero speed). For example, when writing the production line control program, explicit internal status flags (such as M-addresses or DB block variables) are set for different process stages. The data acquisition unit establishes a connection with the aforementioned control system through standard industrial communication protocols (such as OPC UA, PROFINET, EtherNet / IP), periodically reading the values ​​of these status flags to directly and synchronously obtain the current precise extrusion stage information.

[0038] Meanwhile, clear requirements are set regarding the time span and completeness of historical data required for constructing the predictive model. "Replacing aluminum rods" and "online cutting" are two of the most typical periodic events in aluminum extrusion production, significantly impacting temperature fluctuations. The requirement that historical data cover "multiple complete" such cycles means that a sufficient number of data samples are needed to ensure that step S2 can extract statistically significant patterns, rather than random fluctuations. For example, data containing only one rod replacement cannot distinguish between general patterns and specific random occurrences; data covering dozens of rod replacements and hundreds of cuts can reliably calculate the average rod replacement interval, the typical temperature drop and recovery time caused by rod replacement, and the instantaneous temperature rise characteristics caused by cutting. For instance, during deployment, a data acquisition start time is set and continuously run until historical data records in the local or cloud database meet the conditions of "covering at least N (e.g., N>10) rod replacement cycles" and "covering at least M (e.g., M>50) cutting cycles," only then is it considered to have reached a valid "preset historical cycle," and only then can the model training and prediction functions based on this historical data be activated.

[0039] The technical solution of this application, by binding to authoritative control system signals, ensures the accuracy and timeliness of operating condition identification, providing a reliable label for subsequent accurate operating condition classification and prediction. The explicit requirement for historical data periods ensures that the data samples used to establish mapping relationships and identify patterns have sufficient representativeness and statistical significance, thereby making the trained prediction model more robust and the generated prediction curves more reliable. These two points together enhance the feasibility and robustness of the entire control scheme in actual industrial environments.

[0040] In some embodiments of this application, such as Figure 2 As shown, S2 preferably includes: S21. Based on the multi-source temperature data and working status data within the preset historical period, the data is classified according to the identified extrusion stages, and the corresponding multi-source temperature data for each extrusion stage is statistically analyzed to generate typical temperature responses of the mold and profile under various extrusion stages, so as to establish a mapping relationship between working status and temperature response. S22, Analyze the working status data within the preset historical period, identify the periodic patterns of operations such as replacing aluminum rods and online cutting, and generate a sequence of working status changes within the preset future time period based on the periodic patterns.

[0041] S21 specifically involves: Based on multi-source data collected within a preset historical period, the historical data is first categorized according to the identified extrusion stages (such as normal extrusion, aluminum rod replacement, online cutting, etc.). Next, for each category of data within various extrusion stages, the corresponding multi-source temperature data is statistically analyzed to extract the typical temperature response patterns of the die and profile under each working condition. For example, the average value and fluctuation range of the die temperature decrease rate during the "rod replacement stage" can be calculated, or the stable value of the profile exit temperature during the "steady-state extrusion stage." Finally, all these classification and statistical results are correlated and stored, thus establishing the "mapping relationship between working state and temperature response." For example, clustering and regression analysis algorithms from data mining can be used to extract features from the categorized historical temperature time series, and the extracted features (such as mean curves, response amplitudes, and time constants) and the labels for that type of working condition are stored in a database to form a queryable mapping table.

[0042] The function of S22 is to achieve forward-looking prediction of production rhythm. Specifically, it involves analyzing the working status data within the preset historical period to identify periodic patterns in key operations such as aluminum rod replacement and online cutting. For example, by statistically analyzing historical data, it calculates the average number of aluminum rods extruded or the interval between rod replacement operations, as well as the average interval of online cutting actions. Subsequently, based on these identified periodic patterns and combined with the current production progress (such as the current number of aluminum rods extruded and the time since the last cut), a sequence of working status changes within a preset future time period can be derived and generated.

[0043] The above technical solutions ensure that the mapping relationship is established based on scientific statistical analysis, rather than fuzzy experience; the generation of the working state change sequence originates from quantifiable periodic patterns, rather than arbitrary estimations. This enhances the reliability, consistency, and repeatability of the entire predictive feedforward control logic, laying a solid model foundation for ultimately achieving high-precision temperature control.

[0044] The following is a specific example to illustrate this.

[0045] Taking an extrusion production line producing 6063 aluminum alloy profiles as an example, the preset historical period is the production data of the past 24 hours, covering more than 30 complete aluminum bar replacement operations and more than 500 online cuts. In S21, the method divides the data into categories such as "steady-state extrusion (speed 8mm / s)," "bar change transition period," and "online cutting disturbance period." Statistical analysis revealed that the typical temperature response during the "bar change transition period" is as follows: the die temperature drops by an average of about 15°C within 20 seconds after the bar change begins, and then rises by about 20°C within 30 seconds after the new aluminum bar begins extrusion; during the "online cutting disturbance period," the profile exit temperature typically experiences a momentary rise of about 5°C lasting about 3 seconds. These patterns are stored in a mapping library. In S22, the method analysis found that the production line must change bars after extruding every 6 aluminum bars (each about 6 meters), and an online cut is performed every 1.2 meters of extruded profile. Based on this, when the fifth aluminum bar is extruded to 5.8 meters, the method predicts the following sequence of working state changes: "After approximately 20 seconds, the bar changing stage will begin, lasting approximately 50 seconds; after the bar changing stage, extrusion will resume, and an online cut is expected to occur at the 1.2-meter mark of the subsequent extrusion (approximately 150 seconds later)." This predicted sequence provides accurate input for the generation of the precise temperature profile for subsequent S3.

