Method, device and equipment for monitoring and controlling leakage of heating furnace and storage medium

Abstract

The application discloses a heating furnace leakage monitoring and control method, device, equipment and storage medium, relates to the technical field of industrial automation and intelligent control, and discloses a heating furnace leakage monitoring and control method, which comprises the following steps: acquiring multi-dimensional state data collected by a preset multi-modal sensor array and a target digital twin model; inputting the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix; establishing a multi-dimensional residual state matrix according to the multi-dimensional state data and the global theoretical state data matrix; inputting the multi-dimensional residual state matrix into a preset leakage identification model to obtain a leakage type, a leakage severity level and a leakage determination result; when the leakage determination result is that the heating furnace has a target leakage, generating a leakage control instruction according to the leakage type and the leakage severity level, and performing heating furnace leakage control according to the leakage control instruction. The technical scheme of the application can improve the sensitivity of heating furnace leakage monitoring.

Description

Technical Field

[0001] This application relates to the field of industrial automation and intelligent control technology, and in particular to methods, devices, equipment and storage media for monitoring and controlling leaks in heating furnaces. Background Technology

[0002] During long-term operation, the furnace structure, lining materials, and connecting parts of the heating furnace are continuously subjected to complex thermal stress, mechanical stress, and media corrosion, which can easily lead to gradual damage in welds, sealing interfaces, or weak areas of the refractory material. When media leakage occurs, it is often accompanied by the coupled evolution of abnormal temperature distribution, pressure field disturbances, changes in gas composition, and structural vibration characteristics. In the early stages, these anomalies are usually small in amplitude, spatially dispersed, and easily superimposed on fluctuations in normal operating conditions.

[0003] Against this technological backdrop, traditional furnace leak monitoring solutions typically employ manual inspections in conjunction with single-type sensors. They rely on fixed thresholds to trigger alarms based on temperature, pressure, or combustible gas concentration. When monitored values ​​exceed these thresholds, safety interlocks or manual intervention are activated. While simple to implement in engineering practice, these solutions rely primarily on static empirical rules, making it difficult to comprehensively reflect the interrelationships between multiple physical quantities and to distinguish between normal fluctuations caused by process adjustments and abnormal states resulting from leaks.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and storage medium for monitoring and controlling leakage in heating furnaces, which aims to solve the technical problem of difficulty in distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in heating furnaces.

[0006] To achieve the above objectives, this application proposes a method for monitoring and controlling leakage in a heating furnace, the method comprising:

[0007] Acquire multi-dimensional state data and target digital twin model collected by a preset multimodal sensor array;

[0008] The multi-dimensional state data is input into the target digital twin model to obtain the global theoretical state data matrix;

[0009] Based on the multi-dimensional state data and the global theoretical state data matrix, a multi-dimensional residual state matrix is ​​established;

[0010] The multidimensional residual state matrix is ​​input into a preset leak identification model to obtain the leak type, leak severity level, and leak determination result.

[0011] When the leakage determination result indicates that there is a target leakage in the heating furnace, a leakage control instruction is generated based on the leakage type and the leakage severity level, and leakage control of the heating furnace is performed according to the leakage control instruction.

[0012] In one embodiment, the step of obtaining the target digital twin model includes:

[0013] A three-dimensional mesh model of the heating furnace is established based on the finite element method. The three-dimensional mesh model of the heating furnace includes the geometric structure and material property distribution of the furnace shell, insulation layer, refractory bricks and cooling water pipelines.

[0014] In the three-dimensional mesh model of the heating furnace, the fluid flow equation, the energy conservation equation, and the component transport equation are solved in a coupled manner to obtain the target solution result, and an initial digital twin model is generated based on the target solution result;

[0015] Historical operating data is input into a multi-layer fully connected neural network correction module to generate actual operating condition deviations.

[0016] The digital twin model is corrected based on the actual working condition deviation to obtain the target digital twin model.

[0017] In one embodiment, the step of establishing a multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix includes:

[0018] A preset time step is obtained, and the multi-dimensional state data and the global theoretical state data matrix are spatially aligned at the preset time step to obtain an initial residual signal, wherein the initial residual signal includes temperature signal, sound signal, gas signal and pressure signal;

[0019] The initial residual signal is normalized to obtain a normalized residual signal;

[0020] The normalized residual signals are stacked according to the time sequence to obtain a time-series data structure, wherein the dimensions of the time-series data structure are the number of time steps, the total number of sensors, the number of physical quantity types, and the feature dimension.

[0021] A multidimensional residual state matrix is ​​constructed based on the time-series data structure.

[0022] In one embodiment, the step of inputting the multidimensional residual state matrix into a preset leak identification model to obtain the leak type, leak severity level, and leak determination result includes:

[0023] A sensor spatial adjacency matrix is ​​constructed based on the sensor location, and a thermodynamic coupling weight matrix is ​​constructed based on the heat transfer coefficient between sensor nodes.

[0024] The multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix are input into a preset leak identification model to obtain an aggregated feature sequence.

[0025] The aggregated feature sequences are classified and mapped to obtain the leakage type, leakage severity level, and leakage confidence level.

[0026] The presence of a leak in the heating furnace is determined based on the leak confidence level. If a leak is found in the heating furnace, the leak determination result is determined based on the preset leak identification model.

[0027] In one embodiment, the step of inputting the multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix into the preset leak identification model to obtain the aggregated feature sequence includes:

[0028] Obtain the node feature matrix, learnable parameter matrix, and modified linear unit activation function;

[0029] The attribute feature vectors of the graph nodes are determined based on the multidimensional residual state matrix;

[0030] The attribute feature vector, the sensor spatial adjacency matrix, the thermodynamic coupling weight matrix, the node feature matrix, the learnable parameter matrix, and the modified linear unit activation function are input into the preset leak identification model and aggregated to obtain an aggregated feature sequence.

[0031] In one embodiment, the step of determining whether the heating furnace has a leak based on the leak confidence level, and determining the leak determination result based on the preset leak identification model when the heating furnace has a leak, includes:

[0032] Obtain the leakage confidence threshold and preset decay index;

[0033] When the leakage confidence level is greater than the leakage confidence threshold, it is determined that the heating furnace has an initial leakage.

[0034] When the heating furnace has an initial leak, the node activation intensity is determined according to the preset leak identification model;

[0035] A target sensor node set with a preset number of nodes is identified based on the node activation intensity.

[0036] Determine the node coordinates and the distance to the leaking node based on the target sensor node set;

[0037] The coordinates of the leakage source are calculated based on the node coordinates, the distance to the leaking node, and the preset attenuation index.

[0038] When the coordinates of the leak source are present, it is determined that there is a target leak in the heating furnace, and the leak determination result is determined based on the coordinates of the leak source.

[0039] In one embodiment, the step of generating a leak control instruction based on the leak type and the leak severity level when the leak determination result indicates that a target leak exists in the heating furnace, and then performing leak control on the heating furnace according to the leak control instruction, includes:

[0040] When the leakage determination result indicates that there is a target leakage in the heating furnace, a target state vector is generated based on the leakage type, the leakage severity level, the leakage source coordinates in the leakage determination result, and the current process operation parameters of the heating furnace.

[0041] The target state vector is input into a deep network to generate leakage volume change rate, energy loss, equipment damage index, and response delay time.

[0042] The long-term return on control effectiveness is calculated based on the leakage volume change rate, the energy loss, the equipment damage index, and the response delay time.

[0043] The target control action is determined based on the long-term return value of the control effect, and a leakage control instruction is generated based on the target control action. The leakage control instruction is then used to control the leakage of the heating furnace.

[0044] In addition, to achieve the above objectives, this application also proposes a furnace leakage monitoring and control device, which includes: a data acquisition module for acquiring multi-dimensional state data and a target digital twin model collected by a preset multi-modal sensor array;

[0045] The matrix generation module is used to input the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix.

[0046] The matrix generation module is also used to establish a multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix.

[0047] The leakage determination module is used to input the multidimensional residual state matrix into a preset leakage identification model to obtain the leakage type, leakage severity level and leakage determination result.

[0048] The leakage control module is used to generate a leakage control instruction based on the leakage type and the leakage severity level when the leakage determination result indicates that there is a target leakage in the heating furnace, and to perform leakage control of the heating furnace according to the leakage control instruction.

