A mold caking and breakout intelligent prediction method based on thermal-mechanical collaborative sensing

Abstract

The present application relates to a kind of based on heat-power collaborative perception's crystallizer sticking breakout intelligent prediction method, belong to continuous casting safety monitoring and intelligent manufacturing technical field.It includes: in the matrix arrangement MEMS thermal sensor array on crystallizer copper plate, copper plate local temperature and heat flux density are synchronously detected online;Online detection crystallizer friction force;Single time network model and group space network model are constructed, and thermal quantity dynamic characteristic and space propagation characteristic are extracted;Thermal quantity neural network model is constructed to the mode recognition of characteristic waveform;Friction force signal is decomposed by empirical mode, and trend quantity change characteristic is extracted;Friction force neural network prediction model is constructed, and abnormal state is identified by forecast deviation;Steel grade composition crack sensitive index and other auxiliary judgment index are calculated;Finally, through the weighted score and the multi-source information fusion decision maker of dynamic threshold correction output sticking breakout prediction grade.

Description

Technical Field

[0001] This invention belongs to the field of continuous casting safety monitoring and intelligent manufacturing technology, and relates to an intelligent prediction method for crystallizer sticking and leakage based on thermo-mechanical collaborative sensing. Background Technology

[0002] Sticking and leaking steel is the most dangerous and catastrophic accident in continuous casting production. It directly leads to reduced casting machine operating rates, significantly increased production costs, equipment damage, and even personnel injuries. Current technologies for predicting sticking and leaking steel mainly employ logical judgment methods based on thermocouple temperature detection or single neural network methods. These methods mainly suffer from the following technical defects: (1) Single perception dimension: Existing technologies rely solely on temperature signals detected by thermocouples, but thermocouples have inherent thermal hysteresis, making it difficult to capture early signs of rapid bonding changes; (2) Weak anti-interference capability: Single thermal quantity detection is easily affected by working conditions such as fluctuations in the liquid level of the crystallizer and uneven inflow of protective slag, resulting in a high false alarm rate; (3) Lack of mechanical information: Existing prediction systems do not introduce mechanical quantities such as friction for collaborative judgment, but friction is extremely sensitive to abnormal contact between the billet shell and the copper plate, and the lack of mechanical quantity assistance leads to a high risk of missed detection; (4) Rigid decision-making logic: Traditional logical judgment methods use fixed thresholds and lack dynamic consideration of key influencing factors such as steel grade crack sensitivity and meniscus metallurgical state, resulting in poor adaptability; (5) Limited model generalization ability: Pure neural network methods rely on a large number of training samples, which are insufficient for small samples or new working conditions, and lack physical interpretability. These problems make it difficult for the existing steel leakage prediction system to meet the stringent production requirements of high-speed continuous casting in terms of accuracy and timeliness. Summary of the Invention

[0003] In view of this, the purpose of this invention is to solve the technical problems of low accuracy, poor timeliness, and high false alarm rate in the prediction of sticking and leakage in existing technologies, and the inability of single physical field detection to adapt to the complex working conditions of high-speed continuous casting. This invention provides an intelligent prediction method for sticking and leakage in the crystallizer based on thermo-mechanical collaborative sensing. By performing system-level collaborative sensing of thermal and mechanical quantities and constructing a multi-model parallel processing and multi-source information fusion decision architecture, this invention achieves early capture and highly reliable identification of sticking and leakage symptoms.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A smart prediction method for crystallizer sticking leakage based on thermo-mechanical co-sensing includes the following steps: Step 1: Arrange a MEMS thermal sensor array in a matrix on the copper plate of the crystallizer to simultaneously detect the local temperature and heat flux density of the copper plate online; Step 2: Detect the friction force of the crystallizer online and calculate the root mean square value of the friction force and the vibration frequency; Step 3: Perform time series analysis on the temperature and heat flux density of a single sensor, construct a single-unit time network model, and extract dynamic characteristics of thermal quantities; Step 4: Perform spatial sequence analysis on multiple sensors that are adjacent in the horizontal and vertical directions, construct a pair spatial network model, and extract the spatial propagation characteristics of thermal quantities; Step 5: Construct a thermal quantity neural network model, perform pattern recognition on the feature waveforms extracted by the single-unit time network model and the group-pair spatial network model, and output the thermal quantity anomaly assessment results; Step 6: Construct an empirical mode decomposition model for friction force, perform empirical mode decomposition on the friction force signal of the crystallizer, obtain the inherent mode function and trend quantity, and extract the trend quantity change characteristics as an abnormal sign of friction force; Step 7: Construct a friction force neural network prediction model to predict friction force values ​​in real time based on historical root mean square friction force values ​​and vibration frequencies, and identify abnormal friction force states by the deviation rate between predicted and measured values; Step 8: Calculate multiple auxiliary judgment indices to dynamically adjust the forecast decision threshold; Step 9: Based on the combined results of the thermal anomaly assessment, the abnormal signs and abnormal state identification, and the auxiliary judgment index, the bonding leakage prediction level is output through the multi-source information fusion decision-maker.

