A temperature control method based on spatiotemporal curves and multimodal databases

Abstract

This application relates to the field of temperature control technology, and in particular to a temperature control method based on spatiotemporal curves and a multimodal database. The method includes: arranging a multi-type sensor network within a heat treatment furnace to collect temperature data at various spatial locations, forming a spatiotemporal temperature curve; constructing a multimodal database containing a process knowledge base, a historical data pool, and an equipment model library; determining the workpiece parameters and process requirements of the workpiece to be processed, retrieving a matching basic process curve from the process knowledge base, and correcting the process parameters using the historical data pool; placing the workpiece to be processed into the heat treatment furnace and starting heating, obtaining the target process curve using a pre-trained LSTM prediction model; comparing the current spatiotemporal temperature curve with the target process curve in real time, identifying deviations in the time and spatial dimensions, and generating corresponding control commands for temperature control, thus achieving high-precision, low-energy-consumption, and adaptive control of the heat treatment process.

Description

Technical Field

[0001] The embodiments of this application relate to the field of temperature control technology, and in particular to a temperature control method based on spatiotemporal curves and a multimodal database. Background Technology

[0002] Heat treatment is a key process for regulating the properties of metallic materials. Its core principle is to achieve the required microstructure and properties of the workpiece through precise control of the temperature field within the furnace. It is widely applied in processes such as quenching, tempering, and carburizing. In actual heat treatment, precise temperature control within the furnace is necessary to ensure that the workpiece achieves the expected performance indicators such as hardness, strength, and wear resistance. However, most existing heat treatment furnace temperature control technologies combine single-point feedback with fixed process curves, which has the following significant drawbacks.

[0003] First, single-point control struggles to achieve uniform global temperature. Traditional methods rely primarily on a single-point temperature sensor located at the center of the furnace for feedback control, adjusting the overall heating power to maintain the set temperature. This approach only senses the average temperature inside the furnace and cannot effectively capture temperature distribution differences across the furnace's spatial dimensions (such as temperature differences between upper and lower layers, corners, and the furnace center). In large heat treatment furnaces, significant temperature differences often exist at different locations within the furnace due to factors such as heat radiation, convection, and furnace wall heat dissipation. For example, in a pit furnace, the upper layer of the workpiece is prone to overheating due to concentrated heat radiation. (Above), the lower layer is prone to forming under-temperature areas due to delayed heat dissipation. (The following), which leads to a hardness deviation of more than HRC±4 in the same batch of workpieces, seriously affecting the yield.

[0004] Second, the rigidity of process curves leads to insufficient adaptability. Most existing control systems use preset, fixed process curves, such as the "heating rate". → Insulation →Cooling rate "It cannot be dynamically adjusted according to the material, size, furnace load, and other parameters of different workpieces. When faced with new materials, new products, or changes in furnace load, it is necessary to adjust the process parameters through 3 to 5 heats of trial and error, and it relies on the experience and judgment of senior technicians, which results in a long debugging cycle and high cost."

[0005] Third, delayed response to deviations leads to energy waste. Most existing control methods are passive, adjusting only after a deviation occurs, meaning the heating power is adjusted only when the temperature exceeds the set range. This response mechanism has a significant time lag, causing overheating or underheating in localized areas. To compensate for insufficient temperature uniformity, a strategy of overall overheating is often adopted, such as setting the insulation temperature higher than the target value. to This results in a 15% to 20% increase in energy consumption per furnace, while also increasing the risk of workpiece deformation and oxidation.

[0006] Due to the aforementioned technical limitations, existing heat treatment furnace temperature control methods are unable to meet the industrial production requirements of high precision, high consistency, and low energy consumption. There is an urgent need in this field for an intelligent temperature control method that can globally sense the temperature field, dynamically adapt to process requirements, and actively predict and correct deviations. Summary of the Invention

[0007] To address the aforementioned technical issues, embodiments of this application propose a temperature control method based on spatiotemporal curves and a multimodal database. This method aims to achieve global perception of the furnace temperature field through a multi-source sensor network, perform intelligent calibration using a process knowledge base, historical data pool, and equipment model library, predict deviations in advance using an LSTM prediction model, and achieve precise regulation of the furnace temperature based on a zoned closed-loop control strategy. Ultimately, this achieves high-precision, low-energy-consumption, and adaptive control of the heat treatment process.

[0008] To achieve the above objectives, embodiments of this application propose a temperature control method based on spatiotemporal curves and a multimodal database, applicable to heat treatment furnaces. The method includes the following steps: arranging a multi-type sensor network along the height, width, and depth of the furnace chamber within the heat treatment furnace to collect temperature data at various spatial locations, synchronously recording auxiliary parameters, aligning them by timestamps to construct a temperature field, and using an interpolation algorithm to complete the temperature distribution in areas without sensor coverage, forming a spatiotemporal temperature curve; wherein the auxiliary parameters include heating power, fan speed, furnace door opening / closing status, and ambient temperature; constructing a multimodal database containing a process knowledge base, a historical data pool, and an equipment model library; wherein the historical data pool stores spatiotemporal temperature curves from previous furnace cycles; obtaining the workpiece parameters and process requirements of the workpiece to be processed, and from the process knowledge... The system retrieves a matching basic process curve from the knowledge base and corrects the process parameters of the workpiece to be processed by combining them with historical data. The workpiece is then placed in the heat treatment furnace and heating begins. After the preset control start time is reached, the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power, and fan speed are input into a pre-trained LSTM prediction model to obtain the target process curve. The system compares the current spatiotemporal temperature curve with the target process curve in real time, identifies time dimension deviations and spatial dimension deviations, and generates corresponding control commands for temperature control. The control command for time dimension deviation is to adjust the heating power of the corresponding area, and the control command for spatial dimension deviation is to adjust the fan speed. The control commands have a control upper limit designed based on the equipment model library.

[0009] To achieve the above objectives, embodiments of this application also propose a temperature control system based on spatiotemporal curves and a multimodal database, applicable to heat treatment furnaces. The system includes: a spatiotemporal temperature curve acquisition module, used to arrange a multi-type sensor network along the height, width, and depth of the furnace chamber within the heat treatment furnace, collect temperature data at various spatial locations, synchronously record auxiliary parameters, align them by timestamps to construct a temperature field, and complete the temperature distribution in undetected areas using an interpolation algorithm to form a spatiotemporal temperature curve; wherein the auxiliary parameters include heating power, fan speed, furnace door opening / closing status, and ambient temperature; a multimodal database construction module, used to construct a multimodal database containing a process knowledge base, a historical data pool, and an equipment model library; wherein the historical data pool stores spatiotemporal temperature curves from previous furnace cycles; and a process parameter correction module, used to acquire the workpiece parameters and process requirements of the workpiece to be processed, and based on the workpiece parameters... The system retrieves matching basic process curves from the process knowledge base based on process requirements and modifies the process parameters of the workpiece to be processed using historical data. The heating execution module places the workpiece into the heat treatment furnace and begins heating. The target prediction module, upon reaching a preset control start time, inputs the basic process curve, process requirements, modified process parameters, current temperature field matrix, heating power, and fan speed into a pre-trained LSTM prediction model to obtain the target process curve. The control execution module compares the current spatiotemporal temperature curve with the target process curve in real time, identifies time and spatial dimension deviations, and generates corresponding control commands for temperature control. The control command for time dimension deviation adjusts the heating power of the corresponding area, while the control command for spatial dimension deviation adjusts the fan speed. The control commands have upper limits designed based on the equipment model library.