[0049] In addition, to achieve the above objectives, this application also proposes a furnace leakage monitoring and control device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the furnace leakage monitoring and control method described above.

[0050] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the heating furnace leakage monitoring and control method described above.

[0051] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the heating furnace leakage monitoring and control method described above.

[0052] One or more technical solutions proposed in this application have at least the following technical effects:

[0053] By employing multi-dimensional state data acquired from a pre-set multi-modal sensor array and a target digital twin model, the multi-dimensional state data is input into the target digital twin model to obtain a global theoretical state data matrix. A multi-dimensional residual state matrix is ​​then established based on the multi-dimensional state data and the global theoretical state data matrix. This multi-dimensional residual state matrix is ​​input into a pre-set leak identification model to obtain the leak type, leak severity level, and leak determination result. When the leak determination result indicates that a target leak exists in the heating furnace, a leak control command is generated based on the leak type and leak severity level. The leak control command is then used to manage the leak in the heating furnace. This effectively highlights the deviation between actual and ideal operating conditions, fundamentally eliminating the interference of normal process fluctuations. It successfully solves the technical problem of distinguishing between normal process fluctuations and actual leak anomalies under complex operating conditions and multi-source disturbances in the heating furnace. This achieves early warning of leaks, significantly improves monitoring sensitivity, reduces false alarm rates, ensures production continuity and personnel and equipment safety, and balances safety with production energy efficiency optimization. Attached Figure Description

[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart illustrating an embodiment of the heating furnace leakage monitoring and control method of this application.

[0057] Figure 2 This is a schematic diagram of the structure of the heating furnace leakage monitoring and control system provided in Embodiment 1 of the heating furnace leakage monitoring and control method of this application;

[0058] Figure 3 This is a schematic diagram of the leakage feature identification and location process provided in Embodiment 1 of the heating furnace leakage monitoring and control method of this application;

[0059] Figure 4 This is a flowchart illustrating Embodiment 2 of the heating furnace leakage monitoring and control method of this application;

[0060] Figure 5 A simplified flowchart illustrating the furnace leakage monitoring and control method provided in Embodiment 2 of this application;

[0061] Figure 6 This is a schematic diagram of the module structure of the heating furnace leakage monitoring and control device according to an embodiment of this application;

[0062] Figure 7 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the furnace leakage monitoring and control method in the embodiments of this application.

[0063] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0065] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0066] The main solution of this application embodiment is as follows: acquire multi-dimensional state data and a target digital twin model collected by a preset multi-modal sensor array; input the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix; establish a multi-dimensional residual state matrix based on the multi-dimensional state data and the global theoretical state data matrix; input the multi-dimensional residual state matrix into a preset leak identification model to obtain the leak type, leak severity level, and leak determination result; when the leak determination result indicates that the heating furnace has a target leak, generate a leak control instruction based on the leak type and the leak severity level, and perform leak control of the heating furnace based on the leak control instruction.

[0067] In this embodiment, for ease of description, the following description will focus on the detection and control equipment for leaking heating furnaces.

[0068] Because existing technologies struggle to distinguish between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in heating furnaces, this application provides a solution. This solution utilizes multi-dimensional state data collected by a pre-defined multi-modal sensor array and a target digital twin model. The multi-dimensional state data is input into the target digital twin model to obtain a global theoretical state data matrix. A multi-dimensional residual state matrix is ​​then established based on the multi-dimensional state data and the global theoretical state data matrix. This residual state matrix is ​​input into a pre-defined leakage identification model to obtain the leakage type, leakage severity level, and leakage determination result. When the leakage determination result indicates a target leakage in the heating furnace, a leakage control command is generated based on the leakage type and severity level. Based on this command, leakage control of the heating furnace is implemented. This effectively highlights the deviation between actual and ideal operating conditions, fundamentally eliminating the interference of normal process fluctuations. It successfully solves the technical problem of distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in heating furnaces. This achieves advanced leakage warning, significantly improves monitoring sensitivity, reduces false alarm rates, ensures production continuity and personnel and equipment safety, and balances safety with optimized production energy efficiency.

[0069] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a heating furnace leakage monitoring and control device. The following description uses a heating furnace leakage monitoring and control device as an example to illustrate this embodiment and the subsequent embodiments.

[0070] Based on this, embodiments of this application provide a method for monitoring and controlling leakage in a heating furnace, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the heating furnace leakage monitoring and control method of this application.

[0071] In this embodiment, the method for monitoring and controlling leakage in the heating furnace includes steps S10 to S50:

[0072] Step S10: Acquire multi-dimensional state data and target digital twin model collected by a preset multi-modal sensor array;

[0073] It should be noted that the multimodal sensor array is a collection of various types of sensors pre-deployed inside and at specific locations outside the heating furnace, including a distributed fiber optic temperature measurement system, a piezoelectric ceramic acoustic emission sensor, a tunable diode laser absorption spectroscopy analyzer, and a micro differential pressure transmitter, etc., used to comprehensively capture various status information of the heating furnace during operation.

[0074] In addition, multi-dimensional status data are various data reflecting the operating status of the heating furnace, which are collected in real time by a multi-modal sensor array. These include temperature field data distributed along the depth of the furnace lining refractory material, acoustic emission signal data distributed on the outer surface of the furnace shell, gas component concentration data at specific locations in the furnace and flue, and pressure field data inside the furnace chamber. These data can present the furnace's operating status from multiple perspectives.

[0075] In this embodiment, the multimodal sensor array can be installed by laying a distributed optical fiber temperature measurement system in a spiral winding manner along the radial and axial directions of the furnace lining refractory material; arranging multiple piezoelectric ceramic acoustic emission sensors in an array according to a preset equilateral triangular grid topology on the outer surface of the heating furnace metal shell; installing a tunable diode laser absorption spectrometer at a specific height level inside the furnace and in the flue gas outlet pipe; setting multiple micro differential pressure transmitters in different areas of the furnace to form a pressure sensing grid; and achieving nanosecond-level time synchronization of all sensors through a unified clock source, and transmitting the data stream to the central data processing unit at a fixed sampling frequency.

[0076] Furthermore, the target digital twin model is a simulation model that is synchronized with the physical heating furnace in real time. It is based on the finite element method to construct a three-dimensional mesh model, coupled to solve multiple physicochemical equations and introduces a neural network correction module, which can simulate complex processes inside the furnace and output theoretical state data.

[0077] In this embodiment, the distributed optical fiber temperature measurement system is spirally wound along the radial and axial directions of the furnace lining refractory material, with a sampling frequency of ten times per second, a spatial resolution of one meter, a temperature measurement range of 0 degrees Celsius to 1500 degrees Celsius, and an accuracy of ±0.5 degrees Celsius. One hundred and twenty piezoelectric ceramic acoustic emission sensors are arrayed on the outer surface of the furnace shell in an equilateral triangular grid topology with sides of two meters. Their operating frequency band is 20 kHz to 400 kHz, and the signal sampling frequency is one megahertz. Sensors are also located at heights of three meters, six meters, and nine meters from the furnace bottom inside the furnace, as well as at the flue outlet. A tunable diode laser absorption spectrometer is installed, with a measurement response time of less than 100 milliseconds and a concentration detection limit of 1 part per million, for real-time measurement of the volume concentration of oxygen, carbon monoxide, and methane. Thirty-six micro differential pressure transmitters are installed on the front wall, rear wall, side walls, and top area of ​​the furnace, with a measurement range of -5 kPa to +5 kPa, an accuracy of ±5 Pa, and a sampling frequency of five times per second. All sensors are connected to the same high-stability atomic clock source to achieve nanosecond-level timestamp alignment, and the raw data stream is transmitted to the central data processing unit at fixed intervals via industrial Ethernet.

[0078] Understandably, multimodal sensor arrays are deployed at preset locations inside and outside the heating furnace. All sensors achieve nanosecond-level time synchronization through a unified clock source, collect various status data in real time at a fixed sampling frequency, and transmit them synchronously to obtain multi-dimensional status data during the operation of the heating furnace. At the same time, the completed target digital twin model is called to provide a foundation for subsequent theoretical data generation and deviation analysis.

