An AI-based method and system for energy consumption optimization and control in refractory material production lines
By using an AI-based energy consumption optimization control method for refractory material production lines, and employing a neural network model to predict heat load deviation and calculate heat storage buffer margin, the problems of high energy consumption and poor stability under traditional control methods are solved, achieving stable kiln temperature control and optimized energy consumption.
Patent Information
- Application Number
- CN202611270627.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-20
- Publication Date
- 2026-11-10
AI Technical Summary
Traditional refractory material production line control methods rely on manual experience or simple feedback adjustments, resulting in poor equipment coordination, high energy consumption, and frequent adjustments due to deviations between the predictive control model and actual operating conditions when facing material batch fluctuations, which increases energy consumption and instability.
By employing an AI-based approach, the thermal load deviation is predicted using a neural network model by detecting the rate of change in mechanical resistance. Combined with the internal temperature and physical parameters of the kiln, the thermal storage buffer margin is calculated, and operating parameters are limited or adjusted to optimize the control strategy, thereby reducing energy waste and improving stability.
When switching material batches, the deviation of heat load is obtained in advance. By limiting or adjusting the operating parameters, the frequency of over-adjustment caused by material fluctuations is reduced, thereby improving the stability of kiln temperature control and the overall energy consumption optimization level.
Smart Images

Figure CN123067397A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of computer-aided manufacturing, and in particular relates to an AI-based method and system for optimizing and controlling energy consumption in a refractory material production line. Background Technology
[0002] Refractory materials are fundamental materials in high-temperature industries. Their production lines encompass multiple processes, including crushing, mixing, molding, and firing, making them typical high-energy-consuming manufacturing processes. Traditional refractory production line control typically relies on manual experience to set fixed operating parameters or simple single-loop feedback adjustments. This extensive control method is poorly adaptable to complex and variable production conditions, resulting in low coordination between different pieces of equipment, high overall energy consumption, and increased energy expenditure.
[0003] To reduce energy consumption in production, a predictive control system and method for cement rotary kiln calcination has been adopted. This method is also applicable to the control of refractory material firing kilns. This technology establishes a dynamic mathematical model of the firing kiln, combines it with real-time collected sensor data such as temperature and pressure, and uses a predictive control algorithm to continuously solve for the optimal combination of operating parameters (such as fuel supply and exhaust fan speed), which is then automatically sent to the underlying actuators. This model-based predictive control technology can dynamically adjust control commands according to the real-time status of the kiln, overcoming the lag of traditional manual control and reducing energy consumption during the firing process.
[0004] However, when the feed material fluctuates due to batch variations, its physicochemical properties and thermal load requirements change nonlinearly. Predictive control models in related technologies are primarily based on the established physicochemical mechanisms within the kiln and historical operating data, undergoing continuous optimization. Because the nonlinear changes in the physicochemical properties alter the actual thermodynamic reaction process within the kiln, a dynamic deviation arises between the preset predictive model and the actual operating conditions. This deviation causes a misalignment between the fuel and airflow adjustment commands output by the model and the actual thermal load requirements, requiring the kiln control system to rely on underlying feedback mechanisms for repeated corrections and adjustments when facing material fluctuations. This repeated correction and adjustment leads to frequent actuator movements, reducing the stability of the production line operation, increasing additional energy consumption during dynamic adjustments, and ultimately increasing the energy consumption of the refractory material production line. Summary of the Invention
[0005] This application provides an AI-based energy consumption optimization and control method and system for refractory material production lines, which can improve the stability of production line operation, reduce the additional energy consumption of equipment during dynamic adjustment, and thus reduce the energy consumption of refractory material production lines.
[0006] In the first aspect, this application provides an AI-based energy consumption optimization and control method for refractory material production lines, which collects mechanical operation data of upstream forming equipment of the production line when a change in raw material batch is detected in the refractory material production line.
[0007] Extract the rate of change of mechanical resistance of upstream forming equipment when processing a new batch of materials from mechanical operation data;
[0008] The rate of change of mechanical resistance is input into a pre-trained neural network model to obtain the expected deviation of heat load when a new batch of material enters the firing kiln;
[0009] Based on the current internal temperature field distribution data of the firing kiln and the physical parameters of the kiln lining, the current kiln heat storage buffer margin is calculated. The kiln heat storage buffer margin is the maximum heat value that the kiln lining can release or absorb autonomously to maintain the internal temperature of the kiln within the preset qualified process range, under the premise that the current operating parameters of the firing kiln actuator remain unchanged.
[0010] When the absolute value of the heat load deviation is determined to be less than or equal to the kiln's heat storage buffer margin, a limit instruction for the rate of change of operating parameters of the predictive control model for the firing kiln is generated.
[0011] The command limiting the rate of change of operating parameters is input into the predictive control model to control the magnitude of change of operating parameters output by the predictive control model to the actuator;
[0012] If the absolute value of the heat load deviation is greater than the kiln heat storage buffer margin, calculate the difference between the absolute value of the heat load deviation and the kiln heat storage buffer margin to obtain the feedforward compensation heat.
[0013] The feedforward compensation heat is input into the predictive control model so that the predictive control model outputs adjustment commands for the operating parameters of the actuator based on the feedforward compensation heat.
[0014] By adopting the above technical solution, when the batch switching of raw materials in the refractory material production line is detected, the expected thermal load deviation can be obtained in advance before the new batch of materials enters the kiln by extracting the rate of change of mechanical resistance of the upstream molding equipment and processing it through a neural network model. Combined with the kiln's internal temperature field and the physical parameters of the lining, the calculated kiln heat storage buffer margin allows for a quantitative assessment of the kiln's ability to smooth temperature fluctuations using its own thermal inertia. When the thermal load deviation is less than or equal to the buffer margin, the energy waste caused by excessive adjustment of the actuator is avoided by limiting the rate of change of the predictive control model's operating parameters. When the deviation is greater than the buffer margin, feedforward compensation heat is introduced by calculating the difference, enabling the predictive control model to make precise parameter adjustments in advance. This reduces the frequency of over-adjustment caused by batch switching of materials and improves the stability of the firing kiln temperature control and the overall energy consumption optimization level.
[0015] In conjunction with some implementation methods of the first aspect, in some implementation methods, the current heat storage buffer margin of the kiln is calculated based on the current internal temperature field distribution data of the firing kiln and the physical parameters of the kiln lining, specifically including:
[0016] Based on the internal temperature field distribution data, the kiln lining is divided into multiple virtual heat storage layers along the radial depth, and the current level temperature of each virtual heat storage layer is extracted.
[0017] Based on the physical parameters of the kiln lining and the current level temperature of each virtual heat storage layer, calculate the dynamic thermal diffusivity of each virtual heat storage layer at the current temperature.
[0018] The expected transition time of a new batch of material in the firing kiln is obtained, and combined with the dynamic thermal diffusivity of each virtual heat storage layer, the effective heat storage layer with a thermal response time less than or equal to the expected transition time is selected from multiple virtual heat storage layers.
[0019] Calculate the temperature difference between the current level temperature of each effective heat storage layer and the boundary critical temperature of the preset qualified process range;
[0020] Based on the specific heat capacity and density of the kiln lining, as well as the temperature difference, the available heat value of each effective heat storage layer that can be autonomously released or absorbed under the premise of keeping the current operating parameters of the actuator unchanged is calculated.
[0021] The available heat values of each effective heat storage layer are summed to obtain the current heat storage buffer margin of the kiln.
[0022] By adopting the above technical solution, when calculating the kiln's heat storage buffer margin, the kiln lining is divided into multiple virtual heat storage layers along its radial depth, and the dynamic thermal diffusivity is calculated based on the current temperature of each layer. By introducing the expected transition time of a new batch of material in the kiln, the system can screen out effective heat storage layers that meet the thermal response time requirements, excluding deep lining layers that cannot play a buffering role during material transition due to lag in heat conduction. Furthermore, by combining the physical parameters and temperature differences of the effective heat storage layers to calculate and accumulate the available heat of a single layer, the assessment of the buffer margin is transformed from a static overall estimation to a dynamic quantification based on the time dimension and spatial gradient. This improves the accuracy of the kiln's heat storage buffer margin assessment, thereby enhancing the matching degree between subsequent control strategies and the actual thermodynamic state of the kiln.
[0023] In conjunction with some implementations of the first aspect, in some implementations, the rate of change of mechanical resistance is input into a pre-trained neural network model to obtain the expected deviation of heat load when a new batch of material enters the firing kiln, specifically including:
[0024] The rate of change of mechanical resistance is input into the feature extraction layer of a pre-trained neural network model to extract the initial physical property coupling feature vector;
[0025] Obtain the estimated transfer time of a new batch of materials from the upstream forming equipment to the firing kiln, as well as the ambient temperature and humidity data of the production line during the estimated transfer time;
[0026] The expected transmission time and ambient temperature and humidity data are input into the environmental attenuation mapping layer of the neural network model to calculate the dynamic volatility compensation coefficient.
[0027] The initial physical property coupling feature vector is weighted and corrected using a dynamic volatilization compensation coefficient to obtain the target kiln entry feature vector;
[0028] The target kiln feature vector is input into the heat load decoding layer of the neural network model for decoding, so as to obtain the expected heat load deviation when a new batch of material enters the firing kiln.
[0029] By adopting the above technical solution, when predicting the thermal load deviation of a new batch of materials, the system not only extracts the initial physical property characteristics from the rate of change of mechanical resistance through a neural network, but also incorporates the expected transmission time of the material from the forming equipment to the firing kiln and the ambient temperature and humidity data during that period. By calculating the dynamic volatility compensation coefficient to weight and correct the initial physical property coupled feature vector, the system can quantitatively compensate for the physical property attenuation caused by the natural loss of moisture or volatiles during the transmission process. This makes the target feature vector input to the decoding layer more closely match the actual physical state of the material when it enters the kiln, reducing the prediction error caused by ignoring environmental interference during the transmission process, and improving the accuracy of assessing the expected thermal load deviation of a new batch of materials entering the kiln.
[0030] In conjunction with some implementations of the first aspect, in some implementations, after inputting the operating parameter rate-of-change limit command into the predictive control model, the method further includes:
[0031] Collect real-time pressure data from multiple pressure measurement points distributed along the material travel direction inside the firing kiln;
[0032] Based on real-time pressure data, calculate the pressure distribution gradient inside the firing kiln along the material travel direction;
[0033] Obtain the predicted value of the volatile gas release rate of the new batch of materials in the high-temperature zone of the firing kiln;
[0034] Based on the predicted values of pressure distribution gradient and volatile gas release rate, the expected values of local pressure surges in the high-temperature range are calculated.
[0035] When the expected value of local pressure change is determined to be greater than the preset safe pressure threshold, the secondary airflow distribution adjustment parameters are calculated based on the expected value of local pressure change.
[0036] The secondary air flow distribution adjustment parameters are input into the secondary air control system of the firing kiln to control the secondary air control system to adjust the air supply volume of the secondary air inlet in the corresponding area.