[0010] To achieve the above objectives, embodiments of this application also propose an electronic device, including: a processor and a memory, wherein the memory stores instructions executable by the processor, and the processor is configured to execute the instructions such that the electronic device can implement a temperature control method based on spatiotemporal curves and a multimodal database as described above.

[0011] To achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a computer program that, when executed by a processor, enables a temperature control method based on spatiotemporal curves and a multimodal database as described above.

[0012] Optionally, the process involves arranging a multi-type sensor network along the height, width, and depth of the furnace chamber within the heat treatment furnace to collect temperature data at various spatial locations and simultaneously record auxiliary parameters, including: Inside the heat treatment furnace, at least 5 temperature measuring layers are set along the height of the furnace chamber, at least 3 temperature measuring points are set in the width direction of each layer, and at least 3 temperature measuring points are set in the depth direction of each layer, forming a uniformly distributed temperature measuring grid consisting of at least 45 temperature measuring points. At each temperature measurement point, a K-type thermocouple and a non-contact infrared thermometer are simultaneously installed. The K-type thermocouple is used to accurately measure the temperature at the contact point, while the non-contact infrared thermometer is used to measure the surface temperature. The two complement each other to correct the temperature measurement error, obtain the temperature data at each spatial location, and record auxiliary parameters simultaneously. The acquisition frequency of the auxiliary parameters is synchronized with the acquisition frequency of the temperature data. The process of aligning by timestamp to construct a temperature field and using an interpolation algorithm to complete the temperature distribution in areas without data points, forming a spatiotemporal temperature curve, includes: The temperature data from each spatial location are aligned by timestamp to construct a temperature field matrix, represented as follows: ; in, To construct the temperature field matrix, , , These represent the width, depth, and height directions of the furnace, respectively. Indicates time, , , These represent the number of temperature measurement points in the width, depth, and height directions, respectively. Indicates the first The width position, the first The depth position, the first Temperature measurement points at various heights Temperature data at any given time; For the temperature field matrix The Kriging interpolation algorithm is used to complete the temperature distribution in the unspecified areas of the furnace, forming a spatiotemporal temperature curve.

[0013] Optionally, the construction of a multimodal database comprising a process knowledge base, a historical data pool, and an equipment model library includes: Obtain workpiece parameters, process requirements, and standard process data corresponding to different workpieces. Based on the standard process data, draw basic process curves and establish a mapping relationship between workpiece parameters and process requirements and basic process curves to form a process knowledge base. Among them, workpiece parameters include material type, workpiece size, and furnace loading amount. Process requirements are the performance indicators that the workpiece needs to achieve after processing. Performance indicators include at least one of surface hardness, core hardness, diffusion layer depth, and allowable deformation. Standard process data includes the target temperature, stage duration, target temperature change rate, and temperature threshold corresponding to the heating stage, holding stage, and cooling stage, respectively. Acquire workpiece parameters, process requirements, actual process parameters, spatiotemporal temperature curves, auxiliary parameters, equipment status, and workpiece performance data corresponding to multiple past furnace batches. Associate the data corresponding to the same past furnace batch according to the furnace batch identifier, and align the spatiotemporal temperature curves, auxiliary parameters, and equipment status according to the timestamp to form a historical data pool. Obtain the equipment identification, heat transfer parameters, convective heat transfer parameters, and dynamic response parameters of the heat treatment furnace; establish a heating response model between regional heating power and temperature change, a convective heat transfer model between fan speed and furnace temperature distribution, and a furnace body heat transfer model; and associate and store the heating response model, convective heat transfer model, furnace body heat transfer model, and equipment identification to form an equipment model library. A correlation index is established based on material type, workpiece size, furnace loading quantity, and equipment identification. The process knowledge base, historical data pool, and equipment model library are linked through the correlation index to obtain a multimodal database.

[0014] Optionally, the step of determining the process parameters and process requirements of the workpiece to be processed, retrieving matching basic process curves from the process knowledge base, and correcting the process parameters in conjunction with historical data pools includes: Obtain the workpiece parameters and process requirements of the workpiece to be processed, and retrieve the basic process curve matching the workpiece from the process knowledge base based on the workpiece parameters and process requirements. The target temperature, duration of each process stage, rate of change of target temperature, and temperature threshold are extracted from the basic process curve and used as the initial process parameters for the workpiece to be processed. Using the same material type, workpiece size difference less than 1%, and furnace loading difference less than the preset furnace loading threshold as search conditions, similar past furnace batches matching the workpiece to be processed are filtered from the historical data pool. Acquire the spatiotemporal temperature curves, actual process parameters, and workpiece performance data of each similar previous furnace. Align the spatiotemporal temperature curves of each similar previous furnace with the basic process curve matching the workpiece to be processed according to the process stages. Calculate the temperature deviation of the actual temperature relative to the target temperature, the time deviation of the actual stage duration relative to the target stage duration, and the rate deviation of the actual temperature change rate relative to the target temperature change rate for each process stage. Based on the temperature deviation, time deviation, and rate deviation corresponding to each similar previous furnace, as well as the differences between workpiece performance data and process requirements, the correction amounts corresponding to the target temperature, stage duration, and target temperature change rate are determined, and the initial process parameters are corrected to obtain the corrected process parameters.

[0015] Optionally, the step of inputting the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power, and fan speed into a pre-trained LSTM prediction model to obtain the target process curve includes: Starting from a preset control start time, the basic process curve is discretized according to a preset time interval to obtain multiple prediction times and the basic target temperature corresponding to each prediction time. Then, a corresponding process stage identifier is added according to the process stage at which each prediction time is located. Based on the process requirements and the corrected process parameters, the target temperature, stage duration, target temperature change rate and temperature threshold corresponding to each process stage are extracted and combined to form a process constraint feature sequence. Extract temperature data from each spatial location within a preset sampling time window ending at the current time from the current temperature field matrix, and align it with the heating power and fan speed at the corresponding sampling time according to the timestamp to form a furnace condition feature sequence. The basic target temperature, process constraint feature sequence, and furnace condition feature sequence are concatenated in chronological order to form the input feature sequence of the LSTM prediction model. The LSTM prediction model is used to extract the temperature change features between adjacent sampling times and the temperature change trends between different process stages in the input feature sequence, so as to obtain the furnace temperature corresponding to each prediction time. Based on the difference between the furnace temperature corresponding to each prediction time and the corresponding basic target temperature, the temperature correction amount corresponding to each prediction time is determined. The base target temperature is corrected by the temperature correction amount corresponding to each prediction time. Based on the process constraint feature sequence, the corrected base target temperature at each prediction time is constrained so that the rate of temperature change between adjacent prediction times does not exceed the rate of temperature change of the target temperature of the corresponding process stage, and the target temperature at each prediction time is within the range of the corresponding temperature threshold. By connecting the corrected target temperatures at each predicted time point in chronological order, the target process curve is obtained.