[0079] In one feasible implementation, step S10 may include steps S11 to S14:

[0080] Step S11: Establish a three-dimensional mesh model of the heating furnace according to the finite element strategy. The three-dimensional mesh model of the heating furnace includes the geometric structure and material property distribution of the furnace shell, insulation layer, refractory bricks and cooling water pipelines.

[0081] It should be noted that the finite element method is a numerical calculation method that discretizes complex structures into multiple simple elements for analysis. By establishing equations for each element and solving them in a coupled manner, it can achieve accurate simulation of complex physical problems and is the core methodological basis for constructing a three-dimensional mesh model of a heating furnace.

[0082] In addition, the three-dimensional mesh model of the heating furnace is a three-dimensional simulation model built based on the actual structure of the heating furnace. It accurately reproduces the spatial layout and structural details of the furnace shell, insulation layer, refractory bricks and cooling water pipelines, providing a basic carrier for subsequent physical equation solving.

[0083] Furthermore, the geometric structure refers to the shape, size, and relative position of each component of the heating furnace, including the external dimensions of the furnace shell, the thickness of the insulation layer, the arrangement of the refractory bricks, and the direction and distribution of the cooling water pipes, which is the fundamental guarantee for the accuracy of the model.

[0084] In addition, the distribution of material properties refers to the distribution of the physicochemical properties of the materials used in each component of the heating furnace, such as the thermal conductivity of the furnace shell, the heat insulation performance of the insulation layer, the high temperature resistance parameters of the refractory bricks, and the material properties of the cooling water pipes, which directly affect the simulation accuracy of the model.

[0085] Understandably, by adopting the finite element method, based on the actual engineering drawings of the heating furnace, the furnace structure is discretized into multiple units to construct a three-dimensional mesh model of the heating furnace that includes the geometric structure and material property distribution of the furnace shell, insulation layer, refractory bricks and cooling water pipelines.

[0086] Step S12: Couple the solution of fluid flow equation, energy conservation equation and component transport equation in the three-dimensional mesh model of the heating furnace to obtain the target solution result, and generate an initial digital twin model based on the target solution result;

[0087] It should be noted that the fluid flow equations, namely the Navier-Stokes equations, are mathematical equations that describe the laws of fluid motion. They can reflect the relationship between fluid parameters such as velocity, pressure, and density and their changes over time and space, and are used to simulate the flow state of fluids inside a furnace.

[0088] In addition, the energy conservation equation is a mathematical equation based on the law of conservation of energy. It is used to describe the transfer, transformation and distribution of heat in the furnace, including heat transfer processes such as heat conduction, heat convection and heat radiation. It is the core equation for simulating the temperature field changes in the furnace.

[0089] Furthermore, the component transport equation is a mathematical equation that describes the diffusion, convection, and reaction transfer processes of various gas components in the furnace. It can reflect the concentration distribution and changing trends of key gases such as oxygen, carbon monoxide, and methane, providing support for simulating combustion and mass transfer processes.

[0090] In addition, the objective solution results are the calculation results obtained by coupling the above three types of equations in the three-dimensional mesh model of the heating furnace. They include data such as the fluid flow state, temperature field distribution, and gas component concentration distribution in the furnace, which comprehensively reflect the complex physicochemical processes in the furnace.

[0091] Furthermore, the initial digital twin model is a simulation model generated based on the objective solution results. It can initially simulate the operating state of the heating furnace, but it has not yet taken into account the deviation between the actual operating conditions and the theoretical model, and needs to be corrected in the future to improve accuracy.

[0092] In this embodiment, the three-dimensional mesh of the digital twin model is constructed using unstructured tetrahedral elements, with a total number of nodes of no less than 500,000. It can accurately reproduce the geometric structure and material property distribution of the heating furnace. The model is coupled to solve the Navier-Stokes equations, energy conservation equations, and component transport equations. The turbulence model adopts the standard k-ε model, the radiative heat transfer adopts the discrete coordinate method, and the combustion reaction adopts the finite rate / eddy dissipation model, comprehensively simulating the complex physicochemical processes of combustion, heat transfer, mass transfer, and fluid flow in the furnace.

[0093] Understandably, in the constructed three-dimensional mesh model of the heating furnace, the fluid flow equation, energy conservation equation and component transport equation are coupled and solved to obtain the target solution results of the relevant physicochemical processes in the furnace, and the initial digital twin model is generated accordingly.

[0094] Step S13: Input historical operating data into the multilayer fully connected neural network correction module to generate actual operating condition deviation;

[0095] It should be noted that historical operating data is relevant data accumulated during the past stable operation of the heating furnace, including process operating parameters such as fuel flow rate and combustion air flow rate, as well as the corresponding actual data collected by sensors. It is the basic data for training and correction modules.

[0096] In addition, the multi-layer fully connected neural network correction module is a neural network model composed of multiple fully connected layers. By learning from historical operating data, it can capture the difference between the theoretical model and the actual working conditions and has the ability to generate deviation correction terms.

[0097] In this embodiment, the multilayer fully connected neural network correction module is a three-layer fully connected structure, and its loss function adopts mean squared error. The combination of this structure and the loss function can accurately capture the difference between the theoretical model and the actual working conditions, providing support for generating reliable actual working condition deviations.

[0098] Furthermore, the actual operating condition deviation is a value calculated by the multilayer fully connected neural network correction module based on historical operating data. It is used to characterize the difference between the calculation results of the initial digital twin model and the actual operating state of the heating furnace, providing a basis for model correction.

[0099] Understandably, historical operating data of the heating furnace is collected and input into a pre-built multi-layer fully connected neural network correction module. The module then calculates and analyzes the data to generate actual operating condition deviations that reflect the differences between theory and practice.

[0100] Step S14: Correct the digital twin model according to the actual working condition deviation to obtain the target digital twin model.

[0101] Understandably, the actual operating condition deviation is used as a correction term to adjust the calculation results of the initial digital twin model, thereby completing the model correction and obtaining the target digital twin model.

[0102] Reference Figure 2 , Figure 2 This is a schematic diagram of the structure of the furnace leakage monitoring and control system in the first embodiment of the furnace leakage monitoring and control method of this application.

[0103] like Figure 2As shown, the heating furnace leakage monitoring and control system includes a multimodal sensor data acquisition module, a digital twin simulation calculation module, a multidimensional residual matrix generation module, a leakage feature identification and location module, and an intelligent closed-loop control decision-making module. The multimodal sensor data acquisition module is responsible for real-time acquisition of multidimensional state data during the heating furnace operation, including the internal temperature field of the furnace lining, acoustic emission signals from the outer surface of the furnace shell, gas component concentrations inside the furnace and flue, and pressure field data inside the furnace. All sensors achieve nanosecond-level time synchronization through a unified clock source. The digital twin simulation calculation module, based on computational fluid dynamics and thermo-structure interaction mechanics equations, constructs a digital twin model synchronized with the physical heating furnace in real time. It receives process operation parameter inputs and outputs a global theoretical state data matrix corresponding to the spatial position of the sensor array. The multidimensional residual matrix generation module subtracts the actual acquired data from the theoretical state data point by point to generate a time-seriesd multidimensional residual state matrix, representing the deviation between the actual operating state and the theoretical state. The leak feature identification and localization module inputs the residual matrix into a spatial temporal graph convolutional neural network model to extract and identify leak features. It outputs the leak determination result, type, severity level, and confidence level. If a leak is determined, the three-dimensional spatial coordinates of the leak source are determined using an attention activation map. The intelligent closed-loop control and decision-making module, based on the leak type, severity level, and location coordinates, activates a deep reinforcement learning decision-making model to generate the optimal control command sequence and sends it to the distributed control system for execution, enabling proactive intervention in leak events.

[0104] Step S20: Input the multi-dimensional state data into the target digital twin model to obtain the global theoretical state data matrix;

[0105] It should be noted that the global theoretical state data matrix is ​​the output of the target digital twin model, which corresponds one-to-one with the preset multimodal sensor array in spatial location. It includes data such as theoretical temperature field, theoretical acoustic emission background noise level, theoretical gas component concentration distribution and theoretical pressure field, which can comprehensively reflect the ideal state of the heating furnace during normal operation.