[0046] In the implementation scheme of this application, such as Figure 3 As shown, S4 preferably includes: S41, the future preset time period is divided into multiple consecutive time intervals of equal length; S42, For each time interval, calculate the average temperature or integral temperature of the mold temperature prediction curve and the profile temperature prediction curve within the interval, and compare it with the corresponding mold target temperature and profile target temperature to obtain the average temperature deviation of the mold and the average temperature deviation of the profile for the corresponding time interval. S43, based on the average temperature deviation of the mold and the average temperature deviation of the profile, obtain the initial cooling demand of the mold and the initial cooling demand of the profile for the corresponding time interval from the preset temperature deviation-cooling demand mapping relationship; S44, Based on the working status data of the time interval, determine the mold cooling coefficient and the profile cooling coefficient; S45, Multiply the initial cooling demand of the mold and the initial cooling demand of the profile by the corresponding cooling coefficients of the mold and the profile respectively to obtain the final cooling demand of the mold and the final cooling demand of the profile for the corresponding time interval. S46, connect the final cooling demand of the mold and the final cooling demand of the profile calculated in each time interval along the time axis to generate the mold cooling demand curve and the profile cooling demand curve respectively.

[0047] The time intervals can be divided based on the controller's operating cycle and the process response speed. A fixed duration (e.g., 5 or 10 seconds) can be preset, and the future time period (e.g., the next 3 minutes) can be evenly divided into 36 or 18 consecutive time intervals. This provides a unified time framework for subsequent batch calculations based on intervals.

[0048] Interval temperature deviation calculation involves integrating and evaluating the predicted temperatures within each time interval. For example, for each time interval, the arithmetic mean (or integration of the curves divided by the interval length) of all predicted points on the mold temperature prediction curve and profile temperature prediction curve generated by S3 can be calculated to obtain the predicted average temperature for that time interval. This average temperature is then directly subtracted from the preset, constant, or variable target mold temperature and target profile temperature to obtain the average temperature deviation of the mold and the average temperature deviation of the profile for the corresponding time interval.

[0049] The initial demand is the first step in converting the temperature deviation into a cooling physical quantity. For example, this can be achieved using a pre-calibrated "temperature deviation - cooling demand" lookup table (i.e., mapping relationship). This mapping relationship defines the theoretical liquid nitrogen flow rate required to compensate for a preset temperature deviation. For each time interval, the calculated average temperature deviation of the mold and profile is used as input, and by consulting this mapping table, the initial cooling demand for the mold and profile for the corresponding time interval can be directly obtained.

[0050] The determination of the cooling coefficient involves introducing a dynamic correction factor based on real-time operating conditions. For example, based on the operating state data predicted by S2 that will dominate within the corresponding time interval (e.g., whether it is "steady-state extrusion" or "bar changing transition period"), the corresponding mold cooling coefficient and profile cooling coefficient are determined from a preset operating condition-coefficient lookup table. Different operating conditions correspond to different coefficients, such as during the "bar changing transition period" when the heat load changes drastically. The initial values ​​of the mold cooling coefficient and profile cooling coefficient can be obtained through statistical analysis based on historical production data or through active process experiments. Specifically, the coefficient value characterizing the cooling efficiency under various operating conditions (such as steady-state extrusion and bar changing) in historical cycles can be calculated by analyzing the average liquid nitrogen usage and corresponding temperature control effects; or the coefficient can be directly calibrated by actively adjusting the liquid nitrogen flow rate and observing the temperature response under preset operating conditions through specialized experiments.

[0051] The final demand calculation involves personalized adjustments to the cooling demand. The conventional method for this is simple scalar multiplication. The initial mold cooling demand obtained in S43 is multiplied by the mold cooling coefficient determined in S44 to obtain the final mold cooling demand for that range. The same calculation is performed on the profile side. By multiplying the coefficient by the initial demand, dynamic scaling of the base cooling capacity based on future operating conditions is achieved.

[0052] Demand curve generation integrates the calculation results of all discrete intervals into continuous, forward-looking control commands. For example, the final demand for mold cooling and the final demand for profile cooling calculated for all time intervals are arranged in their respective time sequences. In computer memory or the control system, these time-ordered demand value points are connected sequentially to form two complete time-varying mold cooling demand curves and profile cooling demand curves, which can be directly tracked by the S5 actuator.

[0053] The technical solution of this application transforms the vague "demand generation based on prediction" into clear and definite mathematical operations through a structured calculation process. Dividing the time interval allows for step-by-step execution of future planning, making it more flexible; introducing a cooling coefficient enables dynamic adjustment of demand based on the predicted operating conditions, enhancing the adaptability and precision of control; the entire process is standardized, calculable, and easy to program in industrial controllers, ensuring the consistency and reliability of control logic.

[0054] The following is a specific example to illustrate this.