[0016] Optionally, the real-time comparison of the current spatiotemporal temperature curve with the target process curve, identifying time dimension deviations and spatial dimension deviations, and generating corresponding control commands for temperature control includes: According to the correspondence between each temperature measuring point in the furnace and the heating unit and circulating fan, the furnace is divided into multiple control areas. In each control cycle, the average temperature of each control area at the current moment and the actual temperature change rate within the preset time window are determined based on the current spatiotemporal temperature curve. Extract the target temperature and target temperature change rate corresponding to the current moment from the target process curve, compare the actual temperature change rate of each control region with the target temperature change rate, and obtain the time dimension deviation of each control region. The average temperature of each control area is compared with the target temperature to obtain the temperature deviation of each control area. Based on the temperature deviation of each control area and the temperature difference between different control areas, the spatial dimension deviation of each control area is determined. When the time dimension deviation of any control area exceeds the preset rate deviation threshold, the heating response model and response lag time of the heating unit corresponding to the control area are obtained from the equipment model library. Based on the time dimension deviation, the current heating power, the heating response model and the response lag time, the heating power adjustment amount corresponding to the control area is determined, and a heating power control command to increase or decrease the heating power of the control area is generated according to the direction of the time dimension deviation. When the spatial dimension deviation of any control area exceeds the preset temperature deviation threshold, the convective heat transfer model corresponding to the control area is obtained from the equipment model library. The spatial dimension deviation of the control area, the temperature difference between adjacent control areas and the current fan speed are input into the convective heat transfer model to determine the fan speed adjustment amount to reduce the spatial dimension deviation and generate a fan speed control command to adjust the fan speed of the control area. Heating power control commands and fan speed control commands are sent to the corresponding heating units and circulating fans respectively to achieve temperature control of the furnace.

[0017] Optionally, the upper limit of control commands is obtained through the following steps: The rated power, allowable load rate, and maximum single-cycle power change of each heating unit are obtained from the equipment model library, and the upper limit of heating power control is determined based on the rated power and allowable load rate. Obtain the rated speed, allowable speed range, and maximum single-cycle speed change of each circulating fan from the equipment model library, and determine the upper limit of fan speed control based on the rated speed and allowable speed range; The step of sending heating power control commands and fan speed control commands to the corresponding heating units and circulating fans respectively to achieve temperature control of the furnace includes: If the heating power control command exceeds the heating power control limit, it will be limited to the heating power control limit. Based on the maximum single-cycle power change, the change in heating power control command between adjacent control cycles will be limited, and the limited heating power control command will be sent to the corresponding heating unit. If the fan speed control command exceeds the upper limit of the fan speed control, it will be limited to the upper limit of the fan speed control. Based on the maximum single-cycle speed change, the change of the fan speed control command between adjacent control cycles will be limited, and the limited fan speed control command will be sent to the corresponding circulating fan.

[0018] This application proposes a temperature control method based on spatiotemporal curves and a multimodal database, which achieves the following improvements compared to traditional temperature control methods.

[0019] First, it significantly improves temperature uniformity. This application collects temperature data from different spatial locations within the furnace using a multi-type sensor network, constructs a continuous spatiotemporal temperature curve, and adjusts the heating power and fan speed of the corresponding areas according to the time and spatial deviations. This effectively achieves precise zonal temperature control within the furnace, reduces the temperature difference between the upper and lower layers of the furnace, and between the center and the edge, thereby improving the workpiece hardness, diffusion depth, and consistency of the metallographic structure.

[0020] Secondly, it significantly shortens the process debugging cycle. Based on the material type, size, furnace loading, and process requirements of the workpiece to be processed, this application quickly matches the basic process curve from the process knowledge base and uses temperature curves from similar previous furnaces, actual process parameters, and workpiece performance data for calibration. This reduces reliance on manual experience and repeated furnace trials, improves the accuracy of process parameter settings, and shortens the debugging cycle.

[0021] Third, it significantly reduces energy consumption. This application utilizes an LSTM prediction model to predict the temperature change trend inside the furnace and implements control before the temperature deviation expands, avoiding the lag adjustment in traditional control methods. At the same time, it adjusts the heating power and fan speed according to the actual temperature state of different areas to avoid overall overheating due to local low temperatures, reducing energy waste caused by ineffective heating, excessive heat preservation, and rework.

[0022] Fourth, it possesses adaptive and evolvable capabilities. After each furnace cycle, this application can store the actual spatiotemporal temperature curves, control parameters, equipment status, and workpiece performance data into a multimodal database, and update the process knowledge base and equipment model library accordingly. This allows the system to continuously correct process parameters and equipment response models, adapt to parameter changes caused by different materials, workpiece specifications, furnace loading conditions, and equipment aging, and maintain high temperature control accuracy over the long term. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies of this application will be briefly introduced below. Obviously, the following drawings are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings described herein are only used to explain this application and are not intended to limit this application.

[0024] Figure 1 This is a flowchart of a temperature control method based on spatiotemporal curves and a multimodal database provided in one embodiment of this application; Figure 2 This is a schematic diagram of a temperature control system based on spatiotemporal curves and a multimodal database provided in one embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. Those skilled in the art will understand that many technical details have been provided in the embodiments of this application to facilitate better understanding. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of this application. The following embodiments can be combined with and referenced by each other without contradiction.

[0026] One embodiment of this application proposes a temperature control method based on spatiotemporal curves and a multimodal database, applicable to heat treatment furnaces. The implementation details of the temperature control method based on spatiotemporal curves and a multimodal database proposed in this embodiment are described below. The following content is only for the convenience of understanding the relevant implementation details and is not necessary for implementing this solution.

[0027] This embodiment proposes a temperature control method based on spatiotemporal curves and multimodal databases. By modeling the temperature field in the spatiotemporal dimension, deeply mining process knowledge, and implementing predictive control closed loop, it achieves precise control of furnace temperature, improves workpiece performance consistency, reduces energy consumption, and shortens the process debugging cycle.

[0028] The specific process of the temperature control method based on spatiotemporal curves and multimodal database proposed in this embodiment can be described as follows: Figure 1 As shown, it includes: Step 11: Arrange a network of multiple types of sensors along the height, width and depth of the furnace in the heat treatment furnace to collect temperature data at each spatial location, record auxiliary parameters synchronously, align them according to timestamps to construct a temperature field, and use interpolation algorithms to complete the temperature distribution in areas where no sensors are deployed to form a spatiotemporal temperature curve. The auxiliary parameters include heating power, fan speed, furnace door opening and closing status and ambient temperature.

[0029] Specifically, to achieve temperature control within a heat treatment furnace, a multi-sensor network needs to be deployed along the height, width, and depth of the furnace chamber. This network collects temperature data from various spatial locations and simultaneously records auxiliary parameters. By aligning the temperature data from each location according to timestamps, a temperature field can be constructed. Then, an interpolation algorithm is used to complete the temperature distribution in areas without sensors, forming a spatiotemporal temperature curve. The auxiliary parameters include heating power, fan speed, furnace door open / closed status, and ambient temperature.