[0037] By adopting the above technical solution, after executing the operation parameter limitation command, the pressure distribution gradient is calculated by collecting real-time pressure data along the material's direction of travel inside the firing kiln. Combined with the predicted value of the volatile gas release rate of the new batch of material in the high-temperature zone, the expected value of potential local pressure surges in the high-temperature zone can be predicted in advance. When it is determined that this expected value exceeds the safety threshold, the secondary airflow distribution adjustment parameters are calculated and output to control the secondary air inlet airflow in the corresponding area. This allows the system to balance local pressure accumulation by dynamically distributing the secondary airflow before the actual occurrence of a pressure surge, reducing the risk of local pressure imbalance inside the kiln due to the violent volatilization of new material, and improving the stability of the aerodynamic environment inside the firing kiln and the operational safety of the combustion system.
[0038] In conjunction with some implementation methods of the first aspect, in some implementation methods, the secondary airflow distribution adjustment parameters are calculated based on the expected value of local pressure surges, specifically including:
[0039] The target ejector momentum is calculated based on the expected value of local pressure change and the cross-sectional dimensions of the firing kiln in the high-temperature range.
[0040] Obtain the real-time local temperature of each secondary air inlet within the high-temperature range, and calculate the local temperature drop margin corresponding to each secondary air inlet.
[0041] Based on the local temperature drop margin of each secondary air inlet, the momentum weight of each secondary air inlet is determined.
[0042] Calculate the target air volume for each secondary air inlet based on the target ejector momentum and the momentum weight of each secondary air inlet.
[0043] The target air supply volume set of each secondary air inlet is used as the secondary air flow distribution adjustment parameter.
[0044] By adopting the above technical solution, when calculating the secondary air flow distribution adjustment parameters, the target ejector momentum is calculated based on the expected value of local pressure surge and the kiln cross-sectional dimensions. The local temperature drop margin is then calculated by combining this with the real-time local temperature of each secondary air inlet. Based on this margin, the momentum weight of each inlet is determined to allocate the target air volume. This mechanism closely integrates the pressure regulation requirements with the actual heat-bearing capacity of each area of the kiln. When secondary air needs to be introduced to smooth out local pressure surges, more air volume is preferentially allocated to areas with larger temperature drop margins, avoiding indiscriminate air supply that could lead to overcooling in areas that were originally at lower temperatures. This dynamic flow distribution strategy based on local thermal conditions reduces the risk of drastic temperature fluctuations inside the kiln caused by the introduction of secondary air, and improves the spatial distribution stability of the thermal environment in the high-temperature zone and the consistency of refractory material firing quality.
[0045] In conjunction with some embodiments of the first aspect, in some embodiments, after inputting the secondary air flow distribution adjustment parameters into the secondary air control system of the firing kiln, the method further includes:
[0046] Based on the secondary airflow distribution adjustment parameters and the main exhaust air velocity of the firing kiln, the dynamic migration trajectory of the ejected volatile gas inside the firing kiln is calculated.
[0047] Based on the predicted value of the volatile gas release rate, the expected heat that can be released by the ejected volatile gas along the dynamic migration trajectory is calculated.
[0048] Based on the current internal temperature field distribution data of the firing kiln, a low heat buffer zone is located inside the firing kiln where the heat storage consumption has reached the preset consumption warning value.
[0049] Given that the dynamic migration trajectory passes through a low-heat buffer zone, the target residence time of the ejected volatile gas in the low-heat buffer zone is calculated based on the expected release of heat.
[0050] Based on the target residence time, generate an adjustment command for the opening degree of the exhaust branch valves in the low-heat buffer zone;
[0051] The opening command of the exhaust branch valve is input into the exhaust control system of the firing kiln to control the exhaust control system to reduce the opening of the exhaust branch valve corresponding to the low heat buffer zone.
[0052] By adopting the above technical solution, after adjusting the secondary air supply volume, the dynamic migration trajectory of the ejected volatile gas inside the kiln and its expected release of heat are calculated. Combined with the internal temperature field, a low-heat buffer zone where heat storage consumption reaches a warning value is located. When the gas migration trajectory passes through this low-heat buffer zone, the system calculates the target residence time based on the expected release of heat and accordingly reduces the opening of the corresponding exhaust branch valve. Utilizing the latent calorific value of the volatile gases released by the material itself, and by actively intervening in the local exhaust flow velocity, the residence time of the heat-carrying volatile gas is extended in the heat-deficient area, promoting the full release of its latent heat and absorption by the kiln lining in a low-heat state. This reduces the dependence of the local heat-deficient area on the external main heating energy source and improves the overall thermal energy utilization efficiency of the firing kiln.
[0053] In conjunction with some embodiments of the first aspect, in some embodiments, the target residence time of the ejected volatile gas in the low-heat buffer region is calculated based on the expected release of heat, specifically including:
[0054] Acquire real-time oxygen concentration data in the low-heat buffer zone and the ignition temperature of the ejected volatile gas;
[0055] Based on real-time oxygen concentration data and ignition temperature, the combustion reaction hysteresis coefficient of the ejected volatile gas in the low-heat buffer region is calculated.
[0056] Calculate the basic heat release time of the ejected volatile gas based on the expected heat release and the current stratum temperature of the low-temperature buffer zone.
[0057] By using the combustion reaction hysteresis coefficient to weight and compensate the basic heat release time, the target residence time of the ejected volatile gas in the low-heat buffer region is obtained.
[0058] By employing the above technical solution, when calculating the target residence time of the ejected volatile gas in the low-heat buffer zone, real-time oxygen concentration data and the ignition temperature of the gas in this zone are obtained to calculate the combustion reaction hysteresis coefficient. This coefficient is then used to weight and compensate the baseline heat release time calculated based on the expected release heat and the current temperature level. This mechanism fully considers the actual chemical reaction kinetics conditions within the low-heat buffer zone, recognizing that insufficient local oxygen concentration or temperature differences can slow down the combustion reaction rate of the volatile gas. This results in the final exhaust valve adjustment command being more closely aligned with the actual combustion heat release process of the gas, reducing the probability of gas being discharged before complete heat release due to insufficient consideration of the combustion environment. This improves the completeness of the release of latent heat of the volatile gas in the low-heat zone and the accuracy of local temperature compensation.
[0059] Secondly, embodiments of this application provide an AI-based energy consumption optimization control system for a refractory material production line. This AI-based energy consumption optimization control system for a refractory material production line includes: one or more processors and a memory; the memory is coupled to one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and one or more processors call the computer instructions to cause the system to perform the method described in the first aspect and any possible implementation thereof.
[0060] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a system, cause the system to perform the method described in the first aspect and any possible implementation thereof.
[0061] Fourthly, embodiments of this application provide a computer program product that, when run on a system, causes the system to execute the method described in any possible implementation of the first aspect.
[0062] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:
[0063] 1. This application provides an AI-based energy consumption optimization control method for refractory material production lines. When a batch change of raw materials is detected in the refractory material production line, the method extracts the rate of change of mechanical resistance of the upstream forming equipment and processes it through a neural network model. This allows for the prediction of the expected heat load deviation before the new batch of materials enters the kiln. Combined with the kiln's internal temperature field and the physical parameters of the lining, the system calculates the kiln's heat storage buffer margin, enabling a quantitative assessment of the kiln's ability to mitigate temperature fluctuations using its own thermal inertia. When the heat load deviation is less than or equal to the buffer margin, limiting the rate of change of the predictive control model's operating parameters avoids energy waste caused by excessive adjustment of the actuator. When the deviation exceeds the buffer margin, calculating the difference introduces feedforward compensation heat, allowing the predictive control model to make precise parameter adjustments in advance. This reduces the frequency of over-adjustment caused by material batch changes and improves the stability of the firing kiln's temperature control and the overall energy consumption optimization level.
[0064] 2. This application provides an AI-based energy consumption optimization and control method for a refractory material production line. After executing the operation parameter limit command, the method calculates the pressure distribution gradient by collecting real-time pressure data along the material travel direction inside the firing kiln, and combines this with the predicted value of the volatile gas release rate of the new batch of material in the high-temperature zone. This allows for the prediction of potential local pressure surges in the high-temperature zone. When the predicted value exceeds the safety threshold, the method calculates and outputs secondary air flow distribution adjustment parameters to control the secondary air inlet air volume in the corresponding area. This enables the system to balance local pressure accumulation by dynamically distributing secondary air flow before the actual pressure surge occurs, reducing the risk of local pressure imbalance inside the kiln due to the violent volatilization of new material, and improving the stability of the aerodynamic environment inside the firing kiln and the operational safety of the combustion system.
[0065] 3. This application provides an AI-based energy consumption optimization and control method for a refractory material production line. After adjusting the secondary air supply volume, the system calculates the dynamic migration trajectory of the ejected volatile gas inside the kiln and its expected release of heat. Combined with the internal temperature field, it locates a low-heat buffer zone where the heat storage consumption reaches a warning value. When the gas migration trajectory passes through this low-heat buffer zone, the system calculates the target residence time based on the expected release of heat and accordingly reduces the opening of the corresponding exhaust branch valve. By utilizing the latent calorific value of the volatile gas released by the material itself and actively intervening in the local exhaust flow rate, the system extends the residence time of the heat-carrying volatile gas in the heat-deficient area, allowing its latent heat to be fully released and absorbed by the kiln lining in a low-heat state. This reduces the dependence of the local heat-deficient area on the external main heating energy source and improves the overall thermal energy utilization efficiency of the firing kiln. Attached Figure Description
[0066] Figure 1 This is a flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line, as described in an embodiment of this application.
[0067] Figure 2 This is another flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line, as described in this application.
[0068] Figure 3 This is another flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line, as described in this application.
[0069] Figure 4 This is a schematic diagram of the physical device structure of an AI-based energy consumption optimization control system for a refractory material production line, as provided in an embodiment of this application. Detailed Implementation
[0070] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.
[0071] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.
[0072] The following example is used in conjunction with Figure 1 This application describes an AI-based energy consumption optimization and control method for a refractory material production line:
[0073] Please see Figure 1 This is a flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line, as described in this application.
[0074] S101. When a change in raw material batch is detected in the refractory material production line, collect the mechanical operation data of the upstream forming equipment of the production line;
[0075] Raw material batches in a refractory material production line refer to the collection of different batches or proportions of raw materials used in the production process, including but not limited to mixtures of different types of refractory materials such as magnesia-carbon brick raw materials and high-alumina brick raw materials. Switching refers to the process of the production line changing its state from processing one batch of raw materials to processing another. Upstream forming equipment refers to the mechanical devices used before the firing process to press, extrude, or cast powdered or slurry-like raw materials into green bodies of specific shapes, including but not limited to hydraulic presses, friction brick presses, or extrusion molding machines. Mechanical operating data refers to the various physical states and electrical parameters generated during the operation of the above-mentioned equipment, including but not limited to spindle speed, motor operating current, hydraulic system pressure, equipment vibration frequency, and torque data.