[0055] Assuming a preset time period of 180 seconds and an equal time interval of 10 seconds, the time is divided into 18 intervals. Taking the 5th interval (40-50 seconds) as an example, S2 predicts that this interval is in the later stage of "aluminum rod replacement". S42 calculates that the predicted average temperature of the mold in this interval is 455°C, and the predicted average temperature of the profile is 505°C; the target temperature of the mold is 460°C, and the target temperature of the profile is 500°C. Therefore, the average temperature deviation of the mold is -5°C, and the average temperature deviation of the profile is +5°C. S43 consults the mapping table: the -5°C deviation corresponds to an initial cooling demand of 8 L / min for the mold, and the +5°C deviation corresponds to an initial cooling demand of 10 L / min for the profile. S44, based on the condition of "late stage of aluminum rod replacement", consults the table to determine that the mold cooling coefficient is 1.2 (cooling needs to be strengthened to prevent excessive heating), and the profile cooling coefficient is 0.9 (at this time, there is no profile production, and the cooling demand is reduced). S45 calculates: Final cooling demand for the mold = 8 L / min × 1.2 = 9.6 L / min; Final cooling demand for the profile = 10 L / min × 0.9 = 9 L / min. After repeating the above calculation for all 18 intervals, S46 connects the starting time of each interval with the corresponding final demand, thus generating two cooling demand command curves that start from the current time and last for 180 seconds.

[0056] In a specific embodiment of this application, prior to step S4, the liquid nitrogen cooling control method for aluminum profile extrusion dies of this application preferably further includes: The temperature of the extrusion die and the temperature of the profile are acquired in real time. During the predetermined future time period, the temperature of the extrusion die and the temperature of the profile are compared with the predicted temperature curves of the die and the profile to obtain a comprehensive temperature deviation. When the overall temperature deviation continues to exceed the preset deviation threshold, the mold cooling coefficient and profile cooling coefficient in the subsequent production stages are corrected based on the overall temperature deviation.

[0057] The system can read the measured values ​​from the thermocouple at the die working zone and the infrared thermometer at the profile exit in real time, obtaining the measured values ​​of the extrusion die temperature and the profile temperature. Simultaneously, it extracts the predicted temperature value that perfectly corresponds to the current measured moment from the generated die temperature prediction curve and profile temperature prediction curve. Subtracting the corresponding predicted value from the measured value yields the die temperature deviation and profile temperature deviation at the corresponding moment. The comprehensive temperature deviation can be obtained using weighted summation, data table lookup, and a pre-trained model. The data table can pre-store mapping relationships based on historical experience or simulation data, such as recommended comprehensive temperature deviation values ​​corresponding to different combinations of die temperature deviation and profile temperature deviation, along with the confidence weight of these recommended values, or additional correction coefficients for preset operating conditions (such as "bar changing period"). The pre-trained model internally stores parameters trained using a large amount of historical production data (including continuous deviation sequences, operating condition labels, and final quality or control effect labels). This model can learn the complex nonlinear relationship between the dynamic patterns of deviations and overall control requirements, and directly output a comprehensive evaluation value. When the overall temperature deviation continues to exceed the preset deviation threshold, it triggers the correction of the mold cooling coefficient and profile cooling coefficient in the subsequent production stages to obtain a more accurate cold zone requirement.

[0058] The above solution achieves online self-learning and long-term stability maintenance of the control system. Traditional feedforward control, once the model parameters are set, experiences a gradual decrease in accuracy due to changes in equipment status and environment. This solution, through real-time comparison of prediction and actual measurement, can automatically detect this performance degradation and proactively and quantitatively correct the control parameters (cooling coefficient) before it affects product quality. This significantly reduces reliance on manual recalibration, ensuring the continuous effectiveness, adaptability, and robustness of the control strategy throughout the entire production cycle.

[0059] In a specific embodiment of this application, the step of comparing the temperature of the extrusion die and the temperature of the profile with the predicted temperature curves of the die and the profile respectively during the predetermined future time period to obtain the comprehensive temperature deviation preferably includes: During the predetermined future time period, the average temperature deviation between the extrusion die temperature and the die temperature prediction curve, and the average temperature deviation between the profile temperature and the profile temperature prediction curve are calculated to obtain the die temperature deviation and the profile temperature deviation. Based on the temperature deviation of the mold and the temperature deviation of the profile, the comprehensive temperature deviation is calculated using the weighted root mean square error method.

[0060] By calculating the corresponding average temperature, a series of instantaneous temperature deviation data, which may contain random noise, are transformed into statistically significant and stable evaluation values. For example, a fixed evaluation time window can be set (e.g., the past 30 seconds or 60 seconds). Within this window, the system continuously aligns and subtracts the measured extrusion die temperature values ​​acquired at each sampling moment (e.g., per second) from the predicted values ​​at the corresponding moments on the die temperature prediction curve, obtaining a sequence of instantaneous die temperature deviations. The same process is applied to the profile temperature to obtain a sequence of instantaneous profile temperature deviations. Subsequently, the arithmetic mean of these two sequences is calculated. The results represent the die temperature deviation and profile temperature deviation within the evaluation window. Calculating the average effectively smooths out random fluctuations and reflects the systematic and average deviation level between the predicted and measured values ​​within this time period, providing reliable input for subsequent comprehensive evaluation.

[0061] By using the weighted root mean square error method, the temperature deviation of the mold and the temperature deviation of the profile are integrated into a single comprehensive temperature deviation, and a comprehensive evaluation index is constructed that can simultaneously consider the control accuracy of both the mold and the profile, and can reflect the focus of the process.

[0062] The above-mentioned techniques enhance the scientific rigor and robustness of model accuracy assessment and decision triggering. Using the average deviation within a time window as the calculation basis avoids drastic fluctuations in assessment results caused by single-point sensor noise, instantaneous interference, or outliers. This ensures the stability and reliability of the "comprehensive temperature deviation" value used to determine whether correction is necessary, making the decision-making process less susceptible to false triggering. The introduction of the weighted root mean square error method as a fusion algorithm not only simultaneously considers the accuracy of the two key control dimensions of mold and profile but also strengthens the response to significant deviations through mathematical squaring, ensuring the control system can promptly and accurately detect systematic deviations in the model.