[0030] In one example, when deploying a multi-type sensor network, at least five temperature sensing layers need to be set along the height direction (Z-axis) of the furnace chamber inside the heat treatment furnace. At least three temperature sensing points need to be set along the width direction (X-axis) of each layer, and at least three temperature sensing points need to be set along the depth direction (Y-axis) of each layer, forming a uniformly distributed temperature sensing grid consisting of at least 45 temperature sensing points. Subsequently, a K-type thermocouple (with an accuracy of [insert accuracy here]) is simultaneously installed at each temperature sensing point. Both a K-type thermocouple and a non-contact infrared thermometer (non-contact, covering the workpiece surface) are used. The K-type thermocouple accurately measures the temperature at the contact point, while the non-contact infrared thermometer measures the surface temperature. These two methods complement each other to correct measurement errors, obtaining temperature data at various spatial locations and simultaneously recording auxiliary parameters. The acquisition frequency of the auxiliary parameters is synchronized with the acquisition frequency of the temperature data and can be set to 10Hz.

[0031] In one example, when constructing a spatiotemporal temperature curve, it is necessary to align the temperature data of each spatial location by timestamp to construct a temperature field matrix. The constructed temperature field matrix is ​​represented as follows: ; in, To construct the temperature field matrix, , , These represent the width, depth, and height directions of the furnace, respectively. Indicates time, , , These represent the number of temperature measurement points in the width, depth, and height directions, respectively. Indicates the first The width position, the first The depth position, the first Temperature measurement points at various heights Temperature data at any given time.

[0032] For the temperature field matrix It is necessary to use the Kriging interpolation algorithm to complete the temperature distribution of the unspecified areas in the furnace, forming a spatiotemporal temperature curve (also known as a spatiotemporal temperature cloud map).

[0033] Step 12: Construct a multimodal database that includes a process knowledge base, a historical data pool, and an equipment model library. The historical data pool stores the spatiotemporal temperature curves of past furnace cycles.

[0034] Specifically, a multimodal database is the data guarantee for achieving furnace temperature control in heat treatment furnaces. We need to build a multimodal database consisting of a process knowledge base, a historical data pool, and an equipment model library. Among them, the historical data pool stores the spatiotemporal temperature curves of past furnace cycles.

[0035] In one example, building a process knowledge base requires acquiring workpiece parameters, process requirements, and standard process data for different workpieces. Based on the standard process data, basic process curves are plotted, establishing a mapping relationship between workpiece parameters, process requirements, and the basic process curves, thus forming the process knowledge base. Workpiece parameters include material type, workpiece size, and furnace loading. Process requirements are the performance indicators that the workpiece must achieve after processing, including at least one of surface hardness, core hardness, diffusion layer depth, and allowable deformation. Standard process data includes the target temperature, duration, rate of change of target temperature, and temperature threshold for each of the heating, holding, and cooling stages.

[0036] In one example, the process knowledge recorded in the process knowledge base, such as the quenching of 42CrMo steel, requires... Continuous heating for 2 hours results in a surface hardness of HRC58 to HRC62.

[0037] In one example, when building a historical data pool, it is necessary to obtain the workpiece parameters, process requirements, actual process parameters, spatiotemporal temperature curves, auxiliary parameters, equipment status and workpiece performance data corresponding to multiple past furnace batches (historical furnace batches). The data corresponding to the same past furnace batch are associated according to the furnace batch identifier, and the spatiotemporal temperature curves, auxiliary parameters and equipment status are aligned according to the timestamp to form a historical data pool.

[0038] In one example, when building an equipment model library, it is necessary to obtain the equipment identifier, heat conduction parameters, convective heat transfer parameters, and dynamic response parameters of the heat treatment furnace, establish a heating response model between regional heating power and temperature change, a convective heat transfer model between fan speed and furnace temperature distribution, and a furnace body heat conduction model, and then associate and store the heating response model, convective heat transfer model, furnace body heat conduction model, and equipment identifier to form an equipment model library.

[0039] In one example, after the process knowledge base, historical data pool, and equipment model library are constructed, it is necessary to establish an association index based on material type, workpiece size, furnace loading quantity, and equipment identifier. The process knowledge base, historical data pool, and equipment model library are then linked through the association index to obtain a multimodal database.

[0040] Step 13: Obtain the workpiece parameters and process requirements of the workpiece to be processed, retrieve the matching basic process curve from the process knowledge base, and modify the process parameters of the workpiece to be processed in combination with the historical data pool.

[0041] Specifically, after completing the preparation work, we can enter the actual temperature control stage. We need to obtain the workpiece parameters and process requirements of the workpiece to be processed, retrieve the matching basic process curve from the process knowledge base, and modify the process parameters of the workpiece to be processed in combination with the historical data pool.

[0042] After identifying the workpiece to be processed, the workpiece parameters and process requirements are first obtained. Based on the workpiece parameters and process requirements, the basic process curve matching the workpiece is retrieved from the process knowledge base.

[0043] Subsequently, the target temperature, stage duration, target temperature change rate, and temperature threshold corresponding to each process stage are extracted from the basic process curve and used as the initial process parameters for the workpiece to be processed.

[0044] Next, using the same material type, workpiece size difference of less than 1%, and furnace loading difference of less than the preset furnace loading threshold as search conditions, similar past furnace batches that match the workpiece to be processed are filtered from the historical data pool.

[0045] Then, the spatiotemporal temperature curves, actual process parameters, and workpiece performance data of each similar previous furnace are obtained. The spatiotemporal temperature curves of each similar previous furnace and the basic process curve matching the workpiece to be processed are aligned according to the process stages. The temperature deviation of the actual temperature relative to the target temperature, the time deviation of the actual stage duration relative to the target stage duration, and the rate deviation of the actual temperature change rate relative to the target temperature change rate are calculated for each process stage.

[0046] Finally, based on the temperature deviation, time deviation, and rate deviation corresponding to each similar previous furnace run, as well as the differences between the workpiece performance data and process requirements, the correction amounts corresponding to the target temperature, stage duration, and target temperature change rate are determined, and the initial process parameters are corrected to obtain the corrected process parameters (e.g., changing the holding temperature from...). Fine-tuning to (to compensate for heat loss at the edges).

[0047] In one example, the material type of the workpiece to be processed is 20CrMnTi, and the workpiece size is... 50mm×100mm, furnace loading capacity of 50 pieces / furnace, specific process requirements: carburized layer depth of 0.8mm to 1.0mm, surface hardness of HRC58 to HRC62, and the retrieved basic process curve is as follows. Carburizing for 8 hours, cooling rate .

[0048] Step 14: Place the workpiece to be processed into the heat treatment furnace and start heating. After the preset control start time is reached, input the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power and fan speed into the pre-trained LSTM prediction model to obtain the target process curve.

[0049] Specifically, after obtaining the corrected process parameters, the workpiece to be processed can be placed in the heat treatment furnace and heating can begin. After the preset control start time is reached, the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power and fan speed are input into the pre-trained LSTM prediction model to obtain the target process curve.

