A system and method for monitoring a quenching process of a metal product

Abstract

The application provides a metal product quenching process monitoring system and method, relates to the field of metal processing, and obtains a finished metal by obtaining a forged metal and preheating, quenching heating, quenching cooling, low-temperature tempering and cleaning and rust-proof treatment of the metal. The quenching cooling needs to obtain a time queue of real-time metal temperature, real-time cooling medium temperature and real-time cooling medium pressure, then perform time sequence implicit coding, feature splicing and bidirectional interaction response representation on input data, automatically recommend a cooling medium flow rate for comparison with a real-time flow rate, and generate a cooling medium flow rate adjustment instruction. In this way, the adaptability and dynamic response capability of the system can be improved, the stability and consistency of the cooling effect can be ensured, the cooling conditions can be adjusted in a timely manner, and the internal stress problem caused by inconsistent cooling speeds can be avoided.

Description

Technical Field

[0001] This application relates to the field of metal processing, and more particularly to a monitoring system and method for the quenching process of metal products in the embodiments of this application. Background Technology

[0002] Metal quenching is a heat treatment process that alters the microstructure of a metallic material through rapid cooling, thereby improving its hardness, strength, and other mechanical properties. During quenching, precise control of the cooling rate and uniformity is crucial for ensuring product quality.

[0003] Patent CN119162432A discloses a quenching process for forged round steel, which includes: preheating treatment; quenching heating to 840-860℃; cooling in rapid quenching oil at 40-80℃ and stirring until the surface temperature reaches 150-200℃; followed by low-temperature tempering at 180-200℃ for 2-3 hours to reduce internal stress and improve toughness; and cleaning and rust prevention treatment to ensure storage and transportation quality.

[0004] The patented method for controlling the cooling of heated metal relies primarily on experience and preset parameters. This approach has several drawbacks. Firstly, it lacks adaptability and dynamic response, failing to adjust flexibly based on real-time changes in temperature and pressure during cooling, resulting in unstable cooling effects and difficulty in ensuring consistency and repeatability. Secondly, due to its non-real-time nature and fixed parameter configuration, the system struggles to respond promptly to rapidly changing cooling conditions. This not only leads to uneven cooling but also causes inconsistent cooling rates across different parts of the metal, potentially inducing internal stress.

[0005] Therefore, an optimized monitoring scheme for the quenching process of metal products is desired. Summary of the Invention

[0006] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a monitoring system and method for the quenching process of metal products. This system acquires forged metal and performs preheating, quenching heating, quenching cooling, low-temperature tempering, and cleaning and rust prevention treatments to obtain the finished metal. The quenching cooling process requires acquiring a time queue of real-time metal temperature, real-time cooling medium temperature, and real-time cooling medium pressure. Then, the input data undergoes temporal implicit encoding, feature concatenation, and bidirectional interactive response representation to automatically recommend a cooling medium flow rate, which is compared with the real-time flow rate to generate a cooling medium flow rate adjustment command. This improves the system's adaptability and dynamic response capability, ensures stable and consistent cooling effects, and allows for timely adjustments to cooling conditions, avoiding internal stress problems caused by inconsistent cooling rates.

[0007] According to one aspect of this application, a method for monitoring the quenching process of metal products is provided, comprising: step S1: placing forged metal into a heating furnace for preheating treatment to obtain preheated metal; step S2: transferring the preheated metal to a salt bath furnace or a controlled atmosphere furnace for quenching heating to obtain heated metal; step S3: subjecting the heated metal to quenching and cooling treatment to obtain cooled metal; step S4: placing the cooled metal into a tempering furnace for low-temperature tempering treatment to obtain tempered metal; step S5: subjecting the tempered metal to cleaning and rust prevention treatment to obtain finished metal; wherein, step S3 includes:

[0008] The system acquires a time queue of real-time metal temperature collected by the first temperature sensor, and a time queue of real-time cooling medium temperature and real-time cooling medium pressure collected by the second temperature sensor and pressure sensor.

[0009] Sequence encoding is performed on the time queues of the real-time metal temperature, the real-time temperature of the cooling medium, and the real-time pressure of the cooling medium to obtain the time-series characteristics of the metal temperature, the cooling medium temperature, and the cooling medium pressure.

[0010] The time-series characteristics of the cooling medium temperature and the time-series characteristics of the cooling medium pressure are combined to obtain the joint time-series characteristics of the cooling medium temperature-pressure.

[0011] The time-series characteristics of the metal temperature and the combined time-series characteristics of the cooling medium temperature-pressure are encoded into a bidirectional interactive response between the quenching object and the quenching medium to obtain the time-series interactive response characteristics of the quenching object and the quenching medium parameters.

[0012] Based on the time-series interactive response characteristics of the quenching object and quenching medium parameters, a cooling medium flow rate adjustment command is generated.

[0013] According to another aspect of this application, a monitoring system for the quenching process of metal products is provided, comprising:

[0014] The metal product quenching treatment data acquisition module is used to acquire the time queue of real-time metal temperature collected by the first temperature sensor, and the time queue of real-time temperature and real-time pressure of the cooling medium collected by the second temperature sensor and the pressure sensor.

[0015] The metal product quenching data sequence encoding module is used to perform sequence encoding on the time queue of the real-time metal temperature, the time queue of the real-time temperature of the cooling medium, and the time queue of the real-time pressure of the cooling medium to obtain the time sequence characteristics of metal temperature, the time sequence characteristics of cooling medium temperature, and the time sequence characteristics of cooling medium pressure.

[0016] a cooling medium temperature-pressure feature joint module configured to jointly the cooling medium temperature time-series feature and the cooling medium pressure time-series feature to obtain a cooling medium temperature-pressure time-series joint feature;

[0017] a quenching object-quenching medium bidirectional interaction response encoding module configured to encode the metal temperature time-series feature and the cooling medium temperature-pressure time-series joint feature in a quenching object-quenching medium bidirectional interaction response to obtain a quenching object-quenching medium parameter time-series interaction response feature;

[0018] a cooling medium flow rate adjustment instruction generation module configured to generate a cooling medium flow rate adjustment instruction based on the quenching object-quenching medium parameter time-series interaction response feature.

[0019] Compared with the prior art, the metal product quenching process monitoring system and method provided by the application can obtain a forged metal, preheat the metal, quench and heat the metal, quench and cool the metal, low-temperature temper the metal, and clean and rust-proof the metal to obtain a finished metal. The quenching and cooling requires obtaining a time queue of real-time metal temperature, real-time cooling medium temperature, and real-time cooling medium pressure, and then performing time-series implicit coding, feature splicing, and bidirectional interaction response representation on the input data to automatically recommend a cooling medium flow rate for comparison with a real-time flow rate to generate a cooling medium flow rate adjustment instruction. In this way, the adaptability and dynamic response capability of the system can be improved, the stability and consistency of the cooling effect can be ensured, and the cooling conditions can be adjusted in a timely manner to avoid internal stress problems caused by inconsistent cooling speeds. BRIEF DESCRIPTION OF DRAWINGS

[0020] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application, when taken in conjunction with the accompanying drawings. The drawings provided in the present application are used to provide further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 is a flowchart of a metal product quenching process monitoring method according to an embodiment of the present application.

[0022] Figure 2 is a flowchart of quenching and cooling the heated metal to obtain a cooled metal in a metal product quenching process monitoring method according to an embodiment of the present application.

[0023] Figure 3A data flow diagram for quenching and cooling the heated metal to obtain a cooled metal in the metal product quenching process monitoring method according to the embodiments of the present application.

[0024] Figure 4 A flowchart for quenching object-quenching medium parameter time sequence interaction response feature coding in the metal product quenching process monitoring method according to the embodiments of the present application.

[0025] Figure 5 A flowchart for generating cooling medium flow rate adjustment instructions based on the quenching object-quenching medium parameter time sequence interaction response feature in the metal product quenching process monitoring method according to the embodiments of the present application.