[0063] The following is a specific example to illustrate this.

[0064] Assuming the evaluation time window is set to the past 60 seconds, the system calculates the average mold temperature deviation within this window to be T1 = 3.5°C, and the average profile temperature deviation to be T2 = 1.0°C. Based on the current production instruction prioritizing mold safety, the weighting coefficients for mold temperature are set to w1 = 0.8 and w2 = 0.2, respectively. The overall temperature deviation calculation is as follows: T = √( 0.8 × (3.5)² + 0.2 × (1.0)² ) = √( 0.8 × 12.25 + 0.2 ×1.0 ) = √( 9.8 + 0.2 ) = √10.0 ≈ 3.16°C Here, w1 and w2 are preset weighting coefficients that satisfy w1 + w2 = 1, used to allocate the importance of mold-side and profile-side deviations in the overall evaluation. For example, if production focuses more on protecting mold life, w1 > w2 can be set; if more attention is paid to profile exit quality, w2 > w1 can be set. The squared term in the formula amplifies the contribution of larger deviations to the final result, making the "Comprehensive Temperature Deviation" index more sensitive to significant prediction inaccuracies.

[0065] The resulting comprehensive temperature deviation T ≈ 3.16°C will be transmitted to the subsequent control logic and compared with the preset deviation threshold to scientifically determine whether to dynamically correct the cooling control coefficient.

[0066] It should be noted that the mold cooling coefficient is corrected using the first formula, and the profile cooling coefficient is corrected using the second formula. The first formula is: C1' = C1 × (1 + a × T1); The second formula is: C2' = C2 × (1 + b × T2); Wherein, C1 and C2 are the mold cooling coefficient and profile cooling coefficient before correction, respectively; C1' and C2' are the mold cooling coefficient and profile cooling coefficient after correction, respectively; T1 and T2 are the mold temperature deviation and profile temperature deviation, respectively; and a and b are the preset mold cooling coefficient correction sensitivity and profile cooling coefficient correction sensitivity, respectively.

[0067] It should be noted that the specific values ​​of the two correction sensitivity coefficients, a and b, can be determined by combining process characteristic analysis and a limited number of on-site calibration experiments. First, based on the understanding of the thermodynamic process of the extrusion system (such as the magnitude of thermal inertia and temperature response speed) and the core requirements for process stability and response speed, a reasonable initial value range for a and b is set (e.g., typically between 0.05 and 0.3 to ensure a smooth correction and avoid oscillations). Subsequently, in actual production, calibration can be performed through a simple closed-loop test: for example, when the system is running stably, a known, small temperature setpoint disturbance is artificially introduced, or a typical natural temperature fluctuation is recorded. The speed at which the actual temperature returns to the target value and the overshoot are observed and recorded under different values ​​of a and b after the system adjusts the cooling coefficient using the correction formula. Through iterative testing, the set of (a, b) values ​​that allows the system to quickly and smoothly eliminate deviations without causing drastic fluctuations in the control quantity can be selected as the preset sensitivity coefficients for this type of process on this production line.

[0068] The above scheme transforms the adaptive, fuzzy concept of "correction" into precise, repeatable mathematical operations, enhancing the intelligence level and long-term performance consistency of the control system. The formulaic correction mechanism ensures predictability and consistency of behavior, eliminating the arbitrariness and instability that may arise from manual experience adjustments or rule-based judgments. Secondly, the introduction of independent correction sensitivity coefficients (a, b) provides process engineers with crucial adjustment "knobs," enabling them to separately adjust the adaptive response intensity of the control loops on the mold and profile sides to achieve optimal matching for different process objectives (such as prioritizing mold protection or profile surface quality). Finally, this deviation-based quantitative proportional correction allows the control system to smoothly and progressively approach optimal parameters, avoiding control oscillations that may be caused by sudden parameter changes, ensuring a smooth transition in the production process and continuous parameter optimization.

[0069] It should be noted that the liquid nitrogen injection device preferably includes a first mechanism for injecting nitrogen gas into the extrusion die, and a second mechanism for injecting nitrogen gas into the aluminum profile leaving the extrusion die. S5 includes: Based on the mold cooling demand curve, adjust the injection flow rate and injection method of the first mechanism in each time interval; Based on the profile cooling demand curve, adjust the injection flow rate and injection method of the second mechanism in each time interval; The nitrogen injection methods include: average injection mode, pulse injection mode, and gradient injection mode.

[0070] Based on the first and second mechanisms, independent, parallel-controllable execution channels were established in hardware to achieve the two process objectives of mold cooling and profile cooling / protection. The first mechanism is dedicated to spraying critical areas of the extrusion mold (such as the mold pad and working zone), primarily to quickly remove heat from the mold and control its operating temperature to extend its lifespan. The second mechanism is used to spray the surface of the aluminum profile leaving the extrusion mold. Its main purposes include not only assisting in cooling to regulate the profile exit temperature, but also using the nitrogen atmosphere formed after liquid nitrogen vaporization to provide inert gas protection to the high-temperature profile surface and prevent aluminum oxidation. Separating the two mechanisms structurally avoids mutual interference in the distribution of cooling media, allowing for simultaneous, independent, and optimal performance of both intense cooling of the mold and precise protection and cooling of the profile.