[0050] In one example, we use a preset control start time as the starting point, discretize the basic process curve according to a preset time interval, obtain multiple prediction times and the basic target temperature corresponding to each prediction time, and add the corresponding process stage identifier according to the process stage at which each prediction time is located.

[0051] Subsequently, based on the process requirements and the corrected process parameters, the target temperature, stage duration, target temperature change rate, and temperature threshold corresponding to each process stage are extracted and combined to form a process constraint feature sequence. Temperature data of each spatial location within a preset sampling time window with the current time as the endpoint are extracted from the current temperature field matrix and aligned with the heating power and fan speed at the corresponding sampling time according to the timestamp to form a furnace condition feature sequence.

[0052] Next, the basic target temperature, process constraint feature sequence, and furnace condition feature sequence are concatenated in chronological order to form the input feature sequence of the LSTM prediction model. The LSTM prediction model is then used to extract the temperature change features between adjacent sampling times and the temperature change trends between different process stages in the input feature sequence to obtain the furnace temperature corresponding to each prediction time. Based on the difference between the furnace temperature corresponding to each prediction time and the corresponding basic target temperature, the temperature correction amount corresponding to each prediction time is determined.

[0053] Next, the base target temperature is corrected using the temperature correction amount corresponding to each prediction time. Based on the process constraint characteristic sequence, the corrected base target temperature at each prediction time is constrained to ensure that the temperature change rate between adjacent prediction times does not exceed the target temperature change rate of the corresponding process stage, and that the target temperature at each prediction time is within the corresponding temperature threshold range. By connecting the corrected target temperatures at each prediction time in chronological order, the target process curve can be obtained. .

[0054] In one example Take 30 minutes.

[0055] In one example, the LSTM prediction model is trained using spatiotemporal temperature curves from past furnace runs stored in a historical data pool. During training, sensor data and model predictions can be fused using Kalman filtering, and model parameters can be dynamically updated (e.g., adjusting the weights of the LSTM's memory cells) to ensure that the prediction error is less than [a certain value]. .

[0056] Step 15: Compare the current spatiotemporal temperature curve with the target process curve in real time, identify the time dimension deviation and spatial dimension deviation, and generate corresponding control commands for temperature control. The control command corresponding to the time dimension deviation is to adjust the heating power of the corresponding area, and the control command corresponding to the spatial dimension deviation is to adjust the fan speed. The control commands have a control upper limit designed based on the equipment model library.

[0057] Specifically, after obtaining the target process curve, the current spatiotemporal temperature curve can be compared with the target process curve in real time to identify time and spatial deviations, and corresponding control commands can be generated for temperature control. Among them, the control command corresponding to the time deviation is to adjust the heating power of the corresponding area (applied to the heating unit), and the control command for the spatial deviation is to adjust the fan speed (applied to the circulating fan). The control commands have a control upper limit designed based on the equipment model library.

[0058] In one example, we divide the furnace into multiple control zones (which can be divided according to the temperature measurement grid) based on the correspondence between each temperature measurement point in the furnace and the heating unit and circulating fan. In each control cycle, we determine the regional average temperature of each control zone at the current moment and the actual temperature change rate within the preset time window based on the current spatiotemporal temperature curve.

[0059] Regarding time dimension deviation, we extract the target temperature and target temperature change rate corresponding to the current moment from the target process curve. We compare the actual temperature change rate of each control region with the target temperature change rate to obtain the time dimension deviation for each control region. If the time dimension deviation of any control region exceeds a preset rate deviation threshold, we retrieve the heating response model and response lag time of the heating unit corresponding to that control region from the equipment model library. Based on the time dimension deviation, the current heating power, the heating response model, and the response lag time, we determine the heating power adjustment amount corresponding to that control region. And based on the direction of the time dimension deviation, we generate a heating power control command to increase or decrease the heating power of that control region.

[0060] Regarding spatial dimensional deviation, we compare the average temperature of each control zone with the target temperature to obtain the regional temperature deviation of each control zone. Based on the regional temperature deviation of each control zone and the temperature differences between different control zones, we determine the spatial dimensional deviation corresponding to each control zone. If the spatial dimensional deviation of any control zone exceeds a preset temperature deviation threshold, we retrieve the corresponding convective heat transfer model for that control zone from the equipment model library. We then input the spatial dimensional deviation of that control zone, the temperature differences between adjacent control zones, and the current fan speed into the convective heat transfer model to determine the fan speed adjustment amount that reduces the spatial dimensional deviation, and generate a fan speed control command to adjust the fan speed of that control zone.

[0061] After receiving the heating power control command and the fan speed control command, the heating power control command and the fan speed control command are sent to the corresponding heating unit and the circulating fan respectively to realize the temperature control of the furnace.

[0062] In one example, the control commands have a control upper limit designed based on the equipment model library. This embodiment obtains the rated power, allowable load rate, and maximum single-cycle power change of each heating unit from the equipment model library, and determines the upper limit of heating power control based on the rated power and allowable load rate. It also obtains the rated speed, allowable speed range, and maximum single-cycle speed change of each circulating fan from the equipment model library, and determines the upper limit of fan speed control based on the rated speed and allowable speed range.

[0063] If the heating power control command exceeds the heating power control limit, it is limited to the heating power control limit. Based on the maximum single-cycle power change, the change in heating power control command between adjacent control cycles is also limited, and the limited heating power control command is sent to the corresponding heating unit. Similarly, if the fan speed control command exceeds the fan speed control limit, it is limited to the fan speed control limit. Based on the maximum single-cycle speed change, the change in fan speed control command between adjacent control cycles is also limited, and the limited fan speed control command is sent to the corresponding circulating fan.

[0064] In one example, after completing the heat treatment of the current furnace, it is necessary to collect workpiece performance data, calculate temperature control effect indicators, store the spatiotemporal temperature curve, process parameters, and effect indicators of this furnace into a multimodal database, update the process knowledge base (such as correcting the optimal holding time for a certain material) and the equipment model library (such as adjusting the lag time of heating elements), and realize feedback self-optimization.

[0065] This embodiment proposes a temperature control method based on spatiotemporal curves and a multimodal database, which achieves the following improvements compared to traditional temperature control methods.

[0066] First, it significantly improves temperature uniformity. This embodiment collects temperature data from different spatial locations within the furnace using a multi-type sensor network, constructs a continuous spatiotemporal temperature curve, and adjusts the heating power and fan speed of the corresponding areas according to the time and spatial deviations. This effectively achieves precise zonal temperature control within the furnace, reduces the temperature difference between the upper and lower layers of the furnace, and between the center and the edge, and improves the workpiece hardness, diffusion depth, and consistency of metallographic structure.

[0067] Secondly, it significantly shortens the process debugging cycle. In this embodiment, based on the material type, size, furnace loading, and process requirements of the workpiece to be processed, the basic process curve is quickly matched from the process knowledge base, and calibration is performed using temperature curves from similar previous furnaces, actual process parameters, and workpiece performance data. This reduces reliance on manual experience and repeated furnace trials, improves the accuracy of process parameter settings, and shortens the debugging cycle.