[0026] Figure 6 A system block diagram of the metal product quenching process monitoring system according to the embodiments of the present application. DETAILED DESCRIPTION

[0027] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the drawings represent functionally identical or similar elements. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless specifically indicated.

[0028] Metal quenching is a heat treatment process that changes the microstructure of metal through rapid cooling, aiming to improve hardness and strength. Patent CN119162432A discloses a quenching process for forged round steel, involving preheating, heating to 840-860℃, rapid cooling in quenching oil, low-temperature tempering and cleaning and rust-proof treatment. However, this process relies on experience and preset parameters for cooling control, lacking self-adaptability and dynamic response capability. As it cannot adjust temperature and pressure changes in the cooling process in real time, the cooling effect is unstable, and internal stress is easily generated.

[0029] To solve the above technical problems, the present application proposes a metal product quenching process monitoring method. Figure 1 A flowchart of the metal product quenching process monitoring method according to the embodiments of the present application. As shown in Figure 1As shown, the metal product quenching process monitoring method according to an embodiment of this application includes: Step S1: placing the forged metal into a heating furnace for preheating treatment to obtain preheated metal; Step S2: transferring the preheated metal to a salt bath furnace or a controlled atmosphere furnace for quenching heating to obtain heated metal; Step S3: subjecting the heated metal to quenching and cooling treatment to obtain cooled metal; Step S4: placing the cooled metal into a tempering furnace for low-temperature tempering treatment to obtain tempered metal; Step S5: subjecting the tempered metal to cleaning and rust prevention treatment to obtain finished metal.

[0030] In the aforementioned method for monitoring the quenching process of metal products, step S1 involves placing the forged metal into a heating furnace for preheating treatment to obtain preheated metal. It should be understood that preheating the forged metal in a heating furnace reduces internal stress, homogenizes the microstructure, and prepares it for subsequent high-temperature quenching. Specifically, firstly, a suitable heating furnace needs to be selected. The choice of heating furnace depends on factors such as the type and size of the metal, and the required heating temperature. Generally, for large or complex-shaped metal parts, salt bath furnaces or controlled atmosphere furnaces are often used. This is because these two types of furnaces can provide a more uniform heating environment, helping to avoid localized overheating and thus ensuring the consistency of the internal microstructure of the metal. Next, the preheating temperature needs to be set. The selection of the preheating temperature is usually based on the specific type of metal being processed and its alloy composition. For example, for carbon steel, the preheating temperature is generally set between 500-700℃; while for some high-alloy steels, a higher preheating temperature, such as 800-900℃, may be required. This helps reduce thermal stress generated during subsequent quenching and promotes the homogenization of the material's internal structure, improving the mechanical properties of the final product. It's worth noting that the heating rate is crucial during the preheating process after forging. Excessive heating can lead to a large temperature difference between the metal surface and the center, causing internal stress. In practice, a slow and stable heating rate should be adopted, typically controlled at around 10-30°C per minute, which can be adjusted according to the specific metal material and size. Slow heating allows heat to gradually penetrate the metal, resulting in a more uniform temperature distribution throughout the workpiece. After preheating, the metal part is transferred to the next process—quenching. Before this, a thorough inspection is necessary to ensure no new defects have been introduced due to the preheating process. Through preheating, the potential of the metal material can be better utilized, producing higher-quality products.

[0031] In the above method for monitoring the quenching process of a metal product, the step S2 of transferring the preheated metal to a salt bath furnace or a controlled atmosphere furnace for quenching heating to obtain a heated metal. It should be understood that in the metal processing process, quenching heating is a key step to improve the hardness and strength of metal materials. This process requires the preheated metal to be transferred to a specially designed salt bath furnace or controlled atmosphere furnace for further processing. Specifically, first, the appropriate heating equipment needs to be selected. Among them, the salt bath furnace is widely used in the quenching process due to its uniform heating characteristics and good heat transfer performance. The salt bath furnace transfers heat through molten salt as a medium, which can provide a very uniform heating environment, which is very important to avoid local overheating and ensure the consistency of the metal internal structure. In addition, the controlled atmosphere furnace is also a commonly used device, which controls the gas composition (such as nitrogen, argon, etc.) in the furnace to reduce the occurrence of oxidation reactions, thereby protecting the metal surface from the influence of the oxide skin. These two devices have their own advantages, and the choice depends on the specific process requirements and metal materials. When the heating equipment is selected, the next step is to set the quenching heating temperature. The selection of quenching heating temperature mainly depends on the type of metal and its alloy composition. For example, for carbon steel, the quenching heating temperature is usually set between 840-860℃; while for some high alloy steel, higher heating temperature may be required. Accurate heating temperature not only helps to obtain the ideal hardness, but also promotes the transformation of the internal structure of the material and improves its mechanical properties. It is worth mentioning that after the quenching heating temperature is determined, appropriate holding time needs to be maintained. Holding time directly affects the degree of change of the internal structure of the metal. If the holding time is too short, it may not be able to fully achieve austenitization (i.e. transformation into high temperature phase), resulting in poor quenching effect; on the contrary, if the holding time is too long, it may cause unnecessary energy waste, and may also cause problems such as grain growth. Generally speaking, the holding time can be estimated according to the thickness of the metal piece, usually according to the principle of holding 1 hour per 25 millimeters of thickness. By accurately controlling the holding time, the internal structure of the metal can be fully transformed. In this way, through the quenching heating process, the internal structure of the metal material can be effectively improved, and its processing performance can be significantly improved.

[0032] Accordingly, after obtaining the forged metal and preheating and quenching heating, the metal needs to be quenched and cooled. In step S3: quenching and cooling the heated metal to obtain the cooled metal, the technical concept of the present application is to obtain a time queue of real-time metal temperature collected by the first temperature sensor, and obtain a time queue of real-time cooling medium temperature and a time queue of real-time cooling medium pressure collected by the second temperature sensor and the pressure sensor, use artificial intelligence-based data analysis and coding to perform time sequence implicit coding of the real-time metal temperature, the real-time cooling medium temperature and the real-time cooling medium pressure, then splice the coded cooling medium temperature time sequence features and cooling medium pressure time sequence features, and automatically recommend the cooling medium flow rate value according to the bidirectional interaction response representation between the cooling medium temperature-pressure time sequence joint features and the metal temperature time sequence features, and compare it with the real-time flow rate to generate a cooling medium flow rate adjustment instruction. The present application can automatically adjust the cooling strategy according to real-time changes. This greatly enhances the adaptability and dynamic response capability of the system, and ensures the stability and consistency of the cooling effect. At the same time, the system can respond to changes in cooling conditions in time, avoiding the problem of internal stress caused by inconsistent cooling speed.

[0033] Figure 2 A flow chart for quenching and cooling the heated metal to obtain the cooled metal in the metal product quenching process monitoring method according to the embodiment of the present application. Figure 3 A data flow diagram for quenching and cooling the heated metal to obtain the cooled metal in the metal product quenching process monitoring method according to the embodiment of the present application. As Figure 2 and Figure 3As shown, in the step S3, the quenched cooling treatment is performed on the heated metal to obtain a cooled metal, including: S31, acquiring a time queue of real-time metal temperature collected by the first temperature sensor, and acquiring a time queue of real-time cooling medium temperature and a time queue of real-time cooling medium pressure collected by the second temperature sensor and the pressure sensor; S32, respectively performing sequence coding on the time queue of real-time metal temperature, the time queue of real-time cooling medium temperature and the time queue of real-time cooling medium pressure to obtain a metal temperature time sequence feature, a cooling medium temperature time sequence feature and a cooling medium pressure time sequence feature; S33, jointly performing the cooling medium temperature time sequence feature and the cooling medium pressure time sequence feature to obtain a cooling medium temperature-pressure time sequence joint feature; S34, performing quenching object-quenching medium bidirectional interaction response coding on the metal temperature time sequence feature and the cooling medium temperature-pressure time sequence joint feature to obtain a quenching object-quenching medium parameter time sequence interaction response feature; and S35, generating a cooling medium flow rate adjustment instruction based on the quenching object-quenching medium parameter time sequence interaction response feature.

