Low-grade hard clay waste heat sintering method and system

Abstract

The application relates to the technical field of firing of refractory materials and utilization of kiln waste heat, and discloses a low-grade hard clay waste heat firing method and system, wherein the low-grade hard clay waste heat firing method comprises the following steps: collecting detection information of current batch of raw materials before kiln loading and normalizing the detection information to obtain a raw material sensitivity index; collecting atmosphere parameters of recycled gas in a waste heat recovery loop during the firing process, and calculating an atmosphere offset amplitude; calculating a firing boundary drift characteristic value, determining a current control mode, generating a control output, and executing linkage control during the firing process; collecting quality feedback information after the current batch of products are discharged from the kiln, and constructing a comprehensive quality deviation; and updating a calibration coefficient based on the comprehensive quality deviation; the firing boundary drift characteristic value is used for linkage adjustment of the control output, so that the stability of the low-grade hard clay firing process and the product performance consistency are improved while the waste heat utilization effect is maintained.

Description

Technical Field

[0001] This application relates to the field of refractory material firing and kiln waste heat utilization technology, and more specifically, it relates to a method and system for firing low-grade hard clay using waste heat. Background Technology

[0002] Hard clay is a commonly used basic raw material in refractory materials, ceramic materials, and high-temperature industrial equipment linings. It can be used to prepare refractory bricks, refractory castables, kiln linings, and related high-temperature materials. Low-grade hard clay typically refers to hard clay raw materials with relatively low alumina content and relatively high content of impurities such as iron oxides and alkali metal oxides. Although this type of raw material has a lower grade, it still possesses certain refractoriness and high-temperature volume stability. After proper firing, it can meet the requirements of some refractory material products, thus having application value in reducing raw material costs and improving the comprehensive utilization rate of mineral resources.

[0003] In the production of low-grade hard clay products, the firing process directly affects the degree of mineral phase transformation, sintering densification, linear shrinkage, water absorption, bulk density, and refractoriness of the products. Due to the high energy consumption of the firing process, current production methods typically introduce recovered flue gas or hot air generated in the kiln cooling section into the preheating section, combustion section, or other heat utilization branches to improve thermal efficiency and reduce energy consumption per unit product. This type of waste heat recovery can improve energy utilization to a certain extent and has therefore become an important direction for optimizing the firing process of low-grade hard clay.

[0004] However, in actual firing processes, the recovered flue gas or hot air is not merely a simple heat carrier; it also carries certain oxygen partial pressure, water vapor content, and atmospheric components such as carbon monoxide and carbon dioxide. When this recycled gas re-enters the preheating or combustion-supporting section, its atmospheric state affects the preheating of the green body, redox reactions, impurity fluxing behavior, mullite formation, and the high-temperature densification process. For low-grade hard clay, due to its low alumina content, high impurity content, and the tendency for fluctuations in chemical composition, mineral composition, water absorption, and particle size distribution between different batches, its firing window is typically narrow, making it more sensitive to changes in temperature regime and atmospheric state.

[0005] Existing waste heat firing methods for low-grade hard clay typically rely on fixed firing curves, fixed holding regimes, or single-parameter adjustments based on kiln temperature. Even when waste heat recovery loops are implemented, the primary objectives are usually to improve heat utilization or maintain kiln temperature stability. These methods fail to correlate the firing sensitivity of the raw material batch with atmospheric shifts in the recycled waste heat gas, making it difficult to promptly identify the degree of deviation of the actual firing operating point from the standard process boundary. When fluctuations in both raw material batches and waste heat atmosphere coexist, the firing boundary is prone to drift. Existing control methods may still operate according to the original waste heat recovery ratio, air supply path, and holding regime, causing the firing process to deviate from the suitable mineral phase transformation and densification range. Summary of the Invention

[0006] The purpose of this application is to address the problem that in the existing waste heat firing process of low-grade hard clay, it is difficult to correlate and identify the fluctuations of raw material batches with the fluctuations of waste heat recovery gas atmosphere, and thus difficult to characterize and suppress firing boundary drift in a timely manner. This application provides a waste heat firing method and system for low-grade hard clay.

[0007] This application provides a method for firing low-grade hard clay using waste heat, applicable to firing equipment with a waste heat recovery loop, comprising the following steps:

[0008] Before loading the kiln, the detection information of the current batch of raw materials is collected and normalized, and the normalized values ​​are weighted and summed to obtain the raw material sensitivity index.

[0009] During the firing process, the atmospheric parameters of the recycled gas in the waste heat recovery loop are collected, the standardized directional deviation of each atmospheric parameter is calculated, and the atmospheric offset amplitude is calculated based on each standardized directional deviation.

[0010] The firing boundary drift characterization value is calculated by the raw material sensitivity index, atmosphere offset amplitude and preset calibration coefficient. The current control mode is determined based on the comparison result of the firing boundary drift characterization value and the preset threshold.

[0011] A control output is generated based on the firing boundary drift characterization value. Based on the control output, control mode and each standardized directional deviation, linkage control is executed during the firing process.

[0012] After the current batch of products leaves the kiln, quality feedback information is collected and a comprehensive quality deviation is constructed. The calibration coefficient is then updated based on the comprehensive quality deviation.

[0013] Furthermore, methods for calculating the atmospheric shift amplitude include:

[0014] Atmospheric parameters include recycled gas temperature, oxygen partial pressure, water vapor volume fraction, carbon monoxide concentration, and carbon dioxide concentration;

[0015] After applying sliding window mean filtering to the atmosphere parameters, the standardized directed biases of the recycled gas temperature, oxygen partial pressure, and water vapor volume fraction relative to the process target values ​​are calculated; the standardized directed biases of the logarithmic ratios are calculated based on the carbon monoxide concentration and carbon dioxide concentration.

[0016] The atmosphere offset amplitude is the weighted sum of the absolute values ​​of the directed deviations of the reuse gas temperature standardization, oxygen partial pressure standardization, water vapor volume fraction standardization, and logarithmic ratio standardization, and the corresponding preset atmosphere weight coefficients.

[0017] Furthermore, methods for calculating the directional bias of the standardized log-ratio include:

[0018] The measured logarithmic ratio was calculated based on the measured values ​​of carbon monoxide and carbon dioxide concentrations.

[0019] Calculate the target logarithmic ratio based on the process target values ​​of carbon monoxide concentration and carbon dioxide concentration;

[0020] The logarithmic ratio is calculated by adding the carbon monoxide concentration to a preset positive bias concentration as the numerator, adding the carbon dioxide concentration to a preset positive bias concentration as the denominator, and taking the natural logarithm of the fraction.

[0021] Calculate the difference between the measured logarithmic ratio and the target logarithmic ratio, and then divide it by the standard deviation of the logarithmic ratio under normal operating conditions to obtain the standardized directional bias of the logarithmic ratio.

[0022] Furthermore, the methods for determining the current control mode include:

[0023] The first term is obtained by multiplying the raw material sensitivity index by a preset first calibration coefficient, the second term is obtained by multiplying the atmosphere offset amplitude by a preset second calibration coefficient, the third term is obtained by multiplying the raw material sensitivity index by the atmosphere offset amplitude and then by a preset third calibration coefficient, and the first, second and third terms are added together to obtain the firing boundary drift characterization value.

[0024] When the firing boundary drift characterization value is less than or equal to the preset first threshold, the control mode is determined to be high recovery mode;

[0025] When the firing boundary drift characterization value is greater than the preset first threshold and less than or equal to the preset second threshold, the control mode is determined to be the balance mode.

[0026] When the value of the firing boundary drift characterization is greater than the preset second threshold, the control mode is determined to be the steady-state protection mode.

[0027] Furthermore, methods for generating control output quantities include:

[0028] Control output includes waste heat recovery ratio, insulation correction time, and air supply path selection;

[0029] The waste heat recovery ratio is obtained by subtracting the ratio term from the preset baseline waste heat recovery ratio and then truncating it. The ratio term is the preset recovery ratio adjustment coefficient multiplied by the firing boundary drift characterization value.

[0030] The heat preservation correction time is obtained by multiplying the preset heat preservation time adjustment coefficient by the firing boundary drift characterization value, adding the preset reference heat preservation time, and then subjecting it to truncation and rate of change limitation.

[0031] The air supply path is selected as one of the following: preheating priority path, combustion priority path, or bypass gas stabilization path.

[0032] Furthermore, the methods for implementing linkage control include:

[0033] Based on the decision-based control mode, the control output quantity is subjected to linkage control, wherein:

[0034] In high recovery mode, the cutoff range of the waste heat recovery ratio is between the preset upper limit of the waste heat recovery ratio and the preset benchmark waste heat recovery ratio. The air supply path is selected as either the preheating priority path or the combustion priority path, and the heat preservation correction time is set to the preset benchmark heat preservation time.

[0035] In balance mode, the cutoff range of the waste heat recovery ratio is between the preset benchmark waste heat recovery ratio and the preset lower limit of the waste heat recovery ratio, and the air supply path is selected as the combustion priority path.

[0036] In steady-state protection mode, the waste heat recovery ratio is set to the preset lower limit of the waste heat recovery ratio, the air supply path is switched to the bypass stable gas path, and the insulation correction time is set to the preset upper limit of the insulation correction time.

[0037] Furthermore, based on the determined control output, linkage control is executed at different stages of the firing process, specifically including:

[0038] The firing process includes a preheating stage, a heating and phase change transition stage, and a high-temperature holding stage;

[0039] During the preheating stage, the standardized directional deviations of water vapor volume fraction and recycled gas temperature are used as constraints, and the fresh air valve, return valve, or auxiliary heating power is adjusted according to the deviation direction of the standardized directional deviations.

[0040] During the heating and phase change transition phase, the oxygen partial pressure standardization directional deviation and the logarithmic ratio of carbon monoxide concentration to carbon dioxide concentration standardization directional deviation are used as constraints. The blower, waste heat recovery ratio or burner power are adjusted according to the deviation direction of the corresponding standardization directional deviation.

[0041] During the high-temperature heat preservation stage, the proportion of waste heat recovery and the heat preservation correction time are finely adjusted based on the firing boundary drift characterization value.