[0068] Third, energy consumption is significantly reduced. This embodiment uses an LSTM prediction model to predict the temperature change trend inside the furnace and implements control before the temperature deviation expands, avoiding the lag adjustment in traditional control methods. At the same time, the heating power and fan speed are adjusted according to the actual temperature state of different areas to avoid overall overheating due to local low temperature, reducing energy waste caused by ineffective heating, excessive heat preservation, and rework.

[0069] Fourth, it possesses adaptive and evolvable capabilities. After each furnace cycle, this embodiment can store the actual spatiotemporal temperature curves, control parameters, equipment status, and workpiece performance data into a multimodal database, and update the process knowledge base and equipment model library accordingly. This allows the system to continuously correct process parameters and equipment response models, adapting to parameter changes caused by different materials, workpiece specifications, furnace loading conditions, and equipment aging, and maintaining high temperature control accuracy over the long term.

[0070] The steps described above are merely for clarity in describing the technical solution. In actual implementation, they can be combined into one step, or certain steps can be broken down into multiple steps, as long as they involve the same logical relationship, they are all within the scope of protection of this application. Any insignificant modifications or designs added to the algorithm or process, as long as they do not change the core of the algorithm or process, are also within the scope of protection of this application.

[0071] In one embodiment, we will provide a detailed description of the temperature control method based on spatiotemporal curves and multimodal database proposed in this application, taking into account a specific application scenario (gear carburizing and quenching).

[0072] The workpiece to be processed is a 20CrMnTi automotive gear (material type: 20CrMnTi, workpiece dimensions: 50mm×100mm, furnace load: 50 pieces / furnace), process requirements: carburized layer depth 0.8mm to 1.0mm, surface hardness HRC58 to HRC62, core hardness HRC30 to HRC35.

[0073] 1) Data perception and spatiotemporal curve construction.

[0074] The furnace is equipped with 32 K-type thermocouples (covering Z-axis 0mm to 1500mm, X-axis -500mm to 500mm, and Y-axis -500mm to 500mm) and 8 infrared thermometers (monitoring the workpiece surface), with a sampling frequency of 5Hz. Upon startup, the temperature at each measurement point is collected in real time to construct an initial spatiotemporal matrix and generate a dynamic thermal map, i.e., a spatiotemporal temperature curve, showing that the temperature in the upper layer of the furnace is slightly higher than that in the lower layer (with a deviation of approximately...). ).

[0075] 2) Database matching and process calibration.

[0076] After inputting the workpiece parameters, the system retrieves the standard curve for "20CrMnTi carburizing" from the process knowledge base. "Carbon potential 1.0%". Further searching of historical data revealed that three months prior, the same model of gear experienced a lower average temperature in the lower layer during the insulation stage due to aging of the furnace door seal. The final hardness deviation was HRC±2. Therefore, the insulation temperature for this batch of processes was slightly adjusted to... This compensates for heat loss.

[0077] 3) Temperature field prediction and deviation diagnosis.

[0078] The LSTM prediction model is based on the current temperature field ( Based on the temperature (t=1h) and heating power (80% of rated value), the temperature field is predicted for the next 30 minutes, and the upper layer will reach... (Exceeded target) ), lower level only (below target) By combining the real-time data from the infrared thermometer, the existence of the deviation was confirmed.

[0079] 4) Closed-loop control execution.

[0080] To address the issue of overheating in the upper layer, the power of the upper heating wire is reduced by 10% (from 80% to 70%). To address the issue of lag in the lower layer, the speed of the lower fan is increased by 25% (from 1200rpm to 1500rpm) to enhance circulation. At the same time, the furnace door status is monitored (no abnormalities are found) to prevent additional heat leakage.

[0081] 5) Effectiveness evaluation and database update: After carburizing, the depth of the carburized layer (0.9±0.05mm) and surface hardness (HRC59 to HRC61) of the workpiece were measured, with a 100% pass rate. The system stored the spatiotemporal curve, control parameters (upper layer power 70%, lower layer fan 1500rpm), and effect indicators into the database, and updated the insulation temperature in the "20CrMnTi gear" process template. The hysteresis time of the heating element has been corrected to 4.8 min.

[0082] Another embodiment of this application proposes a temperature control system based on spatiotemporal curves and a multimodal database, suitable for heat treatment furnaces. The details of the temperature control system based on spatiotemporal curves and a multimodal database proposed in this embodiment are described below. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0083] Figure 2This is a schematic diagram of a temperature control system based on spatiotemporal curves and a multimodal database proposed in this embodiment, including: a spatiotemporal temperature curve acquisition module 21, a multimodal database construction module 22, a process parameter correction module 23, a heating execution module 24, a target prediction module 25, and a control execution module 26.

[0084] The spatiotemporal temperature curve acquisition module 21 is used to arrange a network of multiple types of sensors along the height, width and depth of the furnace in the heat treatment furnace, collect temperature data at each spatial location, record auxiliary parameters synchronously, align them according to timestamps to construct a temperature field, and complete the temperature distribution of the unseen areas through interpolation algorithms to form a spatiotemporal temperature curve; among which, the auxiliary parameters include heating power, fan speed, furnace door opening and closing status and ambient temperature.

[0085] The multimodal database construction module 22 is used to construct a multimodal database that includes a process knowledge base, a historical data pool, and an equipment model library; among which, the historical data pool stores the spatiotemporal temperature curves of past furnace cycles.

[0086] The process parameter correction module 23 is used to obtain the workpiece parameters and process requirements of the workpiece to be processed, retrieve the matching basic process curve from the process knowledge base based on the workpiece parameters and process requirements, and correct the process parameters of the workpiece to be processed in combination with the historical data pool.

[0087] The heating execution module 24 is used to place the workpiece to be processed into the heat treatment furnace and start heating.

[0088] The target prediction module 25 is used to input the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power and fan speed into the pre-trained LSTM prediction model after the preset control start time is reached, so as to obtain the target process curve.

[0089] The control execution module 26 is used to compare the current spatiotemporal temperature curve with the target process curve in real time, identify time dimension deviation and spatial dimension deviation, and generate corresponding control commands for temperature control. Among them, the control command corresponding to the time dimension deviation is to adjust the heating power of the corresponding area, and the control command for the spatial dimension deviation is to adjust the fan speed. The control commands have a control upper limit designed based on the equipment model library.

[0090] It is worth noting that all modules involved in this embodiment are logical modules. In practical applications, a logical module can be a physical module, a part of a physical module, or an organic combination of multiple physical modules. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce modules that are not closely related to solving the technical problems proposed in this application. However, this does not mean that other modules are absent from this embodiment.

[0091] It is not difficult to see that this embodiment is a system embodiment corresponding to the above method embodiments, and this embodiment can be implemented in conjunction with the above method embodiments. The relevant technical details and technical effects mentioned in the above method embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above method embodiments.

[0092] Another embodiment of this application provides an electronic device, such as Figure 3 As shown, it includes a processor 31 and a memory 32. The memory 32 stores instructions that the processor 31 can execute. When the processor 31 is configured to execute the instructions, the electronic device can implement a temperature control method based on spatiotemporal curves and a multimodal database as described in the above method embodiment.