[0034] In the embodiment of the present application, the step S31, a time queue of real-time metal temperature collected by the first temperature sensor is acquired, and a time queue of real-time cooling medium temperature and a time queue of real-time cooling medium pressure collected by the second temperature sensor and the pressure sensor are acquired. It can be understood that the change of metal temperature directly affects the quality and precision of the processing process, so it is necessary to track and record its temperature change in real time. This data is continuously collected by the first temperature sensor, and a time sequence of temperature is generated in the form of time stamp. At the same time, the cooling medium temperature is too high or the pressure is abnormal, which may cause the equipment to overheat or malfunction, so the second temperature sensor and the pressure sensor will synchronously collect the temperature and pressure data of the cooling medium to form independent time queues. These data not only help to judge the timeliness of the cooling effect, but also provide basis for optimizing the cooling system. By synchronously collecting and storing these time sequence data, the system can understand the state of the metal and the cooling medium in real time, timely adjust the working parameters, ensure that the whole system runs in the best state, and avoid faults or damages caused by abnormal data.

[0035] In the embodiments of the present application, the step S32 of respectively performing sequence encoding on the time queue of the real-time metal temperature, the time queue of the real-time cooling medium temperature and the time queue of the real-time cooling medium pressure to obtain the metal temperature time sequence feature, the cooling medium temperature time sequence feature and the cooling medium pressure time sequence feature comprises: respectively performing sequence encoding based on an LSTM model on the time queue of the real-time metal temperature, the time queue of the real-time cooling medium temperature and the time queue of the real-time cooling medium pressure to obtain a metal temperature time sequence feature hidden code vector as the metal temperature time sequence feature, a cooling medium temperature time sequence feature hidden code vector as the cooling medium temperature time sequence feature and a cooling medium pressure time sequence feature hidden code vector as the cooling medium pressure time sequence feature. It can be understood that, considering that the real-time metal temperature, the real-time cooling medium temperature and the real-time cooling medium pressure change with time during the metal quenching process, there is a close relationship between the data at each time. For example, the metal temperature at the current time is affected by the temperature at the previous time and the state of the cooling medium. Therefore, in order to effectively capture these long-term dependencies, the present application performs sequence encoding based on an LSTM model on the time queue of the real-time metal temperature, the time queue of the real-time cooling medium temperature and the time queue of the real-time cooling medium pressure to capture and mine the implicit time sequence correlation between parameters at different time periods, to obtain a metal temperature time sequence feature hidden code vector, a cooling medium temperature time sequence feature hidden code vector and a cooling medium pressure time sequence feature hidden code vector. The gating mechanism (input gate, forget gate and output gate) of the LSTM model can filter and control the information flow, remember the long-term important information and forget the irrelevant information, so as to better represent the important features of each parameter, such as trend, periodicity and abnormal situation, which are crucial for understanding the quenching cooling process.

[0036] In the embodiments of the present application, the step S33 of combining the cooling medium temperature time sequence feature and the cooling medium pressure time sequence feature to obtain a cooling medium temperature-pressure time sequence combined feature includes: feature splicing the cooling medium temperature time sequence feature implicit coding vector and the cooling medium pressure time sequence feature implicit coding vector to obtain a cooling medium temperature-pressure time sequence combined feature implicit coding vector as the cooling medium temperature-pressure time sequence combined feature. It should be understood that, in the process of metal quenching, the temperature and pressure of the cooling medium do not act independently, but affect each other and cooperatively produce effects on metal cooling. For example, cooling medium pressure changes can change its flow rate and distribution, thereby affecting the heat transfer efficiency and causing temperature distribution to change; conversely, temperature changes can also affect the physical properties of the cooling medium and indirectly affect the pressure. Based on this, in the technical solutions of the present application, the cooling medium temperature time sequence feature implicit coding vector and the cooling medium pressure time sequence feature implicit coding vector are spliced to integrate the information of the two key factors and comprehensively reflect the comprehensive characteristics of the cooling medium in the time sequence to obtain the cooling medium temperature-pressure time sequence combined feature implicit coding vector. In this way, the heat exchange law between the cooling medium and the metal and the influence mechanism on the metal organization performance can be more effectively mined.

[0037] In the embodiments of the present application, the step S34 of quenching object-quenching medium bidirectional interaction response coding the metal temperature time sequence feature and the cooling medium temperature-pressure time sequence combined feature to obtain a quenching object-quenching medium parameter time sequence interaction response feature. It should be understood that, in the process of metal quenching, there is a strong interaction between the metal (quenching object) and the cooling medium (quenching medium). The temperature change of the metal changes the temperature distribution and flow state of the cooling medium, and conversely, the temperature, pressure and flow characteristics of the cooling medium directly affect the cooling rate of the metal and the final organization structure. This interaction is the core of the quenching process. Therefore, in order to more finely capture and mine this essential relationship and comprehensively describe the dynamic changes of the mutual influence between the two, the present application codes the metal temperature time sequence feature and the cooling medium temperature-pressure time sequence combined feature in a quenching object-quenching medium bidirectional interaction response to obtain a quenching object-quenching medium parameter time sequence interaction response feature. In this way, the interaction mechanism between the metal and the cooling medium in the entire quenching process can be deeply analyzed, and rich data support is provided for subsequent cooling medium flow rate generation.

[0038] Figure 4This is a flowchart illustrating the process of encoding the time-series characteristics of the metal temperature and the combined temperature-pressure time-series characteristics of the cooling medium into a bidirectional interactive response between the quenching object and the quenching medium, according to an embodiment of this application, to obtain the time-series interactive response characteristics of the quenching object-quenching medium parameters. For example... Figure 4 As shown, in an embodiment of this application, step S34, which encodes the metal temperature time series features and the cooling medium temperature-pressure time series joint features into a bidirectional interactive response feature between the quenching object and the quenching medium to obtain the quenching object-quenching medium parameter time series interactive response features, includes: S341, performing homography projection transformation on the implicit encoding vector of the metal temperature time series features and the implicit encoding vector of the cooling medium temperature-pressure time series joint features to obtain a metal temperature time series homography projection encoding vector and a cooling medium temperature-pressure time series joint homography projection encoding vector; S342, based on the metal temperature time series homography projection encoding vector and the cooling medium temperature-pressure time series joint homography projection encoding vector... A bidirectional attention balance field between the positive and negative quenching object-quenching medium parameters is formed. Feature modulation is applied to the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector to obtain the metal temperature time-series homography projection attention modulation coding vector and the cooling medium temperature-pressure time-series joint homography projection attention modulation coding vector; S343, interactive response processing is performed on the metal temperature time-series homography projection attention modulation coding vector and the cooling medium temperature-pressure time-series joint homography projection attention modulation coding vector to obtain the quenching object-quenching medium parameter time-series interactive response coding vector as the quenching object-quenching medium parameter time-series interactive response feature.

[0039] In an embodiment of this application, step S341, which involves performing a homography projection transformation on the metal temperature time-series feature latent encoding vector and the cooling medium temperature-pressure time-series joint feature latent encoding vector to obtain a metal temperature time-series homography projection encoding vector and a cooling medium temperature-pressure time-series joint homography projection encoding vector, includes: S3411, performing a homography projection transformation on the metal temperature time-series feature latent encoding vector using a metal temperature time-series mapping homography matrix to obtain the metal temperature time-series homography projection encoding vector; and S3412, performing a homography projection transformation on the cooling medium temperature-pressure time-series joint feature latent encoding vector using a cooling medium temperature-pressure time-series joint mapping homography matrix to obtain the cooling medium temperature-pressure time-series joint homography projection encoding vector.