[0042] Furthermore, methods for fine-tuning the waste heat recovery ratio and insulation correction time include:

[0043] Linear fitting is performed on the firing boundary drift characterization value of several control cycles within a continuously preset trend fitting period during the heat preservation stage. When the slope obtained by fitting is greater than the preset trend slope threshold, the firing boundary drift characterization value is determined to be in an upward trend; when the slope obtained by fitting is not greater than the preset trend slope threshold, the firing boundary drift characterization value is determined to be in a stable or downward trend.

[0044] When an upward trend is detected, a command to reduce the opening of the reflux valve is sent to reduce the proportion of waste heat recovery. Within the preset upper limit of the allowable heat preservation correction time, a heat preservation delay command is sent to the firing equipment to extend the heat preservation correction time.

[0045] When the trend is determined to be stable or declining, the current waste heat recovery ratio and insulation correction time should be maintained unchanged.

[0046] Furthermore, the method for updating the calibration coefficients includes:

[0047] Multiply the overall quality deviation by the preset scale mapping coefficient to obtain the mapped overall quality deviation;

[0048] The prediction error is obtained by subtracting the representative value of the firing boundary drift characterization value of the current batch from the overall quality deviation after mapping of the current batch. The representative value of the firing boundary drift characterization value is the time-weighted average value of the firing boundary drift characterization value of the current batch.

[0049] The calibration correction is obtained by multiplying the preset learning rate and prediction error by the corresponding input. The current calibration coefficient is added to the calibration correction to obtain the updated calibration coefficient, and a range constraint is applied to the updated calibration coefficient. The first, second, and third calibration coefficients correspond to the raw material sensitivity index, the atmosphere offset amplitude, and the product of the raw material sensitivity index and the atmosphere offset amplitude, respectively.

[0050] This application provides a waste heat firing system for low-grade hard clay, used to perform the aforementioned waste heat firing method for low-grade hard clay; the system includes:

[0051] The raw material sensitivity module collects and normalizes the detection information of the current batch of raw materials before loading the kiln, and then calculates the weighted sum of the normalized values ​​to obtain the raw material sensitivity index.

[0052] The atmosphere offset module collects the atmosphere parameters of the recycled gas in the waste heat recovery loop during the firing process, calculates the standardized directional deviation of each atmosphere parameter, and calculates the atmosphere offset amplitude based on each standardized directional deviation.

[0053] The boundary drift module calculates the firing boundary drift characterization value by combining the raw material sensitivity index, atmosphere offset amplitude, and preset calibration coefficients, and determines the current control mode based on the comparison result between the firing boundary drift characterization value and the preset threshold.

[0054] The linkage control module generates a control output based on the firing boundary drift characterization value, and performs linkage control during the firing process based on the control output, control mode and each standardized directional deviation.

[0055] The feedback update module collects quality feedback information and constructs a comprehensive quality deviation after the current batch of products leaves the kiln, and updates the calibration coefficients based on the comprehensive quality deviation.

[0056] The beneficial effects of this application are as follows: This application constructs a raw material sensitivity index for low-grade hard clay before kiln loading and identifies the atmosphere offset vector of recycled waste heat gas during firing. It incorporates raw material batch fluctuations and waste heat atmosphere fluctuations into the firing control judgment, avoiding the shortcomings of existing technologies that rely solely on kiln temperature or fixed firing regime for empirical adjustment, which makes it difficult to adapt to the fluctuation characteristics of low-grade hard clay.

[0057] This application further couples the raw material sensitivity index with the atmosphere shift amplitude calculated from the atmosphere shift vector to form a firing boundary drift characterization value, so that the firing boundary shift risk under waste heat recovery conditions can be quantitatively characterized, thereby providing a unified basis for subsequent control mode determination and process parameter adjustment, and improving the pertinence and accuracy of firing control.

[0058] This application adjusts the waste heat recovery ratio, heat preservation correction time, and air supply path in conjunction with the firing boundary drift characterization value. This ensures that the utilization of waste heat gas is no longer solely aimed at energy saving, but is matched with mineral phase transformation, densification, and firing stability, reducing the probability of local under-firing, over-firing, insufficient densification, and linear shrinkage fluctuations. While maintaining the effectiveness of waste heat utilization, this application improves the stability of the low-grade hard clay firing process and the consistency of product performance, achieving a balance between energy saving and firing quality control. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating a method for firing low-grade hard clay using residual heat, as described in this application.

[0060] Figure 2 This is an example diagram of the judgment and control mode of a waste heat firing method for low-grade hard clay according to this application;

[0061] Figure 3 This is a module example diagram of a waste heat firing system for low-grade hard clay according to this application. Detailed Implementation

[0062] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0063] A method and system for waste heat firing of low-grade hard clay, comprising the following embodiments:

[0064] Example 1:

[0065] A waste heat firing method for low-grade hard clay is applied in firing equipment with a waste heat recovery loop. The firing equipment includes a tunnel kiln, roller kiln, intermittent firing kiln, or a firing system with a cooling section and heat recovery branch. The waste heat recovery loop can send the recovered flue gas or hot air generated in the cooling section of the firing equipment to the preheating section, combustion section, or bypass branch via pipeline to achieve cascade utilization of waste heat. The reuse gas pipeline in the waste heat recovery loop is equipped with an extraction sampling interface, a heated sampling tube, and a dust removal and condensation pretreatment device to ensure that the online gas analyzer obtains a representative clean dry basis gas sample. Each online sensor performs automatic zero-point calibration and range calibration according to a preset calibration cycle, and triggers fault safety logic when the sensor output signal shows abnormalities such as over-range, freezing, or jump, automatically switching the corresponding control channel to the safe alternative value pre-stored in the process boundary table.

[0066] The safety substitution value includes the substitution atmosphere parameter value used for calculation when the corresponding sensor fails, the substitution output value of the corresponding actuator, and the protection level of the corresponding control mode; wherein, the safety substitution value of the atmosphere parameter is determined according to the direction that prevents the firing boundary drift characterization value from being underestimated, the safety substitution output value of the actuator is determined according to the principle of reducing the disturbance of recycled gas, maintaining oxygen supply, and maintaining the thermal balance in the kiln, and the protection level of the control mode is not lower than that of the balance mode; when the sensor related to oxygen partial pressure, carbon monoxide concentration, carbon dioxide concentration, or recycled gas temperature fails, the controller is prohibited from entering the high recovery mode; when any of the above-mentioned key sensors fails and cannot be replaced by redundant measuring points, the controller switches to the steady-state protection mode; the low-grade hard clay refers to hard clay raw material with an alumina content lower than the preset alumina content threshold and an impurity content higher than the preset impurity content threshold; the preset alumina content threshold and the preset impurity content threshold are determined according to the refractory material industry's classification standard for hard clay grade, and the impurity content includes the content of ferric oxide, the total amount of alkali metal oxides, and the total amount of other non-alumina and non-silica components; Low-grade hard clay has a wide range of applications in the preparation of refractory materials such as refractory bricks, castables, and kiln linings. The technical advantage of using a waste heat recovery loop to fire low-grade hard clay is twofold: firstly, it fully utilizes the sensible heat and chemical heat carried in the high-temperature gas discharged from the cooling section of the firing equipment, reducing primary energy consumption during the firing process; secondly, by introducing the recovered gas into the preheating and combustion sections, it achieves cascaded utilization of heat, making low-grade hard clay, which originally required higher energy consumption to complete mineral phase transformation and densification due to its low grade, economically viable. It has achieved a firing cost competitiveness comparable to that of medium and high-grade raw materials; however, parameters such as the temperature, oxygen content, water vapor content, and carbon monoxide and carbon dioxide concentration of the recycled gas in the waste heat recovery loop will fluctuate with the changes in the operating conditions of the firing equipment. This fluctuation has a particularly significant impact on the firing quality of low-grade hard clay, because the sintering window of low-grade raw materials is narrow and they are highly sensitive to changes in atmosphere. Therefore, a control method is needed that can sense the fluctuations in waste heat atmosphere in real time and dynamically adjust the firing process parameters accordingly, so as to maintain the stability of firing quality while ensuring the utilization rate of waste heat.

[0067] The method described in this embodiment is initiated when the following conditions are met: the low-grade hard clay of the current batch to be fired has completed raw material testing, including at least the main chemical composition, mineral composition, water absorption, and particle size distribution; the waste heat recovery loop is connected and has the ability to distribute the recovered gas to the preheating section, combustion section, or bypass branch; the kiln temperature, recycled gas temperature, oxygen partial pressure, water vapor volume fraction, carbon monoxide concentration, carbon dioxide concentration, air volume, valve position, burner power, and blower frequency are all capable of online acquisition; the water vapor volume fraction measurement point is set in the heat-traced sampling branch before the condensation pretreatment device, or an in-situ wet basis measurement method is used to ensure that the water vapor data involved in the atmosphere offset calculation is not changed by the condensation process; the carbon monoxide concentration and carbon dioxide concentration measurement points are set after the dust removal condensation pretreatment device to ensure that the flue gas analyzer obtains a clean dry basis gas sample, and all online sensors have been calibrated and are in normal working condition; the system has a pre-stored basic firing regime and corresponding process boundary table.

[0068] The process boundary table is a set of process parameters established and stored in the controller, corresponding to the basic firing regime. It is used to uniformly record the boundary parameters of various detection quantities, atmosphere parameters, control output quantities, and quality evaluation indicators under the current product grade, kiln type, raw material source, and kiln loading density. The boundary parameters include parameter name, parameter symbol, parameter unit, process target value, upper limit, lower limit, historical maximum value, historical minimum value, historical standard deviation, lower limit of standard deviation, initial weight coefficient, atmosphere weight coefficient, calibration coefficient and its non-negative lower limit, high temperature performance index, mode judgment threshold, safety substitute value, allowable input range, and allowable value range. Among them, the process target value is used to calculate the standardized directional deviation of each atmosphere parameter relative to the standard process state; the upper and lower limits are used to truncate or project the normalization, control output quantity, and calibration coefficient update; and the safety substitute value is used for fault safety control when the sensor fails or the actuator malfunctions. The name, symbol, unit, upper limit, lower limit, and value range of the same parameter are consistent in the process boundary table, the instruction manual, and the control algorithm. Parameters not written into the process boundary table are not used as the boundary judgment basis of the controller.