[0076] The system can achieve this step through deep interaction between the Enterprise Resource Planning (ERP) system and the Programmable Logic Controller (PLC). The system monitors production work order instructions issued by the ERP system in real time. When it receives a signal indicating a change in the raw material formula number within the work order, it determines that a raw material batch switch has occurred on the refractory material production line. Subsequently, the system immediately sends a data acquisition trigger command to the PLC connected to the upstream molding equipment via the Industrial Ethernet protocol. Based on the command, the PLC synchronously reads data from the current transformers and torque sensors installed on the upstream molding equipment at a high-frequency sampling rate. The acquired motor operating current and spindle torque are used as mechanical operating data, packaged and uploaded to the system's data processing platform via a message queue telemetry transmission protocol for storage and subsequent analysis.
[0077] The system can also achieve this step using machine vision combined with IoT sensing technology. A high-resolution industrial camera is deployed above the raw material conveyor belt. Using a pre-set image recognition algorithm, the system analyzes the color, texture, and particle size characteristics of the material on the conveyor belt in real time. When the Euclidean distance between the material feature vector extracted from multiple consecutive frames of images and the feature template of the current batch exceeds a set threshold, the system determines that a raw material batch has been switched in the refractory material production line. At this time, the system activates wireless IoT vibration sensors and pressure transmitters deployed on key transmission components of the upstream forming equipment. Through low-power wide-area network technology, it collects the vibration acceleration and hydraulic cylinder working pressure of the equipment in real time. These physical quantities are used as mechanical operation data, and after preliminary filtering and noise reduction by an edge computing gateway, they are pushed to the system's cloud server.
[0078] S102. Extract the rate of change of mechanical resistance of the upstream forming equipment when processing a new batch of materials from the mechanical operation data;
[0079] New batch material refers to a completely new proportion or source of refractory material mixture currently being processed by upstream molding equipment after a raw material batch switch. The rate of change in mechanical resistance refers to the degree of change in the mechanical force consumed by the upstream molding equipment to overcome the plastic deformation or friction of the material during processing, used to characterize the dynamic fluctuations in the rheological properties and molding difficulty of the new batch material.
[0080] The system can implement this step using a sliding time window combined with a time series differencing algorithm. After acquiring mechanical operation data, the system extracts the motor operating current and spindle torque data and constructs a fixed-length sliding time window. As data flows in in real time, the system calculates the average torque data within the current time window as the current mechanical resistance characterization value and extracts the average torque of the adjacent previous time window as the historical mechanical resistance characterization value. The system subtracts the historical mechanical resistance characterization value from the current mechanical resistance characterization value and divides it by the time interval between the two time windows to obtain the first-order time series differencing result. The system uses this differencing result as the rate of change of mechanical resistance of the upstream forming equipment when processing a new batch of material and stores it in the system's real-time database for downstream models to access.
[0081] The system can also achieve this step through frequency domain analysis combined with a Kalman filter estimation algorithm. The system extracts hydraulic system pressure and equipment vibration frequency data from the mechanical operation data. First, it uses a Fast Fourier Transform to convert the time-domain pressure and vibration data to the frequency domain, extracting the energy amplitude of a specific dominant frequency band directly related to the equipment's work. Then, the system constructs a Kalman filter state-space model, inputting the extracted energy amplitude as the observed variable into the model. Through the model's state prediction update equation, high-frequency random noise interference during mechanical operation is filtered out, estimating the actual work power of the equipment overcoming material resistance. The system further differentiates the actual work power over continuous time steps, calculating the slope of power change over time, and outputs this slope as the rate of change of mechanical resistance.
[0082] S103. Input the rate of change of mechanical resistance into the pre-trained neural network model to obtain the expected deviation of heat load when a new batch of material enters the firing kiln;
[0083] The system inputs the rate of change of mechanical resistance into a pre-trained neural network model to obtain the expected thermal load deviation when a new batch of material enters the firing kiln. Specifically, this includes: inputting the rate of change of mechanical resistance into the feature extraction layer of the pre-trained neural network model to extract the initial physical property coupling feature vector; obtaining the estimated transmission time of the new batch of material from the upstream forming equipment to the firing kiln, as well as the ambient temperature and humidity data of the production line during the estimated transmission time; inputting the estimated transmission time and ambient temperature and humidity data into the environmental attenuation mapping layer of the neural network model to calculate the dynamic volatilization compensation coefficient; using the dynamic volatilization compensation coefficient to weight and correct the initial physical property coupling feature vector to obtain the target kiln entry feature vector; and inputting the target kiln entry feature vector into the thermal load decoding layer of the neural network model for decoding to obtain the expected thermal load deviation when a new batch of material enters the firing kiln.
[0084] A pre-trained neural network model refers to a deep learning architecture that has undergone iterative optimization of weight parameters using a large amount of historical production data, enabling it to map heat load relationships from multi-source heterogeneous data. Heat load deviation refers to the expected difference between the actual heat consumed by a new batch of material during firing and the current kiln's basic heating baseline. The feature extraction layer is a network layer in the neural network model used for high-dimensional spatial mapping and latent feature mining of one-dimensional time-series or scalar data. The initial physical property coupling feature vector is a high-dimensional numerical matrix characterizing the correlation between the material's mechanical forming characteristics and potential thermal properties. The expected transport time refers to the physical flow time required for the material to leave the forming process and reach the firing process entrance. Environmental temperature and humidity data refer to the temperature and relative humidity values of the air in the production workshop. The environmental attenuation mapping layer is a network module in the model specifically used to assess the impact of the external environment on the loss of moisture or volatiles from the material. The dynamic volatility compensation coefficient is a correction multiplier used to quantify the degree of calorific value or moisture loss caused by environmental factors during material transport. The target kiln entry feature vector is a comprehensive feature expression that truly reflects the state of the material at the instant it enters the kiln, after correction for environmental factors. The thermal load decoding layer is the output module in the model responsible for restoring high-dimensional abstract features into predicted values of specific physical quantities.
[0085] The system can implement this refined technical solution through a deep learning architecture based on the fusion of multilayer perceptrons and one-dimensional convolutional networks. The underlying structure of this pre-trained neural network model includes a feature extraction layer composed of one-dimensional convolutional kernels, an environmental attenuation mapping layer composed of fully connected networks, and a heat load decoding layer composed of multilayer perceptron regressors. Its training and test sets are derived from a historical production database, containing historical mechanical resistance change rates, historical transmission times, historical environmental temperature and humidity, and labels of actual heat load deviations calculated from thermal instruments. The system inputs the extracted mechanical resistance change rate into the feature extraction layer, and through one-dimensional convolution operations and nonlinear activation function processing, extracts the initial physical property coupling feature vector. The system obtains the estimated transmission time by reading the logistics cycle time from the manufacturing execution system and acquires environmental temperature and humidity data through workshop temperature and humidity sensors. The system concatenates the estimated transmission time and environmental temperature and humidity data and inputs them into the environmental attenuation mapping layer. After weight matrix multiplication by the fully connected layer, a dynamic volatility compensation coefficient is calculated. The system uses this coefficient and the initial physical property coupling feature vector to perform element-wise multiplication operations for weighted correction, obtaining the target kiln entry feature vector. The system inputs the target kiln feature vector to the heat load decoding layer, and outputs the expected heat load deviation when a new batch of material enters the firing kiln through the linear mapping of the multilayer perceptron.
[0086] The system can also implement this refined technical solution using a deep learning model based on a combination of long short-term memory networks and attention mechanisms. The underlying structure of this pre-trained neural network model includes a feature extraction layer composed of long short-term memory units, an environmental attenuation mapping layer composed of self-attention mechanism modules, and a thermal load decoding layer composed of a fully connected network with a dropout layer. The training and test sets contain multi-dimensional sensor sequence data and corresponding thermal load deviations from the true values during multiple batch material switching. The system inputs the rate of change of mechanical resistance as a sequence into the feature extraction layer. The long short-term memory units update the cell state through forget gates and input gates, outputting an initial physical property coupling feature vector containing temporal dependencies. The system tracks and calculates the estimated transmission time using RFID tags and calls the plant's micro-weather station interface to obtain environmental temperature and humidity data. The system inputs the estimated transmission time and environmental temperature and humidity data into the environmental attenuation mapping layer. The self-attention mechanism module calculates the attention weight distribution of environmental factors and material properties, thereby generating a dynamic volatility compensation coefficient. The system uses this coefficient to perform tensor dot product weighted correction on the initial physical property coupling feature vector to obtain the target kiln entry feature vector. The system inputs the target kiln entry feature vector to the heat load decoding layer. After dimensionality reduction and inverse normalization processing by a fully connected network, the expected heat load deviation when a new batch of material enters the firing kiln is decoded.
[0087] S104. Calculate the current heat storage buffer margin of the kiln based on the current internal temperature field distribution data of the firing kiln and the physical parameters of the kiln lining.
[0088] The system calculates the current heat storage buffer margin of the kiln based on the current internal temperature field distribution data of the firing kiln and the physical parameters of the kiln lining. The heat storage buffer margin of the kiln is the maximum heat value that the kiln lining can release or absorb to maintain the internal temperature of the kiln within the preset qualified process range, under the premise that the current operating parameters of the firing kiln actuator remain unchanged.
[0089] Specifically, based on the internal temperature field distribution data, the system divides the kiln lining into multiple virtual heat storage layers along its radial depth and extracts the current level temperature of each virtual heat storage layer. Based on the physical parameters of the kiln lining and the current level temperature of each virtual heat storage layer, the system calculates the dynamic thermal diffusivity of each virtual heat storage layer at the current temperature. It obtains the expected transition time of a new batch of material within the firing kiln and, combined with the dynamic thermal diffusivity of each virtual heat storage layer, selects effective heat storage layers from the multiple virtual heat storage layers whose thermal response time is less than or equal to the expected transition time. The system calculates the temperature difference between the current level temperature of each effective heat storage layer and the boundary critical temperature of the preset qualified process range. Based on the specific heat capacity and density in the physical parameters of the kiln lining, and the temperature difference, the system calculates the single-layer usable heat value that each effective heat storage layer can autonomously release or absorb while keeping the current operating parameters of the actuator unchanged. Finally, the system sums the single-layer usable heat values of all effective heat storage layers to obtain the current kiln heat storage buffer margin.