[0040] Specifically, in the step S3411, the metal temperature time sequence feature implicit encoding vector is homographic projection transformed using the metal temperature time sequence mapping homographic matrix to obtain the metal temperature time sequence homographic projection encoding vector, expressed by a metal temperature homographic projection transformation formula as: ; wherein, is the metal temperature time sequence feature implicit encoding vector, is a metal temperature time sequence mapping homographic matrix, is a metal temperature time sequence homographic projection encoding vector. In this way, in the feature space, the relative positional relationship between features is maintained, but these features can be observed from different perspectives or scales.

[0041] Specifically, in the step S3412, the cooling medium temperature-pressure time sequence joint feature implicit encoding vector is homographic projection transformed using the cooling medium temperature-pressure time sequence joint mapping homographic matrix to obtain the cooling medium temperature-pressure time sequence joint homographic projection encoding vector, expressed by a cooling medium temperature-pressure homographic projection transformation formula as: ; wherein, is the cooling medium temperature-pressure time sequence joint feature implicit encoding vector, is a cooling medium temperature-pressure time sequence joint mapping homographic matrix, is a cooling medium temperature-pressure time sequence joint homographic projection encoding vector. In this way, the original cooling medium temperature-pressure time sequence joint feature implicit encoding vector is homographic projection transformed to map the cooling medium temperature-pressure time sequence joint feature implicit encoding vector to a new feature space.

[0042] In the embodiment of the present application, the step S342, the metal temperature time sequence homographic projection encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection encoding vector are feature modulated based on the positive-negative quenching object-quenching medium parameter bidirectional attention balance field between the metal temperature time sequence homographic projection encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection encoding vector to obtain a metal temperature time sequence homographic projection attention modulation encoding vector and a cooling medium temperature-pressure time sequence joint homographic projection attention modulation encoding vector, comprising: S3421, calculating the positive direction quenching object-quenching medium parameter attention score field of the metal temperature time sequence homographic projection encoding vector relative to the cooling medium temperature-pressure time sequence joint homographic projection encoding vector; S3422, calculating the reverse quenching object-quenching medium parameter attention score field of the cooling medium temperature-pressure time sequence joint homographic projection encoding vector relative to the metal temperature time sequence homographic projection encoding vector; S3423, constructing the positive-negative quenching object-quenching medium parameter bidirectional attention balance field between the positive direction quenching object-quenching medium parameter attention score field and the reverse quenching object-quenching medium parameter attention score field; S3424, mapping the metal temperature time sequence homographic projection encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection encoding vector to the positive-negative quenching object-quenching medium parameter bidirectional attention balance field respectively to obtain the metal temperature time sequence homographic projection attention modulation encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection attention modulation encoding vector.

[0043] Specifically, in the step S3421, the positive direction quenching object-quenching medium parameter attention score field of the metal temperature time sequence homographic projection encoding vector relative to the cooling medium temperature-pressure time sequence joint homographic projection encoding vector is calculated, which is expressed by a positive direction attention score field formula as: ; wherein, is the transpose vector of , is the length of , is the forward quenching object-quenching medium parameter attention score field. It should be understood that the calculation of the forward attention score field of the encoded vector after the homographic projection transformation evaluates the importance distribution of the metal temperature time sequence homographic projection encoded vector relative to the cooling medium temperature-pressure time sequence joint homographic projection encoded vector, which usually involves calculating the position-sensitive similarity score between the two encoded vectors to construct an importance map of the metal temperature time sequence feature implicit encoded vector in the background of the cooling medium temperature-pressure time sequence joint feature implicit encoded vector. The core of the forward attention mechanism is to quantify the correlation between the two encoded vectors, and such a mechanism simulates the characteristics of the human visual system. In this way, the model can "focus" on the part of the metal temperature time sequence feature implicit encoded vector that best reflects the correlation with the cooling medium temperature-pressure time sequence joint feature implicit encoded vector, while serving as a filter to filter out irrelevant noise information.

[0044] Specifically, in the step S3422, the reverse quenching object-quenching medium parameter attention score field of the cooling medium temperature-pressure time sequence joint homographic projection encoded vector relative to the metal temperature time sequence homographic projection encoded vector is calculated, which is represented by the reverse attention score field formula as follows: ; wherein, is the transpose vector of , is the length of , is the reverse quenching object-quenching medium parameter attention score field. It should be understood that in order to supplement the information of the forward attention score field, it is necessary to calculate the reverse attention score field of the cooling medium temperature-pressure time sequence joint homographic projection encoded vector relative to the metal temperature time sequence homographic projection encoded vector. In this way, the forward and reverse attention score fields work together to form a more complete and detailed relationship map, which helps to capture the asymmetric relationship. This process provides the opportunity to understand the relationship between features from the opposite direction, ensuring that important interaction details are not missed.

[0045] Specifically, in the step S3423, the forward quenching object-quenching medium parameter attention score field and the reverse quenching object-quenching medium parameter attention score field are constructed to form the forward and reverse quenching object-quenching medium parameter bidirectional attention balance field, which is represented by the forward and reverse bidirectional attention balance field formula as follows: ; wherein, is a feature concatenation operation, is a convolutional encoding with a convolution kernel of 3x3, is the forward-backward quenching object-quenching medium parameter bidirectional attention balance field. It should be understood that in order to integrate the information of the forward and backward attention score fields, the forward-backward quenching object-quenching medium parameter bidirectional attention balance field is constructed. The design of the balance field needs to consider the specific needs of the task, and dynamically adjusting the weight coefficient can automatically optimize the balance effect according to the characteristics of the input data, thereby improving the adaptability and generalization ability of the model. In an ideal case, the forward-backward quenching object-quenching medium parameter bidirectional attention balance field should complement each other rather than cancel each other out. This operation takes into account the mutual influence between the two feature vectors, ensuring that neither side is overly biased in the final interactive response encoding, nor is the contribution of either side ignored.

[0046] Specifically, in the step S3424, the metal temperature time sequence homographic projection encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection encoding vector are respectively mapped to the forward-backward quenching object-quenching medium parameter bidirectional attention balance field to obtain the metal temperature time sequence homographic projection attention modulation encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection attention modulation encoding vector, which are expressed by the homographic projection attention modulation formula as follows: ; wherein, is the metal temperature time sequence homographic projection attention modulation encoding vector, is the cooling medium temperature-pressure time sequence joint homographic projection attention modulation encoding vector. It should be understood that when the encoding vector is mapped to the aforementioned constructed forward-backward quenching object-quenching medium parameter bidirectional attention balance field, it is actually performing a reweighting operation. The modulated encoding vector not only contains the information of the original feature, but also integrates the influence from the feature of the other party. This process is similar to the feedback mechanism in human communication - one party will adjust their expression according to the other party's response. In this way, the model can understand and represent the complex interaction between features at a higher level of abstraction, ensuring that each encoding vector adjusts itself according to the guidance provided by the balance field to generate a new encoding vector.

[0047] Specifically, in the step S343, the metal temperature time sequence homographic projection attention modulation encoding vector and the cooling medium temperature-pressure time sequence joint homographic projection attention modulation encoding vector are subjected to interactive response processing to obtain a quenching object-quenching medium parameter time sequence interactive response encoding vector as the quenching object-quenching medium parameter time sequence interactive response feature, which is expressed by the per-position point division formula as follows: ; wherein, is the quenching object-quenching medium parameter time series interaction response encoding vector. It should be understood that the final quenching object-quenching medium parameter time series interaction response encoding vector is obtained by performing point-by-point division on the metal temperature time series projection attention modulation encoding vector and the cooling medium temperature-pressure time series joint projection attention modulation encoding vector. The point-by-point division operation is used here as a comparison mechanism to reveal the new relationship between the two feature vectors after modulation, providing an information-rich representation for subsequent tasks. This deep feature comparison and integration reveals subtle proportional relationships between features, which are crucial for interpreting model decisions.