[0071] Meanwhile, it provides two adjustable dimensions for feedforward control commands: "intensity" and "mode," greatly enriching the strategic and adaptive capabilities at the execution level. Adjusting the injection flow rate is fundamental, directly responding to the magnitude of the cooling demand curve, and is achieved by adjusting the valve opening or duty cycle. Adjusting the injection mode is a higher-order strategy, defining the distribution pattern of the flow rate in time or space.

[0072] Each of the three modes has its own applicable scenarios: Average injection mode: Its function is to provide uniform and continuous cooling during periods of stable operation. The conventional method to achieve this is to control the liquid nitrogen valve to maintain a constant opening, thus stabilizing the flow rate.

[0073] Pulse jet mode: This mode provides short-duration, high-efficiency intermittent cooling to cope with transient thermal shocks (such as the moment a new aluminum rod is introduced) or to prevent localized overcooling. The conventional method is to control a high-speed on / off valve, which periodically opens and closes according to a preset frequency and duty cycle.

[0074] Gradient spraying mode: Its function is to provide spatially differentiated cooling based on the uneven temperature distribution on the surface of the mold or profile. Conventional methods include using an array of multiple independently controlled nozzles, or a single oscillating nozzle, to allocate different spray flow rates or residence times to different areas based on a predicted temperature distribution curve.

[0075] Based on the above scheme, flexibility and control precision at the cooling execution level are achieved. The independent control of the dual mechanisms allows the system to fully implement the concept of "cooperative control." It can increase the flow rate of the first mechanism to address the risk of mold overheating, while simultaneously adjusting the protective air curtain of the second mechanism to address the risk of profile oxidation. These two mechanisms operate independently, solving the problem of a single nozzle being unable to handle both simultaneously. Introducing the "spraying mode" control dimension allows the system to go beyond simple flow adjustment and dynamically switch cooling strategies based on predicted operating conditions. For example, "average spraying" is used during steady-state extrusion to conserve media and reduce fluctuations; when a temperature surge is predicted after bar changing, the first mechanism can be instructed to switch to "pulse spraying" for stronger instantaneous cooling to suppress the temperature rise; when a prediction indicates that the temperature on one side of the mold is too high, "gradient spraying" can be used to focus cooling on that side. This intelligent execution strategy significantly improves temperature control capabilities and media utilization efficiency under complex disturbance conditions.

[0076] The following is a specific example to illustrate this.

[0077] In an online cutting cycle, the system predicts that the cutting action will cause a short-term spike in the profile exit temperature. At this point, step S5 instructs the second mechanism as follows: Two seconds before cutting, the spray mode is switched from "average spray mode" to "pulse spray mode," and the peak flow rate in pulse mode is increased to 15 L / min based on the "profile cooling demand curve," with a pulse frequency of 2 Hz and a duty cycle of 60%. This allows the second mechanism to provide intermittent, high-intensity liquid nitrogen flow for impact cooling and enhanced protection of the profile surface during the brief window before the profile temperature rises, effectively suppressing the predicted temperature spike. After cutting, the flow rate and mode return to steady-state settings. Meanwhile, the first mechanism, based on the "mold cooling demand curve," maintains the "average spray mode" during the speed fluctuations caused by cutting, but fine-tunes the flow rate from 12 L / min to 10 L / min to address the slight changes in mold heat input caused by speed variations. This example fully demonstrates the effectiveness and accuracy of the dual-mechanism, dual-dimensional (flow rate and mode) independent control strategy in dealing with coordinated and complex process disturbances.

[0078] In a specific embodiment of this application, the liquid nitrogen cooling control method for aluminum profile extrusion dies preferably further includes the following steps: Obtain the surface quality test results of the extruded aluminum profile, including the oxidation index and the tensile index; If the oxidation index continues to exceed a preset first preset threshold for a first preset number of consecutive production cycles, a downward adjustment to the cooling coefficient of the profile will be triggered. If the tear index continues to exceed the preset second preset threshold for a second preset number of consecutive production cycles, an upward adjustment of the mold cooling coefficient will be triggered.

[0079] For example, a surface quality inspection device (such as a machine vision-based surface defect detection system) can be deployed downstream of the production line, before the profiles are cooled or aged. This device scans and analyzes the surface of the extruded aluminum profiles online and outputs quantified "surface quality inspection results." The "oxidation index" quantifies the degree of oxidation and discoloration on the profile surface (usually manifested as darkening or graying); a higher value indicates more severe oxidation. The "scratch index" quantifies the degree of defects such as scratches and streaks on the profile surface caused by abnormal friction with the die's working zone. These two indices are calculated using image processing algorithms (such as contrast analysis and texture recognition) and uploaded to the control system database along with the corresponding profile production batch or time period information, providing a clear and quantifiable evaluation basis for optimizing the cycle.

[0080] The correction logic for oxidation defects involves continuously monitoring the "oxidation index." When this index consistently exceeds the "first preset threshold" within a "first preset number of production cycles" (e.g., three consecutive aluminum bar production cycles), it is "diagnosed" that insufficient profile cooling / protection is the primary cause of the oxidation problem. The system will then "trigger a downward adjustment correction to the profile cooling coefficient." This "downward adjustment" is a key technical detail because the "profile cooling coefficient," used as a multiplier in step S45 of claim 4, means that its value will be "downwarded" when calculating the "final profile cooling requirement." This may seem counterintuitive to enhancing cooling, but it needs to be understood in conjunction with the formula in claim 7: In the correction formula C2' = C2 × (1 + b × T2), to increase C2' (enhance cooling), T2 (profile temperature deviation) needs to be positive. The "trigger downward adjustment" instruction essentially uses an external command to directly set or forcibly intervene in a virtual, larger "profile temperature deviation T2" used for subsequent correction calculations, or directly set a product factor greater than 1, thereby increasing the actual value of the new coefficient C2' calculated using the formula in claim 7. In terms of control effect, this ultimately manifests as enhanced cooling and protection of the profile to combat oxidation.