[0005] Furthermore, the MEMS thermal sensor mentioned in step 1 is a microelectromechanical sensor that can simultaneously output temperature and heat flux density values.

[0006] Furthermore, the dynamic characteristics of thermal quantities mentioned in step 3 include the mean deviation of thermal quantities. Rise rate and rise / fall check marks; mean deviation of the thermal quantity Calculate using the following formula:

[0007] in, This represents the maximum thermal quantity detected at the current moment. The current moment within the previous analysis period. Each sample value.

[0008] Furthermore, the spatial propagation characteristics of thermal quantity mentioned in step 4 include the fluctuation of the mean deviation of thermal quantity of laterally adjacent sensors and the delay time, wherein the lateral distribution judgment condition is:

[0009] in, for The mean deviation of the thermal properties of time sensor C for The mean deviation of the thermal properties of time sensor B. To delay time, This is the threshold for the fluctuation of the mean deviation of the transverse thermal quantity.

[0010] Furthermore, the empirical mode decomposition model of friction force in step 6 specifically includes: Step 6.1: Process the original friction force signal Perform empirical mode decomposition to obtain One IMF component and a residual trend quantity ; Step 6.2: Calculate the trend quantity Slope within the sliding time window ; Step 6.3: When continuous The number of sampling periods exceeds the preset threshold. When this occurs, it is determined that the frictional force exhibits an abnormal trend.

[0011] Furthermore, the auxiliary judgment index mentioned in step 8 includes: Steel grade cracking sensitivity index :

[0012] Cracking Sensitivity Index under Stress During Billet Pulling :

[0013] Liquid level fluctuation slag sensitivity index :

[0014] Hook-shaped vibration mark formation sensitivity index :

[0015] Thick slag layer and slag-rolling sensitivity index :

[0016] in, , and These represent the mass percentages of carbon, manganese, and sulfur in the steel, respectively. To protect the viscosity of the slag, The effective height of the crystallizer, To increase speed, For the oscillation frequency, For amplitude, For the high-temperature tensile strength of steel, The thickness of the liquid slag film in the gap of the crystallizer. The solidification coefficient of steel; This represents the fluctuation value of the molten steel level at the meniscus. The critical slag spooling speed; The local heat flux density at the meniscus of the copper plate in the crystallizer; For the thickness of the liquid slag layer, This is the minimum allowable thickness of the liquid slag layer.

[0017] Furthermore, the multi-source information fusion decision-maker described in step 9 employs a weighted scoring and dynamic threshold correction mechanism, including the following steps: Step 9.1: Assign weight coefficients to the auxiliary judgment index, the thermal quantity neural network model, the friction force empirical mode decomposition model, and the friction force neural network prediction model, respectively. , , , ,satisfy ; Step 9.2: Each model outputs a normalized anomaly score. ; Step 9.3: Calculate the comprehensive anomaly index ; Step 9.4: Introduce an auxiliary judgment index to dynamically adjust the alarm threshold:

[0018] Step 9.5: When When the time comes, a forecast of the corresponding level will be triggered.