[0106] Understandably, the collected multi-dimensional state data is input into the target digital twin model, and the model compensates for the deviation through coupled calculation and correction modules, outputting a global theoretical state data matrix corresponding to the spatial position of the sensor.

[0107] Step S30: Establish a multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix;

[0108] It should be noted that the multidimensional residual state matrix is ​​a matrix constructed according to a preset time step. Its dimensions are consistent with the spatial layout of the sensor array. Each element corresponds to the residual value of a sensor position, and after normalization, it falls into the range of negative one to positive one, which is used to eliminate dimensional differences.

[0109] Understandably, at each preset time step, the multi-dimensional state data is subtracted point by point from the global theoretical state data matrix, and the generated residual signal is normalized to establish a time-series multi-dimensional residual state matrix.

[0110] In one feasible implementation, step S30 may include steps S31 to S34:

[0111] Step S31: Obtain a preset time step, and perform spatial alignment of the multi-dimensional state data and the global theoretical state data matrix on the preset time step to obtain an initial residual signal, wherein the initial residual signal includes temperature signal, sound signal, gas signal and pressure signal;

[0112] It should be noted that the preset time step is a fixed time interval set in advance for data comparison and analysis. It serves as the time reference for constructing the time-series residual matrix, ensuring the timeliness and consistency of data comparison.

[0113] In addition, spatial alignment is the process of matching multi-dimensional state data with the global theoretical state data matrix one by one according to the actual spatial deployment location of the sensors, so as to ensure that the actual data collected by each sensor can be compared with the corresponding theoretical data output by the model.

[0114] Furthermore, the initial residual signal is the original deviation signal obtained by subtracting the multi-dimensional state data and the global theoretical state data matrix point by point after spatial alignment. It directly reflects the initial difference between the actual working conditions and the ideal working conditions.

[0115] Additionally, the temperature signal is the deviation signal of the temperature field data corresponding to the initial residual signal. It is obtained by subtracting the theoretical temperature data from the actual temperature data and is used to reflect the abnormal deviation of the furnace body temperature distribution.

[0116] Furthermore, the sound signal is the deviation signal corresponding to the acoustic emission signal data in the initial residual signal. It is obtained by subtracting the actual acoustic emission signal energy from the theoretical background noise energy, and can capture the sound wave deviation caused by abnormal furnace structure.

[0117] Additionally, the gas signal is the deviation signal of the gas component concentration data corresponding to the initial residual signal. It is obtained by subtracting the theoretical gas concentration data from the actual gas concentration data and is used to reflect abnormal changes in the gas components inside the furnace.

[0118] Furthermore, the pressure signal is the deviation signal of the pressure field data corresponding to the initial residual signal. It is obtained by subtracting the theoretical pressure data from the actual pressure data and can reflect the abnormal fluctuations in the pressure distribution inside the furnace.

[0119] Understandably, a preset time step is first obtained, and at this time step, the multi-dimensional state data is matched and aligned with the global theoretical state data matrix one by one according to the spatial location of the sensor. The initial residual signal containing temperature signal, sound signal, gas signal and pressure signal is obtained by point-by-point subtraction.

[0120] Step S32: Normalize the initial residual signal to obtain a normalized residual signal;

[0121] It should be noted that normalization is a process that converts initial residual signals with different dimensions into a unified numerical range. In this embodiment, the initial residual signal is adjusted to the range of negative one to positive one. The core purpose is to eliminate the dimensional differences between different physical quantities such as temperature, sound, gas, and pressure.

[0122] In addition, the normalized residual signal is a standardized signal obtained by normalizing the initial residual signal. All signals are in the same numerical range, which provides a basis for horizontal comparison and can avoid the imbalance of feature weights caused by different units.

[0123] Understandably, a preset normalization algorithm is used to process the initial residual signal, mapping the values ​​of various signals uniformly to the range of negative one to positive one, thus obtaining a standardized normalized residual signal.

[0124] Step S33: Stack the normalized residual signals according to the time order to obtain a time-series data structure, wherein the dimensions of the time-series data structure are the number of time steps, the total number of sensors, the number of physical quantity types, and the feature dimension.

[0125] It should be noted that the time-series data structure is a four-dimensional data structure formed by superimposing normalized residual signals in chronological order, which can simultaneously reflect the spatial distribution characteristics and temporal evolution of the residual signals.

[0126] Additionally, the number of time steps is a time dimension parameter that constitutes the time-series data structure. It corresponds to the cumulative number of preset time steps and reflects the time span of the data.

[0127] Furthermore, the total number of sensors is a spatial dimension parameter that constitutes the time-series data structure, corresponding to the total number of sensors in the preset multimodal sensor array, reflecting the spatial coverage of the data.

[0128] Additionally, the number of physical quantity types is a physical dimension parameter that constitutes the time-series data structure. It corresponds to the quantity of four physical quantities: temperature, sound, gas, and pressure, reflecting the physical attribute categories of the data.

[0129] Furthermore, the feature dimension is a feature description parameter that constitutes the temporal data structure. It is used to quantify the specific feature information of each physical quantity, providing detailed support for subsequent leakage feature extraction.

[0130] It is understandable that the normalized residual signals at each preset time step are stacked sequentially according to the time sequence to form a time-series data structure with dimensions of time step number, total number of sensors, number of physical quantity types and feature dimension.

[0131] Step S34: Construct a multidimensional residual state matrix based on the time-series data structure.

[0132] Understandably, based on a time-series data structure, four-dimensional data is organized and arranged according to matrix rules to construct a multi-dimensional residual state matrix that can comprehensively characterize the spatiotemporal features of the residual signal.

[0133] Step S40: Input the multidimensional residual state matrix into the preset leakage identification model to obtain the leakage type, leakage severity level and leakage judgment result;

[0134] It should be noted that the preset leak identification model is a pre-trained spatial-temporal graph convolutional neural network model, which abstracts the sensor into a weighted undirected graph, captures the spatial correlation of the residual signal through graph convolution operation, and captures the temporal evolution trend through temporal convolution operation.

[0135] Additionally, the leakage type is a predefined category of possible leaks in the heating furnace, including three types: micropore leakage, crack leakage, and weld failure, used to specify the specific form of leakage.

[0136] Furthermore, the severity level of a leak is a classification system that characterizes the extent of a leak, divided into Level 1, Level 2, and Level 3, used to reflect the degree of impact of the leak on production safety.

[0137] Additionally, the leak determination result is a conclusion from the model output indicating whether a leak exists in the heating furnace, accompanied by a confidence level expressed as a percentage to quantify the reliability of the identification result.

[0138] Understandably, the time-series multidimensional residual state matrix is ​​input into the preset leakage identification model. The model deeply extracts and identifies leakage features, and outputs the leakage judgment result, leakage type, leakage severity level and confidence level.

[0139] In one feasible implementation, step S40 may include steps S41 to S44:

[0140] Step S41: Construct a sensor spatial adjacency matrix based on the sensor location, and construct a thermodynamic coupling weight matrix based on the heat transfer coefficient between sensor nodes.

[0141] It should be noted that the sensor spatial adjacency matrix is ​​a matrix constructed based on the physical positional relationship of each sensor in the multimodal sensor array. If the Euclidean distance between two sensors in physical space is less than a preset threshold, the corresponding element in the matrix is ​​set to one; otherwise, it is set to zero. It is used to characterize the physical adjacency association between sensors.

[0142] In addition, the thermodynamic coupling weight matrix is ​​a matrix constructed based on the heat transfer coefficient between sensor nodes. The heat transfer coefficient is calculated through finite element thermal analysis and is used to quantify the thermodynamic coupling strength between any two sensor nodes, serving as the weight of the edges in the matrix.

[0143] Furthermore, the heat transfer coefficient is a parameter that reflects the heat transfer capability between sensor nodes. It is obtained by applying a unit thermal perturbation in the digital twin model and calculating the steady-state heat flow distribution, and can reflect the thermodynamic correlation characteristics of the sensor location.

[0144] Understandably, the physical location information of each sensor is first obtained, the Euclidean distance between any two sensors is calculated and compared with a preset threshold, a sensor spatial adjacency matrix is ​​constructed, and then the heat transfer coefficient between sensor nodes is calculated through finite element thermal analysis to construct a thermodynamic coupling weight matrix.