[0069] The core control logic of the method is executed according to a three-layer architecture: a raw material sensitivity index layer, an atmosphere offset vector layer, and a firing boundary drift characterization value layer. The reason for using a three-layer architecture instead of a single-layer or two-layer architecture is that the factors affecting the firing quality during the waste heat firing of low-grade hard clay can be clearly divided into two independent sources: raw material-side factors and atmosphere-side factors. Furthermore, the influence of these two factors on the firing boundary exhibits a non-linear coupling effect, and considering either end in isolation cannot accurately assess the actual offset risk of the firing boundary. The raw material sensitivity index layer quantifies and encodes the firing sensitivity of the current batch of raw materials before kiln loading. Its effect is to compress multi-dimensional raw material detection information into a single scalar index, reducing the computational complexity of subsequent control logic and enabling rapid assessment of the raw material sensitivity. The risk contribution from the material side; the atmosphere offset vector layer identifies the degree and direction of deviation of the recycled waste heat gas from the standard process state in real time during the firing process. Its effect is to transform the multi-parameter fluctuation information of the waste heat gas into a directional offset description, providing a basis for subsequent direction-sensitive control; the firing boundary drift characterization layer couples the raw material sensitivity index with the atmosphere offset vector to obtain a dimensionless index characterizing the comprehensive offset risk of the actual operating point of the firing process relative to the standard process boundary, and determines the control mode and generates control output based on this. Its effect is to integrate the risk information from the raw material side and the atmosphere side into a unified decision-making basis, avoiding decision conflicts that may occur when multiple indicators are judged in parallel; the method is as follows. Figure 1 As shown, it includes the following steps:

[0070] Step S1: Before loading the kiln, collect and normalize the detection information of the current batch of raw materials, and then sum the normalized values ​​by weight to obtain the raw material sensitivity index. The technical effect of this step is that by quantifying and fusing the multi-dimensional detection information of the raw materials before loading the kiln, the comprehensive sensitivity of the raw materials to changes in the residual heat atmosphere is encoded into a scalar index. This allows the control logic in the subsequent firing process to quickly obtain risk information from the raw materials without repeatedly accessing the original detection data, thereby reducing the computational burden of real-time control and improving the control response speed. The reason for using a pre-coding method instead of calculating the raw material sensitivity in real time during firing is that the chemical composition, physical properties, and mineral composition of the raw materials do not change after loading the kiln and are batch-level static parameters. Encoding them once before loading the kiln can be reused throughout the entire firing cycle, avoiding unnecessary repeated calculations.

[0071] Before loading the kiln, raw material testing information for the current batch of low-grade hard clay is collected. This information includes three categories: chemical composition parameters, physical property parameters, and mineral composition parameters. Chemical composition parameters include alumina content, silica content, ferric oxide content, total alkali metal oxides, and loss on ignition. Alumina content, silica content, ferric oxide content, and total alkali metal oxides are obtained through X-ray fluorescence spectroscopy analysis, while loss on ignition is obtained using the standard loss on ignition test method. Physical property parameters include water absorption rate, bulk density, median particle size, and plasticity index. Water absorption rate is obtained using the standard water absorption rate test method, bulk density is obtained using a bulk density measurement method, median particle size is obtained using a laser particle size analyzer, and plasticity index is obtained using the standard plasticity test method. Mineral composition parameters include clay mineral content, obtained through X-ray diffraction analysis, characterizing... The total proportion of clay minerals such as kaolinite and illite in the raw materials; the selection of the above parameters is based on the following: the alumina content and silica content together determine the stoichiometric basis of the mullite reaction and the amount of high-temperature liquid phase generated, which are the core chemical factors affecting the firing temperature range and mineral phase transformation path; the ferric oxide content and the total amount of alkali metal oxides, as the main fluxing impurities, directly affect the liquid phase appearance temperature and the sintering window width; the loss on ignition reflects the total content of organic matter, carbonates and structural water in the raw materials, which determines the amount of gas released during firing and the endogenous contribution of the reducing atmosphere; the water absorption rate and bulk density reflect the pore structure and compactness of the green body, which determine the atmosphere penetration depth and heat and mass transfer rate; the median particle size and clay mineral content jointly affect the reactivity and sintering kinetics of the green body; the plasticity index reflects the forming moisture content and the microstructure characteristics of the green body, which indirectly affects the drying shrinkage behavior and cracking sensitivity during the preheating stage.

[0072] The above-mentioned measured quantities were normalized. Normalization was used because the physical dimensions and numerical ranges of these quantities differ significantly. For example, alumina content, expressed as a mass percentage, is typically in the tens, while median particle size, measured in micrometers, can range from hundreds to thousands. If the raw values ​​were directly used in the weighted summation, parameters with larger numerical ranges would dominate the summation result, masking the contribution of parameters with smaller numerical ranges but equally important for firing sensitivity. This would prevent the raw material sensitivity index from accurately reflecting the relative influence of each parameter. The normalization method employed was a minimum-maximum normalization approach. For any measured quantity, the measured value was subtracted from the minimum value recorded in historical batch data or the process boundary table, and then divided by the minimum value recorded in the historical batch data or process boundary table. The difference between the recorded maximum and minimum values ​​is used to map the detected quantity to a range of zero to one. The reason for using the min-maximum normalization method instead of the standardization method or other normalization methods is that the min-maximum normalization method can strictly map all detected quantities to a closed range of zero to one, so that the raw material sensitivity index obtained by subsequent weighted summation has a clear value boundary, which is convenient for comparison with the preset threshold and pattern determination. At the same time, the calculation process of this method is simple and intuitive, requiring only the maximum and minimum values ​​from historical data, which is suitable for real-time execution in industrial controllers. The above normalization processing is performed on alumina content, silica content, ferric oxide content, total alkali metal oxides, loss on ignition, water absorption rate, bulk density, median particle size, plasticity index and clay mineral content respectively to obtain their corresponding normalized values.

[0073] When the maximum value of a certain detection quantity recorded in the historical batch data or process boundary table is equal to the minimum value, division normalization is not performed. The controller sets the normalized value of the detection quantity to the fixed normalized value pre-stored in the process boundary table and marks the detection quantity as a parameter that does not distinguish the current batch. When the measured value of a certain detection quantity exceeds the maximum or minimum value recorded in the historical batch data or process boundary table, the controller first truncates it according to the allowable detection range in the process boundary table and then participates in the normalization calculation. The selection of the normalization upper and lower limits takes priority from the upper and lower limits confirmed by the process in the process boundary table. When the process boundary table does not have corresponding upper and lower limits, the maximum and minimum values ​​in the historical qualified batches of the same firing equipment, the same raw material source, and the same product brand are used.

[0074] Based on the above normalized values, a raw material sensitivity index is constructed. The reason for using a weighted summation method to construct the raw material sensitivity index is that weighted summation is a linear combination method with a transparent calculation process and clear physical meaning. Each weight coefficient directly reflects the contribution of the corresponding raw material parameter to the firing sensitivity, making it easy for process engineers to understand and verify. Furthermore, the computational complexity of weighted summation is extremely low, suitable for execution in industrial controllers at millisecond intervals without consuming excessive computing resources. The raw material sensitivity index is equal to the weighted sum of the following terms: the first term is the preset first weight coefficient multiplied by the complementary value of the normalized alumina content, i.e., the difference obtained by subtracting the normalized alumina content from one and then multiplying it by the preset first weight coefficient; the second term is the preset second weight coefficient multiplied by the normalized silica content; the third term is the preset... The third weighting coefficient is multiplied by the normalized value of ferric oxide content; the fourth term is the preset fourth weighting coefficient multiplied by the normalized value of total alkali metal oxides; the fifth term is the preset fifth weighting coefficient multiplied by the normalized value of loss on ignition; the sixth term is the preset sixth weighting coefficient multiplied by the normalized value of water absorption; the seventh term is the preset seventh weighting coefficient multiplied by the complementary value of the normalized value of bulk density, that is, the difference obtained by subtracting the normalized value of bulk density from one and then multiplying it by the preset seventh weighting coefficient; the eighth term is the preset eighth weighting coefficient multiplied by the complementary value of the normalized value of median particle size, that is, the difference obtained by subtracting the normalized value of median particle size from one and then multiplying it by the preset eighth weighting coefficient; the ninth term is the preset ninth weighting coefficient multiplied by the normalized value of plasticity index; and the tenth term is the preset tenth weighting coefficient multiplied by the normalized value of clay mineral content.

[0075] The aforementioned preset weighting coefficients are determined based on the contribution of each raw material parameter to the firing sensitivity. They are obtained by performing a multiple regression analysis on historical batch firing data of the same firing equipment and similar low-grade hard clay, using the raw material detection value of each batch as the independent variable and the overall firing quality deviation of the corresponding batch as the dependent variable. The regression coefficients of each raw material parameter on the overall firing quality deviation are extracted and then normalized before being used as the values ​​of each weighting coefficient. When historical data does not meet the model's identifiability conditions, each weighting coefficient is determined by the initial weighting coefficients in the process boundary table. These initial weighting coefficients are obtained from process test records. The initial weighting coefficients are generated from historical qualified batch records or internal control process documents, and satisfy the constraint that each weighting coefficient is non-negative and the sum of the weighting coefficients is one. Initial weighting coefficients that have not been verified by quality feedback are only used as initial control parameters. The training process for these weighting coefficients involves first collecting complete historical batch data from the production records of the same firing equipment that meets the model's identifiability conditions. The model's identifiability conditions are that the number of historical batches is not less than the number of weighting coefficients to be determined and the number of data required for the intercept term, and that there is no perfectly linear correlation among the columns of the independent variable matrix. When the independent variable matrix does not meet the identifiability conditions, regression training results are not used, and the controller adopts... The initial weighting coefficients confirmed by the process in the process boundary table are used, and these weighting coefficients are marked as parameters to be corrected by subsequent batch quality feedback. Data for each batch includes raw material test values ​​before kiln loading and comprehensive firing quality deviation values ​​after kiln exit. Then, the collected raw material test values ​​for each batch are subjected to the same minimum-maximum normalization process as in step S1, so that all parameters are on the same numerical scale. Next, using the normalized raw material parameter values ​​as the independent variable matrix and the comprehensive firing quality deviation of the corresponding batch as the dependent variable vector, the regression coefficients of the multiple linear regression equation are solved using the least squares method. The least squares method minimizes the predicted values ​​for each batch. The optimal regression coefficients are determined by the sum of squared residuals between the quality deviation and the actual quality deviation. Finally, the absolute values ​​of each regression coefficient are taken and normalized so that the sum of each weight coefficient is one. The normalized regression coefficients are the values ​​of each preset weight coefficient. The advantage of using the multiple regression analysis method to obtain the weight coefficients is that this method can objectively extract the actual influence of each raw material parameter on the firing quality from historical data, avoiding the subjective bias that may be caused by purely relying on manual experience to assign values. At the same time, multiple regression analysis can separate the independent contributions of each parameter to a certain extent and reduce the interference of the correlation between parameters on the determination of weights.