[0081] The correction logic for tear defects involves continuously monitoring the "tear index." When this index consistently exceeds the "second preset threshold" within a "second consecutive preset number of production cycles," it is "diagnosed" that the primary cause is likely a low mold temperature (leading to poor metal flowability and increased friction). The system will then "trigger an upward adjustment to the mold cooling coefficient." Similarly, this "upward adjustment" will reduce the subsequently calculated new mold cooling coefficient C1', effectively weakening the cooling intensity on the mold, maintaining a higher operating temperature, improving metal flowability, and thus mitigating tear defects.

[0082] Through the above scheme, the control objective has been elevated from "stabilizing process parameters" to "ensuring final quality." A result-oriented adaptive capability has been established; the system no longer merely focuses on controlling the temperature within a set range, but proactively detects continuous quality degradation trends and automatically adjusts its strategies to curb defects at their source. Intelligent cross-process correlation has been achieved, seamlessly feeding downstream quality information back to the upstream cooling control unit, breaking down information silos. Finally, preset continuous cycle counts and threshold judgments provide a buffer against interference, avoiding erroneous adjustments caused by single, accidental quality fluctuations and ensuring the robustness of optimization decisions.

[0083] The following is a specific example to illustrate this.

[0084] Suppose the surface inspection system reports that the profile "oxidation index" for the most recent four consecutive production cycles ("first preset quantity" set to 4) is 52, 55, 58, and 60, respectively, while the preset "first preset threshold" is 50. Since the index exceeds the threshold four times consecutively, the system triggers a "downward correction of the profile cooling coefficient." One possible implementation for the control system is to ignore the currently calculated profile temperature deviation T2 and directly apply a strengthening correction command. For example, the virtual T2 value used for subsequent correction calculations could be set to +3.0°C, or the product factor (1+b×T2) in the formula of claim 7 could be directly set to a value greater than 1, such as 1.2. After calculation using the formula of claim 7, the profile cooling coefficient C2 will therefore be increased (e.g., from 1.0 to 1.2). In subsequent production, this increased coefficient will increase the "final profile cooling demand" calculated by S45, thereby instructing the second mechanism to spray more liquid nitrogen to enhance the cooling and protection of the profile, ultimately causing the oxidation index to fall back below the threshold. The "upward adjustment correction" of the mold cooling coefficient acts on the first mechanism with the opposite logic to solve the tearing problem.

[0085] like Figure 4 As shown, this application also discloses a liquid nitrogen cooling control system for aluminum profile extrusion dies, which uses a liquid nitrogen injection device to coordinately control the temperature of the extrusion die and the profile temperature. The system includes: Data acquisition module 210 is used to continuously collect multi-source temperature data and working status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum rod feeding temperature; The state prediction module 220 is used to establish a mapping relationship between working state and temperature response based on multi-source temperature data and working state data within the preset historical period, and to predict the sequence of working state changes that will occur in the extrusion production line in the future preset period. Temperature prediction module 230 is used to generate mold temperature prediction curve and profile temperature prediction curve within the preset future time period based on the working state change sequence and the mapping relationship between the working state and temperature response. The demand generation module 240 is used to calculate the deviation from the corresponding preset temperature control target based on the mold temperature prediction curve and the profile temperature prediction curve, and generate the mold cooling demand curve and the profile cooling demand curve in the preset future time period. The strategy adjustment module 250 is used to adjust the injection strategy of the liquid nitrogen injection device according to the mold cooling demand curve and the profile cooling demand curve, so as to perform cooling intervention in advance.

[0086] The data acquisition module 210 is used to perform continuous data acquisition tasks. Its function is to provide real-time and historical multi-source information input for the entire system. In specific implementation, the data acquisition module 210 may include a series of physical interfaces for connecting sensors deployed at key locations on the production line. For example, it may include analog input interfaces for receiving thermocouple signals from the mold and aluminum rod feed inlet, and a dedicated interface for receiving digital signals from the infrared thermometer at the profile outlet. Simultaneously, the data acquisition module 210 also integrates industrial communication interfaces (such as Ethernet, PROFINET, or EtherCAT master stations) for communicating with the extruder's main control PLC to read "working status data," such as extrusion stage flags defined by the PLC program and real-time extrusion speed values ​​from the encoder. Internally, the data acquisition module 210 runs data acquisition drivers and software, responsible for synchronously sampling all interfaces at fixed intervals (e.g., 100 milliseconds), converting analog signals into digital quantities, accurately timestamping all data, and writing it to a local time-series database or sending it to a central server for storage. In a preferred embodiment, the data acquisition module 210 can be implemented based on an industrial edge computing gateway, which completes on-site data acquisition, caching and preprocessing at the network edge close to the production equipment, and then uploads it uniformly. This helps to reduce network load and improve data timeliness.