[0019] The beneficial effects of this invention are as follows: For the first time, a system-level collaborative sensing and fusion decision-making of thermal quantities (temperature, heat flux density) and mechanical quantities (friction force) has been achieved. Existing technologies are all based on judgments based on a single physical field. This invention significantly improves the accuracy and robustness of predicting sticky steel leakage by complementing and verifying thermal-mechanical multi-field coupled information. The prediction accuracy rate reaches over 96%, and the false alarm rate is less than 4%. Compared with single thermal prediction methods, the false alarm rate is reduced by about 60%.

[0020] The use of MEMS sensors to simultaneously detect temperature and heat flux density overcomes the thermal hysteresis limitation of traditional thermocouples, which can only measure temperature. Heat flux density has a faster dynamic response to changes in heat transfer caused by bonding (response time less than 0.5 seconds), and can detect bonding signs 5 to 10 seconds in advance, gaining valuable time for taking control measures such as slowing down. It is especially suitable for high-speed (>1.8 m / min) continuous casting production.

[0021] By combining Empirical Mode Decomposition (EMD) with neural network prediction in friction signal processing, friction anomalies can be monitored simultaneously from two dimensions: trend change and prediction deviation. This approach retains EMD's adaptive decomposition capability for non-stationary signals while utilizing the neural network's learning and prediction capabilities based on historical data. This achieves highly sensitive identification of abnormal friction states, providing an early warning approximately 15 to 20 seconds earlier than simple threshold judgment.

[0022] The system introduces a steel grade crack sensitivity index and a meniscus metallurgical condition-related auxiliary judgment index. These indices comprehensively consider factors that have an important impact on steel adhesion leakage but are ignored by existing prediction systems, such as steel composition, protective slag characteristics, liquid level fluctuations, and vibration mark formation. By dynamically correcting the alarm threshold, the system effectively suppresses false alarms caused by operating condition fluctuations and improves the system's adaptability to different steel grades and process conditions.

[0023] A decision architecture was constructed that combines logical judgment with parallel processing of neural networks and weighted fusion of multiple modules. This architecture retains the physical interpretability of logical judgment while leveraging the self-learning and pattern recognition advantages of neural networks. It overcomes the shortcomings of single neural network methods, which rely on a large number of training samples and have limited generalization ability, and achieves an organic unity of physical meaning and data-driven approach.

[0024] The overall technical solution provides a reliable safety guarantee for high-speed continuous casting with leakage, which helps to improve the casting machine's operating rate and reduce production and maintenance costs. It has significant economic benefits and engineering application value, and is worthy of widespread promotion and application in the fields of intelligent monitoring of continuous casting process and crystallizer expert system.

[0025] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0026] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 A schematic diagram of the arrangement of thermal sensor matrix on the wide copper plate of the crystallizer; Figure 2 The characteristics of temperature and heat flux density changes of the thermal sensor in the crystallizer during bonding; Figure 3 This is a schematic diagram of the analysis of a single-unit time network model; Figure 4 This is a schematic diagram of the distribution of sensors in a spatial network. Figure 5This is a schematic diagram of the time series BP neural network model structure; Figure 6 This is a schematic diagram of the empirical mode decomposition process of the friction force signal; Figure 7 This refers to the abnormal change in the frictional force of the crystallizer before the sticking and leakage of steel. Figure 8 This is a flowchart for multi-source information fusion decision-making. Detailed Implementation

[0027] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0028] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0029] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0030] Example 1: like Figure 1 As shown, this invention provides an intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing, comprising the following steps: Step 1: Arrange a MEMS thermal sensor array in a matrix on the copper plate of the crystallizer. The MEMS thermal sensors can simultaneously detect the local temperature and heat flux density of the copper plate online, with a sampling frequency of not less than 1Hz. The MEMS thermal sensors are manufactured using microelectromechanical systems technology and can simultaneously sense temperature and heat flux density. Its heat flux density detection response time constant is less than 0.5 seconds, which captures the heat flux change caused by bonding 3 to 8 seconds earlier than traditional thermocouple temperature detection.

[0031] Step 2: Real-time online detection of crystallizer friction force based on hydraulic cylinder pressure difference method or power difference method, and calculation of root mean square value of friction force and vibration frequency.