[0145] Step S42: Input the multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix into the preset leak identification model to obtain the aggregated feature sequence;

[0146] It should be noted that the aggregated feature sequence is a feature sequence obtained by processing the three input matrices by the preset leakage identification model. It integrates the spatial correlation information of the residual signal and can highlight the diffusion pattern of the leakage signal in the spatial dimension.

[0147] In addition, the preset leak identification model, namely the spatial-temporal graph convolutional neural network model, can aggregate spatial features through graph convolution operations and capture temporal features through temporal convolution operations, thereby achieving in-depth extraction of leak features.

[0148] In this embodiment, the training of the spatial temporal graph convolutional neural network adopts the cross-entropy loss function, and the focus loss is introduced to alleviate the problem of class imbalance in the leakage scenario, ensuring that the model can effectively learn the features of a few leakage types. Its training data comes from thousands of different types, severity levels and locations of leakage scenarios simulated in the digital twin environment, providing the model with rich and realistic learning samples, ensuring the accuracy and generalization ability of the model's recognition.

[0149] It is understandable that the multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix are input into the preset leak identification model. The model uses the latter two matrices to aggregate the residual features of adjacent nodes through the spatial graph convolutional layer, and obtains an aggregated feature sequence containing spatial correlation.

[0150] In one feasible implementation, step S42 may include steps S421 to S423:

[0151] Step S421: Obtain the node feature matrix, learnable parameter matrix, and modified linear unit activation function;

[0152] It should be noted that the node feature matrix is ​​a set of feature data of each layer node in a spatial temporal graph convolutional neural network. The node feature matrix contains the residual feature information corresponding to all sensor nodes in that layer and is the basic data for the model to perform feature aggregation.

[0153] Additionally, the learnable parameter matrix is ​​a set of adjustable parameters used for feature transformation in a spatial temporal graph convolutional neural network. The learnable parameter matrix can be continuously optimized through model training to adapt to the extraction requirements of leaked features.

[0154] Furthermore, the modified linear unit activation function is a commonly used neural network activation function, used to perform nonlinear transformations on aggregated features, enhancing the model's ability to fit complex leakage features and improving the depth of feature extraction.

[0155] Understandably, the node feature matrix, learnable parameter matrix, and modified linear unit activation function are directly obtained from the pre-trained leak detection model after training, in order to prepare for subsequent feature aggregation.

[0156] Step S422: Determine the attribute feature vectors of the graph nodes based on the multidimensional residual state matrix;

[0157] It should be noted that the attribute feature vector of the graph node is vector data representing the state of each sensor node. Its data comes from the residual data at each moment in the multidimensional residual state matrix, which can accurately reflect the deviation between the actual working conditions and the theoretical working conditions at the location of the corresponding sensor.

[0158] It is understandable that residual data at each time step is extracted from the temporally sequenced multidimensional residual state matrix and used as the attribute feature vector of the corresponding sensor node in the spatial temporal graph convolutional neural network.

[0159] Step S423: The attribute feature vector, the sensor spatial adjacency matrix, the thermodynamic coupling weight matrix, the node feature matrix, the learnable parameter matrix, and the modified linear unit activation function are input into the preset leak identification model for aggregation to obtain an aggregated feature sequence.

[0160] Understandably, all input parameters are fed into a pre-defined leak detection model. The model performs aggregation operations through a spatial graph convolutional layer, fusing the spatial correlation features of the residual signals to obtain an aggregated feature sequence containing spatial correlation information. The formula for calculating the aggregated feature sequence is as follows:

[0161]

[0162] In the formula, =A⊙W+I, which is a weighted adjacency matrix with a self-loop, where A is the sensor space adjacency matrix, W is the thermodynamic coupling weight matrix, and I is the identity matrix; Its degree matrix, Let l be the node feature matrix of the l-th layer. For learnable parameter matrix, To modify the activation function of the linear unit.

[0163] Step S43: Classify and map the aggregated feature sequence to obtain the leakage type, leakage severity level, and leakage confidence.

[0164] It should be noted that classification mapping is the process of transforming aggregated feature sequences into preset category results. This is achieved through temporal convolutional layers and fully connected classification layers in the model. The core is to match the learned spatiotemporal features with preset leakage-related categories.

[0165] In addition, the leak confidence level is a percentage value used to quantify the reliability of the leak type and severity level determination results. The higher the value, the more reliable the determination results.

[0166] It is understandable that stacked one-dimensional causal convolutional layers are used to process aggregated feature sequences to capture temporal evolution trends. Then, through global average pooling layers and fully connected classification layers, the fused spatiotemporal features are mapped onto the probability distributions of preset leakage types, severity levels, and confidence levels to obtain the corresponding results.

[0167] Step S44: Determine whether the heating furnace has a leak based on the leak confidence level. If the heating furnace has a leak, determine the leak determination result based on the preset leak identification model.

[0168] Understandably, the obtained leakage confidence level is compared with a preset threshold. If the leakage confidence level exceeds the preset threshold, it is determined that the heating furnace has a leak, and a leakage judgment result is generated. If it does not exceed the preset threshold, it is determined that the heating furnace does not have a leak.

[0169] In one feasible implementation, step S44 may include steps S441 to S447:

[0170] Step S441: Obtain the leakage confidence threshold and the preset decay index;

[0171] It should be noted that the leakage confidence threshold is a pre-set critical value used to determine whether the heating furnace has an initial leakage. Additionally, the preset decay index is a key parameter used to calculate the coordinates of the leakage source.

[0172] In this embodiment, the leakage confidence threshold is set to 90%, which is a key critical value for distinguishing between real leakage and false anomalies; the preset number of nodes is five, which is used to screen the candidate sensor nodes with the highest activation intensity to locate the leakage source; the preset attenuation index is represented by α in the three-dimensional inverse distance weighted interpolation method, and takes a value of two, which is used to adjust the influence of the distance from the sensor node to the leakage source on the weight.

[0173] Understandably, pre-set leakage confidence thresholds and preset decay indices can be directly obtained from the preset parameter library to provide parameter support for subsequent leakage judgment and coordinate calculation.

[0174] Step S442: When the leakage confidence level is greater than the leakage confidence level threshold, it is determined that the heating furnace has an initial leakage.

[0175] It should be noted that the initial leakage situation refers to the leakage situation initially determined after the leakage confidence exceeds the preset threshold. This determination is based on the identification results of leakage characteristics and has not yet accurately located the leakage source, which needs further verification and confirmation.

[0176] Understandably, the leakage confidence level is compared with the obtained leakage confidence level threshold. If the leakage confidence level is greater than the threshold, it is preliminarily determined that there is an initial leakage in the heating furnace.

[0177] Step S443: When the heating furnace has an initial leakage, determine the node activation intensity according to the preset leakage identification model;

[0178] It should be noted that the node activation intensity is the activation degree of each sensor node in the last spatial graph convolutional layer of the preset leak identification model. It is calculated by the gradient weighted class activation mapping method and is directly related to the distance to the leak source and the leak intensity.

[0179] Understandably, after determining that there is an initial leak in the heating furnace, the node activation intensity data of the last spatial graph convolutional layer is extracted from the preset leak identification model.

[0180] Step S444: Identify a target sensor node set with a preset number of nodes based on the node activation intensity;

[0181] It should be noted that the preset number of nodes is a pre-set number of sensor nodes used to screen candidate leakage areas. In this embodiment, it is set to five to ensure the relevance and representativeness of the candidate areas.

[0182] In addition, the target sensor node set is a preset number of nodes selected from all sensor nodes with the highest node activation intensity, which is the core reference for locating the leakage source.

[0183] It is understandable that the activation intensity of all sensor nodes is sorted, and the nodes with the highest activation intensity (a preset number of nodes) are selected to form the target sensor node set.

[0184] Step S445: Determine the node coordinates and the distance to the leaking node based on the target sensor node set;

[0185] It should be noted that node coordinates are the actual position data of each sensor in the target sensor node set in physical space, including longitude, latitude and altitude information in three-dimensional space, and are the reference for calculating the coordinates of the leakage source.

[0186] Additionally, the leak node distance is the assumed spatial distance from the leak point to each target sensor node, which is a key parameter for calculating the weights and needs to be estimated by combining the node coordinates and the assumed location of the leak point.