[0076] The physical meaning of the raw material sensitivity index is the overall sensitivity of the batch of raw materials to changes in the residual heat atmosphere. A higher raw material sensitivity index value indicates that the batch of raw materials is more prone to firing boundary shifts when the residual heat atmosphere fluctuates, i.e., a higher risk of under-firing or over-firing. In the above calculations, alumina content is included in the calculation as its complementary normalized value because a lower alumina content indicates a lower raw material grade, a weaker chemical driving force for the high-temperature mullitization reaction, and greater sensitivity to atmosphere changes. Bulk density is included in the calculation as its complementary normalized value because a lower bulk density results in a more porous green body, a greater atmosphere penetration depth, and a greater susceptibility to changes during firing. The material is susceptible to atmospheric influences. The median particle size is included in the calculation as a complementary value of its normalized value because the finer the particle size, the larger the specific surface area, the higher the reactivity, and the more sensitive it is to atmospheric and temperature fluctuations. The silica content, ferric oxide content, total alkali metal oxides, loss on ignition, water absorption, plasticity index, and clay mineral content are all directly included in the calculation as their normalized values ​​because the larger the values ​​of these parameters, the more they correspond to increased liquid phase generation, enhanced fluxing effect, narrower sintering window, increased gas release, increased porosity, higher forming moisture content, and enhanced mineral transformation activity, all of which make the raw material more sensitive to changes in atmosphere.

[0077] Step S2 involves collecting atmospheric parameters of the recycled gas in the waste heat recovery loop during the firing process, calculating the standardized directional deviation of each atmospheric parameter, and calculating the atmospheric offset amplitude based on each standardized directional deviation. The technical effect of this step is that it transforms the multidimensional atmospheric parameter fluctuation information of the recycled gas in the waste heat recovery loop into a structured offset description. The atmospheric offset vector retains the directional information of each parameter's deviation from the process target, while the atmospheric offset amplitude compresses the multidimensional offset into a scalar for quantifying the overall deviation degree. The reason for using atmospheric offset identification instead of simple single-parameter threshold alarm is that there is an inherent thermodynamic and chemical kinetic relationship between parameters such as temperature, oxygen content, water vapor content, and carbon monoxide and carbon dioxide concentrations of the recycled gas. The offset of a single parameter is often accompanied by the coordinated change of other parameters. Setting an alarm threshold only for a single parameter cannot capture the comprehensive impact of the coordinated offset of multiple parameters on the firing quality. The atmospheric offset vector and atmospheric offset amplitude can completely describe the offset state of the waste heat atmosphere from both directional and amplitude dimensions, providing sufficient information support for subsequent coupled calculations and direction-sensitive control.

[0078] After firing begins, atmospheric parameters of the recycled gas in the waste heat recovery loop are collected in real time using a detection device. These atmospheric parameters include: recycled gas temperature (collected by a temperature sensor); oxygen partial pressure (collected by an oxygen probe); water vapor volume fraction (collected by a dew point meter or humidity sensor); and carbon monoxide and carbon dioxide concentrations (collected by a flue gas analyzer). The reason for selecting these five atmospheric parameters as inputs is that the recycled gas temperature directly determines the contribution of waste heat recovery to the kiln's thermal balance and the heating rate of the billet, making it the primary parameter affecting the firing thermodynamic process; the oxygen partial pressure determines the redox atmosphere state within the kiln, directly affecting the valence state transformation of iron components and the mullite formation reaction pathway; and the water vapor volume fraction affects the billet... The moisture absorption and vapor pressure gradient on the surface of the clay affect the drying behavior during the preheating stage and the mineral dehydration process during the heating stage, and have a greater impact on the firing quality than on medium and high grade hard clay. Carbon monoxide concentration and carbon dioxide concentration together reflect the completeness of combustion of recycled gas and the degree of residual reducing gas, which are key parameters for characterizing the chemical properties of the atmosphere. The above five parameters comprehensively cover the main influence channels of waste heat recycled gas on the firing process from both thermodynamic state and chemical composition dimensions, and can provide sufficient information input for subsequent atmosphere deviation identification. Before participating in subsequent calculations, the sampling signals of the above sensors are first processed by sliding window mean filtering to suppress the interference of instrument drift and transient noise on control decisions.

[0079] The reason for using sliding window mean filtering instead of median filtering or low-pass filtering is that sliding window mean filtering has a smaller computational load, making it suitable for execution at fixed sampling periods in industrial controllers. Furthermore, its filtering effect has good suppression capability for Gaussian random noise, and the measurement noise of industrial sensors approximately follows a Gaussian distribution under normal operating conditions. Therefore, sliding window mean filtering can effectively reduce the interference of noise on subsequent deviation calculations while maintaining signal response speed. The sliding window mean filtering process involves taking the arithmetic mean of the sampled values ​​at each sampling time, subtracting the preset number of samples from the current time and previous preset filtering window samples, as the filtered output value for that time. The preset number of samples in the filtering window is determined based on the sensor response. The time constant, sampling period, and preset allowable control delay are determined. The preset allowable control delay is jointly determined by the control period of the firing equipment, the response time of the actuator, and the allowable rate of change of the corresponding atmosphere parameters in the process boundary table. The signal delay generated by the filtering process does not affect the effective control of the fresh air valve, return valve, air supply fan, burner power regulation unit, and bypass valve. When determining the number of samples in the filtering window, the controller prioritizes ensuring that the signal delay corresponding to the filtering window does not exceed the preset allowable control delay and ensures that the filtered atmosphere parameters cover a continuous number of valid samples of no less than the preset number of valid samples. When there is a conflict between the sensor response time constant and the firing process control period, not exceeding the preset allowable control delay is taken as the priority constraint.

[0080] The atmospheric parameters, collected in real time and filtered, are compared with the corresponding process target values ​​in the system's pre-stored process boundary table. The standardized directional deviation and atmospheric offset amplitude of each atmospheric parameter are calculated. The reason for using standardized directional deviation instead of simple absolute or percentage deviation is that the physical dimensions and normal fluctuation ranges of each atmospheric parameter differ significantly. For example, the normal fluctuation range of recycled gas temperature may be tens of degrees Celsius, while the normal fluctuation range of carbon monoxide concentration may only be tens of parts per million. Directly comparing absolute deviations would fail to reflect the relative severity of each parameter's offset. Using the historical standard deviation of each parameter under normal operating conditions as the standardization benchmark can unify the deviations of parameters with different dimensions and fluctuation ranges. The deviation is converted into a dimensionless deviation value scaled by the normal fluctuation range of each parameter, making the degree of deviation of each parameter comparable. At the same time, the standardized directional deviation retains the positive and negative direction information of the deviation, which is crucial for the control logic in the subsequent step S4 to determine the adjustment direction of the actuator based on the deviation direction. The standardized directional deviation is calculated as follows: for each atmosphere parameter, first calculate the difference between the measured value of the parameter and the process target value, and then divide it by the standard deviation of the parameter in the historical operating data to obtain the dimensionless standardized directional deviation value, which retains the positive and negative direction information of the deviation. The normal operating condition is the historical batch operating condition in which the sensor does not alarm, the product quality feedback information meets the preset stability conditions, and the firing equipment does not trigger the fault safety logic.

[0081] Specifically, the directional bias of oxygen partial pressure standardization is equal to the difference between the measured oxygen partial pressure and the process target value, divided by the historical standard deviation of oxygen partial pressure; the directional bias of water vapor volume fraction standardization is equal to the difference between the measured water vapor volume fraction and the process target value, divided by the historical standard deviation of water vapor volume fraction; the directional bias of recycled gas temperature standardization is equal to the difference between the measured recycled gas temperature and the process target value, divided by the historical standard deviation of recycled gas temperature; for carbon monoxide and carbon dioxide concentrations, standardization is performed using a logarithmic ratio, rather than calculating the standard deviation for carbon monoxide and carbon dioxide concentrations independently. The standardization bias is addressed by using a logarithmic ratio because carbon monoxide and carbon dioxide concentrations are not thermodynamically independent variables. They are coupled through chemical equilibrium relationships such as combustion and Boudouin reactions, and their ratio directly reflects the redox characteristics of the recycled gas. Considering the shift in either carbon monoxide or carbon dioxide concentration alone cannot accurately characterize changes in the redox state of the atmosphere. The logarithmic form is also beneficial because the ratio of carbon monoxide to carbon dioxide concentrations can span several orders of magnitude in actual operating conditions. Directly using a linear ratio would result in a numerically larger shift in the high-concentration range. The shift occurs in the low concentration range, and logarithmic transformation can compress the scale of large-scale numerical changes, making the shifts at different concentration levels numerically comparable, thereby improving the stability and robustness of shift identification. The specific calculation method is as follows: first, calculate the measured logarithmic ratio, that is, add the measured carbon monoxide concentration value to a preset positive bias concentration value as the numerator, add the measured carbon dioxide concentration value to the preset positive bias concentration value as the denominator, and take the natural logarithm of the fraction; then calculate the target logarithmic ratio, that is, add the process target carbon monoxide concentration value to the ..., and add the measured carbon dioxide concentration value to the preset positive bias concentration value as the denominator, and take the natural logarithm of the fraction. The target carbon concentration value is added to the preset positive bias concentration value as the denominator, and the natural logarithm of the fraction is taken. Then, the target logarithmic ratio value is subtracted from the measured logarithmic ratio value to obtain the logarithmic ratio deviation, which is then divided by the standard deviation of the logarithmic ratio under normal operating conditions to obtain the standardized directed deviation of the logarithmic ratio of carbon monoxide and carbon dioxide concentrations. The logarithmic ratio under normal operating conditions is calculated by adding the carbon monoxide concentration under normal operating conditions to the preset positive bias concentration value as the numerator, adding the carbon dioxide concentration under normal operating conditions to the preset positive bias concentration value as the denominator, and taking the natural logarithm of the fraction.