[0093] The memory and processor are connected via a bus, which includes any number of interconnecting buses and bridges. The bus can connect various circuits of one or more processors and memories, as well as other circuits such as peripherals, voltage regulators, and power management circuits—all well-known in the art and therefore not described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which also receives and transmits data to the processor.

[0094] The processor manages the bus and handles general processing, providing various functions, including but not limited to timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory, on the other hand, is used to store data used by the processor during operation.

[0095] Another embodiment of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, can implement a temperature control method based on spatiotemporal curves and a multimodal database as described in the above method embodiments.

[0096] That is, those skilled in the art will understand that all or part of the steps in the above method embodiments can be implemented by a program instructing related hardware. The program is stored in a storage medium and includes several instructions to cause a device (such as a microcontroller, chip, etc.) or processor to execute all or part of the steps of the method described in the method embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0097] It will be understood by those skilled in the art that the above embodiments are specific implementations of this application, and various changes in form and detail can be made in practical applications without departing from the spirit and scope of this application. For those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A temperature control method based on spatiotemporal curves and multimodal databases, suitable for a heat treatment furnace, characterized in that, The method comprises: Arranging a multi-type sensor network in the height, width and depth directions of the hearth in the heat treatment furnace, collecting temperature data at each spatial position, synchronously recording auxiliary parameters, aligning by time stamp to construct a temperature field, and completing the temperature distribution of the un-distributed area by an interpolation algorithm to form a space-time temperature curve; wherein the auxiliary parameters include heating power, fan speed, furnace door opening and closing state, and ambient temperature; A multi-modal database containing a process knowledge base, a historical data pool and a device model library is constructed; wherein the historical data pool stores the space-time temperature curve of the past furnace; The workpiece parameters and process requirements of the workpiece to be processed are obtained, the matched basic process curve is retrieved from the process knowledge base, and the process parameters of the workpiece to be processed are corrected in combination with the historical data pool; The workpiece to be processed is placed in the heat treatment furnace and heating is started, and after reaching the preset control start time, the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power and fan speed are input into the pre-trained LSTM prediction model to obtain the target process curve; The current space-time temperature curve is compared with the target process curve in real time, the time dimension deviation and the space dimension deviation are identified, and the corresponding control instructions are generated for temperature control; wherein the control instruction corresponding to the time dimension deviation is to adjust the heating power of the corresponding area, and the control instruction of the space dimension deviation is to adjust the fan speed, and the control instruction has a control upper limit designed based on the device model library.

2. The temperature control method based on spacetime curve and multi-modal database according to claim 1, wherein, The multi-type sensor network is arranged in the height, width and depth directions of the hearth in the heat treatment furnace, and the temperature data at each spatial position is collected, and the auxiliary parameters are synchronously recorded, comprising: In the heat treatment furnace, at least 5 temperature measuring layers are arranged in the height direction of the hearth, at least 3 temperature measuring points are arranged in the width direction of each layer, and at least 3 temperature measuring points are arranged in the depth direction of each layer, forming a uniformly distributed temperature measuring grid composed of at least 45 temperature measuring points; A K-type thermocouple and a non-contact infrared temperature measuring instrument are arranged at each temperature measuring point at the same time, the K-type thermocouple is used to accurately measure the contact point temperature, the non-contact infrared temperature measuring instrument is used to measure the surface temperature, and the two are complementary to correct the temperature measurement error, thereby obtaining the temperature data at each spatial position, and the auxiliary parameters are synchronously recorded; wherein the collection frequency of the auxiliary parameters is synchronized with the collection frequency of the temperature data; The temperature data at each spatial position is aligned by time stamp to construct a temperature field matrix, which is represented as: The multi-modal database containing a process knowledge base, a historical data pool and a device model library is constructed, comprising: ； wherein, is a temperature field matrix constructed, , , respectively represent a width direction, a depth direction, a height direction of the hearth, represents time, , , respectively represent the number of temperature measurement points in the width direction, the depth direction, the height direction, represents temperature data of a temperature measurement point at a th width position, a th depth position, a th height position at a th time. For the temperature field matrix , the Kriging interpolation algorithm is used to complete the temperature distribution of the un-distributed area of the furnace, forming the space-time temperature curve.

3. The temperature control method based on spacetime curves and multimodal databases according to claim 2, characterized in that, ​ The workpiece parameters, process requirements and standard process data corresponding to different workpieces are acquired, a basic process curve is drawn based on the standard process data, a mapping relationship between the workpiece parameters and process requirements and the basic process curve is established, and a process knowledge base is formed; wherein, the workpiece parameters include material type, workpiece size and furnace charging amount, the process requirements are performance indexes that need to be reached after the workpiece processing is completed, and the performance indexes at least include at least one of surface hardness, core hardness, case depth and allowable deformation amount, and the standard process data includes target temperature, stage duration, target temperature change rate and temperature threshold value corresponding to each of the heating-up stage, holding stage and cooling stage; The workpiece parameters, process requirements, actual execution process parameters, space-time temperature curve, auxiliary parameters, equipment state and workpiece performance data corresponding to multiple past furnace batches are acquired, the data corresponding to the same past furnace batch are associated according to the furnace batch identification, and the space-time temperature curve, auxiliary parameters and equipment state are aligned according to the time stamp, so as to form a historical data pool; The equipment identification, heat conduction parameter, convection heat transfer parameter and dynamic response parameter of the heat treatment furnace are acquired, a heating response model between the regional heating power and temperature change, a convection heat transfer model between the fan speed and the temperature distribution in the furnace and a furnace body heat conduction model are established, and the heating response model, convection heat transfer model, furnace body heat conduction model and equipment identification are stored in association, so as to form an equipment model library; The association index is established based on the material type, workpiece size, furnace charging amount and equipment identification, the process knowledge base, historical data pool and equipment model library are associated through the association index, and a multi-modal database is obtained.

4. The temperature control method based on spacetime curve and multi-modal database according to claim 3, wherein, The process parameters and process requirements of the workpiece to be processed are determined, the matched basic process curve is retrieved from the process knowledge base, and the process parameters are corrected in combination with the historical data pool, including: The workpiece parameters and process requirements of the workpiece to be processed are acquired, and the basic process curve matched with the workpiece to be processed is retrieved from the process knowledge base according to the workpiece parameters and process requirements; The target temperature, stage duration, target temperature change rate and temperature threshold value corresponding to each process stage are extracted from the basic process curve as the initial process parameters of the workpiece to be processed; The similar past furnace batches matched with the workpiece to be processed are screened out from the historical data pool by taking the same material type, workpiece size difference less than 1% and furnace charging amount difference less than the preset furnace charging amount threshold value as the retrieval conditions; The space-time temperature curve, actual execution process parameters and workpiece performance data of each similar past furnace batch are acquired, the space-time temperature curves of each similar past furnace batch and the basic process curve matched with the workpiece to be processed are aligned according to the process stages, and the temperature deviation of the actual temperature relative to the target temperature, the time deviation of the actual stage duration relative to the target stage duration and the rate deviation of the actual temperature change rate relative to the target temperature change rate of each process stage are calculated respectively; Based on the temperature deviation, time deviation, and rate deviation corresponding to each similar previous furnace, as well as the differences between workpiece performance data and process requirements, the correction amounts corresponding to the target temperature, stage duration, and target temperature change rate are determined, and the initial process parameters are corrected to obtain the corrected process parameters.