[0187] Understandably, the physical location information corresponding to the target sensor node set is retrieved, the node coordinates of each node are determined, and the distance from the assumed leak point to each leak node is initially estimated.

[0188] Step S446: Calculate the coordinates of the leakage source based on the node coordinates, the distance to the leaking node, and the preset attenuation index;

[0189] It should be noted that the leak source coordinates are the actual three-dimensional spatial location data of the leak source calculated by the three-dimensional inverse distance weighted interpolation method. This data can accurately indicate the specific location where the leak occurred, with a positioning accuracy of within half a meter.

[0190] Understandably, using the node coordinates of the target sensor node as a reference, combined with the distance to the leaking node and a preset attenuation index, the precise three-dimensional spatial coordinates of the leak source are calculated using a three-dimensional inverse distance weighted interpolation method. The formula for calculating the leak source coordinates is as follows:

[0191]

[0192] In the formula, Let be the physical coordinates of the i-th highly active node; As weight; The coordinates of the leak source serve as the initial location result for subsequent control decisions.

[0193] Step S447: When the coordinates of the leakage source are present, it is determined that there is a target leakage in the heating furnace, and the leakage determination result is determined based on the coordinates of the leakage source.

[0194] It should be noted that the target leakage situation is the actual leakage situation confirmed after the leakage source has been accurately located. It is different from the initial leakage situation and has clear leakage location information, providing accurate basis for subsequent control.

[0195] In addition, the leakage determination result at this time not only indicates that there is a leak in the heating furnace, but also includes complete information such as the coordinates of the leak source, the type of leak and the severity level, which is the core basis for initiating closed-loop management.

[0196] Understandably, after successfully calculating the coordinates of the leak source, it is formally determined that there is a target leak in the heating furnace. Information such as the coordinates of the leak source, the type of leak, and the severity level are integrated to form a complete leak determination result.

[0197] Reference Figure 3 , Figure 3 This is a schematic diagram of the leakage feature identification and location process of the first embodiment of the heating furnace leakage monitoring and control method of this application.

[0198] like Figure 3 As shown, the actual multimodal sensing data is collected by a sensor array, including temperature field, acoustic emission signal, gas concentration, and pressure field data. The digital twin model outputs a theoretical state data matrix. Subtracting the two generates a time-seriesd multidimensional residual state matrix, representing the deviation between the actual and theoretical states. This matrix is ​​input into a spatial-temporal graph convolutional neural network. The spatial graph convolutional layer learns the spatial correlation of the leakage signal, and the one-dimensional causal convolutional layer captures the temporal evolution trend. Finally, the leakage determination result and the three-dimensional spatial coordinates of the leakage source are output.

[0199] Step S50: When the leakage determination result indicates that there is a target leakage in the heating furnace, a leakage control instruction is generated according to the leakage type and the leakage severity level, and leakage control of the heating furnace is performed according to the leakage control instruction.

[0200] It should be noted that the target leakage situation refers to a real leakage situation with a certain degree of confidence as determined by the preset leakage identification model, which is different from false anomalies caused by normal process fluctuations.

[0201] In addition, leakage control instructions are targeted operational instructions generated based on the type and severity of the leakage. They include discrete, executable atomic operations such as reducing the fuel supply and starting the backup cooling water pump, which are used to suppress the spread of leakage or guide the furnace to a safe state.

[0202] Understandably, when the leakage determination result indicates that there is a target leakage in the heating furnace, a leakage control instruction is generated based on the leakage type and leakage severity level, and the instruction is sent to the distributed control system for execution and control.

[0203] This embodiment provides a method for monitoring and controlling leaks in a heating furnace. It involves deploying a multimodal sensor array to collect multi-dimensional state data of the heating furnace in real time, constructing a digital twin model synchronized with the physical furnace, outputting a global theoretical state data matrix, establishing a time-seriesd multi-dimensional residual state matrix to represent the deviation between the actual and theoretical states, inputting the residual matrix into a spatial temporal graph convolutional neural network for in-depth extraction and identification of leak features, determining the three-dimensional spatial coordinates of the leak source, and generating and executing control commands based on the leak type and severity level. This method solves the technical problems of low leak monitoring sensitivity, high false alarm rate, inaccurate location, and lack of intelligent closed-loop control in existing technologies. It achieves the beneficial effects of enabling early warning, accurate location, and proactive intervention for leaks, significantly improving monitoring sensitivity, reducing false alarm rate, and ensuring production safety.

[0204] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 The step S50 of the heating furnace leakage monitoring and control method further includes steps S51 to S54:

[0205] Step S51: When the leakage determination result indicates that there is a target leakage in the heating furnace, a target state vector is generated based on the leakage type, the leakage severity level, the leakage source coordinates in the leakage determination result, and the current process operation parameters of the heating furnace.

[0206] It should be noted that process operation parameters are key data reflecting the current operating status of the heating furnace, including fuel flow rate, combustion air flow rate, material input rate, and flue damper opening, which are important inputs for generating the target state vector.

[0207] In addition, the target state vector is a comprehensive vector that integrates core leakage information with current operating data. It integrates leakage type, leakage severity level, leakage source coordinates and process operation parameters to provide comprehensive state support for deep network analysis.

[0208] Understandably, when a target leak is determined to exist in the heating furnace, the leak type, leak severity level, leak source coordinates, and current process operation parameters are extracted, and these data are integrated and processed to generate a target state vector.

[0209] Step S52: Input the target state vector into the deep network to generate leakage volume change rate, energy loss, equipment damage index and response delay time;

[0210] It should be noted that the deep network, specifically the deep Q-network algorithm structure, is the core component of a pre-trained deep reinforcement learning decision model. Through extensive offline simulation training in a digital twin environment, it has the ability to predict relevant parameters based on the input state.

[0211] In this embodiment, the specific atomic operations of the deep reinforcement learning decision model include reducing the fuel supply to a specific area by 10%, decreasing the opening of the combustion air valve in the corresponding area by 5%, starting the backup cooling water pump, increasing the opening of the flue damper by 3%, cutting off the fuel supply around the leak area, and starting the inert gas purging system. The weight coefficients λ1 to λ4 of the corresponding leak volume change rate, energy loss, equipment damage index, and response delay time in the long-term reward function are 0.4, 0.2, 0.3, and 0.1, respectively. The training architecture adopts a double-Q learning architecture, the target network is updated every 500 steps, and the experience replay buffer capacity is set to 100,000. The model is trained by simulating different leak scenarios offline in a digital twin environment to ensure that the generated control instructions have the best long-term benefits.

[0212] In addition, the leakage volume change rate is a parameter characterizing the change of leakage degree over time, reflecting the speed of leakage diffusion; the energy loss is the energy loss of the heating furnace caused by leakage, reflecting the impact of leakage on energy efficiency.

[0213] Furthermore, the equipment damage index is an indicator that quantifies the degree of damage caused by a leak to the heating furnace equipment, and the response delay time is the time interval between detecting a leak and implementing control actions. Both are key indicators for evaluating the effectiveness of control measures.

[0214] Understandably, the generated target state vector is input into a deep network, and through the network's calculation and analysis, the leakage volume change rate, energy loss, equipment damage index, and response delay time are output.

[0215] Step S53: Calculate the long-term return value of the control effect based on the leakage volume change rate, the energy loss, the equipment damage index, and the response delay time;

[0216] It should be noted that the long-term return on control effectiveness is a quantitative indicator that comprehensively evaluates the long-term benefits of control actions. Its calculation takes into account multiple objectives such as suppressing leakage and diffusion, reducing material and energy loss, avoiding secondary damage to equipment, and ensuring personnel safety, and can reflect the quality of control actions.

[0217] Understandably, preset weighting coefficients are assigned to the leakage volume change rate, energy loss, equipment damage index, and response delay time, and the long-term return value of the control effect is obtained through weighted calculation.

[0218] Step S54: Determine the target control action based on the long-term return value of the control effect, generate a leakage control instruction based on the target control action, and perform leakage control of the heating furnace based on the leakage control instruction.

[0219] It should be noted that the target control action is the optimal operation selected from a preset set of atomic operation instructions. Atomic operation instructions include discrete executable instructions such as reducing the fuel supply in a specific area, reducing the opening of the combustion air valve, and starting the standby cooling water pump. The target control action is a single instruction or combination of instructions that has the highest long-term return value for control effect.