[0032] Step 3: Construct a single-unit time network model to perform time series analysis on the thermal quantities (temperature and heat flux density) of a single sensor, calculating characteristic parameters including the mean deviation of thermal quantities, rate of rise, and rise / fall checks. The dynamic characteristics of thermal quantities include the mean deviation of thermal quantities. Rise rate and rise / fall check marks; mean deviation of the thermal quantity Calculate using the following formula:

[0033] in, This represents the maximum thermal quantity detected at the current moment. The current moment within the previous analysis period. Each sample value.

[0034] Step 4: Construct a paired spatial network model, perform spatial sequence analysis on multiple horizontally and vertically adjacent sensors, and calculate characteristic parameters including thermal quantity change delay time, inversion bias, and mean deviation fluctuation. The spatial propagation characteristics of thermal quantities include the mean deviation fluctuation and delay time of thermal quantities of horizontally adjacent sensors. The lateral distribution judgment condition is:

[0035] in, for The mean deviation of the thermal properties of time sensor C for The mean deviation of the thermal properties of time sensor B. To delay time, This is the threshold for the fluctuation of the mean deviation of the transverse thermal quantity.

[0036] Step 5: Construct a time series neural network model and a spatial sequence neural network model for thermal quantities, and perform pattern recognition on the waveforms of thermal quantity changes from a single sensor and the spatial propagation waveforms from multiple sensors, respectively, and output the probability of thermal quantity anomalies.

[0037] Step 6: Construct an empirical mode decomposition model for friction force, adaptively decomposing the real-time detected friction force signal into a series of intrinsic mode functions (IMFs) and trend quantities. The slope of change and energy distribution characteristics of the trend quantities are extracted as parameters indicating adhesion. Specifically, this includes: 1) Regarding the original friction force signal Perform empirical mode decomposition to obtain One IMF component and a residual trend quantity ; 2) Calculate the trend quantity Slope within the sliding time window ; 3) When continuous The number of sampling periods exceeds the preset threshold. When this occurs, it is determined that the frictional force exhibits an abnormal trend.

[0038] Step 7: Construct a friction force neural network prediction model. This model takes the historical root mean square value of friction force and vibration frequency as input, predicts the current friction force value in real time, and identifies abnormal states by calculating the deviation rate between the predicted value and the measured value.

[0039] Step 8: Calculate auxiliary judgment indices, including the steel grade cracking sensitivity index, billet stress cracking sensitivity index, liquid level fluctuation slag entrapment sensitivity index, hook-shaped vibration mark formation sensitivity index, and liquid slag layer thickness slag entrapment sensitivity index, to correct the prediction threshold and suppress false alarms. The auxiliary judgment indices specifically include: Steel grade cracking sensitivity index:

[0040] Cracking sensitivity index under stress during billet pulling:

[0041] Liquid level fluctuation slag sensitivity index:

[0042] Hook-shaped vibration mark formation sensitivity index:

[0043] Liquid slag layer thickness and slag coating sensitivity index:

[0044] in, , and These represent the mass percentages of carbon, manganese, and sulfur in the steel, respectively. To protect the viscosity of the slag, The effective height of the crystallizer, To increase speed, For the oscillation frequency, For amplitude, For the high-temperature tensile strength of steel, The thickness of the liquid slag film in the gap of the crystallizer. The solidification coefficient of steel; This represents the fluctuation value of the molten steel level at the meniscus. The critical slag spooling speed; The local heat flux density at the meniscus of the copper plate in the crystallizer; For the thickness of the liquid slag layer, This is the minimum allowable thickness of the liquid slag layer.