[0087] The state prediction module 220 is used to build a prediction model based on historical data. In its specific implementation, the state prediction module 220 can run as an independent software service or algorithm container. It periodically (e.g., after each production batch is completed) retrieves data from the storage of the data acquisition module 210 within a "preset historical period". The state prediction module 220 can contain two processing units: one is a "mapping relationship construction unit", which can use data mining algorithms (such as K-means clustering combined with multiple regression analysis) to classify historical data according to "extrusion stage" and calculate the statistical characteristics (mean, variance, and trend) of the corresponding mold and profile temperature for each type of working condition, forming a "typical temperature response" feature vector, and finally constructing a mapping relationship database of "working condition type - feature vector"; the other is a "sequence prediction unit", which can use time series analysis (such as autoregressive models) or rule-based state machines to analyze the periodic patterns in the historical working state sequence (such as the statistical distribution of bar changing intervals) and combine them with the current production progress to generate a "sequence of working state changes within a future preset period". In a preferred embodiment, the state prediction module 220 can use machine learning frameworks (such as TensorFlow or PyTorch) to build more complex prediction models and perform offline training and online fine-tuning using historical data to improve prediction accuracy.

[0088] The temperature prediction module 230 generates temperature curves based on predicted operating conditions. The temperature prediction module 230 receives a "sequence of operating state changes" from the state prediction module 220 as input. Specifically, the temperature prediction module 230 internally includes a "curve synthesis engine." This engine retrieves the corresponding "typical temperature response" features from the "mapping relationship database" maintained by the state prediction module 220 for each future operating condition segment in the input sequence. Then, the engine unfolds and stitches the temperature change patterns represented by these features along a time axis in chronological order. At the junctions of operating condition segments, linear interpolation or smoothing filtering algorithms can be used to ensure that the two generated curves (mold temperature prediction curve and profile temperature prediction curve) are continuous and smooth. The output of the temperature prediction module 230 is two discrete or parameterized temperature curves defined in the future time domain.

[0089] The demand generation module 240 is used to convert temperature prediction into control commands. The demand generation module 240 receives two prediction curves output by the temperature prediction module 230 and a preset temperature control target. In specific implementation, the demand generation module 240 may include a "deviation calculation unit," an "initial demand query unit," a "coefficient correction unit," and a "curve generation unit." The "deviation calculation unit" is responsible for comparing the prediction curve with the target curve and calculating the deviation; the "initial demand query unit" has a built-in or connected "temperature deviation-cooling demand" lookup table, mapping the deviation value to the initial cooling demand; the "coefficient correction unit" can determine the dynamic cooling coefficient based on future operating condition information obtained from the state prediction module; and the "curve generation unit" finally completes the calculation of the initial demand and the coefficient, and outputs a continuous cooling demand curve. In one specific embodiment, the demand generation module 240 can be implemented as a set of function blocks running in a real-time controller (such as a PLC or industrial PC), executing calculations in the order of sub-steps S41-S46 as described in claim 4.

[0090] The strategy adjustment module 250 is the final execution layer of the control commands. It receives the "mold cooling demand curve" and "profile cooling demand curve" output by the demand generation module 240. In its implementation, the strategy adjustment module 250 interacts directly with the drive components of the liquid nitrogen injection device. For an injection device including a "first mechanism" and a "second mechanism," this module typically contains two independent control channels. Each channel includes: a "feedforward setpoint tracker" to parse the input demand curve into the set flow rate at the current moment; a "control algorithm generator" (usually a composite algorithm combining feedforward and PID) to calculate the control output; and a "signal output interface" to send the control quantity (such as a 4-20mA analog signal or PWM pulse) to the corresponding liquid nitrogen proportional valve or high-speed switching valve. The strategy adjustment module 250 also manages the switching of the "injection mode," for example, switching to pulse control or gradient control routines in the control logic based on the characteristics of the operating conditions or demand curve. In a preferred embodiment, the strategy adjustment module 250 can be integrated into a programmable automation controller with high-speed motion control function to ensure accurate, low-latency tracking of the flow curve.

[0091] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0092] The preferred embodiments of this application have been described in detail above, but this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application.

Claims

1. A method for controlling liquid nitrogen cooling of aluminum profile extrusion dies, characterized in that, the temperature of the extrusion die and the temperature of the profile are controlled in a coordinated manner by a liquid nitrogen injection device, wherein, The method includes the following steps: S1, continuously collect multi-source temperature data and working status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum bar feeding temperature; S2, Based on the multi-source temperature data and working status data within the preset historical period, establish a mapping relationship between working status and temperature response, and predict the sequence of working status changes that the extrusion production line will undergo in the future preset period. S3, Based on the working state change sequence and the mapping relationship between the working state and temperature response, generate the mold temperature prediction curve and the profile temperature prediction curve for the future preset time period. S4. Based on the mold temperature prediction curve and the profile temperature prediction curve, calculate the deviation from the corresponding preset temperature control target, and generate the mold cooling demand curve and the profile cooling demand curve within the preset future time period. S5. Based on the mold cooling demand curve and the profile cooling demand curve, adjust the injection strategy of the liquid nitrogen injection device to perform cooling intervention in advance.

2. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 1, characterized in that, The working status data includes: the extrusion stage based on the control system signal recognition of the extrusion production line; the preset historical cycle covers multiple complete aluminum rod replacement operation cycles and online cutting operation cycles.

3. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 2, characterized in that, S2 includes: S21. Based on the multi-source temperature data and working status data within the preset historical period, the data is classified according to the identified extrusion stages, and the corresponding multi-source temperature data for each extrusion stage is statistically analyzed to generate typical temperature responses of the mold and profile under various extrusion stages, so as to establish a mapping relationship between working status and temperature response. S22, Analyze the working status data within the preset historical period, identify the periodic patterns of operations such as replacing aluminum rods and online cutting, and generate a sequence of working status changes within the preset future time period based on the periodic patterns.

4. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 1, characterized in that, S4 includes: S41, the future preset time period is divided into multiple consecutive time intervals of equal length; S42, For each time interval, calculate the average temperature or integral temperature of the mold temperature prediction curve and the profile temperature prediction curve within the interval, and compare it with the corresponding mold target temperature and profile target temperature to obtain the average temperature deviation of the mold and the average temperature deviation of the profile for the corresponding time interval. S43, based on the average temperature deviation of the mold and the average temperature deviation of the profile, obtain the initial cooling demand of the mold and the initial cooling demand of the profile for the corresponding time interval from the preset temperature deviation-cooling demand mapping relationship; S44, Based on the working status data of the time interval, determine the mold cooling coefficient and the profile cooling coefficient; S45, Multiply the initial cooling demand of the mold and the initial cooling demand of the profile by the corresponding cooling coefficients of the mold and the profile respectively to obtain the final cooling demand of the mold and the final cooling demand of the profile for the corresponding time interval. S46, connect the final cooling demand of the mold and the final cooling demand of the profile calculated in each time interval along the time axis to generate the mold cooling demand curve and the profile cooling demand curve respectively.

5. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 4, characterized in that, Before S4, it also includes: The temperature of the extrusion die and the temperature of the profile are acquired in real time. During the predetermined future time period, the temperature of the extrusion die and the temperature of the profile are compared with the predicted temperature curves of the die and the profile to obtain a comprehensive temperature deviation. When the overall temperature deviation continues to exceed the preset deviation threshold, the mold cooling coefficient and profile cooling coefficient in the subsequent production stages are corrected based on the overall temperature deviation.

6. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 5, characterized in that, The step of comparing the temperature of the extrusion die and the temperature of the profile with the predicted temperature curves of the die and profile, respectively, to obtain the comprehensive temperature deviation during the predetermined future time period includes: During the predetermined future time period, the average temperature deviation between the extrusion die temperature and the die temperature prediction curve, and the average temperature deviation between the profile temperature and the profile temperature prediction curve are calculated to obtain the die temperature deviation and the profile temperature deviation. Based on the temperature deviation of the mold and the temperature deviation of the profile, the comprehensive temperature deviation is calculated using the weighted root mean square error method.

7. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 6, characterized in that, The mold cooling coefficient is corrected using a first formula, and the profile cooling coefficient is corrected using a second formula. The first formula is: C1' = C1 × (1 + a × T1); The second formula is: C2' = C2 × (1 + b × T2); Wherein, C1 and C2 are the mold cooling coefficient and profile cooling coefficient before correction, respectively; C1' and C2' are the mold cooling coefficient and profile cooling coefficient after correction, respectively; T1 and T2 are the mold temperature deviation and profile temperature deviation, respectively; and a and b are the preset mold cooling coefficient correction sensitivity and profile cooling coefficient correction sensitivity, respectively.

8. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 1, characterized in that, The liquid nitrogen injection device includes a first mechanism for injecting nitrogen gas into the extrusion die, and a second mechanism for injecting nitrogen gas into the aluminum profile leaving the extrusion die. S5 includes: Based on the mold cooling demand curve, adjust the injection flow rate and injection method of the first mechanism in each time interval; Based on the profile cooling demand curve, adjust the injection flow rate and injection method of the second mechanism in each time interval; The nitrogen injection methods include: average injection mode, pulse injection mode, and gradient injection mode.

9. The method for controlling liquid nitrogen cooling of aluminum profile extrusion dies according to claim 4, characterized in that, It also includes the following steps: Obtain the surface quality test results of the extruded aluminum profile, including the oxidation index and the tensile index; If the oxidation index continues to exceed a preset first preset threshold for a first preset number of consecutive production cycles, a downward adjustment to the cooling coefficient of the profile will be triggered. If the tear index continues to exceed the preset second preset threshold for a second preset number of consecutive production cycles, an upward adjustment of the mold cooling coefficient will be triggered.

10. A liquid nitrogen cooling control system for aluminum profile extrusion dies, characterized in that, the temperature of the extrusion die and the temperature of the profile are controlled in a coordinated manner by a liquid nitrogen injection device. The system includes: The data acquisition module is used to continuously collect multi-source temperature data and operating status data of the extrusion production line within a preset historical period; the multi-source temperature data includes extrusion die temperature, profile temperature and aluminum rod feeding temperature; The status prediction module is used to establish a mapping relationship between working status and temperature response based on multi-source temperature data and working status data within the preset historical period, and to predict the sequence of working status changes that the extrusion production line will undergo in the future preset period. The temperature prediction module is used to generate mold temperature prediction curves and profile temperature prediction curves within the preset future time period based on the working state change sequence and the mapping relationship between the working state and temperature response. The demand generation module is used to calculate the deviation from the corresponding preset temperature control target based on the mold temperature prediction curve and the profile temperature prediction curve, and generate the mold cooling demand curve and the profile cooling demand curve in the preset future time period. The strategy adjustment module is used to adjust the injection strategy of the liquid nitrogen injection device according to the mold cooling demand curve and the profile cooling demand curve, so as to perform cooling intervention in advance.

Images