[0082] The preset positive bias concentration value is a positive number much smaller than the carbon monoxide and carbon dioxide concentrations under normal operating conditions. Its value is not lower than the detection limit of the flue gas analyzer, which is used to prevent the value from being unstable when the denominator is zero or close to zero in the logarithmic calculation. When the measured value of carbon monoxide or carbon dioxide concentration is lower than the detection limit of the flue gas analyzer, the detection limit value is used to replace the measured value in the calculation, so as to avoid the instrument noise being amplified into false control disturbances by logarithmic calculation in the low concentration range. The historical standard deviation of the above parameters is obtained by statistical calculation of historical operating data accumulated by the same firing equipment under normal operating conditions, reflecting the degree of dispersion of each atmosphere parameter within the normal fluctuation range.

[0083] When the standard deviation of a certain atmosphere parameter in historical operating data is zero, or lower than the preset lower limit of standard deviation, the controller uses the lower limit of standard deviation as the standardization denominator. The lower limit of standard deviation is determined by the corresponding sensor detection resolution, the allowable fluctuation range of the process target, and the fluctuation records of historical qualified batches, and is written into the process boundary table. If an effective standardized directional deviation cannot be obtained after using the lower limit of standard deviation, the sensor channel corresponding to the parameter is determined to be unusable, and fault safety logic is triggered.

[0084] The aforementioned standardized directional biases collectively constitute the atmosphere offset vector. This vector retains the directional information of each atmosphere parameter's deviation from the process target, where a positive value indicates that the measured value is higher than the process target value, and a negative value indicates that the measured value is lower than the process target value. The significance of retaining the deviation directional information lies in the fact that high and low oxygen partial pressures have different effects on mineral phase transformation and redox states. A high oxygen partial pressure is conducive to an oxidizing atmosphere but may lead to excessive oxidation of iron components, affecting the color and strength of the product. A low oxygen partial pressure may form a reducing atmosphere, resulting in dark core defects and carbonaceous residues. The compensation methods required for high and low recycled gas temperatures also differ. A high temperature requires increasing the proportion of fresh air mixing. Alternatively, the waste heat recovery ratio can be reduced to prevent the surface of the billet from heating up too quickly. If the ratio is too low, the burner power or the waste heat recovery ratio needs to be increased to maintain thermal balance. If the water vapor volume fraction is too high, dehumidification needs to be strengthened or the waste heat recovery ratio needs to be reduced to prevent the billet from absorbing moisture and retaining steam. If the water vapor volume fraction is too low, the impact on the firing process is within a controllable range but still needs to be monitored. A positive logarithmic ratio of carbon monoxide to carbon dioxide concentration indicates enhanced atmosphere reducing power, while a negative logarithmic ratio indicates enhanced atmosphere oxidizing power. The two have opposite effects on the mineral phase transformation path and densification behavior. The above directional information is used to determine the adjustment direction of each actuator in the phased differentiated control in step S4.

[0085] Based on the atmosphere offset vector, the atmosphere offset amplitude is calculated to quantify the overall deviation of the current recycled waste heat gas from the standard process state. The reason for using a weighted summation method to combine the absolute values ​​of the four standardized directed deviations into a single scalar amplitude is that the calculation of the firing boundary drift characterization value in subsequent step S3 requires a scalar input that reflects the overall offset degree at the atmosphere end. Since the atmosphere offset vector, as a four-dimensional vector, cannot be directly coupled with the scalar form of the raw material sensitivity index, a weighted summation is needed to compress the vector information into a scalar. The reason for including the absolute values ​​of each standardized directed deviation in the weighted summation is that the atmosphere offset amplitude aims to measure the severity of the offset rather than its direction. Regardless of whether a certain atmosphere parameter is high or low, as long as the degree of deviation from the process target value is the same, its risk contribution to firing quality should be the same. Therefore, directional information needs to be eliminated, retaining only amplitude information. The atmosphere offset amplitude is equal to the weighted sum of the following four items: the first item is the preset first atmosphere weight coefficient multiplied by the absolute value of the oxygen partial pressure standardized directed deviation; the second item is the preset second atmosphere weight coefficient multiplied by the water vapor volume fraction standard. The first term is the absolute value of the directional deviation; the second term is the absolute value of the directional deviation standardized by the logarithmic ratio of carbon monoxide and carbon dioxide concentrations, multiplied by the preset third atmosphere weighting coefficient; the third term is the absolute value of the directional deviation standardized by the preset fourth atmosphere weighting coefficients multiplied by the absolute value of the directional deviation standardized by the temperature of the recycled gas; since each standardized directional deviation is dimensionless, the above weighted sum is dimensionally consistent, and each atmosphere weighting coefficient is a dimensionless coefficient, reflecting only the relative importance of the deviation of each atmosphere parameter on the firing quality; the preset first atmosphere weighting coefficient, the preset second atmosphere weighting coefficient, and the preset... The third atmosphere weighting coefficient and the preset fourth atmosphere weighting coefficient are obtained by performing a multiple regression analysis based on historical operating data of the same firing equipment, with the standardized offset of each atmosphere parameter as the independent variable and the overall firing quality deviation of the corresponding batch as the dependent variable. The regression coefficients of each standardized offset on the overall firing quality deviation are extracted, and then normalized so that the sum of each weighting coefficient is one, which is then used as the value of each atmosphere weighting coefficient. When historical data is insufficient, process technicians can also calibrate and assign values ​​based on their engineering experience in the influence of each atmosphere parameter on the firing process.

[0086] The training process for the aforementioned atmosphere weighting coefficients is as follows: First, complete data from at least a preset minimum number of batches is collected from the historical operation records of the same firing equipment. Each batch's data includes the measured values ​​of atmosphere parameters at each sampling time during the firing process and the overall firing quality deviation value after exiting the kiln. Then, the standardized directed deviations for each atmosphere parameter in each batch are calculated according to the method described in step S2, and their absolute values ​​are taken. Next, the time average of the absolute values ​​of the standardized deviations within each batch is taken as the standardized offset of each atmosphere parameter in that batch. Finally, the standardized offsets of each batch are used as the independent variable matrix, and the values ​​of the standardized deviations within each batch are used as the standardization offsets for each atmosphere parameter. The overall firing quality deviation of each batch is taken as the dependent variable vector. The regression coefficients of the multiple linear regression equation are solved using the least squares method. Finally, the absolute values ​​of each regression coefficient are taken and normalized so that the sum of each weight coefficient is one. The normalized regression coefficients are the values ​​of each preset atmosphere weight coefficient. The effect of obtaining the atmosphere weight coefficients by using the multiple regression analysis method is consistent with the effect of obtaining the raw material weight coefficients in step S1. That is, it can objectively extract the actual impact of the offset of each atmosphere parameter on the firing quality from historical data, and make the calculation results of the atmosphere offset amplitude closer to the actual process law.

[0087] In the above calculations, carbon monoxide and carbon dioxide concentrations are used in the calculation in the form of a logarithmic ratio. The physical meaning of this ratio is that it reflects the redox characteristics of the recycled gas. The larger the ratio of carbon monoxide to carbon dioxide concentration, the stronger the reducing power of the atmosphere, and the greater the impact on the mineral phase transformation and densification process. Using the logarithmic form can compress the numerical scale of large-scale concentration changes, making the calculation of the offset more stable.

[0088] The atmosphere offset amplitude treats the waste heat gas as a process medium with chemical properties rather than a simple heat source, comprehensively reflecting the degree of deviation of the waste heat recovery gas from the process target in terms of both thermodynamic state and chemical composition. The larger the atmosphere offset amplitude, the farther the current waste heat atmosphere deviates from the standard process state. The atmosphere offset vector provides additional deviation direction information in each dimension on the basis of the atmosphere offset amplitude, which can be used for the phased differentiated control in the subsequent step S4.

[0089] Step S3: Calculate the firing boundary drift characterization value using the raw material sensitivity index, atmosphere offset amplitude, and preset calibration coefficients. Determine the current control mode based on the comparison result of the firing boundary drift characterization value and a preset threshold. Specifically, as follows... Figure 2As shown, the technical effect of this step is that by coupling and fusing the static sensitivity information of the raw materials with the dynamic offset information of the atmosphere, a unified index that can comprehensively reflect the degree of risk of the actual working point of the firing process deviating from the standard process boundary is constructed. This allows subsequent control mode determination and control output calculation to be based on a single index, avoiding the decision contradictions that may occur when making independent judgments based on the raw material sensitivity index and the atmosphere offset amplitude. The reason for using coupled calculation instead of simple superposition is that the influence of raw material sensitivity and atmosphere offset on firing quality is not linearly additive. When the raw material sensitivity is high, the same degree of atmosphere offset will lead to a larger quality deviation, and vice versa. This interactive amplification effect can only be accurately characterized by introducing a coupling term.

[0090] The firing boundary drift characterization value is obtained by coupling the raw material sensitivity index and the atmosphere offset amplitude. The firing boundary drift characterization value is a dimensionless comprehensive risk index used to characterize the comprehensive risk degree of the actual working point of the firing process deviating from the standard process boundary under the current raw material batch characteristics and current waste heat atmosphere conditions. This characterization value is not a directly measurable physical quantity, but a risk assessment index constructed based on the raw material sensitivity index and the atmosphere offset amplitude through an empirical coupling model. The larger the value, the higher the risk of firing boundary drift.