[0045] Step 9: Construct a multi-source information fusion decision-maker. This decision-maker integrates the thermal quantity logical judgment results and neural network recognition results output from steps 3-5, the friction force EMD analysis results and neural network prediction deviation results output from steps 6-7, and the auxiliary judgment index calculated in step 8. It performs fusion decision-making according to a preset weight matrix and dynamic threshold, outputting the predicted level of steel leakage (normal, attention, warning, severe). The multi-source information fusion decision-maker adopts a weighted scoring and dynamic threshold correction mechanism as follows: 1) Assign weight coefficients to the four modules: thermal quantity logical judgment, thermal quantity neural network identification, friction force EMD analysis, and friction force neural network prediction. , , , ,satisfy ; 2) Normalized anomaly scores output by each module ; 3) Calculate the comprehensive anomaly index ; 4) Introduce an auxiliary judgment index to dynamically adjust the alarm threshold:

[0046] 5) When When the time comes, a forecast of the corresponding level will be triggered.

[0047] Example 2: Using a 1500mm×230mm slab continuous casting machine at a steel plant as an example, with a casting speed of 1.2m / min, the method of this invention is used for intelligent prediction of sticking and leakage. First, 48 MEMS thermal sensors, arranged in 3 rows and 16 columns, are placed on the wide copper plate of the crystallizer to simultaneously detect temperature and heat flux density at a sampling frequency of 1Hz.

[0048] Secondly, the frictional force of the crystallizer is detected in real time using the hydraulic cylinder pressure difference method, with a sampling frequency of 20Hz, and the root mean square value of the frictional force per second is calculated. The characteristics of temperature and heat flux density changes of the crystallizer thermal sensors when adhesion occurs are as follows: Figure 2 As shown.

[0049] Then, perform individual temporal network analysis as per step 3. For example... Figure 3 As shown, taking a certain sensor as an example, five sampling times (5 seconds) are taken as one analysis cycle to calculate the mean temperature deviation. mean heat flux deviation , respectively with threshold , The comparisons showed that none exceeded the threshold. For example... Figure 5The time series analysis of temperature and heat flux density of a single sensor is shown using a time series BP neural network.

[0050] Next, perform pairwise spatial network analysis as per step 4. For example... Figure 4 As shown, taking sensors B, C, and D as examples, the lateral mean deviation fluctuation is calculated. less than the threshold Longitudinal inversion deviation less than the threshold None of them triggered an alarm.

[0051] Simultaneously, perform empirical mode decomposition of frictional force according to step 6, such as... Figure 6 As shown, 60 seconds of friction data are taken and decomposed to obtain the trend quantity. Calculate the difference between the current trend volume and the trend volume at the previous moment. less than the threshold .

[0052] Perform friction force neural network prediction in step 7. Predict the current value based on the root mean square value of friction force and vibration frequency in the previous 30-45 seconds. The prediction error is 8%, which is less than the threshold of 10%.

[0053] Calculate the auxiliary judgment index according to step 8: , , , , All values ​​were less than the baseline value of 1.0, and threshold correction was not triggered.

[0054] Based on the overall assessment, no alarms were triggered in any of the modules, and the system determined that the casting process was in normal condition.

[0055] like Figure 7 As shown, when adhesion symptoms occur, the temperature of one sensor rises by 22°C within 15 seconds, decreases by -18°C, and rises at a rate of 1.8°C / s, exceeding the threshold; the lateral delay time between adjacent sensors is 5.2 seconds, consistent with adhesion propagation characteristics; the frictional trend... Exceeding the threshold; such as Figure 8 As shown, a comprehensive judgment triggered a serious alarm, and the system automatically reduced its speed to 0.7 m / min, successfully preventing a steel leakage accident.

[0056] Example 3: An electronic device, comprising a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the method described in Embodiment 1 when executing the computer program.

[0057] Example 4: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0058] Example 5: A computer program product includes a computer program that, when executed by a processor, implements the method described in Example 1.

[0059] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily refer to the same embodiment.