[0048] Figure 5 A flowchart for generating a cooling medium flow rate adjustment instruction based on the quenching object-quenching medium parameter time series interaction response feature in the metal product quenching process monitoring method according to the embodiments of the present application is shown. As shown in Figure 5 the step S35 of generating a cooling medium flow rate adjustment instruction based on the quenching object-quenching medium parameter time series interaction response feature in the embodiments of the present application includes: S351, obtaining an optimization result based on the quenching object-quenching medium parameter time series interaction response encoding vector, the optimization result being a cooling medium flow rate recommended value; S352, generating the cooling medium flow rate adjustment instruction based on a comparison between the cooling medium flow rate recommended value and a real-time cooling medium flow rate value.

[0049] In the embodiments of the present application, the step S351 of obtaining an optimization result based on the quenching object-quenching medium parameter time series interaction response encoding vector includes: inputting the quenching object-quenching medium parameter time series interaction response encoding vector into a decoder-based cooling medium flow rate optimization module to obtain the optimization result. That is, the quenching object-quenching medium parameter time series interaction response feature obtained by bidirectional interaction of the metal temperature time series feature implicit encoding vector and the cooling medium temperature-pressure time series joint feature implicit encoding vector is decoded to learn the complex nonlinear relationship using the powerful nonlinear mapping capability of the decoder-based module. For example, the decoder can understand, through learning on a large amount of historical data, what cooling medium flow rate can achieve the best quenching effect under different quenching object-quenching medium parameter interactions, thereby accurately giving a flow rate recommended value.

[0050] In a preferred example, inputting the quenching object-quenching medium parameter time series interaction response encoding vector into a decoder-based cooling medium flow rate optimization module to obtain an optimization result includes:

[0051] The first characteristic value and the first distances between characteristic values distances, thereby obtaining a first quenched object-quenching medium parameter time series interaction response distance matrix and a second quenched object-quenching medium parameter time series interaction response distance matrix: ; wherein, represents the first characteristic value of the quenched object-quenching medium parameter time series interaction response encoding vector, represents the first characteristic value of the quenched object-quenching medium parameter time series interaction response encoding vector, represents distance, represents distance, represents the first characteristic value of the first quenched object-quenching medium parameter time series interaction response distance matrix, represents the first characteristic value of the second quenched object-quenching medium parameter time series interaction response distance matrix;

[0052] calculating a weighted sum of the first quenched object-quenching medium parameter time series interaction response distance matrix and the second quenched object-quenching medium parameter time series interaction response distance matrix, and determining each eigenvalue of the distance weighted sum matrix to , to arrange the each eigenvalue to obtain a quenched object-quenching medium parameter time series interaction response distance eigen vector: ; wherein, represents the first eigenvalue of the distance weighted sum matrix, represents the second eigenvalue of the distance weighted sum matrix, represents the first eigenvalue of the distance weighted sum matrix, represents a quenched object-quenching medium parameter time series interaction response distance eigen vector;

[0053] interpolating the quenched object-quenching medium parameter time series interaction response distance eigen vector to obtain a quenched object-quenching medium parameter time series interaction response interpolated eigen vector with the same length as the quenched object-quenching medium parameter time series interaction response encoding vector;

[0054] matrix multiplication of the quenching object-quenching medium parameter time sequence interaction response distance matrix and the quenching object-quenching medium parameter time sequence interaction response coding vector to obtain a quenching object-quenching medium parameter time sequence interaction response intermediate feature vector, and matrix multiplication of the self-association matrix of the quenching object-quenching medium parameter time sequence interaction response coding vector and the second quenching object-quenching medium parameter time sequence interaction response distance matrix, that is, matrix multiplication to obtain a quenching object-quenching medium parameter time sequence interaction response intermediate feature matrix, that is: ; wherein, the quenching object-quenching medium parameter time sequence interaction response coding vector, is the first quenching object-quenching medium parameter time sequence interaction response distance matrix, is the quenching object-quenching medium parameter time sequence interaction response intermediate feature vector, denotes the transposed vector of the vector, denotes matrix multiplication, is the second quenching object-quenching medium parameter time sequence interaction response distance matrix, denotes the quenching object-quenching medium parameter time sequence interaction response intermediate feature matrix;

[0055] After matrix multiplication of the quenching object-quenching medium parameter time sequence interaction response intermediate feature vector and the quenching object-quenching medium parameter time sequence interaction response intermediate feature matrix, further dot multiplication with the quenching object-quenching medium parameter time sequence interaction response interpolation eigen vector is performed to obtain an optimized quenching object-quenching medium parameter time sequence interaction response coding vector.

[0056] The optimized quenching object-quenching medium parameter time sequence interaction response coding vector is input into the decoder-based cooling medium flow rate optimization module to obtain an optimization result.

[0057] Here, in the case that the metal temperature time sequence feature implicit coding vector and the cooling medium temperature-pressure time sequence joint feature implicit coding vector respectively represent the short-range-long-range bidirectional time sequence correlation features of the real-time metal temperature and the spliced time sequence correlation features of the real-time cooling medium temperature and real-time pressure, when the quenching object-quenching medium bidirectional interaction response coding is performed, the insufficient correlation correspondence of the positive and negative attention field reinforcement caused by the difference in the dimension of the source time sequence distribution feature sample will cause the sparse correlation interaction response of the quenching object-quenching medium parameter time sequence interaction response coding vector, thereby reducing the accuracy of the optimization result obtained by inputting the decoder-based cooling medium flow rate optimization module due to the lack of decoding inference degree.

[0058] Therefore, the distance matrix and the two-distance matrix of the quenching object-quenching medium parameter time sequence interaction response coding vector are taken as the fine-grained metric correlation cluster representation of the quenching object-quenching medium parameter time sequence interaction response coding vector, the dynamic programming of the inter-association cluster relationship of different association clusters is performed on the quenching object-quenching medium parameter time sequence interaction response coding vector and its self-association representation, respectively, to simulate the neuron cluster-based sparse activation of the association system, and the fine-grained predictable sparsity of the quenching object-quenching medium parameter time sequence interaction response coding vector is matched by using the metric correlation cluster eigen-representation of the distance matrix and the two-distance matrix of the quenching object-quenching medium parameter time sequence interaction response coding vector, so as to avoid the lack of association caused by sparsity affecting the missing of decoding reasoning degree, and improve the accuracy of the optimization result of the quenching object-quenching medium parameter time sequence interaction response coding vector input based on the decoder-based cooling medium flow rate optimization module.

[0059] Specifically, in the step S352, the cooling medium flow rate adjustment instruction is generated based on the comparison between the cooling medium flow rate recommended value and the cooling medium real-time flow rate value. That is, by comparing the cooling medium flow rate recommended value and the real-time flow rate value, the deviation can be found in time, and the basis for generating the adjustment instruction is provided, so as to realize real-time feedback adjustment of the cooling medium flow rate and ensure that the quenching process is always in the best state. It should be understood that, in one specific embodiment of the present application, the flow rate of the cooling medium is monitored in real time by a sensor and compared with the recommended value set in advance. If the real-time flow rate is lower than the recommended value, the system will determine that the cooling effect may be insufficient and the flow rate needs to be increased; if the real-time flow rate is higher than the recommended value, the system will determine that the cooling is too fast and may affect the quenching quality, and the flow rate needs to be reduced. Next, the system generates specific adjustment instructions, such as increasing the pump speed or adjusting the valve opening, according to the direction and amplitude of the deviation, to indicate how the cooling system adjusts to restore to the ideal flow rate. These adjustment instructions can be executed by an automatic control system without human intervention, ensuring that the cooling medium flow rate is always maintained within the optimal range, thereby ensuring the stability of the quenching process and the product quality.