[0091] The firing boundary drift characteristic value is equal to the sum of the following three terms: the first term is the preset first calibration coefficient multiplied by the raw material sensitivity index; the second term is the preset second calibration coefficient multiplied by the atmosphere shift amplitude; and the third term is the preset third calibration coefficient multiplied by the product of the raw material sensitivity index and the atmosphere shift amplitude. The reason for using a three-term sum structure instead of a linear superposition of only two terms is that the first two terms characterize the independent contributions of raw material sensitivity and atmosphere shift to the risk of firing boundary drift, while the third term, as a cross term, characterizes the interactive coupling effect between the two. This interactive coupling effect has a clear physical background in the actual firing process, namely, the mass deviation produced by highly sensitive raw materials when encountering a large atmosphere shift is much greater than the sum of the two independent effects. If the cross term is ignored, it will lead to problems with the raw material sensitivity index. When both the index and atmosphere offset amplitudes are large, the risk of firing boundary drift is underestimated, causing the controller to fail to switch to a more conservative control mode in time. Among them, the preset first calibration coefficient, the preset second calibration coefficient, and the preset third calibration coefficient represent the contributions of raw material sensitivity alone, atmosphere offset alone, and the interaction and coupling of the two to the risk of firing boundary drift, respectively. The initial values ​​of the preset calibration coefficients are obtained by collecting historical firing data of the same firing equipment under different raw material batches and different waste heat atmosphere conditions, using the raw material sensitivity index, atmosphere offset amplitude, and their product for each batch as independent variables, and the actual comprehensive firing quality deviation of the corresponding batch as the dependent variable, and using a multiple linear regression method to fit the initial values ​​of each calibration coefficient.

[0092] The training process for the initial values ​​of the aforementioned calibration coefficients is as follows: First, historical firing data covering different raw material batches and different waste heat atmosphere conditions is collected from the historical production records of the same firing equipment. The collected data must have sufficient coverage in the value space of the raw material sensitivity index and atmosphere offset amplitude to ensure the representativeness of the regression results. Then, for each historical batch, the representative value of the raw material sensitivity index is calculated according to step S1, and the representative value of the atmosphere offset amplitude is calculated according to step S2, and their product is calculated. Next, using the raw material sensitivity index, the representative value of the atmosphere offset amplitude, and their product for each batch as the independent variable matrix, and the actual comprehensive firing quality deviation of the corresponding batch as the dependent variable vector, the regression coefficients of the multiple linear regression equation are solved using the least squares method. The three regression coefficients obtained are the initial values ​​of the preset first calibration coefficient, the preset second calibration coefficient, and the preset third calibration coefficient. The effect of using the multiple linear regression method to train the initial values ​​of the calibration coefficients is that this method can determine the values ​​of each batch while minimizing the prediction error. The optimal initial value of the calibration coefficients enables the firing boundary drift characterization value to predict the actual overall firing quality deviation in the initial stage, thereby reducing the number of control mode misjudgments caused by inaccurate calibration coefficients in the early stage of system operation. The calibration coefficients are iteratively updated in subsequent step S5 based on quality feedback. The third term in the above calculation, namely the preset third calibration coefficient multiplied by the product of the raw material sensitivity index and the atmosphere offset amplitude, is a cross term used to characterize the nonlinear coupling effect between raw material sensitivity and atmosphere offset. That is, when the raw material sensitivity index and the atmosphere offset amplitude increase simultaneously, the superposition effect of the two will exceed the sum of their individual effects. The first calibration coefficient, the second calibration coefficient, and the third calibration coefficient are all non-negative coefficients. When any initial calibration coefficient obtained through regression is negative, the calibration coefficient is set to the corresponding non-negative lower limit in the process boundary table, and the remaining calibration coefficients are renormalized or refitted. Each calibration coefficient is configured with an allowable value range, and the controller performs range verification on the calibration coefficients during initialization and batch updates.

[0093] The current control mode is determined based on the comparison between the firing boundary drift characterization value and two preset thresholds, where the first preset threshold is less than the second preset threshold. The specific determination rules are as follows:

[0094] When the value of the firing boundary drift characterization is less than or equal to the preset first threshold, the control mode is determined to be the high recovery mode, which means that the current raw material sensitivity is low and the waste heat atmosphere deviation is small, the risk of firing boundary drift is within an acceptable range, and the waste heat utilization rate can be improved first.

[0095] When the value of the firing boundary drift is greater than the preset first threshold and less than or equal to the preset second threshold, the control mode is determined to be the balance mode, indicating that there is a certain risk of the firing boundary shifting, and a balance adjustment needs to be made between the waste heat utilization rate and the firing stability.

[0096] When the value of the firing boundary drift is greater than the preset second threshold, the control mode is determined to be the steady-state protection mode, indicating that the risk of firing boundary offset exceeds the balance adjustment capability, and the stability of firing quality must be prioritized.

[0097] To avoid frequent switching of the control mode near the threshold due to minor fluctuations in the firing boundary drift value, a hysteresis mechanism is introduced into the controller's mode determination. The technical effect of this hysteresis mechanism is that when the firing boundary drift value is near the threshold, factors such as sensor noise, short-term fluctuations in atmosphere parameters, and rounding during calculation may cause the value to repeatedly cross the threshold. Without hysteresis, the control mode would frequently switch between adjacent control cycles, leading to drastic changes in control outputs such as the waste heat recovery ratio, air supply path, and heat preservation correction time. These changes not only exacerbate fluctuations in the kiln's temperature and atmosphere fields but also cause frequent operation of actuators such as valves and fans, accelerating mechanical wear and potentially even causing thermal instability within the kiln. The hysteresis mechanism, by setting different switching criteria between positively crossing the threshold and negatively falling back to the threshold, ensures that once the control mode switches, the firing boundary drift value must fall back to a position one hysteresis width below the threshold before switching back, thus effectively suppressing mode chattering near the threshold.

[0098] The hysteresis mechanism is set independently for a preset first threshold and a preset second threshold. For the preset first threshold, when the firing boundary drift value increases from a range less than or equal to the preset first threshold to a range greater than the preset first threshold, exceeding the preset first threshold, the preset first threshold itself is used as the criterion for switching from high recovery mode to balanced mode. When the firing boundary drift value falls back from a range greater than the preset first threshold to a range less than the preset first threshold, it must fall back to the value obtained by subtracting the preset first hysteresis width from the preset first threshold before switching back from balanced mode to high recovery mode. For the preset second threshold, when the firing boundary drift value increases from a range less than or equal to the preset second threshold to a range greater than the preset second threshold, exceeding the preset second threshold, the preset first threshold itself is used as the criterion for switching from high recovery mode to balanced mode. The second threshold is set as the criterion for switching from the balanced mode to the steady-state protection mode. When the firing boundary drift characterization value falls from the range greater than the preset second threshold to the range less than the preset second threshold, it needs to fall back to the value obtained by subtracting the preset second hysteresis width from the preset second threshold before switching back from the steady-state protection mode to the balanced mode. The preset first hysteresis width and the preset second hysteresis width are determined according to the short-term fluctuation amplitude of the firing boundary drift characterization value under normal operating conditions. Each hysteresis width is required to be greater than the normal fluctuation amplitude to avoid false switching. At the same time, the preset first hysteresis width is less than the difference between the preset first threshold and zero to ensure the effective coverage range of the high recovery mode, and the preset second hysteresis width is less than the difference between the preset second threshold and the preset first threshold to ensure the effective coverage range of the balanced mode.

[0099] The preset first threshold and preset second threshold are obtained by statistically analyzing the product quality pass rate and waste heat utilization rate at different firing boundary drift characterization value levels in historical firing data, based on the type of firing equipment, the range of raw material characteristics, and product quality requirements. The firing boundary drift characterization value corresponding to the product quality pass rate starting to decrease to the preset first pass rate limit is selected as the preset first threshold, and the firing boundary drift characterization value corresponding to the product quality pass rate decreasing to the preset second pass rate limit is selected as the preset second threshold. The preset first pass rate limit is higher than the preset second pass rate limit. The preset first pass rate limit and the preset second pass rate limit are determined according to product standards and production management requirements, thereby dividing the value space of the firing boundary drift characterization value into three control intervals: high recovery, balance, and steady-state protection.

[0100] Step S4: Generate control output quantities based on the firing boundary drift characterization values. Based on the control output quantities, control modes, and standardized directional deviations, perform coordinated control during the firing process to achieve coordinated adjustment of the thermal path, gas path, and time regime. The technical effect of this step is that by linking the three control output quantities—waste heat recovery ratio, heat preservation correction time, and air supply path selection—instead of adjusting them independently, the heat supply, atmosphere control, and time regime during the firing process can respond synergistically to changes in the risk of firing boundary drift, avoiding imbalances in other process parameters that may be caused by adjusting a single control quantity. The reason for adopting a phased differentiated control strategy instead of a uniform control strategy throughout the entire process is that low-grade hard clay faces different main contradictions at different stages of the firing process. The main contradiction in the preheating stage is the influence of water vapor and temperature fluctuations on the drying and thermal stress of the green body, while the main contradiction in the heating and phase transformation transition stage is the redox atmosphere. Regarding the impact on the mineral phase transformation path, the main contradiction in the high-temperature insulation stage is the influence of the firing boundary drift trend on densification and the completion of mineral phase transformation. If a unified control strategy is adopted throughout the process, it is impossible to focus on the main contradictions in each stage, which may lead to the accumulation of irreversible quality defects due to the failure to correct the deviations of key parameters in some stages in a timely manner. The linkage control is driven by an industrial controller to coordinate the execution of the return valve, bypass valve, fresh air valve, air supply fan and burner power adjustment unit. The operating parameters of the above-mentioned actuators, such as valve position, speed and power, as well as the three control outputs, constitute the process parameters of the firing process. The process parameters refer to the set of all adjustable operating parameters determined by the controller according to the control mode and the firing boundary drift characterization value during the firing process and output to each actuator, including the waste heat recovery ratio, insulation correction time, air supply path selection and the specific set values ​​of each actuator in the linkage control.