5. The temperature control method based on spacetime curve and multimodal database according to claim 4, wherein, The process involves inputting the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power, and fan speed into a pre-trained LSTM prediction model to obtain the target process curve, including: Starting from a preset control start time, the basic process curve is discretized according to a preset time interval to obtain multiple prediction times and the basic target temperature corresponding to each prediction time. Then, a corresponding process stage identifier is added according to the process stage at which each prediction time is located. Based on the process requirements and the corrected process parameters, the target temperature, stage duration, target temperature change rate and temperature threshold corresponding to each process stage are extracted and combined to form a process constraint feature sequence. Extract temperature data from each spatial location within a preset sampling time window ending at the current time from the current temperature field matrix, and align it with the heating power and fan speed at the corresponding sampling time according to the timestamp to form a furnace condition feature sequence. The basic target temperature, process constraint feature sequence, and furnace condition feature sequence are concatenated in chronological order to form the input feature sequence of the LSTM prediction model. The LSTM prediction model is used to extract the temperature change features between adjacent sampling times and the temperature change trends between different process stages in the input feature sequence, so as to obtain the furnace temperature corresponding to each prediction time. Based on the difference between the furnace temperature corresponding to each prediction time and the corresponding basic target temperature, the temperature correction amount corresponding to each prediction time is determined. The base target temperature is corrected by the temperature correction amount corresponding to each prediction time. Based on the process constraint feature sequence, the corrected base target temperature at each prediction time is constrained so that the rate of temperature change between adjacent prediction times does not exceed the rate of temperature change of the target temperature of the corresponding process stage, and the target temperature at each prediction time is within the range of the corresponding temperature threshold. By connecting the corrected target temperatures at each predicted time point in chronological order, the target process curve is obtained.

6. The temperature control method based on spacetime curve and multi-modal database according to claim 5, wherein, The real-time comparison of the current spatiotemporal temperature curve with the target process curve, identification of time dimension deviation and spatial dimension deviation, and generation of corresponding control commands for temperature control include: According to the correspondence between each temperature measuring point in the furnace and the heating unit and circulating fan, the furnace is divided into multiple control areas. In each control cycle, the average temperature of each control area at the current moment and the actual temperature change rate within the preset time window are determined based on the current spatiotemporal temperature curve. Extract the target temperature and target temperature change rate corresponding to the current moment from the target process curve, compare the actual temperature change rate of each control region with the target temperature change rate, and obtain the time dimension deviation of each control region. The average temperature of each control area is compared with the target temperature to obtain the temperature deviation of each control area. Based on the temperature deviation of each control area and the temperature difference between different control areas, the spatial dimension deviation of each control area is determined. When the time dimension deviation of any control area exceeds the preset rate deviation threshold, the heating response model and response lag time of the heating unit corresponding to the control area are obtained from the equipment model library. Based on the time dimension deviation, the current heating power, the heating response model and the response lag time, the heating power adjustment amount corresponding to the control area is determined, and a heating power control command to increase or decrease the heating power of the control area is generated according to the direction of the time dimension deviation. When the spatial dimension deviation of any control area exceeds the preset temperature deviation threshold, the convective heat transfer model corresponding to the control area is obtained from the equipment model library. The spatial dimension deviation of the control area, the temperature difference between adjacent control areas and the current fan speed are input into the convective heat transfer model to determine the fan speed adjustment amount to reduce the spatial dimension deviation and generate a fan speed control command to adjust the fan speed of the control area. Heating power control commands and fan speed control commands are sent to the corresponding heating units and circulating fans respectively to achieve temperature control of the furnace.

7. The temperature control method based on spacetime curves and multimodal databases according to claim 6, wherein, The upper limit of control commands is obtained through the following steps: Obtain the rated power, allowable load rate, and maximum single-cycle power change of each heating unit from the equipment model library, and determine the upper limit of heating power control based on the rated power and allowable load rate; Obtain the rated speed, allowable speed range, and maximum single-cycle speed change of each circulating fan from the equipment model library, and determine the upper limit of fan speed control based on the rated speed and allowable speed range; The step of sending heating power control commands and fan speed control commands to the corresponding heating units and circulating fans respectively to achieve temperature control of the furnace includes: If the heating power control command exceeds the heating power control limit, it will be limited to the heating power control limit. Based on the maximum single-cycle power change, the change in heating power control command between adjacent control cycles will be limited, and the limited heating power control command will be sent to the corresponding heating unit. If the fan speed control command exceeds the upper limit of the fan speed control, it will be limited to the upper limit of the fan speed control. Based on the maximum single-cycle speed change, the change of the fan speed control command between adjacent control cycles will be limited, and the limited fan speed control command will be sent to the corresponding circulating fan.

8. A temperature control system based on spatiotemporal curves and multimodal databases, suitable for heat treatment furnaces, characterized by, The system includes: The spatiotemporal temperature curve acquisition module is used to arrange a network of multiple types of sensors along the height, width, and depth of the furnace in the heat treatment furnace to collect temperature data at each spatial location, synchronously record auxiliary parameters, align them according to timestamps to construct a temperature field, and use interpolation algorithms to complete the temperature distribution in areas without sensors to form a spatiotemporal temperature curve; among which, auxiliary parameters include heating power, fan speed, furnace door opening and closing status, and ambient temperature. The multimodal database construction module is used to build a multimodal database that includes a process knowledge base, a historical data pool, and an equipment model library; among which, the historical data pool stores the spatiotemporal temperature curves of past furnace cycles; The process parameter correction module is used to obtain the workpiece parameters and process requirements of the workpiece to be processed, retrieve the matching basic process curve from the process knowledge base based on the workpiece parameters and process requirements, and correct the process parameters of the workpiece to be processed in combination with the historical data pool. The heating execution module is used to place the workpiece to be processed into the heat treatment furnace and start heating; The target prediction module is used to input the basic process curve, process requirements, corrected process parameters, current temperature field matrix, heating power and fan speed into the pre-trained LSTM prediction model after the preset control start time is reached, so as to obtain the target process curve. The control execution module is used to compare the current spatiotemporal temperature curve with the target process curve in real time, identify time dimension deviation and spatial dimension deviation, and generate corresponding control commands for temperature control. Among them, the control command corresponding to the time dimension deviation is to adjust the heating power of the corresponding area, and the control command corresponding to the spatial dimension deviation is to adjust the fan speed. The control commands have a control upper limit designed based on the equipment model library.

9. An electronic device, comprising: include: A processor and a memory, wherein the memory stores instructions executable by the processor, and the processor is configured to, when executing the instructions, enable the electronic device to implement a temperature control method based on spatiotemporal curves and a multimodal database as described in any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it can implement a temperature control method based on spatiotemporal curves and a multimodal database as described in any one of claims 1 to 7.

Images