[0090] Internal temperature field distribution data refers to the set of real-time temperature measurements at various spatial nodes and different depths of the lining within the firing kiln. The kiln lining refers to the refractory material masonry laid inside the kiln's metal shell for heat insulation and to withstand high temperatures. Physical parameters refer to the inherent thermal properties of the kiln lining material, including but not limited to thermal conductivity, specific heat capacity, and density. The kiln's heat storage buffer margin refers to the maximum capacity of the kiln to mitigate temperature fluctuations by relying on its own stored heat energy without changing the external heat source input. The actuator refers to the physical device used to regulate the kiln's heat supply, including but not limited to gas regulating valves, combustion air dampers, or electric heating power controllers. Operating parameters refer to the specific numerical commands controlling the actuator's actions, such as valve opening percentage. The preset qualified process range refers to the normal upper and lower temperature limits required to ensure the firing quality of the refractory material. The maximum heat value refers to the limit of heat energy that can be utilized within the allowable temperature fluctuation range. The virtual heat storage layer is a set of thin-layer computational units discretized along the thickness direction of the continuous kiln lining in a mathematical model. The current layer temperature is the average temperature of each thin-layer computational unit. Dynamic thermal diffusivity is a physical quantity characterizing a material's ability to transfer temperature changes at a specific temperature. Expected transition time is the estimated duration for a new batch of material to remain or pass through a specific temperature zone in the kiln. Thermal response time is the time required for heat to be conducted within the heat storage layer and released into the kiln. An effective heat storage layer refers to a virtual heat storage layer whose thermal response time is less than or equal to the expected transition time. This threshold relationship stems from the time-scale matching principle of heat transfer to ensure timely heat exchange during material passage. Boundary critical temperature is the upper or lower limit temperature value of a preset qualified process range. Temperature difference is the absolute difference between the current layer temperature and the boundary critical temperature. Available heat value per layer is the thermal energy that a single effective heat storage layer can contribute or absorb.
[0091] The system can implement this refinement technique using finite difference numerical calculations combined with discrete integral algorithms. The system acquires internal temperature field distribution data and uses a one-dimensional finite difference mesh generation method to divide the kiln lining into multiple virtual heat storage layers at equal intervals along its radial depth. A spatial interpolation algorithm is then used to extract the current layer temperature of the center node of each virtual heat storage layer. The system retrieves the physical parameters of the kiln lining from a material property database. Based on the empirical formula for the change of the lining's thermal conductivity with temperature, and combined with the current layer temperature of each virtual heat storage layer, the system calculates the dynamic thermal diffusivity of each virtual heat storage layer at the current temperature. The system calculates the expected transition time of a new batch of material in the firing kiln according to the production scheduling system. Using the heat penetration depth formula derived from Fourier's law of thermal conductivity, the system calculates the thermal response time of each virtual heat storage layer and selects effective heat storage layers whose thermal response time is less than or equal to the expected transition time. Finally, the system calculates the temperature difference between the current layer temperature of each effective heat storage layer and the boundary critical temperature of the preset qualified process range. Based on the specific heat capacity and density of the kiln lining, the system uses thermodynamic sensible heat calculation formulas to multiply density, specific heat capacity, virtual heat storage layer volume, and temperature difference to calculate the usable heat value of each effective heat storage layer, while keeping the current operating parameters of the actuator constant. The system then uses a discrete summation algorithm to accumulate the usable heat values of all effective heat storage layers to obtain the current kiln heat storage buffer margin.
[0092] The system can also achieve this refinement technique through a finite element analysis model combined with matrix operations. Based on internal temperature field distribution data, the system divides the mesh into multiple virtual heat storage layers along the radial depth in a pre-constructed three-dimensional finite element geometric model of the kiln lining, and extracts the average value of each virtual heat storage layer's mesh nodes as the current layer temperature. The system reads the kiln lining physical parameters from the configuration file and calculates the dynamic thermal diffusivity of each virtual heat storage layer at the current temperature using a lookup table method combined with polynomial fitting. The system obtains the expected transition time of a new batch of material by analyzing the running speed of the kiln conveyor belt, simulates the heat transfer process using a finite element transient thermal analysis solver, extracts the thermal response time of each layer, and thus filters out effective heat storage layers whose thermal response time is less than or equal to the expected transition time. The system calculates the temperature difference between the current layer temperature of each effective heat storage layer and the boundary critical temperature of the preset qualified process range. The system constructs a parameter vector containing the specific heat capacity, density, and volume of each effective heat storage layer. It then uses matrix multiplication to perform a dot product operation between the parameter vector and the temperature difference vector to calculate the usable heat value that each effective heat storage layer can autonomously release or absorb while keeping the current operating parameters of the actuator constant. The system then performs a global scalar summation operation on the calculated vector of usable heat values for each layer to obtain the current heat storage buffer margin of the kiln.
[0093] S105. When the absolute value of the heat load deviation is less than or equal to the kiln heat storage buffer margin, generate a command to limit the rate of change of the operating parameters of the predictive control model for the firing kiln.
[0094] Absolute value refers to a non-negative value regardless of its sign; here, it is used to measure the true magnitude of heat load deviation. Predictive control model refers to an advanced control algorithm architecture based on a system dynamic model, which calculates future control sequences through rolling optimization. Operating parameter change rate limit command refers to a control signal sent to the control system to constrain the actuator's action speed or single adjustment step size. The threshold relationship involved in this step is: the absolute value of the heat load deviation is less than or equal to the kiln's heat storage buffer margin. This threshold relationship originates from the system's assessment of the kiln's own thermal inertia mitigation capability. When this relationship is met, it indicates that the kiln's own heat storage is sufficient to cope with heat load fluctuations, and no significant intervention from the actuator is required.
[0095] The system can implement this step through a rule engine combined with instruction generation templates. The system obtains the calculated heat load deviation and calls the absolute value calculation function to obtain its non-negative magnitude. The system then compares this absolute value with the calculated kiln heat storage buffer margin in the built-in logic judgment rule engine. When the rule engine determines that the absolute value of the heat load deviation is less than or equal to the kiln heat storage buffer margin, it triggers the limit instruction generation logic. The system reads a preset operating parameter change rate limit instruction template, sets the limit coefficient to a small safety threshold, for example, limiting the maximum allowable opening change rate of the gas valve to within two percent per minute. The system fills this limit coefficient into the instruction template, encapsulates it into a standard Extensible Markup Language (EXPLAIN) format operating parameter change rate limit instruction, and pushes it into the input message queue of the predictive control model.
[0096] The system can also implement this step through state machine logic combined with a dynamic parameter configuration interface. The system maintains a kiln thermal state monitoring state machine in memory, using the absolute value of the heat load deviation and the kiln's heat storage buffer margin as state transition conditions. The system compares these two values in real time. If the absolute value of the heat load deviation is less than or equal to the kiln's heat storage buffer margin, the state machine automatically jumps to a self-buffering easing state. In this state, the system calls the dynamic parameter configuration application interface of the predictive control model to dynamically calculate a decay factor based on the current buffer margin ratio. The system uses this decay factor to generate a rate-of-change constraint instruction for the operating parameters of the predictive control model for the firing kiln. This instruction is written directly into the runtime memory of the predictive control model in a key-value pair structure, thereby constraining subsequent control actions.
[0097] S106. Input the operating parameter change rate limit command into the predictive control model to control the change range of the operating parameters output by the predictive control model to the actuator;
[0098] Variation range refers to the maximum range or percentage span by which the operating parameters of an actuator are allowed to change between two adjacent control cycles.
[0099] The system can achieve this step by modifying the objective function constraints of the predictive control model. The system inputs the generated operating parameter change rate limit command into the predictive control model, which then parses the command to extract the specific limit values. During rolling optimization calculations, the core of the predictive control model is solving a constrained quadratic programming problem. The system dynamically adds the extracted limit values as inequality constraints to the constraint matrix of the quadratic programming solver. During the solver's search for the optimal control sequence, any solution that causes the actuator's operating parameter change rate to exceed the limit value is deemed infeasible and discarded. In this way, the final operating parameter sequence output by the predictive control model to the actuator is strictly limited to the change rate specified by the command.
[0100] The system can also achieve this step by connecting a saturation limiting filter in series at the output of the predictive control model. The system inputs the operating parameter rate-of-change limit command to the post-processing module of the predictive control model. The predictive control model calculates the initial operating parameter control sequence according to conventional unconstrained or relaxed constraints. Before outputting this sequence to the actuator, the system uses the saturation limiting filter in the post-processing module to intercept the initial control sequence. The filter reads the upper limit value of the change amplitude in the operating parameter rate-of-change limit command and calculates the difference between the initial output value of the current control cycle and the actual output value of the previous cycle. If the difference exceeds the upper limit value, the filter forcibly truncates the output value to the actual output value of the previous cycle plus or minus the upper limit value. Through this hard truncation by the filter, the system successfully controls the change amplitude of the operating parameters output from the predictive control model to the actuator.
[0101] S107. When the absolute value of the heat load deviation is greater than the kiln heat storage buffer margin, calculate the difference between the absolute value of the heat load deviation and the kiln heat storage buffer margin to obtain the feedforward compensation heat.
[0102] The difference refers to the result of subtracting two values, specifically the heat shortfall or excess that exceeds the kiln's own heat regulation capacity. Feedforward compensation heat refers to the heat demand value that the system pre-calculates and actively injects into the control loop to offset external disturbances before feedback control takes effect. The threshold relationship involved in this step is: the absolute value of the heat load deviation must be greater than the kiln's heat storage buffer margin. This threshold relationship stems from the prediction that the kiln's thermal balance will be disrupted. When this relationship is met, it indicates that external actuators must actively adjust the temperature to maintain stability.
[0103] The system can implement this step using a combination of basic arithmetic logic units and a sign-preserving algorithm. The system monitors the relationship between the absolute value of the heat load deviation and the kiln's heat storage buffer margin in real time. When the absolute value of the heat load deviation is determined to be greater than the kiln's heat storage buffer margin, the system calls the basic arithmetic logic unit of the central processing unit to perform a high-precision floating-point subtraction operation. The system subtracts the kiln's heat storage buffer margin from the absolute value of the heat load deviation to calculate the initial difference. Subsequently, the system extracts the sign of the original heat load deviation. If the original deviation is positive, it indicates that additional heating is needed, and the system assigns a positive sign to the difference; if the original deviation is negative, it indicates that heating needs to be reduced, and the system assigns a negative sign to the difference. The signed difference is then recognized by the system as feedforward compensation heat and stored in the shared memory space.
[0104] The system can also implement this step using a difference calculation model with a safety margin coefficient. When the absolute value of the heat load deviation is determined to be greater than the kiln's heat storage buffer margin, a preset safety margin coefficient is introduced to prevent overcompensation. The system first calculates the direct difference between the absolute value of the heat load deviation and the kiln's heat storage buffer margin. Then, the system multiplies this direct difference by the safety margin coefficient to appropriately scale the difference. Based on the direction of the original heat load deviation, the system determines the directional attribute of the scaled difference, thus obtaining the final feedforward compensation heat. This calculation method treats the feedforward compensation heat as a smoothed disturbance compensation term, ensuring that it can compensate for insufficient kiln heat storage capacity while avoiding actuator overshoot due to calculation errors.
[0105] S108. Input the feedforward compensation heat into the predictive control model so that the predictive control model outputs adjustment instructions for the operating parameters of the actuator based on the feedforward compensation heat.