[0220] Understandably, based on the magnitude of the long-term return value of the control effect, the target control action is determined from the set of atomic operation instructions, leakage control instructions are generated based on the target control action, and the instructions are sent to the distributed control system for execution and control.

[0221] This embodiment provides a method for monitoring and controlling leaks in a heating furnace. By integrating the leak type, leak severity level, leak source coordinates, and process operation parameters to generate a target state vector, and inputting it into a deep network to output the leak volume change rate, energy loss, equipment damage index, and response delay time, the method calculates the long-term return value of the control effect and determines the optimal target control action to generate control instructions. This solves the technical problems in the prior art where the control strategy and monitoring results lack deep coupling and the control action cannot be adaptively adjusted according to the dynamic parameters of the leak. It achieves the beneficial effects of realizing intelligent decision-making based on multi-objective optimization, improving the long-term benefits of control actions, and taking into account both safety and production continuity.

[0222] For example, to help understand the implementation process of the furnace leakage monitoring and control method obtained in this embodiment combined with the above embodiment one, please refer to... Figure 5 , Figure 5 A simplified flowchart of a method for monitoring and controlling leaks in a heating furnace is provided, specifically:

[0223] The multimodal sensor array deployment and data acquisition are responsible for collecting temperature, acoustic emission, gas concentration, and pressure data. It constructs and runs a digital twin model to output a theoretical state data matrix, generating a time-series multidimensional residual state matrix. The deviation signal is obtained by subtracting the theoretical data from the actual data. Leakage feature extraction and identification utilizes a spatial-temporal graph convolutional neural network to output leakage judgment results and type levels, determining the three-dimensional spatial coordinates of the leak source. The leak location is determined analytically using an attention activation graph. Closed-loop control commands are generated and executed. Based on the leak information, a deep reinforcement learning model is activated to generate and issue control commands to the control system for execution.

[0224] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the heating furnace leakage monitoring and control method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0225] This application also provides a heating furnace leakage monitoring and control device; please refer to... Figure 6 The furnace leakage monitoring and control device includes:

[0226] Data acquisition module 10 is used to acquire multi-dimensional state data and target digital twin model collected by a preset multimodal sensor array;

[0227] Matrix generation module 20 is used to input the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix;

[0228] The matrix generation module 20 is further configured to establish a multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix.

[0229] Leakage determination module 30 is used to input the multidimensional residual state matrix into a preset leakage identification model to obtain leakage type, leakage severity level and leakage determination result;

[0230] The leakage control module 40 is used to generate a leakage control instruction based on the leakage type and the leakage severity level when the leakage determination result is that there is a target leakage in the heating furnace, and to perform leakage control of the heating furnace according to the leakage control instruction.

[0231] The furnace leakage monitoring and control device provided in this application, employing the furnace leakage monitoring and control method described in the above embodiments, can solve the technical problem of difficulty in distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in the furnace. Compared with the prior art, the beneficial effects of the furnace leakage monitoring and control device provided in this application are the same as those of the furnace leakage monitoring and control method provided in the above embodiments, and other technical features of the furnace leakage monitoring and control device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0232] In one embodiment, the data acquisition module 10 is further configured to establish a three-dimensional mesh model of the heating furnace according to the finite element strategy, wherein the three-dimensional mesh model of the heating furnace includes the geometric structure and material property distribution of the furnace shell, insulation layer, refractory bricks and cooling water pipelines;

[0233] In the three-dimensional mesh model of the heating furnace, the fluid flow equation, the energy conservation equation, and the component transport equation are solved in a coupled manner to obtain the target solution result, and an initial digital twin model is generated based on the target solution result;

[0234] Historical operating data is input into a multi-layer fully connected neural network correction module to generate actual operating condition deviations.

[0235] The digital twin model is corrected based on the actual working condition deviation to obtain the target digital twin model.

[0236] In one embodiment, the matrix generation module 20 is further configured to obtain a preset time step, and to perform spatial alignment of the multi-dimensional state data and the global theoretical state data matrix on the preset time step to obtain an initial residual signal, wherein the initial residual signal includes a temperature signal, a sound signal, a gas signal, and a pressure signal.

[0237] The initial residual signal is normalized to obtain a normalized residual signal;

[0238] The normalized residual signals are stacked according to the time sequence to obtain a time-series data structure, wherein the dimensions of the time-series data structure are the number of time steps, the total number of sensors, the number of physical quantity types, and the feature dimension.

[0239] A multidimensional residual state matrix is ​​constructed based on the time-series data structure.

[0240] In one embodiment, the leakage determination module 30 is further configured to construct a sensor spatial adjacency matrix based on the sensor location and a thermodynamic coupling weight matrix based on the heat transfer coefficient between sensor nodes.

[0241] The multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix are input into a preset leak identification model to obtain an aggregated feature sequence.

[0242] The aggregated feature sequences are classified and mapped to obtain the leakage type, leakage severity level, and leakage confidence level.

[0243] The presence of a leak in the heating furnace is determined based on the leak confidence level. If a leak is found in the heating furnace, the leak determination result is determined based on the preset leak identification model.

[0244] In one embodiment, the leakage determination module 30 is further configured to acquire a node feature matrix, a learnable parameter matrix, and a modified linear unit activation function;

[0245] The attribute feature vectors of the graph nodes are determined based on the multidimensional residual state matrix;

[0246] The attribute feature vector, the sensor spatial adjacency matrix, the thermodynamic coupling weight matrix, the node feature matrix, the learnable parameter matrix, and the modified linear unit activation function are input into the preset leak identification model and aggregated to obtain an aggregated feature sequence.

[0247] In one embodiment, the leakage determination module 30 is further configured to obtain a leakage confidence threshold and a preset decay index;

[0248] When the leakage confidence level is greater than the leakage confidence threshold, it is determined that the heating furnace has an initial leakage.

[0249] When the heating furnace has an initial leak, the node activation intensity is determined according to the preset leak identification model;

[0250] A target sensor node set with a preset number of nodes is identified based on the node activation intensity.

[0251] Determine the node coordinates and the distance to the leaking node based on the target sensor node set;

[0252] The coordinates of the leakage source are calculated based on the node coordinates, the distance to the leaking node, and the preset attenuation index.

[0253] When the coordinates of the leak source are present, it is determined that there is a target leak in the heating furnace, and the leak determination result is determined based on the coordinates of the leak source.

[0254] In one embodiment, the leakage control module 40 is further configured to generate a target state vector based on the leakage type, the leakage severity level, the leakage source coordinates in the leakage determination result, and the current process operation parameters of the heating furnace when the leakage determination result indicates that there is a target leakage in the heating furnace;

[0255] The target state vector is input into a deep network to generate leakage volume change rate, energy loss, equipment damage index, and response delay time.

[0256] The long-term return on control effectiveness is calculated based on the leakage volume change rate, the energy loss, the equipment damage index, and the response delay time.

[0257] The target control action is determined based on the long-term return value of the control effect, and a leakage control instruction is generated based on the target control action. The leakage control instruction is then used to control the leakage of the heating furnace.

[0258] This application provides a furnace leakage monitoring and control device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the furnace leakage monitoring and control method in the above embodiment 1.

[0259] The following is for reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing the furnace leakage monitoring and control device in the embodiments of this application. The furnace leakage monitoring and control device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The heating furnace leakage monitoring and control equipment shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0260] like Figure 7As shown, the furnace leak monitoring and control device may include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in ROM (Read Only Memory) 1002 or the program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the furnace leak monitoring and control device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the furnace leak monitoring and control equipment to exchange data wirelessly or via wired communication with other devices. Although the figure shows furnace leak monitoring and control equipment with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0261] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0262] The furnace leakage monitoring and control equipment provided in this application, employing the furnace leakage monitoring and control method described in the above embodiments, can solve the technical problem of difficulty in distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in the furnace. Compared with the prior art, the beneficial effects of the furnace leakage monitoring and control equipment provided in this application are the same as those of the furnace leakage monitoring and control method provided in the above embodiments, and other technical features of this furnace leakage monitoring and control equipment are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0263] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0264] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0265] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the furnace leakage monitoring and control method in the above embodiments.