[0060] In summary, step S3 is explained as follows: It acquires the time-series data of real-time metal temperature collected by the first temperature sensor, and the time-series data of real-time cooling medium temperature and pressure collected by the second temperature sensor and pressure sensor. Using artificial intelligence-based data analysis and encoding methods, it performs temporal implicit encoding of the real-time metal temperature, the real-time cooling medium temperature, and the real-time cooling medium pressure. Then, it concatenates the encoded temporal features of the cooling medium temperature and pressure, automatically recommending a cooling medium flow rate value based on the bidirectional interactive response between the combined temporal features of the cooling medium temperature-pressure and the metal temperature, and comparing this value with the real-time flow rate to generate a cooling medium flow rate adjustment command. This application can automatically adjust the cooling strategy according to real-time changes. This greatly enhances the system's adaptability and dynamic response capability, ensuring the stability and consistency of the cooling effect. Simultaneously, the system can respond promptly to changes in cooling conditions, avoiding internal stress problems caused by inconsistent cooling rates.

[0061] In the above-mentioned monitoring method for the quenching process of metal products, step S4 involves placing the cooled metal into a tempering furnace for low-temperature tempering to obtain tempered metal. It should be understood that common tempering equipment includes electric tempering furnaces, gas-fired tempering furnaces, and vacuum tempering furnaces. Among them, vacuum tempering furnaces, because they can operate in a low-oxygen environment, help reduce oxidation reactions, thereby protecting the metal surface from the effects of oxide scale. The selection of tempering temperature mainly depends on the type of metal and its alloy composition. For example, for carbon steel, the low-temperature tempering temperature is usually set between 180-200℃; while for some high-alloy steels, a higher tempering temperature may be required. Accurate tempering temperature not only helps restore some plasticity but also effectively reduces the internal stress generated during quenching, improving the toughness of the material. Through the tempering process, the internal structure of metal materials can be effectively improved, and their processing performance can be significantly enhanced.

[0062] In the above-mentioned metal product quenching process monitoring method, the step S5 is to clean and rust-proof the tempered metal to obtain a finished metal. It should be understood that in the last stage of the metal processing process, cleaning and rust-proofing can remove various contaminants and oxides generated during the tempering process, and can also prevent further corrosion of the metal surface through appropriate rust-proof measures. Among them, common cleaning methods include alkaline cleaning, acid cleaning, water cleaning, and ultrasonic cleaning, etc. Each method has its scope of application and characteristics, and the most suitable solution needs to be selected according to the specific metal material and surface state. For example, for carbon steel and low alloy steel, the method of combining alkaline cleaning and acid cleaning is usually adopted; while for stainless steel and other materials with strong corrosion resistance, neutral or weak acid cleaning agents can be used. In addition, ultrasonic cleaning, as an efficient cleaning means, is particularly suitable for metal parts with complex shapes, as it can remove dirt in hard-to-reach places through high-frequency vibration. After the cleaning method is selected, the corresponding cleaning solution needs to be prepared. Alkaline cleaning solution is usually prepared from sodium hydroxide (NaOH) or sodium carbonate (Na2CO3) to remove grease and other organic contaminants; while acid cleaning solution is usually selected from hydrochloric acid (HCl), sulfuric acid (H2SO4) or phosphoric acid (H3PO4) to remove scale and rust. It should be noted that the concentration and temperature of the acid solution should be strictly controlled during the acid cleaning process to avoid excessive corrosion of the metal surface. In addition, to improve cleaning efficiency, a suitable amount of surfactant can be added to the cleaning solution, which helps to reduce the surface tension of the solution and enhance the dirt removal capacity. Further, after the preparation of the cleaning solution is completed, the metal parts are placed in the cleaning equipment for operation. In the alkaline cleaning process, the metal parts usually need to be soaked in an alkaline solution at 60-80°C for 10-30 minutes, and the specific time depends on the type and thickness of the contaminants. Subsequently, the metal parts need to be thoroughly washed with water to completely remove the residual alkaline solution. Similarly, in the acid cleaning process, the concentration and temperature of the acid solution also need to be accurately controlled, and the soaking time needs to be adjusted according to the actual situation. After acid cleaning, thorough water washing is again performed to ensure that all acid solution is completely removed. Ultrasonic cleaning utilizes the cavitation effect generated by high-frequency vibration to remove stubborn contaminants on the metal surface in a short time. This cleaning method is particularly suitable for metal parts with complex shapes and precise structures, such as gears, bearings, etc. During ultrasonic cleaning, the metal parts are usually placed in a tank filled with cleaning solution, and then the ultrasonic generator is started. High-frequency vibration generates micro-bubbles in the liquid, which burst to produce local high pressure, effectively removing contaminants from the metal surface. The ultrasonic cleaning time is generally controlled between 5-30 minutes. After the cleaning process is completed, the next step is to perform rust-proofing treatment. The main purpose of rust-proofing treatment is to form a protective film on the metal surface to prevent corrosion during storage and transportation. Common rust-proofing methods include oiling, phosphating, passivation, etc.Oiling is one of the simplest and most direct methods of rust prevention, suitable for most metal parts. Commonly used rust-proof oils include mineral oil, synthetic oil, etc. They can form a thin oil film on the metal surface, effectively isolating moisture and oxygen in the air. Phosphating is a chemical conversion treatment method that generates a protective phosphate film on the metal surface, improving its corrosion resistance. Passivation is a chemical reaction that forms a dense oxide film on the metal surface, particularly suitable for materials with strong corrosion resistance such as stainless steel. It is worth mentioning that during rust prevention treatment, the metal part surface must be clean and free of water marks. If there is residual moisture on the surface, it will greatly reduce the rust prevention effect. Therefore, before oiling or other rust prevention treatment, the metal part usually needs to be dried. Common drying methods include natural air drying, hot air drying, and vacuum drying. Natural air drying is suitable for small metal parts, but for large workpieces, hot air drying or vacuum drying is more efficient. Hot air drying accelerates moisture evaporation by heating air, while vacuum drying quickly removes moisture in a low-pressure environment. Both methods can significantly shorten drying time and improve production efficiency. After cleaning and rust prevention treatment, the surface quality of the metal material can be effectively improved, and its corrosion resistance can be significantly improved.

[0063] In summary, the metal product quenching process monitoring method based on the embodiments of the present application is illustrated, which obtains forged metal and performs preheating, quenching heating, quenching cooling, low-temperature tempering, and cleaning and rust prevention treatment to obtain finished metal. Among them, quenching cooling needs to obtain the time queue of real-time metal temperature, cooling medium real-time temperature and cooling medium real-time pressure, then perform time sequence implicit coding, feature splicing and bidirectional interaction response representation on the input data, to automatically recommend the cooling medium flow rate for comparison with the real-time flow rate, thereby generating a cooling medium flow rate adjustment instruction. In this way, the adaptability and dynamic response capability of the system can be improved, ensuring the stability and consistency of the cooling effect, and timely responding to adjust the cooling conditions to avoid internal stress problems caused by inconsistent cooling speed.