[0106] Operating parameter adjustment commands refer to the final control commands issued by the predictive control model to the actuator after comprehensive calculation, which are used to change its current operating state. These commands include, but are not limited to, specific action commands such as increasing gas flow and decreasing combustion air volume.
[0107] The system can achieve this step by using the feedforward compensation heat as a measurable disturbance input to the predictive control model. The system inputs the calculated feedforward compensation heat into the state-space equations of the predictive control model. The predictive control model defines this feedforward compensation heat as a known feedforward measurable disturbance variable. When predicting future states, the model not only utilizes the current internal temperature state of the kiln and historical control inputs, but also superimposes this feedforward compensation heat into the system's heat balance prediction equations. By considering the impact of this disturbance on the future temperature trajectory in advance in the rolling optimization objective function, the solver calculates the optimal control sequence that can offset the heat deviation. The system extracts the first control action of this control sequence, converts it into an operating parameter adjustment command for the actuator, and outputs it.
[0108] The system can also achieve this step by dynamically modifying the reference trajectory of the predictive control model. The system inputs the feedforward compensation heat into the reference trajectory generation module of the predictive control model. Based on the magnitude and sign of the feedforward compensation heat, the reference trajectory generation module dynamically calculates the equivalent actuator reference offset required to maintain stable kiln temperature. The system superimposes this reference offset onto the original steady-state operating parameter reference trajectory set by the predictive control model, generating a new dynamic reference trajectory. When executing the optimization algorithm, the predictive control model aims to track this new dynamic reference trajectory. By tracking the reference trajectory superimposed with the feedforward compensation heat, the predictive control model naturally calculates the control sequence containing compensation actions and converts it into operating parameter adjustment commands for the actuators for output.
[0109] In the above embodiments, when a batch change of raw materials in the refractory material production line is detected, the expected thermal load deviation can be obtained in advance before the new batch of materials enters the kiln by extracting the rate of change of mechanical resistance of the upstream molding equipment and processing it through a neural network model. Combined with the kiln's internal temperature field and the physical parameters of the lining, the calculated kiln heat storage buffer margin allows the system to quantitatively assess the kiln's ability to smooth temperature fluctuations using its own thermal inertia. When the thermal load deviation is less than or equal to the buffer margin, the energy waste caused by excessive adjustment of the actuator is avoided by limiting the rate of change of the predictive control model's operating parameters. When the deviation is greater than the buffer margin, feedforward compensation heat is introduced by calculating the difference, enabling the predictive control model to make precise parameter adjustments in advance. This reduces the frequency of over-adjustment caused by material batch changes and improves the stability of the firing kiln temperature control and the overall energy consumption optimization level.
[0110] In the above embodiments, by combining the kiln's heat storage buffer margin with a feedforward compensation mechanism, the heat load fluctuations caused by material batch switching were mitigated, achieving stable temperature control and energy consumption optimization in the firing kiln. However, after a new batch of material enters the high-temperature zone of the kiln, in addition to causing changes in heat load, the gases generated by its violent volatilization can disrupt the original aerodynamic balance inside the kiln, easily causing sudden changes in local gas pressure, which in turn affects the safe operation of the combustion system. Therefore, after completing the basic operating parameter limits and heat compensation, the system still needs to further proactively monitor and dynamically adjust the gas pressure distribution inside the kiln. Based on this, this application provides another AI-based energy consumption optimization control method for refractory material production lines. The following combines... Figure 2 Another AI-based energy consumption optimization and control method for refractory material production lines is described in the embodiments of this application:
[0111] Please see Figure 2 This is another flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line, as described in this application.
[0112] S201. Collect real-time pressure data from multiple pressure measurement points distributed along the material travel direction inside the firing kiln;
[0113] Material movement direction refers to the physical trajectory of refractory green billets moving from the feed end to the discharge end within the firing kiln. Pressure measurement point refers to a pre-defined physical spatial location on the side wall or top of the firing kiln for installing pressure sensing equipment. Real-time pressure data refers to the set of pressure values generated by the gas inside the kiln against the kiln wall at the current moment, including but not limited to absolute pressure values, relative static pressure values, or differential pressure fluctuation values.
[0114] The system can implement this step using an industrial fieldbus combined with distributed pressure transmitters. High-temperature resistant differential pressure transmitters are installed at multiple pressure measurement points along the material travel direction in the firing kiln. The system sends high-frequency synchronous sampling commands to these transmitters via a programmable logic controller (PLC). The piezoresistive sensors inside the transmitters convert the detected minute pressure deformations into standard analog electrical signals, which are then processed by an analog-to-digital converter (ADC) to generate digitized real-time pressure data. The system uses a fieldbus protocol to package the real-time pressure data transmitted from each measurement point according to spatial location and store it in the system's real-time time-series database.
[0115] The system can also achieve this step using a wireless sensor network combined with pressure tap measurement technology. A wireless sensor gateway is deployed outside the firing kiln, and at each pressure measurement point, corundum pressure taps draw gas from inside the kiln to a safe external temperature range. A capacitive pressure sensor with a wireless communication module is connected to the end of each pressure tap. The system broadcasts a data acquisition synchronization clock signal to each sensor node via a low-power wide-area network. Upon receiving the signal, the sensor measures the current gas pressure using the characteristic that the distance between the capacitor plates changes with gas pressure, generating real-time pressure data. The system receives this real-time pressure data, which includes timestamps and pressure measurement point location identifiers, and performs data cleaning and outlier removal before storing it in an in-memory database.
[0116] S202. Calculate the pressure distribution gradient along the material travel direction inside the firing kiln based on real-time pressure data;
[0117] Pressure distribution gradient refers to the rate of change of gas pressure inside the kiln along the material travel direction per unit distance. It is used to characterize the degree of spatial non-uniformity of the airflow driving force inside the kiln and the potential trend of gas flow.
[0118] The system can implement this step using spatial difference calculation combined with a smoothing filtering algorithm. The system extracts real-time pressure data and the physical coordinates of each pressure measuring point along the material travel direction from a real-time time-series database. The system sequentially selects two adjacent pressure measuring points according to the material travel direction, subtracting the real-time pressure data of the previous measuring point from the real-time pressure data of the subsequent measuring point to obtain the pressure difference between the two points. The system then subtracts the physical coordinates of the previous measuring point from the physical coordinates of the subsequent measuring point to obtain the spatial distance between the two points. The system divides the pressure difference by the spatial distance to calculate the local pressure change rate between adjacent measuring points. The system performs moving average filtering on all local pressure change rates to eliminate high-frequency measurement noise, ultimately obtaining the pressure distribution gradient along the material travel direction inside the firing kiln.
[0119] The system can also achieve this step through polynomial curve fitting combined with differentiation. The system acquires real-time pressure data and corresponding spatial coordinates of all pressure measurement points, constructing a discrete data point set with spatial coordinates as the independent variable and real-time pressure data as the dependent variable. Using the least squares method, the system selects a polynomial function of appropriate order to perform nonlinear curve fitting on this discrete data point set, obtaining a pressure distribution function that continuously describes the pressure inside the kiln as it changes with spatial position. The system then calls the symbolic computation engine or numerical differentiation module to solve for the first derivative of this pressure distribution function with respect to the spatial position independent variable. The system outputs the function values of the derivative function obtained after differentiation at each spatial position node as the pressure distribution gradient along the material travel direction inside the firing kiln.
[0120] S203. Obtain the predicted value of the volatile gas release rate of the new batch of materials in the high-temperature zone of the firing kiln.
[0121] The high-temperature zone refers to a specific spatial section inside the firing kiln where the temperature is sufficient to cause violent physicochemical reactions in the refractory materials and the large-scale decomposition and gasification of organic matter and binders. The predicted value of the volatile gas release rate refers to the estimated mass or volume of gas released into the kiln's internal space per unit time due to thermal decomposition or evaporation after a new batch of material enters the high-temperature zone.
[0122] The system can perform this step by calling a thermogravimetric analysis database and combining it with the material mass conservation equation. The system reads the formulation information of the new batch of material, including the type and mass percentage of the organic binder. The system connects to a pre-established thermogravimetric analysis database in the laboratory and retrieves the thermogravimetric loss rate curve for this formulation under the corresponding high-temperature range. The system obtains the infeed mass flow rate of the new batch of material from the upstream process, and, combined with the expected residence time of the material in the high-temperature range, calculates the mass loss of material per unit time using the material mass conservation equation and the thermogravimetric loss rate curve. The system converts this mass loss into gas release, and then calculates the predicted volatile gas release rate of the new batch of material in the high-temperature range of the firing kiln.
[0123] The system can also implement this step using a pre-trained volatile dynamics neural network model. The underlying structure of this model is a deep learning architecture containing multiple layers of long short-term memory (LSTM) networks and fully connected regression layers. The training and testing sets contain historical formulation parameters, kiln feed mass, high-temperature zone temperature curves, and the actual volatile gas release rate calculated from the exhaust gas flow meter for different batches of material. The system inputs the formulation feature vector of the new batch of material, the current kiln feed speed, and the real-time temperature field data of the high-temperature zone into the neural network model. The model extracts the dynamic features of temperature changes over time and their temporal correlation with material reaction kinetics through the LSM network layers, then performs numerical mapping through the fully connected regression layer, ultimately outputting a predicted value for the volatile gas release rate of the new batch of material in the high-temperature zone of the firing kiln, which is completely consistent with the steps described in the previous step.
[0124] S204. Based on the pressure distribution gradient and the predicted value of the volatile gas release rate, calculate the expected value of the local pressure change in the high temperature range;
[0125] The expected value of local pressure change refers to the magnitude of the possible sharp increase or decrease in air pressure in a specific local area within a high-temperature range in a short period of time, taking into account the existing airflow drive distribution and the impact of newly added volatile gases.
[0126] The system can achieve this step by combining the one-dimensional steady-state pipe flow equation of fluid mechanics with the gas equation of state. The system obtains the calculated pressure distribution gradient and the predicted value of the volatile gas release rate. The system treats the high-temperature region as an equivalent one-dimensional variable-mass flow pipe model, and uses the pressure distribution gradient to calculate the baseline gas velocity and baseline pressure of the high-temperature region under the current state. The system introduces the predicted volatile gas release rate as a mass source term into the continuity equation of the one-dimensional steady-state pipe flow to calculate the impact of the increased gas mass on the local gas density. Combining the ideal gas equation of state, and assuming a constant temperature in the high-temperature region, the system calculates the pressure increment caused by the increase in density. The system superimposes this pressure increment with the baseline pressure value and extracts the difference exceeding the normal fluctuation range as the expected value of the local pressure surge in the high-temperature region.