[0060] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0061] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0062] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0063] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0064] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0065] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0066] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing, characterized in that: Includes the following steps: Step 1: Arrange a MEMS thermal sensor array in a matrix on the copper plate of the crystallizer to simultaneously detect the local temperature and heat flux density of the copper plate online; Step 2: Detect the friction force of the crystallizer online and calculate the root mean square value of the friction force and the vibration frequency; Step 3: Perform time series analysis on the temperature and heat flux density of a single sensor, construct a single-unit time network model, and extract dynamic characteristics of thermal quantities; Step 4: Perform spatial sequence analysis on multiple sensors that are adjacent in the horizontal and vertical directions, construct a pair spatial network model, and extract the spatial propagation characteristics of thermal quantities; Step 5: Construct a thermal quantity neural network model, perform pattern recognition on the feature waveforms extracted by the single-unit time network model and the group-pair spatial network model, and output the thermal quantity anomaly assessment results; Step 6: Construct an empirical mode decomposition model for friction force, perform empirical mode decomposition on the friction force signal of the crystallizer, obtain the inherent mode function and trend quantity, and extract the trend quantity change characteristics as an abnormal sign of friction force; Step 7: Construct a friction neural network prediction model to predict friction values ​​in real time based on historical root mean square friction values ​​and vibration frequencies, and identify abnormal friction states by the deviation rate between predicted and measured values; Step 8: Calculate multiple auxiliary judgment indices to dynamically adjust the forecast decision threshold; Step 9: Based on the combined results of the thermal anomaly assessment, the abnormal signs and abnormal state identification, and the auxiliary judgment index, the adhesion leakage prediction level is output through the multi-source information fusion decision-maker.

2. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The MEMS thermal sensor mentioned in step 1 is a microelectromechanical sensor that can simultaneously output temperature and heat flux density values.

3. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The dynamic characteristics of thermal quantities mentioned in step 3 include the mean deviation of thermal quantities. Rise rate and rise / fall check marks; mean deviation of the thermal quantity Calculate using the following formula: in, This represents the maximum thermal quantity detected at the current moment. The current moment within the previous analysis period. Each sample value.

4. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The spatial propagation characteristics of thermal quantity mentioned in step 4 include the fluctuation of the mean deviation of thermal quantity of laterally adjacent sensors and the delay time, wherein the lateral distribution judgment condition is: in, for The mean deviation of the thermal properties of time sensor C for The mean deviation of the thermal properties of time sensor B. To delay time, This is the threshold for the fluctuation of the mean deviation of the transverse thermal quantity.

5. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The empirical mode decomposition model of friction force in step 6 specifically includes: Step 6.1: Process the original friction force signal Perform empirical mode decomposition to obtain One IMF component and one residual trend quantity ; Step 6.2: Calculate the trend quantity Slope within the sliding time window ; Step 6.3: When continuous The number of sampling periods exceeds the preset threshold. When this occurs, it is determined that the frictional force exhibits an abnormal trend.

6. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The auxiliary judgment index mentioned in step 8 includes: Steel grade cracking sensitivity index : Cracking Sensitivity Index under Stress During Billet Pulling : Liquid level fluctuation slag sensitivity index : Hook-shaped vibration mark formation sensitivity index : Liquid slag layer thickness and slag coating sensitivity index : in, , and These represent the mass percentages of carbon, manganese, and sulfur in the steel, respectively. To protect the viscosity of the slag, The effective height of the crystallizer, To increase speed, For the oscillation frequency, For amplitude, For the high-temperature tensile strength of steel, The thickness of the liquid slag film in the gap of the crystallizer. The solidification coefficient of steel; This represents the fluctuation value of the molten steel level at the meniscus. The critical slag entrainment speed; The local heat flux density at the meniscus of the copper plate in the crystallizer; For the thickness of the liquid slag layer, This is the minimum allowable thickness of the liquid slag layer.

7. The intelligent prediction method for crystallizer sticking leakage based on thermo-mechanical synergistic sensing according to claim 1, characterized in that: The multi-source information fusion decision-maker described in step 9 employs a weighted scoring and dynamic threshold correction mechanism, including the following steps: Step 9.1: Assign weight coefficients to the auxiliary judgment index, the thermal quantity neural network model, the friction force empirical mode decomposition model, and the friction force neural network prediction model, respectively. , , , ,satisfy ; Step 9.2: Each model outputs a normalized anomaly score. ; Step 9.3: Calculate the comprehensive anomaly index ; Step 9.4: Introduce an auxiliary judgment index to dynamically adjust the alarm threshold: Step 9.5: When When the time comes, a forecast of the corresponding level will be triggered.

Images