[0064] Figure 6 The system block diagram of the metal product quenching process monitoring system according to the embodiments of the present application is shown in FIG. 1. As shown in FIG. 1, the system includes a preheating module 101, a quenching heating module 102, a quenching cooling module 103, a low-temperature tempering module 104, a cleaning and rust prevention treatment module 105, a data acquisition module 106, a data preprocessing module 107, a feature extraction module 108, a model training module 109, a model prediction module 110, and a system control module 111. Figure 6As shown, according to the metal product quenching process monitoring system 100 of the embodiments of the present application, it comprises: a metal product quenching data acquisition module 110, configured to acquire a time queue of real-time metal temperature collected by a first temperature sensor, and acquire a time queue of real-time cooling medium temperature and a time queue of real-time cooling medium pressure collected by a second temperature sensor and a pressure sensor; a metal product quenching data sequence encoding module 120, configured to respectively perform sequence encoding on the time queue of real-time metal temperature, the time queue of real-time cooling medium temperature and the time queue of real-time cooling medium pressure to obtain a metal temperature time sequence feature, a cooling medium temperature time sequence feature and a cooling medium pressure time sequence feature; a cooling medium temperature-pressure feature jointing module 130, configured to joint the cooling medium temperature time sequence feature and the cooling medium pressure time sequence feature to obtain a cooling medium temperature-pressure time sequence jointing feature; a quenching object-quenching medium bidirectional interaction response encoding module 140, configured to perform quenching object-quenching medium bidirectional interaction response encoding on the metal temperature time sequence feature and the cooling medium temperature-pressure time sequence jointing feature to obtain a quenching object-quenching medium parameter time sequence interaction response feature; and a cooling medium flow rate adjusting instruction generating module 150, configured to generate a cooling medium flow rate adjusting instruction based on the quenching object-quenching medium parameter time sequence interaction response feature.

[0065] Here, those skilled in the art can understand that the specific operations of each step in the above metal product quenching process monitoring system have been described in detail above with reference to the description of the metal product quenching process monitoring method Figures 1 to 5 , and thus the repeated description thereof will be omitted.

[0066] As described above, the metal product quenching process monitoring system 100 according to the embodiments of the present application can be implemented in various terminal devices, and in one example, the metal product quenching process monitoring system 100 can be integrated into a terminal device as a software module and / or a hardware module. For example, the metal product quenching process monitoring system 100 can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the metal product quenching process monitoring system 100 can also be one of the many hardware modules of the terminal device.

[0067] In summary, the metal product quenching process monitoring system based on the embodiments of the present application is illustrated, which obtains the forged metal and performs preheating, quenching heating, quenching cooling, low-temperature tempering and cleaning and rust-proof treatment to obtain the finished metal. Among them, the quenching cooling needs to obtain the time queue of real-time metal temperature, real-time cooling medium temperature and real-time cooling medium pressure, and then perform time sequence implicit coding, feature splicing and bidirectional interaction response representation on the input data to automatically recommend the cooling medium flow rate for comparison with the real-time flow rate, thereby generating a cooling medium flow rate adjustment instruction. In this way, the adaptability and dynamic response capability of the system can be improved, the stability and consistency of the cooling effect can be ensured, and the cooling conditions can be adjusted in time to avoid the internal stress problem caused by inconsistent cooling speed.

Claims

1. A method for monitoring the quenching process of metal products, comprising: Step S1: Place the forged metal into a heating furnace for preheating treatment to obtain preheated metal; Step S2: Transfer the preheated metal to a salt bath furnace or a controlled atmosphere furnace for quenching and heating to obtain heated metal; Step S3: Quench and cool the heated metal to obtain cooled metal; Step S4: Place the cooled metal into a tempering furnace for low-temperature tempering to obtain tempered metal; Step S5: Cleaning and rust-preventing treatment of the tempered metal to obtain the finished metal; characterized in that step S3 includes: The system acquires a time queue of real-time metal temperature collected by the first temperature sensor, and a time queue of real-time cooling medium temperature and real-time cooling medium pressure collected by the second temperature sensor and pressure sensor. Sequence encoding is performed on the time queues of the real-time metal temperature, the real-time temperature of the cooling medium, and the real-time pressure of the cooling medium to obtain the time-series characteristics of the metal temperature, the cooling medium temperature, and the cooling medium pressure. The time-series characteristics of the cooling medium temperature and the time-series characteristics of the cooling medium pressure are combined to obtain the joint time-series characteristics of the cooling medium temperature-pressure. The time-series characteristics of the metal temperature and the combined time-series characteristics of the cooling medium temperature-pressure are encoded into a bidirectional interactive response between the quenching object and the quenching medium to obtain the time-series interactive response characteristics of the quenching object and the quenching medium parameters. Based on the time-series interactive response characteristics of the quenching object-quenching medium parameters, a cooling medium flow rate adjustment command is generated. Specifically, the metal temperature time series characteristics and the cooling medium temperature-pressure time series joint characteristics are encoded using a bidirectional interactive response encoding method between the quenching object and the quenching medium to obtain the time series interactive response characteristics of the quenching object-quenching medium parameters, including: Homography projection transformation is performed on the latent coding vector of metal temperature time series features and the latent coding vector of cooling medium temperature-pressure time series joint features to obtain the homography projection coding vector of metal temperature time series and the homography projection coding vector of cooling medium temperature-pressure time series joint features. Based on the bidirectional attention balance field of the forward and reverse quenching object-quenching medium parameters between the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector, feature modulation is performed on the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector to obtain the metal temperature time-series homography projection attention modulation coding vector and the cooling medium temperature-pressure time-series joint homography projection attention modulation coding vector; Interactive response processing is performed on the metal temperature temporal homography projection attention modulation coding vector and the cooling medium temperature-pressure temporal joint homography projection attention modulation coding vector to obtain the quenching object-quenching medium parameter temporal interactive response coding vector as the quenching object-quenching medium parameter temporal interactive response feature. Specifically, inputting the temporal interaction response encoding vector of the quenching object-quenching medium parameters into the decoder-based cooling medium flow rate optimization module to obtain the optimization result includes: calculating the first... Eigenvalues ​​and the first Between eigenvalues Distance and The distance is calculated to obtain the time-series interaction response distance matrix of the first quenching object-quenching medium parameter and the second quenching object-quenching medium parameter; the weighted sum of the time-series interaction response distance matrix of the first quenching object-quenching medium parameter and the second quenching object-quenching medium parameter is calculated, and the eigenvalues ​​of the weighted sum of the distances are determined. arrive The process involves arranging the eigenvalues ​​to obtain the eigenvector of the temporal interaction response distance between the quenching object and the quenching medium parameters; interpolating the eigenvector to obtain an interpolated eigenvector of the temporal interaction response of the quenching object and the quenching medium parameters with the same length as the encoded vector; performing a matrix multiplication between the encoded vector and the first distance matrix to obtain an intermediate feature vector of the temporal interaction response of the quenching object and the quenching medium parameters; and then multiplying the second distance matrix of the temporal interaction response of the quenching object and the quenching medium parameters. The autocorrelation matrix of the quenching object-quenching medium parameter time-series interactive response encoding vector is multiplied to obtain the intermediate feature matrix of the quenching object-quenching medium parameter time-series interactive response. The intermediate feature vector of the quenching object-quenching medium parameter time-series interactive response is multiplied with the intermediate feature matrix of the quenching object-quenching medium parameter time-series interactive response, and then multiplied by the dot product of the interpolation eigenvector of the quenching object-quenching medium parameter time-series interactive response to obtain the optimized quenching object-quenching medium parameter time-series interactive response encoding vector. The optimized quenching object-quenching medium parameter time-series interactive response encoding vector is input into the decoder-based cooling medium flow rate optimization module to obtain the optimization result.

2. The method for monitoring the quenching process of metal products according to claim 1, characterized in that, Sequence encoding is performed on the time queues of the real-time metal temperature, the real-time cooling medium temperature, and the real-time cooling medium pressure to obtain the time-series features of metal temperature, cooling medium temperature, and cooling medium pressure, respectively. This includes performing LSTM-based sequence encoding on the time queues of the real-time metal temperature, the real-time cooling medium temperature, and the real-time cooling medium pressure to obtain the hidden encoding vectors of the metal temperature time sequence features, the hidden encoding vectors of the cooling medium temperature time sequence features, and the hidden encoding vectors of the cooling medium pressure time sequence features as the cooling medium pressure time sequence features.