[0127] The system can also achieve this step using a multivariable state-space observer combined with empirical compensation coefficients. The system constructs a state-space observer model describing the dynamic changes in air pressure within a high-temperature range. The pressure distribution gradient is used as the state variable input to the observer, reflecting the potential energy distribution within the system, while the predicted volatile gas release rate is used as the input to the observer as an external measurable disturbance. Internally, the observer uses a pre-defined aerodynamic transfer function matrix to calculate the transient impact response amplitude of the local pressure state node under the current pressure gradient background. The system introduces empirical compensation coefficients derived from historical operating data to nonlinearly correct this transient impact response amplitude, compensating for the calculation bias of the one-dimensional model regarding complex eddy effects. The system directly outputs the corrected response amplitude as the expected value of local pressure surges within the high-temperature range.
[0128] S205. When it is determined that the expected value of local pressure change is greater than the preset safe pressure threshold, calculate the secondary airflow distribution adjustment parameters based on the expected value of local pressure change.
[0129] When the expected value of a local pressure surge is determined to be greater than a preset safe pressure threshold, the system calculates the secondary airflow distribution adjustment parameters based on the expected value of the local pressure surge, specifically including:
[0130] Based on the expected value of local pressure change and the cross-sectional dimensions of the kiln in the high-temperature zone, the target ejector momentum is calculated; the real-time local temperature of each secondary air inlet in the high-temperature zone is obtained, and the local temperature drop margin corresponding to each secondary air inlet is calculated; based on the local temperature drop margin of each secondary air inlet, the momentum weight of each secondary air inlet is determined; based on the target ejector momentum and the momentum weight of each secondary air inlet, the target air supply volume of each secondary air inlet is calculated; the set of target air supply volumes of each secondary air inlet is used as the secondary air flow distribution adjustment parameter.
[0131] The preset safety gas pressure threshold refers to the maximum allowable limit of local gas pressure fluctuations set to ensure the structural safety of the kiln and prevent backfire or flameout during combustion. This threshold is derived from kiln design specifications and historical safe operation data statistics. Its relationship with the expected value of local gas pressure surges is comparative; when the expected value exceeds this threshold, the adjustment mechanism is triggered. Cross-sectional dimensions refer to the internal spatial geometric parameters of the kiln perpendicular to the material travel direction, including but not limited to width and height. Target ejector momentum refers to the total hydrodynamic momentum required by the secondary air injection system to offset gas pressure surges and re-establish aerodynamic equilibrium. Real-time local temperature refers to the current actual temperature value of the area near each secondary air nozzle. Local temperature drop margin refers to the maximum temperature drop allowed in the local area without affecting the normal firing quality of the material. Momentum weight refers to the distribution ratio coefficient of the total target ejector momentum among the various secondary air inlets. Target air supply volume refers to the expected gas injection flow rate allocated to a specific secondary air inlet. Secondary air flow distribution adjustment parameters refer to a comprehensive control dataset containing target air supply volume commands for all inlets.
[0132] The system can implement this refined technical solution using a momentum conservation algorithm combined with a normalized allocation function. When the expected value of a local pressure surge exceeds a preset safe pressure threshold, the system reads the cross-sectional dimensions of the kiln in the high-temperature zone. Using the momentum theorem, it multiplies the expected local pressure surge value by the cross-sectional dimensions to calculate the additional pressure load that needs to be offset, which is then converted into the target ejector momentum. The system obtains the real-time local temperature of each secondary air inlet within the high-temperature zone using a thermocouple array. It calculates the local temperature reduction margin for each secondary air inlet by subtracting the absolute value of the real-time local temperature from the preset lower limit temperature. The system processes the local temperature reduction margins of all inlets using a linear normalization function to calculate the momentum weight of each secondary air inlet. The system multiplies the target ejector momentum by the momentum weight of each secondary air inlet and, combined with the aerodynamic characteristic curve of the secondary air nozzle, calculates the target airflow for each secondary air inlet. The system packages the target airflow for each secondary air inlet into an array set, which serves as the secondary air flow distribution adjustment parameter.
[0133] The system can also implement this refined technical solution by combining a computational fluid dynamics surrogate model with a nonlinear programming algorithm. The system inputs the expected value of local pressure surges and the cross-sectional dimensions of the kiln in the high-temperature zone into a pre-constructed fluid dynamics inversion algorithm to calculate the target ejector momentum required to maintain flow field stability. The system calls the infrared thermometer interface to obtain the real-time local temperature of each secondary air inlet within the high-temperature zone, and calculates the local temperature drop margin corresponding to each secondary air inlet by combining it with the heat capacity characteristic curve of material sintering. The system constructs a nonlinear programming model with maximizing momentum distribution efficiency as the objective function, using the local temperature drop margin as a constraint, and solves the model using the interior point method to determine the optimal momentum weight for each secondary air inlet. Based on the target ejector momentum and the momentum weights of each secondary air inlet, the system calculates the target air volume for each secondary air inlet using the gas flow rate to momentum conversion equation. The system encapsulates the set of target air volumes for each secondary air inlet into a control command frame, which serves as the secondary air flow distribution adjustment parameter.
[0134] S206. Input the secondary air flow distribution adjustment parameters into the secondary air control system of the firing kiln to control the secondary air control system to adjust the air supply volume of the secondary air inlet in the corresponding area.
[0135] A secondary air control system refers to an auxiliary automated air supply control device, independent of the main burner air supply system, specifically used to regulate the atmosphere inside the kiln, supplement combustion oxygen, or turbulent airflow. This includes, but is not limited to, variable frequency fans, electric regulating butterfly valves, and corresponding servo controllers. Air supply volume refers to the volume or mass of air injected into the kiln through the secondary air inlet per unit time.
[0136] The system can implement this step using industrial Ethernet communication combined with a proportional-integral-derivative (PID) control algorithm. The system parses the generated secondary air flow distribution adjustment parameters according to a register address mapping table recognizable by the programmable logic controller (PLC), extracting the target air volume values for each secondary air inlet. The system then sends these values to the target value register of the secondary air control system in the firing kiln via the industrial Ethernet protocol. The microprocessor inside the secondary air control system reads the target air volume and collects the actual air volume in the current duct as a feedback signal. The microprocessor runs the PID control algorithm, calculates the deviation between the target air volume and the actual air volume, and outputs a control voltage signal to drive the actuator of the electric regulating butterfly valve, changing the valve opening and thus precisely adjusting the air volume at the corresponding secondary air inlet area.
[0137] The system can also achieve this step by combining analog signal output with the frequency adjustment mechanism of the frequency converter. The system transmits the target air volume values from the secondary air flow distribution adjustment parameters to the digital-to-analog conversion module. Based on the pre-calibrated correspondence between air volume and control signals, the digital-to-analog conversion module converts the target air volume into a standard industrial analog signal. This analog signal is then directly input to the control terminals of the frequency converters in the secondary air control system of the firing kiln via hardwiring. Upon receiving the analog signal, the frequency converter's internal control logic converts it into a corresponding output frequency command, changing the AC frequency driving the secondary air fan motor. The motor speed changes accordingly, thereby altering the fan's air supply pressure and flow rate, achieving dynamic adjustment of the air supply volume at the secondary air inlet of the corresponding area.
[0138] In the above embodiments, after executing the operation parameter limitation command, the pressure distribution gradient is calculated by collecting real-time pressure data along the material travel direction inside the firing kiln. Combined with the predicted value of the volatile gas release rate of the new batch of material in the high-temperature zone, the expected value of potential local pressure surges in the high-temperature zone can be predicted in advance. When it is determined that the expected value exceeds the safety threshold, the secondary air flow distribution adjustment parameters are calculated and output to control the secondary air inlet air volume in the corresponding area. This allows the system to balance local pressure accumulation by dynamically distributing the secondary air flow before the actual occurrence of a pressure surge, reducing the risk of local pressure imbalance inside the kiln due to the violent volatilization of new material, and improving the stability of the aerodynamic environment inside the firing kiln and the operational safety of the combustion system.
[0139] Through the above embodiment, the system successfully mitigated the risk of local pressure imbalance caused by the vigorous volatilization of new materials by dynamically allocating secondary airflow, ensuring the stability of the aerodynamic environment inside the kiln. However, during this adjustment process, the volatilized gases ejected by the secondary air contain a large amount of releaseable latent heat. If they are allowed to be directly discharged outside the kiln with the main exhaust air, it will result in energy waste. In order to deeply integrate pressure regulation with waste heat recovery, the system also needs to accurately track the migration trajectory of these volatilized gases and actively intervene in their residence state inside the kiln to achieve the directional transfer and full utilization of internal heat. Based on this, this application embodiment provides another AI-based energy consumption optimization control method for refractory material production lines. The following is combined with... Figure 3 This application describes yet another AI-based energy consumption optimization and control method for a refractory material production line:
[0140] Please see Figure 3 This is another flowchart illustrating an AI-based energy consumption optimization and control method for a refractory material production line in this application.
[0141] S301. Calculate the dynamic migration trajectory of the ejected volatile gas inside the firing kiln based on the secondary air flow distribution adjustment parameters and the main exhaust air velocity of the firing kiln.
[0142] The main exhaust velocity refers to the macroscopic average velocity of the overall gas flow in the kiln towards the exhaust port under the suction of the main exhaust fan. The dynamic migration trajectory refers to the continuous movement path formed by volatile gas molecules or gas masses carried by the secondary air within the three-dimensional space of the kiln over time.
[0143] This step can be achieved using a Lagrange particle tracking algorithm combined with flow field interpolation. The system reads the secondary air flow distribution adjustment parameters from the control system and transforms them into local velocity boundary conditions for each secondary air inlet. The system obtains the main exhaust air velocity of the firing kiln using a Pitot tube anemometer installed on the main exhaust duct and uses it as the global outlet boundary condition. The system constructs a three-dimensional mesh model inside the kiln and uses the Navier-Stokes equations to solve for the steady-state flow field distribution under the current boundary conditions. The system discretizes the volatile gas into a large number of massless virtual particles, using the volatile gas release position as the initial coordinates, and performs step integration in the pre-calculated steady-state flow field using the Lagrange particle tracking algorithm. By continuously calculating the displacement vector of the particles in each time step and performing spatial interpolation, the system tracks the position changes of these virtual particles in three-dimensional space, and finally fits the motion paths of all particles as the dynamic migration trajectory of the ejected volatile gas inside the firing kiln.
[0144] This step can also be achieved using a pre-trained graph neural network fluid prediction model. The underlying structure of this model includes graph convolutional network layers and gated recurrent unit layers. The training and test sets are derived from historical computational fluid dynamics simulation data, containing gas diffusion path coordinate sequences under different secondary airflow rates and main exhaust air velocities. The system divides the internal space of the firing kiln into multiple graph nodes, inputting the secondary airflow distribution adjustment parameters and the main exhaust air velocity of the firing kiln as node features into the graph neural network fluid prediction model. The graph convolutional network layers extract the spatial topological features of the flow field by aggregating information from adjacent nodes, while the gated recurrent unit layers infer the temporal dependence of gas evolution over time based on the spatial features. The model ultimately outputs a series of three-dimensional spatial coordinate points arranged in chronological order. The system connects these coordinate point sets into a smooth curve, thus obtaining the dynamic migration trajectory of the ejected volatile gas inside the firing kiln.