[0101] The stages of the firing process are determined based on the measured temperature of the kiln temperature field and the characteristic temperature of the mineral phase transformation of this type of low-grade hard clay. The preheating stage is the interval from when the green body enters the kiln until the kiln temperature reaches the preset preheating termination temperature. The preset preheating termination temperature is determined based on the temperature at which free water and structural water are completely discharged and organic matter is basically burned out in this type of low-grade hard clay. This temperature is obtained through differential thermal analysis and thermogravimetric analysis. The heating and phase transformation transition stage is the interval from the preset preheating termination temperature to when the kiln temperature reaches the preset high-temperature holding start temperature. The preset high-temperature holding start temperature is determined based on the temperature at which the mulliteification reaction in this type of low-grade hard clay begins to enter the main reaction zone. The temperature is obtained by analyzing the onset temperature of the mullite exothermic peak in high-temperature X-ray diffraction or differential thermal analysis. The high-temperature holding stage is the interval from the preset high-temperature holding start temperature to the holding end temperature. During this stage, the kiln temperature is maintained between the preset high-temperature holding start temperature and the preset high-temperature holding end temperature. The preset high-temperature holding end temperature is determined based on the highest firing temperature required for this type of low-grade hard clay to complete mullite formation and densification. This temperature is obtained by observing the temperature at which the product's bulk density and linear shrinkage rate reach the target values ​​during firing tests. The temperature limits of each of the above stages are written into the process boundary table during system initialization. The controller determines the current firing stage based on the real-time readings of the kiln temperature sensors.

[0102] The first control output is the waste heat recovery ratio, which represents the volume proportion of recovered waste heat gas actually reused in the firing system. The reason for choosing the waste heat recovery ratio as the control output is that it directly determines the flow rate of the reused gas entering the firing system. It is the most direct means of adjusting the balance between waste heat utilization rate and the impact of atmosphere fluctuations. Reducing the waste heat recovery ratio can reduce the interference of highly fluctuating reused gas on the kiln atmosphere, but it will also reduce the waste heat utilization rate. Increasing the waste heat recovery ratio can increase the waste heat utilization rate, but it will increase the risk of atmosphere fluctuations affecting firing quality. The reason for using a linear decreasing relationship with the firing boundary drift value as the independent variable to calculate the waste heat recovery ratio is that a larger firing boundary drift value indicates a higher firing risk. If the temperature is high, the waste heat recovery ratio should be reduced to minimize atmosphere interference. The calculation process of the linear decreasing relationship is simple and the adjustment behavior is predictable, making it easy for process technicians to understand and adjust. The waste heat recovery ratio is calculated by subtracting the ratio term from the preset baseline waste heat recovery ratio to obtain a preliminary calculated value. The ratio term is the preset recovery ratio adjustment coefficient multiplied by the firing boundary drift characterization value. Then, the preliminary calculated value is truncated. If the preliminary calculated value is lower than the preset allowable lower limit of the waste heat recovery ratio, the allowable lower limit value is taken. If the preliminary calculated value is higher than the preset allowable upper limit of the waste heat recovery ratio, the allowable upper limit value is taken. If the preliminary calculated value is between the preset allowable lower limit and the preset allowable upper limit, the preliminary calculated value itself is taken.

[0103] The preset baseline waste heat recovery ratio is a set value for waste heat recovery ratio under standard process conditions. It is obtained by determining, through firing tests or historical operating data statistics, a waste heat recovery ratio that simultaneously meets product quality requirements and energy-saving targets under standard operating conditions where the raw material sensitivity index is less than or equal to a preset raw material sensitivity index baseline threshold and the atmosphere deviation amplitude is less than or equal to a preset atmosphere deviation amplitude baseline threshold. The preset raw material sensitivity index baseline threshold is determined based on the low fluctuation range of the raw material sensitivity index in historical qualified batches, and the preset atmosphere deviation amplitude baseline threshold is determined based on the normal fluctuation upper limit of the atmosphere deviation amplitude in historical qualified batches. When the current batch simultaneously meets both baseline threshold conditions, it indicates that the raw material batch fluctuation and the recovered gas atmosphere fluctuation are both within the acceptable range of the standard process, and can be used as a standard operating condition sample for determining the baseline waste heat recovery ratio. The waste heat recovery ratio adjustment coefficient is defined as the sensitivity of the firing boundary drift value to the adjustment of the waste heat recovery ratio. It is obtained by analyzing the required adjustment range of the waste heat recovery ratio to maintain product quality at different firing boundary drift value levels based on historical firing data. Linear regression fitting is used to obtain the adjustment range of the waste heat recovery ratio for each unit change in the firing boundary drift value. The preset lower and upper limits of the waste heat recovery ratio are boundary values ​​determined based on the thermal design parameters and safe operation requirements of the firing equipment. The lower limit is determined based on the waste heat replenishment required to maintain the minimum thermal balance in the kiln, and the upper limit is determined based on the maximum reclaimed gas flow rate and atmosphere fluctuation tolerance that the firing system can withstand. The above calculations show that the larger the firing boundary drift value, the lower the waste heat recovery ratio, thereby reducing the interference of highly fluctuating reclaimed gas on the firing process.

[0104] The second control output is the heat preservation correction time, which represents the actual heat preservation duration during the high-temperature heat preservation stage. The reason for selecting the heat preservation correction time as the control output is that when the residual heat atmosphere shift causes interference with the mineral phase transformation and densification processes during the heating stage, appropriately extending the high-temperature heat preservation time can provide more sufficient time compensation for the mineral phase transformation reaction. This allows the murolite formation and densification processes that were not fully completed during the heating stage due to the atmosphere shift to be compensated for during the heat preservation stage, thereby reducing the risk of underfiring caused by atmosphere fluctuations. The reason for using a linear increasing relationship with the firing boundary drift value as the independent variable to calculate the heat preservation correction time is that a larger firing boundary drift value indicates a greater disturbance to the mineral phase transformation and densification. The longer the time compensation required, the more linear the relationship can achieve this positive correlation adjustment in the simplest way. The calculation method of the heat preservation correction time is to add the preset base heat preservation time to the product of the preset heat preservation time adjustment coefficient and the firing boundary drift characterization value to obtain a preliminary calculation value. Then, the preliminary calculation value is truncated and the rate of change is limited. The truncation method is as follows: if the preliminary calculation value is lower than the preset base heat preservation time, the preset base heat preservation time is taken; if the preliminary calculation value is higher than the preset upper limit of the heat preservation correction time, the upper limit value is taken; if the preliminary calculation value is between the preset base heat preservation time and the preset upper limit of the heat preservation correction time, the preliminary calculation value itself is taken.

[0105] The rate of change is limited by the following method: when the change in the holding correction time between two adjacent control cycles exceeds the preset maximum rate of change in the holding time, the holding correction time for the current cycle is limited to the holding correction time of the previous cycle plus or minus the change corresponding to the maximum rate of change, to prevent drastic jumps in the holding time within a short period. The preset baseline holding time is the set value of the holding time under standard process conditions. It is obtained by combining the minimum holding time required for the low-grade hard clay to complete mullitization and densification under standard operating conditions with the thermal characteristics of the firing equipment and the statistical values ​​of the holding time of historical qualified batches. The preset holding time adjustment coefficient characterizes the sensitivity of the holding time adjustment to the firing boundary drift value. It is obtained by analyzing the extension of the holding time required to maintain the product's mineral phase transformation and densification at different firing boundary drift value levels based on historical firing data, and obtaining the adjustment of the holding time for each unit change in the firing boundary drift value through linear regression fitting. The maximum allowable holding time correction is determined based on the most stringent of the following constraints: production line cycle time constraint, i.e., extending the holding time must not cause conflicts with the loading and progress plans of subsequent batches; over-burning risk constraint, i.e., the holding time must not exceed the critical holding time at which the raw material begins to generate excessive liquid phase or coarsen crystal phase at the current firing temperature, which is obtained through high-temperature microscopy or statistical analysis of the holding time of historical over-burned batches; energy consumption constraint, i.e., the increased unit product energy consumption due to extending the holding time must not exceed the maximum preset energy consumption increment; the maximum rate of change of the preset holding time is determined based on the thermal inertia of the firing equipment and the response characteristics of the kiln temperature field, requiring that the adjustment rate of the holding time does not exceed the limit at which the kiln temperature field can smoothly follow; the above cutoff treatment and rate of change limit ensure that the holding time correction can effectively compensate for the risk of insufficient mineral phase transformation and densification caused by residual heat atmosphere shift and increased raw material sensitivity, and will not cause liquid phase abnormalities, crystal phase coarsening, over-burning, or production line cycle disorder due to excessive extension.

[0106] The third control output is the air supply path selection, used to determine the distribution path of the recovered waste heat gas in the firing system. The reason for choosing air supply path selection as the control output is that different air supply paths have different impact mechanisms on the firing process. Sending the recycled gas into the preheating section mainly affects the preheating and drying process of the green body, sending it into the combustion section mainly affects the combustion efficiency of the burner and the oxidation-reduction atmosphere in the kiln, while discharging it through the bypass branch completely isolates the recycled gas from interfering with the firing process. By selecting different air supply paths under different control modes, the waste heat utilization efficiency can be maximized while ensuring firing quality. The air supply path selection value is one of the following three paths: preheating priority path, combustion priority path, or bypass stabilization path. Among them, the preheating priority path means that the recovered waste heat gas is given priority to be sent into the preheating section for green body preheating and drying; the combustion priority path means that the recovered waste heat gas is given priority to be sent into the combustion section for use as combustion gas for the burner; the bypass stabilization path means that the recovered waste heat gas is discharged through the bypass branch to avoid highly fluctuating recycled gas directly entering the critical firing section.