3. The method for monitoring the quenching process of metal products according to claim 2, characterized in that, The process of combining the time-series features of the cooling medium temperature and the time-series features of the cooling medium pressure to obtain the joint time-series features of the cooling medium temperature includes: performing feature concatenation on the hidden encoding vector of the time-series features of the cooling medium temperature and the hidden encoding vector of the time-series features of the cooling medium pressure to obtain the hidden encoding vector of the joint time-series features of the cooling medium temperature and pressure as the joint time-series features of the cooling medium temperature and pressure.

4. The method for monitoring the quenching process of metal products according to claim 3, characterized in that, The metal temperature time-series latent coding vector and the cooling medium temperature-pressure time-series joint latent coding vector are subjected to homography projection transformation to obtain the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector, including: The metal temperature time-series feature latent encoding vector is subjected to homography projection transformation using the metal temperature time-series mapping homography matrix to obtain the metal temperature time-series homography projection encoding vector. The cooling medium temperature-pressure temporal joint mapping homography matrix is ​​used to perform homography projection transformation on the cooling medium temperature-pressure temporal joint feature latent encoding vector to obtain the cooling medium temperature-pressure temporal joint homography projection encoding vector.

5. The method for monitoring the quenching process of metal products according to claim 4, characterized in that, Based on the bidirectional attention balance field of the forward and reverse quenching object-quenching medium parameters between the metal temperature-time homography projection coding vector and the cooling medium temperature-pressure time-time joint homography projection coding vector, feature modulation is performed on the metal temperature-time homography projection coding vector and the cooling medium temperature-pressure time-time joint homography projection coding vector to obtain the metal temperature-time homography projection attention modulation coding vector and the cooling medium temperature-pressure time-time joint homography projection attention modulation coding vector, including: Calculate the positive quenching object-quenching medium parameter attention score field of the metal temperature temporal homography projection coding vector relative to the cooling medium temperature-pressure temporal joint homography projection coding vector; Calculate the inverse quenching object-quenching medium parameter attention score field of the cooling medium temperature-pressure temporal joint homography projection encoding vector relative to the metal temperature temporal homography projection encoding vector; Construct a bidirectional attention balance field between the forward quenching object-quenching medium parameter attention score field and the reverse quenching object-quenching medium parameter attention score field; The metal temperature temporal homography projection coding vector and the cooling medium temperature-pressure temporal joint homography projection coding vector are respectively mapped to the bidirectional attention balance field of the positive and negative quenching object-quenching medium parameters to obtain the metal temperature temporal homography projection attention modulation coding vector and the cooling medium temperature-pressure temporal joint homography projection attention modulation coding vector.

6. The method for monitoring the quenching process of metal products according to claim 5, characterized in that, Based on the time-series interactive response characteristics of the quenching object and quenching medium parameters, a cooling medium flow rate adjustment command is generated, including: Based on the temporal interaction response encoding vector of the quenching object-quenching medium parameters, the optimization result is obtained, and the optimization result is the recommended value of the cooling medium flow rate. Based on the comparison between the recommended value of the cooling medium flow rate and the real-time value of the cooling medium flow rate, the cooling medium flow rate adjustment command is generated.

7. A monitoring system for the quenching process of metal products, characterized in that, include: The metal product quenching treatment data acquisition module is used to acquire the time queue of real-time metal temperature collected by the first temperature sensor, and the time queue of real-time temperature and real-time pressure of the cooling medium collected by the second temperature sensor and the pressure sensor. The metal product quenching data sequence encoding module is used to perform sequence encoding on the time queue of the real-time metal temperature, the time queue of the real-time temperature of the cooling medium, and the time queue of the real-time pressure of the cooling medium to obtain the time sequence characteristics of metal temperature, the time sequence characteristics of cooling medium temperature, and the time sequence characteristics of cooling medium pressure. A cooling medium temperature-pressure characteristic joint module is used to combine the time-series characteristics of the cooling medium temperature and the time-series characteristics of the cooling medium pressure to obtain the time-series joint characteristics of the cooling medium temperature-pressure. The bidirectional interactive response encoding module for quenching object-quenching medium is used to encode the time-series characteristics of the metal temperature and the combined time-series characteristics of the cooling medium temperature-pressure to obtain the time-series interactive response characteristics of the quenching object-quenching medium parameters. The cooling medium flow rate adjustment command generation module is used to generate a cooling medium flow rate adjustment command based on the time-series interaction response characteristics of the quenching object-quenching medium parameters. Specifically, the metal temperature time series characteristics and the cooling medium temperature-pressure time series joint characteristics are encoded using a bidirectional interactive response encoding method between the quenching object and the quenching medium to obtain the time series interactive response characteristics of the quenching object-quenching medium parameters, including: Homography projection transformation is performed on the latent coding vector of metal temperature time series features and the latent coding vector of cooling medium temperature-pressure time series joint features to obtain the homography projection coding vector of metal temperature time series and the homography projection coding vector of cooling medium temperature-pressure time series joint features. Based on the bidirectional attention balance field of the forward and reverse quenching object-quenching medium parameters between the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector, feature modulation is performed on the metal temperature time-series homography projection coding vector and the cooling medium temperature-pressure time-series joint homography projection coding vector to obtain the metal temperature time-series homography projection attention modulation coding vector and the cooling medium temperature-pressure time-series joint homography projection attention modulation coding vector; Interactive response processing is performed on the metal temperature temporal homography projection attention modulation coding vector and the cooling medium temperature-pressure temporal joint homography projection attention modulation coding vector to obtain the quenching object-quenching medium parameter temporal interactive response coding vector as the quenching object-quenching medium parameter temporal interactive response feature. Specifically, inputting the temporal interaction response encoding vector of the quenching object-quenching medium parameters into the decoder-based cooling medium flow rate optimization module to obtain the optimization result includes: calculating the first... Eigenvalues ​​and the first Between eigenvalues Distance and The distance is calculated to obtain the time-series interaction response distance matrix of the first quenching object-quenching medium parameter and the second quenching object-quenching medium parameter; the weighted sum of the time-series interaction response distance matrix of the first quenching object-quenching medium parameter and the second quenching object-quenching medium parameter is calculated, and the eigenvalues ​​of the weighted sum of the distances are determined. arrive The process involves arranging the eigenvalues ​​to obtain the eigenvector of the temporal interaction response distance between the quenching object and the quenching medium parameters; interpolating the eigenvector to obtain an interpolated eigenvector of the temporal interaction response of the quenching object and the quenching medium parameters with the same length as the encoded vector; performing a matrix multiplication between the encoded vector and the first distance matrix to obtain an intermediate feature vector of the temporal interaction response of the quenching object and the quenching medium parameters; and then multiplying the second distance matrix of the temporal interaction response of the quenching object and the quenching medium parameters. The autocorrelation matrix of the quenching object-quenching medium parameter time-series interactive response encoding vector is multiplied to obtain the intermediate feature matrix of the quenching object-quenching medium parameter time-series interactive response. The intermediate feature vector of the quenching object-quenching medium parameter time-series interactive response is multiplied with the intermediate feature matrix of the quenching object-quenching medium parameter time-series interactive response, and then multiplied by the dot product of the interpolation eigenvector of the quenching object-quenching medium parameter time-series interactive response to obtain the optimized quenching object-quenching medium parameter time-series interactive response encoding vector. The optimized quenching object-quenching medium parameter time-series interactive response encoding vector is input into the decoder-based cooling medium flow rate optimization module to obtain the optimization result.

8. The metal product quenching process monitoring system according to claim 7, characterized in that, The metal product quenching data sequence encoding module is used to: perform LSTM-based sequence encoding on the time queue of the real-time metal temperature, the time queue of the real-time temperature of the cooling medium, and the time queue of the real-time pressure of the cooling medium to obtain the hidden encoding vector of the metal temperature time sequence as the metal temperature time sequence feature, the hidden encoding vector of the cooling medium temperature time sequence as the cooling medium temperature time sequence feature, and the hidden encoding vector of the cooling medium pressure time sequence as the cooling medium pressure time sequence feature.

Images