[0145] S302. Based on the predicted value of the volatile gas release rate, calculate the expected heat that can be released by the ejected volatile gas along the dynamic migration trajectory.
[0146] The expected heat release is the total amount of heat energy that is estimated to be released to the surrounding environment when the gas undergoes oxidation, combustion, or physical condensation reactions along a specific path inside the kiln.
[0147] This step can be implemented using a thermodynamic enthalpy integral algorithm combined with a component concentration decay model. The system obtains the predicted value of the volatile gas release rate and, in conjunction with material formulation data, determines the main combustible components of the volatile gas and their corresponding standard enthalpy of combustion. The system discretizes the calculated dynamic migration trajectory into multiple small spatial steps. Using the component concentration decay model, the system calculates the rate of change of each component concentration due to diffusion and initial oxidation as the volatile gas moves along the dynamic migration trajectory. At each spatial step, the system multiplies the current remaining combustible component concentration, the predicted value of the volatile gas release rate, and the corresponding standard enthalpy of combustion to calculate the local heat release rate within that step. The system performs discrete integration and summation of the local heat release rates for all spatial steps along the entire dynamic migration trajectory to calculate the expected release heat of the entrained volatile gas along the dynamic migration trajectory.
[0148] This step can also be achieved using a pre-trained multilayer perceptron energy assessment model. The underlying structure of this model is a fully connected neural network with multiple hidden layers. The hidden layers use a linear rectified function as the activation function, and the output layer uses a linear activation function. The training and test sets contain historical data on different volatile gas release rates, different migration trajectory lengths and morphological characteristics, and the corresponding actual released heat. The system extracts the geometric features of the dynamic migration trajectory, including the total trajectory length, rate of curvature change, and residence distance in the high-temperature range. The system concatenates these geometric features with the predicted volatile gas release rate to form a one-dimensional input feature vector, which is then input into the multilayer perceptron energy assessment model. The model uses a forward propagation algorithm to perform nonlinear mapping calculations using the weight matrices and bias terms of each layer of neurons. The output directly outputs a scalar value, which is the expected release heat of the ejected volatile gas along the dynamic migration trajectory.
[0149] S303. Based on the current internal temperature field distribution data of the firing kiln, locate the low heat buffer zone inside the firing kiln where the heat storage consumption has reached the preset consumption warning value.
[0150] Heat storage consumption refers to the cumulative amount of heat energy released by the kiln lining to the materials or cold air within the kiln over a period of time, representing the degree of reduction in heat reserves in that area. The preset consumption warning value is a critical threshold set to determine whether the local heat reserves of the kiln are in a state of scarcity. This threshold is derived from the kiln thermal design specifications and historical heat balance test data. Its relationship with heat storage consumption is comparative; when heat storage consumption is greater than or equal to the preset consumption warning value, the area is determined to be in a low-heat state. The low-heat buffer zone refers to the three-dimensional spatial section within the kiln where the internal heat storage has significantly decreased and urgently requires replenishment of heat energy to maintain the process temperature.
[0151] This step can be achieved by combining three-dimensional temperature field reconstruction with energy conservation differential calculation. The system acquires the current internal temperature field distribution data of the firing kiln and uses the Kriging spatial interpolation algorithm to reconstruct the discrete measuring point temperatures into a continuous three-dimensional temperature field model. The system extracts the temperature values of each spatial grid in the three-dimensional temperature field model at the current moment and retrieves the corresponding grid temperature values at the previous reference moment. The system uses the energy conservation differential calculation formula, combined with the density, specific heat capacity, and grid volume of the kiln lining, to calculate the difference in heat change between two moments for each grid, and uses the absolute value of this difference as the heat storage consumption. The system compares the heat storage consumption of each grid with the preset consumption warning value one by one, and marks the grids whose heat storage consumption is greater than or equal to the preset consumption warning value. The system uses a region growing algorithm to merge spatially adjacent marked grids into a connected region, which is then located as the low-heat buffer region inside the firing kiln.
[0152] This step can also be achieved by combining density clustering algorithms with thermal fluid finite element analysis. The system imports the current internal temperature field distribution data of the firing kiln into a pre-established thermal fluid finite element analysis software. Using the transient heat conduction module in the finite element solver, the system calculates the heat flux density vector of each node in the kiln lining and integrates the heat flux density over a set time window to obtain the heat storage consumption of each node. The system filters out the set of node coordinates whose heat storage consumption is greater than or equal to a preset consumption warning value. These node coordinate sets are used as input data and processed using a density-based noise-based spatial clustering algorithm. Based on the set neighborhood radius and minimum number of contained points, the system divides high-density node clusters into different clusters. The system extracts the cluster with the most contained nodes, calculates the geometric bounding box of the cluster, and the spatial range defined by this bounding box is positioned as the low-heat buffer zone inside the firing kiln.
[0153] S304. When it is determined that the dynamic migration trajectory passes through a low-heat buffer zone, calculate the target residence time of the ejected volatile gas in the low-heat buffer zone based on the expected release of heat.
[0154] When the dynamic migration trajectory is determined to pass through a low-heat buffer zone, the system calculates the target residence time of the entrained volatile gas within the low-heat buffer zone based on the expected release of heat, specifically including:
[0155] Acquire real-time oxygen concentration data and ignition temperature of the ejected volatile gas in the low-heat buffer zone; calculate the combustion reaction hysteresis coefficient of the ejected volatile gas in the low-heat buffer zone based on the real-time oxygen concentration data and ignition temperature; calculate the basic heat release time of the ejected volatile gas based on the expected heat release and the current level temperature of the low-heat buffer zone; and obtain the target residence time of the ejected volatile gas in the low-heat buffer zone by weighting the basic heat release time using the combustion reaction hysteresis coefficient.
[0156] Real-time oxygen concentration data refers to the current volume percentage of combustion-supporting gases within the low-heat buffer zone. Ignition temperature is the minimum ambient temperature at which combustion-supporting gases can spontaneously ignite and burn under a specific oxygen concentration. The combustion hysteresis coefficient is a dimensionless correction factor that slows the combustion reaction rate of volatile gases and prolongs the heat release process due to insufficient oxygen concentration or unfavorable temperature conditions. The basic heat release time is the theoretically shortest time required for volatile gases to release the expected heat under ideal, complete combustion conditions. The target residence time is the desired duration for volatile gases to remain in the low-heat buffer zone through control measures in order to fully complete delayed combustion and release heat within the zone.
[0157] After determining that the dynamic migration trajectory passes through the low-heat buffer zone, the system acquires real-time oxygen concentration data using a zirconia oxygen analyzer installed on the sidewall of the low-heat buffer zone, and consults a combustible gas property handbook to obtain the ignition temperature of the entrained volatile gas. Utilizing the Arrhenius chemical kinetic equation, the system substitutes the real-time oxygen concentration data and ignition temperature into the reaction rate constant calculation formula to obtain the actual reaction rate under current conditions. The maximum reaction rate under ideal conditions is then divided by this actual reaction rate to calculate a combustion reaction hysteresis coefficient greater than one. Based on the expected release of heat and the current temperature of the low-heat buffer zone, the system calculates the average heat absorption rate of the zone. Dividing the expected release of heat by the average heat absorption rate yields the basic heat release time of the entrained volatile gas. The system then uses a weighted compensation method, directly multiplying the combustion reaction hysteresis coefficient and the basic heat release time, to finally obtain the target residence time of the entrained volatile gas within the low-heat buffer zone.
[0158] After confirming the spatial intersection between the dynamic migration trajectory and the low-heat buffer zone, the system acquires real-time oxygen concentration data of the low-heat buffer zone using a tunable semiconductor laser absorption spectroscopy sensor and reads the ignition temperature of the entrained volatile gas from a built-in volatile property database. The system constructs a two-dimensional interpolation lookup table with real-time oxygen concentration data and ignition temperature as input, and retrieves the corresponding combustion reaction hysteresis coefficient from the table using a bilinear interpolation algorithm. The system uses the expected release heat and the current layer temperature of the low-heat buffer zone as input features, feeding them into a pre-trained radial basis function neural network. The underlying structure of this neural network includes an input layer, hidden layers, and an output layer. The activation function of the hidden layers is a Gaussian kernel function, and the training and test sets contain historical heat release test data. The neural network outputs the basic heat release time of the entrained volatile gas. The system multiplies the basic heat release time by the combustion reaction hysteresis coefficient to perform weighted compensation, calculating the target residence time of the entrained volatile gas within the low-heat buffer zone.
[0159] S305. Based on the target residence time, generate an instruction to adjust the opening of the exhaust branch valve for the low-heat buffer zone;
[0160] The smoke exhaust branch valve opening adjustment command refers to the digital control signal sent to the actuator to change the cross-sectional area of a specific smoke exhaust duct to adjust the local air volume.
[0161] This step can be implemented by mapping the fluid resistance characteristic curve and combining it with proportional control logic. The system obtains the calculated target residence time and extracts the current spatial volume of the low-heat buffer zone. The system divides the spatial volume by the target residence time to calculate the maximum allowable local exhaust flow rate of the low-heat buffer zone to meet that residence time. The system calls the fluid resistance characteristic curve of the exhaust branch valve, which reflects the nonlinear mapping relationship between valve opening and flow rate. The system uses a reverse interpolation algorithm to find the target valve opening percentage corresponding to the maximum local exhaust flow rate on the characteristic curve. The system uses proportional control logic to encapsulate the target valve opening percentage into a standard control message format, generates an adjustment command for the exhaust branch valve opening of the low-heat buffer zone, and stores it in the distribution queue.
[0162] This step can also be implemented using a fuzzy logic controller combined with an empirical rule base. The system compares the target residence time with the current actual residence time, calculating the residence time deviation and the rate of change of the residence time deviation. The system uses the residence time deviation and the rate of change of the deviation as input variables, inputting them into a pre-designed fuzzy logic controller. The fuzzy logic controller first fuzzifies the input variables, mapping them to membership degrees in a fuzzy set. Then, the system calls the built-in empirical rule base; for example, if the residence time deviation is large positive, then the valve opening adjustment amount is large negative, performing fuzzy inference to obtain the fuzzy control output. The system uses the centroid method to defuzzify the fuzzy control output, calculating the specific valve opening adjustment increment. The system adds this increment to the current valve opening to obtain the final opening value, and encodes it to generate an opening adjustment command for the exhaust branch valve of the low-heat buffer area.
[0163] S306. Input the flue gas branch valve opening adjustment command into the flue gas control system of the firing kiln to control the flue gas control system to reduce the opening of the flue gas branch valve corresponding to the low heat buffer area.
[0164] A flue gas control system refers to a collection of automated devices used to manage and regulate the velocity, flow rate, and pressure of exhaust gas from kilns, including but not limited to induced draft fans, frequency converters, electric regulating valves, and programmable logic controllers.