[0266] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), Erasable Programmable Read Only Memory (EPROM), optical fiber, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0267] The aforementioned computer-readable storage medium may be included in the furnace leak monitoring and control equipment; or it may exist independently and not be assembled into the furnace leak monitoring and control equipment.

[0268] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the heating furnace leakage monitoring and control equipment, the heating furnace leakage monitoring and control equipment: acquires multi-dimensional state data collected by a preset multi-modal sensor array and a target digital twin model; inputs the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix; establishes a multi-dimensional residual state matrix based on the multi-dimensional state data and the global theoretical state data matrix; inputs the multi-dimensional residual state matrix into a preset leakage identification model to obtain the leakage type, leakage severity level, and leakage determination result; when the leakage determination result indicates that the heating furnace has a target leakage situation, generates a leakage control instruction based on the leakage type and the leakage severity level, and performs heating furnace leakage control based on the leakage control instruction.

[0269] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0270] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0271] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0272] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described furnace leakage monitoring and control method. This solves the technical problem of difficulty in distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in the furnace. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the furnace leakage monitoring and control method provided in the above embodiments, and will not be elaborated upon here.

[0273] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the heating furnace leakage monitoring and control method described above.

[0274] The computer program product provided in this application can solve the technical problem of difficulty in distinguishing between normal process fluctuations and actual leakage anomalies under complex operating conditions and multi-source disturbances in heating furnaces. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the heating furnace leakage monitoring and control method provided in the above embodiments, and will not be repeated here.

[0275] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for monitoring and controlling leakage in a heating furnace, characterized in that, The method includes: The system acquires multi-dimensional state data and target digital twin model collected by a preset multi-modal sensor array. The multi-dimensional state data is a variety of data reflecting the operating status of the heating furnace, which are collected in real time by the multi-modal sensor array. These include temperature field data distributed along the depth of the furnace lining refractory material, acoustic emission signal data distributed on the outer surface of the furnace shell, gas component concentration data in the furnace and flue, and pressure field data inside the furnace. The multi-dimensional state data is input into the target digital twin model to obtain the global theoretical state data matrix; Based on the multi-dimensional state data and the global theoretical state data matrix, a multi-dimensional residual state matrix is ​​established; The multidimensional residual state matrix is ​​input into a preset leak identification model to obtain the leak type, leak severity level, and leak determination result. When the leakage determination result indicates that there is a target leakage in the heating furnace, a leakage control instruction is generated based on the leakage type and the leakage severity level, and leakage control of the heating furnace is performed according to the leakage control instruction.

2. The method as described in claim 1, characterized in that, The steps to obtain the target digital twin model include: A three-dimensional mesh model of the heating furnace is established based on the finite element method. The three-dimensional mesh model of the heating furnace includes the geometric structure and material property distribution of the furnace shell, insulation layer, refractory bricks and cooling water pipelines. The fluid flow equation, energy conservation equation, and component transport equation are coupled and solved in the three-dimensional mesh model of the heating furnace to obtain the target solution result, and an initial digital twin model is generated based on the target solution result. Historical operating data is input into a multi-layer fully connected neural network correction module to generate actual operating condition deviations. The digital twin model is corrected based on the actual working condition deviation to obtain the target digital twin model.

3. The method as described in claim 1, characterized in that, The step of establishing the multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix includes: A preset time step is obtained, and the multi-dimensional state data and the global theoretical state data matrix are spatially aligned at the preset time step to obtain an initial residual signal, wherein the initial residual signal includes temperature signal, sound signal, gas signal and pressure signal; The initial residual signal is normalized to obtain a normalized residual signal; The normalized residual signals are stacked according to the time sequence to obtain a time-series data structure, wherein the dimensions of the time-series data structure are the number of time steps, the total number of sensors, the number of physical quantity types, and the feature dimension. A multidimensional residual state matrix is ​​constructed based on the time-series data structure.

4. The method as described in claim 1, characterized in that, The step of inputting the multidimensional residual state matrix into a preset leak identification model to obtain the leak type, leak severity level, and leak determination result includes: A sensor spatial adjacency matrix is ​​constructed based on the sensor location, and a thermodynamic coupling weight matrix is ​​constructed based on the heat transfer coefficient between sensor nodes. The multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix are input into a preset leak identification model to obtain an aggregated feature sequence. The aggregated feature sequences are classified and mapped to obtain the leakage type, leakage severity level, and leakage confidence level. The presence of a leak in the heating furnace is determined based on the leak confidence level. If a leak is found in the heating furnace, the leak determination result is determined based on the preset leak identification model.

5. The method as described in claim 4, characterized in that, The step of inputting the multidimensional residual state matrix, the sensor spatial adjacency matrix, and the thermodynamic coupling weight matrix into the preset leak identification model to obtain the aggregated feature sequence includes: Obtain the node feature matrix, learnable parameter matrix, and modified linear unit activation function; The attribute feature vectors of the graph nodes are determined based on the multidimensional residual state matrix; The attribute feature vector, the sensor spatial adjacency matrix, the thermodynamic coupling weight matrix, the node feature matrix, the learnable parameter matrix, and the modified linear unit activation function are input into the preset leak identification model and aggregated to obtain an aggregated feature sequence.

6. The method as described in claim 4, characterized in that, The steps of determining whether the heating furnace has a leak based on the leak confidence level, and determining the leak determination result based on the preset leak identification model when the heating furnace has a leak, include: Obtain the leakage confidence threshold and preset decay index; When the leakage confidence level is greater than the leakage confidence threshold, it is determined that the heating furnace has an initial leakage. When the heating furnace has an initial leak, the node activation intensity is determined according to the preset leak identification model; A target sensor node set with a preset number of nodes is identified based on the node activation intensity. Determine the node coordinates and the distance to the leaking node based on the target sensor node set; The coordinates of the leakage source are calculated based on the node coordinates, the distance to the leaking node, and the preset attenuation index. When the coordinates of the leak source are present, it is determined that there is a target leak in the heating furnace, and the leak determination result is determined based on the coordinates of the leak source.

7. The method as described in claim 1, characterized in that, The step of generating a leak control instruction based on the leak type and the leak severity level when the leak determination result indicates that a target leak exists in the heating furnace, and then performing leak control on the heating furnace according to the leak control instruction, includes: When the leakage determination result indicates that there is a target leakage in the heating furnace, a target state vector is generated based on the leakage type, the leakage severity level, the leakage source coordinates in the leakage determination result, and the current process operation parameters of the heating furnace. The target state vector is input into a deep network to generate leakage volume change rate, energy loss, equipment damage index, and response delay time. The long-term return on control effectiveness is calculated based on the leakage volume change rate, the energy loss, the equipment damage index, and the response delay time. The target control action is determined based on the long-term return value of the control effect, and a leakage control instruction is generated based on the target control action. The leakage control instruction is then used to control the leakage of the heating furnace.

8. A device for monitoring and controlling leakage in a heating furnace, characterized in that, The device includes: The data acquisition module is used to acquire multi-dimensional state data and target digital twin model collected by a preset multi-modal sensor array. The multi-dimensional state data is a variety of data reflecting the operating status of the heating furnace, which are collected in real time by the multi-modal sensor array. These include temperature field data distributed along the depth direction of the furnace lining refractory material, acoustic emission signal data distributed on the outer surface of the furnace shell, gas component concentration data in the furnace and flue, and pressure field data inside the furnace. The matrix generation module is used to input the multi-dimensional state data into the target digital twin model to obtain a global theoretical state data matrix. The matrix generation module is also used to establish a multidimensional residual state matrix based on the multidimensional state data and the global theoretical state data matrix. The leakage determination module is used to input the multidimensional residual state matrix into a preset leakage identification model to obtain the leakage type, leakage severity level and leakage determination result. The leakage control module is used to generate a leakage control instruction based on the leakage type and the leakage severity level when the leakage determination result indicates that there is a target leakage in the heating furnace, and to perform leakage control of the heating furnace according to the leakage control instruction.

9. A device for monitoring and controlling leakage in a heating furnace, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the furnace leakage monitoring and control method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the heating furnace leakage monitoring and control method as described in any one of claims 1 to 7.

Images