[0165] This step can be achieved using the Industrial Ethernet protocol combined with programmable logic controller (PLC) register writing. The system parses the generated exhaust branch valve opening adjustment command, extracting the target opening value and the device address code of the exhaust branch valve corresponding to the low-heat buffer zone. The system uses the Industrial Ethernet protocol to establish a transmission control protocol connection with the main control PLC of the kiln exhaust control system. The system packages the target opening value into a standard data frame and sends it to the designated holding register address of the main control PLC. The main program of the exhaust control system reads the register value in the next scan cycle and converts it into a corresponding analog voltage signal to drive the servo motor of the electric regulating valve of the exhaust branch corresponding to the low-heat buffer zone, thereby reducing the opening of the exhaust branch valve.
[0166] This step can also be achieved through a combination of an open platform communication unified architecture protocol and a distributed control system interface. The system encapsulates the flue gas branch valve opening adjustment command into a data object conforming to the open platform communication unified architecture protocol standard. The system, acting as a client, connects to the server side of the distributed control system where the kiln flue gas control system resides. By calling the write service interface provided by the server, the system directly writes the data object containing the target opening degree into the control node variable of the flue gas branch valve corresponding to the low-heat buffer zone. Upon receiving the variable update notification, the distributed control system triggers the underlying control loop, sending digital control commands to the corresponding intelligent valve positioner via the fieldbus. The intelligent valve positioner adjusts the air source pressure of the pneumatic actuator according to the command, pushing the valve stem to move, thereby precisely reducing the opening degree of the flue gas branch valve corresponding to the low-heat buffer zone.
[0167] In the above embodiments, after adjusting the secondary air supply volume, the dynamic migration trajectory of the ejected volatile gas inside the kiln and its expected release of heat are calculated, and the low-heat buffer zone where the heat storage consumption reaches the warning value is located in conjunction with the internal temperature field. When the gas migration trajectory passes through this low-heat buffer zone, the system calculates the target residence time based on the expected release of heat and accordingly reduces the opening of the exhaust branch valve in the corresponding area. By utilizing the latent calorific value of the volatile gas released by the material itself and actively intervening in the local exhaust flow rate, the residence time of the heat-carrying volatile gas in the heat-deficient area is extended, promoting the full release of its latent heat and absorption by the kiln lining in a low-heat state. This reduces the dependence of the local heat-deficient area on the external main heating energy source and improves the overall thermal energy utilization efficiency of the firing kiln.
[0168] The system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 4 This is a schematic diagram of the physical device structure of an AI-based energy consumption optimization control system for a refractory material production line, provided in an embodiment of this application.
[0169] It should be noted that, Figure 4 The structure of the system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.
[0170] like Figure 4As shown, the system includes a Central Processing Unit (CPU) 401, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 402 or a program loaded from storage portion 408 into Random Access Memory (RAM) 403, such as executing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An Input / Output (I / O) interface 405 is also connected to the bus 404.
[0171] The following components are connected to I / O interface 405: input section 406 including a camera, infrared sensor, etc.; output section 407 including a liquid crystal display (LCD) and speakers, etc.; storage section 408 including a hard disk, etc.; and communication section 409 including a network interface card such as a LAN (Local Area Network) card and a modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from it can be installed into storage section 408 as needed.
[0172] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the various functions defined in the present invention.
[0173] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. The transmitted data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof.
[0174] 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 the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains 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 a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may 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.
[0175] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the system described in the above embodiments; or it may exist independently and not assembled into the system. The storage medium carries one or more computer programs that, when executed by a processor of a system, cause the system to implement the methods provided in the above embodiments.
[0176] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
[0177] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".
[0178] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.
[0179] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.
Claims
1. An AI-based energy consumption optimization and control method for a refractory material production line, characterized in that, include: When a change in raw material batch is detected in the refractory material production line, mechanical operation data of the upstream forming equipment of the production line is collected; Extract the rate of change of mechanical resistance of the upstream forming equipment when processing a new batch of material from the mechanical operation data; The mechanical resistance change rate is input into a pre-trained neural network model to obtain the expected thermal load deviation when the new batch of material enters the firing kiln; Based on the current internal temperature field distribution data of the firing kiln and the physical parameters of the kiln lining, the current kiln heat storage buffer margin is calculated. The kiln heat storage buffer margin is the maximum heat value that the kiln lining can release or absorb to maintain the internal temperature of the kiln within the preset qualified process range, under the premise that the current operating parameters of the firing kiln actuator remain unchanged. If the absolute value of the heat load deviation is less than or equal to the kiln heat storage buffer margin, a command to limit the rate of change of operating parameters for the predictive control model of the firing kiln is generated. The command limiting the rate of change of the operating parameters is input into the predictive control model to control the magnitude of change of the operating parameters output by the predictive control model to the actuator. If the absolute value of the heat load deviation is greater than the kiln heat storage buffer margin, the difference between the absolute value of the heat load deviation and the kiln heat storage buffer margin is calculated to obtain the feedforward compensation heat. The feedforward compensation heat is input into the predictive control model so that the predictive control model outputs adjustment instructions for the operating parameters of the actuator based on the feedforward compensation heat.
2. The method according to claim 1, characterized in that, The calculation of the current heat storage buffer margin of the kiln based on the current internal temperature field distribution data and the physical parameters of the kiln lining specifically includes: Based on the internal temperature field distribution data, the kiln lining is divided into multiple virtual heat storage layers along the radial depth, and the current level temperature of each virtual heat storage layer is extracted. Based on the physical parameters of the kiln lining and the current layer temperature of each virtual heat storage layer, calculate the dynamic thermal diffusivity of each virtual heat storage layer at the current temperature; The expected transition time of the new batch of material in the firing kiln is obtained, and combined with the dynamic thermal diffusivity of each virtual heat storage layer, an effective heat storage layer with a thermal response time less than or equal to the expected transition time is selected from the multiple virtual heat storage layers. Calculate the temperature difference between the current level temperature of each effective heat storage layer and the boundary critical temperature of the preset qualified process range; Based on the specific heat capacity and density of the physical parameters of the kiln lining, and the temperature difference, calculate the single-layer usable heat value that each effective heat storage layer can autonomously release or absorb while keeping the current operating parameters of the actuator unchanged. The available heat values of each single layer of the effective heat storage layer are summed to obtain the current heat storage buffer margin of the kiln.
3. The method according to claim 1, characterized in that, The step of inputting the rate of change of mechanical resistance into a pre-trained neural network model to obtain the expected deviation of heat load when the new batch of material enters the firing kiln specifically includes: The rate of change of mechanical resistance is input into the feature extraction layer of the pre-trained neural network model to extract the initial physical property coupling feature vector; The estimated transfer time of the new batch of materials from the upstream forming equipment to the firing kiln is obtained, as well as the ambient temperature and humidity data of the production line during the estimated transfer time; The expected transmission time and the ambient temperature and humidity data are input into the environmental attenuation mapping layer of the neural network model to calculate the dynamic volatile compensation coefficient. The initial physical property coupling feature vector is weighted and corrected using the dynamic volatilization compensation coefficient to obtain the target kiln entry feature vector; The target kiln entry feature vector is input into the heat load decoding layer of the neural network model for decoding to obtain the expected heat load deviation when the new batch of material enters the firing kiln.
4. The method according to claim 1, characterized in that, After inputting the operating parameter change rate limit command into the predictive control model, the method further includes: Real-time pressure data are collected from multiple pressure measuring points distributed along the material travel direction inside the firing kiln. Based on the real-time pressure data, the pressure distribution gradient inside the firing kiln along the material travel direction is calculated; Obtain the predicted value of the volatile gas release rate of the new batch of material in the high-temperature zone of the firing kiln; Based on the pressure distribution gradient and the predicted value of the volatile gas release rate, the expected value of the local pressure change in the high temperature range is calculated. If the expected value of the local pressure change is determined to be greater than the preset safe pressure threshold, the secondary airflow distribution adjustment parameters are calculated based on the expected value of the local pressure change. The secondary air flow distribution adjustment parameters are input into the secondary air control system of the firing kiln to control the secondary air control system to adjust the air supply volume of the secondary air inlet in the corresponding area.
5. The method according to claim 4, characterized in that, The calculation of secondary airflow distribution adjustment parameters based on the expected value of local air pressure change specifically includes: The target ejector momentum is calculated based on the expected value of the local pressure change and the cross-sectional dimensions of the firing kiln in the high-temperature range. The real-time local temperature of each secondary air inlet within the high-temperature range is obtained, and the local temperature reduction margin corresponding to each secondary air inlet is calculated. Based on the local temperature drop margin of each of the secondary air inlets, the momentum weight of each of the secondary air inlets is determined. The target air volume of each secondary air inlet is calculated based on the target ejector momentum and the momentum weight of each secondary air inlet. The target air supply volume set of each of the secondary air inlets is used as the secondary air flow distribution adjustment parameter.
6. The method according to claim 4, characterized in that, After inputting the secondary air flow distribution adjustment parameters into the secondary air control system of the firing kiln, the method further includes: Based on the secondary airflow distribution adjustment parameters and the main exhaust airflow velocity of the firing kiln, the dynamic migration trajectory of the ejected volatile gas inside the firing kiln is calculated. Based on the predicted value of the volatile gas release rate, the expected heat that can be released by the ejected volatile gas along the dynamic migration trajectory is calculated; Based on the current internal temperature field distribution data of the firing kiln, a low heat buffer zone where the heat storage consumption reaches the preset consumption warning value is located inside the firing kiln. If it is determined that the dynamic migration trajectory passes through the low-heat buffer zone, the target residence time of the ejected volatile gas in the low-heat buffer zone is calculated based on the expected release of heat. Based on the target residence time, a command is generated to adjust the opening degree of the exhaust branch valve for the low-heat buffer zone. The opening adjustment command of the exhaust branch valve is input to the exhaust control system of the firing kiln to control the exhaust control system to reduce the opening of the exhaust branch valve corresponding to the low heat buffer zone.
7. The method according to claim 6, characterized in that, The calculation of the target residence time of the ejected volatile gas in the low-heat buffer zone based on the expected release of heat specifically includes: Acquire real-time oxygen concentration data of the low-heat buffer zone and ignition temperature of the ejected volatile gas; Based on the real-time oxygen concentration data and the ignition temperature, the combustion reaction hysteresis coefficient of the ejected volatile gas in the low-heat buffer zone is calculated. The basic heat release time of the ejected volatile gas is calculated based on the expected heat release and the current stratum temperature of the low-temperature buffer zone. By using the combustion reaction hysteresis coefficient to weight and compensate the basic heat release time, the target residence time of the ejected volatile gas in the low-heat buffer region is obtained.
8. An AI-based energy consumption optimization control system for a refractory material production line, characterized in that, The system includes: One or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.
9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the system, the system performs the method as described in any one of claims 1-7.
10. A computer program product, characterized in that, When the computer program product is run on the system, the system performs the method as described in any one of claims 1-7.