[0107] Based on the control mode determined in step S3, the linkage control logic for the above control output is as follows:

[0108] In high recovery mode, the waste heat recovery ratio is taken as the value between the preset upper limit of the waste heat recovery ratio and the preset benchmark waste heat recovery ratio. The air supply path is selected as either the preheating priority path or the combustion priority path, and the heat preservation correction time is equal to the preset benchmark heat preservation time, so that more recovered waste heat gas enters the preheating section or the combustion section to achieve energy-saving effect. In high recovery mode, the selection of the air supply path is determined based on the standardized directional deviation of the current firing stage and the corresponding stage. When the current firing stage is the preheating stage, and the standardized directional deviation of water vapor volume fraction and the standardized directional deviation of recovered gas temperature are both within the preset allowable range of the preheating stage, the controller selects the preheating priority path. When the current firing stage is the heating and phase change transition stage or the high-temperature heat preservation stage, and the standardized directional deviation of oxygen partial pressure and logarithmic ratio are both within the preset allowable range of the preheating stage, the controller selects the preheating priority path. When the standardized directional deviations are all within the allowable range of the corresponding stage, the controller selects the combustion priority path. The allowable range of the preset preheating stage is determined based on the statistical distribution of the standardized directional deviations of water vapor volume fraction and recycled gas temperature of historical qualified batches in the preheating stage. The allowable range of the corresponding stage is determined based on the statistical distribution of the standardized directional deviations of oxygen partial pressure and logarithmic ratio of historical qualified batches in the heating and phase change transition stage and the high temperature heat preservation stage, and is checked in conjunction with the allowable upper limit, allowable lower limit and product quality qualification criteria in the process boundary table. If any corresponding standardized directional deviation exceeds its allowable range, the controller does not select the priority air supply path in the high recovery mode, but switches the air supply path to the combustion priority path or the bypass gas stabilization path.

[0109] In balance mode, the waste heat recovery ratio is taken as the middle range between the preset benchmark waste heat recovery ratio and the preset allowable lower limit of waste heat recovery ratio. The air supply path selects the combustion priority path. The heat preservation correction time is extended based on the preset benchmark heat preservation time. The extension range is determined by the product of the firing boundary drift characterization value and the preset heat preservation time adjustment coefficient. A balance is achieved between waste heat utilization rate and firing stability to prevent excessive atmosphere deviation.

[0110] In steady-state protection mode, the waste heat recovery ratio is set to the preset lower limit of the waste heat recovery ratio, the air supply path is switched to the bypass stable gas path, and the heat preservation correction time is extended to the preset upper limit of the heat preservation correction time. This reduces the direct entry of highly fluctuating recovered gas into the key firing section and simultaneously extends the heat preservation time to maintain the stability of mineral phase transformation and densification.

[0111] Based on the determination of the above three control output quantities, the controller performs linkage control at different stages of the firing process, and determines the adjustment direction of each actuator according to the direction information of each component in the atmosphere offset vector obtained in step S2.

[0112] During the preheating phase, the controller prioritizes controlling moisture content deviation and reclaimed gas temperature fluctuation. Specifically, the controller uses the standardized directional deviation of moisture volume fraction in the atmosphere offset vector and the standardized directional deviation of reclaimed gas temperature as the main constraint objects. When the standardized directional deviation of moisture volume fraction is positive and its absolute value exceeds the preset allowable range for moisture content deviation during the preheating phase, it indicates that the moisture content in the reclaimed gas is too high. The controller sends an opening increase command to the fresh air valve to adjust its valve position towards the open direction to increase the introduction of external dry fresh air. At the same time, it sends an opening decrease command to the return valve to adjust its valve position towards the closed direction to reduce the entry of high-humidity reclaimed gas. By increasing the mixing ratio of fresh air and reclaimed gas, the moisture content in the gas entering the preheating section is reduced. When the standardized directional deviation of moisture volume fraction is negative and its absolute value exceeds the above allowable range, it indicates that the moisture content is too low. The impact on the preheating phase is within a controllable range. The controller maintains the current valve positions of the fresh air valve and the return valve unchanged but continuously monitors the temperature. The controller monitors the trend of water vapor volume fraction change. When the directional deviation of the temperature standardization of the recycled gas is positive and its absolute value exceeds the preset allowable range of temperature deviation in the preheating stage, it indicates that the inlet air temperature is too high. The controller sends an opening increase command to the fresh air valve to adjust the valve position to the opening direction to introduce more low-temperature fresh air. By increasing the mixing ratio of fresh air and recycled gas, the temperature of the mixed gas entering the preheating section is reduced, avoiding cracking or excessive internal and external temperature differences caused by excessively rapid surface heating of high water absorption batches of billets. When the directional deviation of the temperature standardization of the recycled gas is negative and its absolute value exceeds the above allowable range, it indicates that the inlet air temperature is too low. The controller sends an opening increase command to the reflux valve to adjust the valve position to the opening direction to increase the introduction of high-temperature recycled gas and thus increase the waste heat recovery ratio. If adjusting the reflux valve alone cannot restore the inlet air temperature to near the process target value, the controller simultaneously sends a power increase command to the auxiliary heating power adjustment unit to increase the heat supply of the auxiliary heat source and maintain the thermal balance of the preheating section.

[0113] The preset allowable ranges for water vapor deviation and temperature deviation during the preheating stage are determined based on the thermal shock resistance and drying shrinkage characteristics of this type of low-grade hard clay body during the preheating stage. Specifically, they are obtained through process tests during the preheating stage or statistical values ​​of atmospheric parameter fluctuations from historical qualified batches. During this stage, the waste heat recovery ratio and air supply path selection serve as basic control variables. Based on these variables, the controller performs directional fine-tuning of the fresh air valve, return valve, and auxiliary heating power adjustment unit according to the deviation direction.

[0114] During the heating and phase transition stages, the controller prioritizes constraining the oxygen partial pressure deviation and the deviation of the logarithmic ratio of carbon monoxide to carbon dioxide concentrations. Specifically, the controller uses the standardized directional deviations of oxygen partial pressure and carbon monoxide to carbon dioxide concentrations in the atmosphere offset vector as the main constraint objects. When the standardized directional deviation of oxygen partial pressure is negative and its absolute value exceeds the preset allowable range for oxygen partial pressure deviation during the heating stage, it indicates that the oxygen partial pressure in the kiln is too low and the reducing atmosphere is enhanced. The controller sends a frequency increase command to the inverter of the blower to increase the blower speed, thereby increasing the oxygen supply to the kiln. At the same time, it sends an opening decrease command to the reflux valve to adjust the valve position towards the closed direction to reduce the proportion of waste heat recovery and thus reduce the amount of reducing gas introduced. When the standardized directional deviation of oxygen partial pressure is positive and its absolute value exceeds the above allowable range, it indicates that the oxygen partial pressure is too high and the oxidizing atmosphere is too strong. The controller sends a frequency decrease command to the inverter of the blower to decrease the blower speed, thereby reducing the oxygen supply, or sends an opening increase command to the reflux valve. The controller adjusts the reflux valve to the open direction to increase the waste heat recovery ratio, thereby restoring the kiln's redox atmosphere to the target process state. When the standardized directional deviation of the logarithmic ratio of carbon monoxide to carbon dioxide concentration is positive and its absolute value exceeds the preset allowable range for the logarithmic ratio deviation during the heating stage, it indicates that the atmosphere is too reducing. The controller sends a power increase command to the burner power adjustment unit to increase the combustion air volume or adjust the air-fuel ratio, and adjusts the burner power as needed to improve combustion completeness and reduce carbon monoxide concentration. At the same time, it sends a frequency increase command to the frequency converter of the blower to increase oxygen supply. When the deviation is negative and its absolute value exceeds the above-mentioned allowable range, it indicates that the atmosphere is too oxidizing. The controller sends a power decrease command to the burner power adjustment unit to reduce the gas supply to the burner, or adjusts the air-fuel ratio to restore the ratio of combustion air volume to gas supply to the process target set value, thereby restoring the kiln atmosphere to the target redox state. The above adjustments reduce the risk of mineral phase transformation boundary drift.

[0115] The preset allowable ranges for oxygen partial pressure deviation and logarithmic ratio deviation during the heating stage are determined based on the sensitivity of this type of low-grade hard clay to redox atmosphere during the heating and phase transformation transition stages, involving mineral phase transformation reactions such as mullitization and glass phase formation. Specifically, these ranges are obtained through process tests during the heating stage or statistical values ​​of atmospheric parameter fluctuations in historical qualified batches during this stage. Mineral phase transformation boundary drift refers to the phenomenon where the temperature range and reaction degree of phase transformation reactions such as mullitization and glass phase formation in clay minerals during firing deviate from the expected process settings due to changes in atmosphere.

[0116] During the high-temperature holding stage, the controller fine-tunes the waste heat recovery ratio and holding correction time based on the current firing boundary drift characterization value and its changing trend during the holding stage. The reason for introducing trend judgment during the high-temperature holding stage is that this stage is the main completion stage of the mullite formation reaction and densification process. If the firing boundary drift characterization value continues to rise during the holding stage, it indicates that the waste heat atmosphere shift is intensifying or the firing sensitivity effect of the raw materials is accumulating and amplifying. In this case, controlling solely based on the current firing boundary drift characterization value may be lagging. However, by analyzing the firing boundary drift value over multiple consecutive control cycles... Linear fitting of eigenvalues ​​to determine trends can identify rising risks in advance and take preventative measures, avoiding the discovery of insufficient densification or liquid phase anomalies at the end of the holding stage, which would be too late. The reason for using linear fitting to determine trends instead of simply comparing the differences between two adjacent control cycles is that the firing boundary drift characterization values ​​may fluctuate in the short term during the holding stage, and the difference between two adjacent control cycles is easily affected by transient fluctuations, leading to misjudgments. Linear fitting, on the other hand, utilizes data from multiple consecutive control cycles, which can effectively smooth out the impact of short-term fluctuations and make the trend determination results more reliable.

[0117] The method for determining the trend is as follows: the controller performs linear fitting on the firing boundary drift characterization value for a number of control cycles within a continuously preset trend fitting period during the heat preservation stage. The preset trend fitting period number is determined based on the fluctuation frequency and control cycle of the firing boundary drift characterization value during the heat preservation stage. When the fitted slope is greater than a preset trend slope threshold, the firing boundary drift characterization value is determined to be in an upward trend. When the fitted slope is not greater than the preset trend slope threshold, the firing boundary drift characterization value is determined to be in a stable or downward trend. The preset trend slope threshold is determined based on the fluctuation data of the firing boundary drift characterization value of historical qualified batches during the high-temperature heat preservation stage. The controller selects historical batches with qualified product quality and no heat preservation delay intervention, calculates their fitting slope distribution under the same number of trend fitting periods, and uses the upper limit value that can cover normal short-term fluctuations as the preset trend slope threshold. When the fitting slope of the firing boundary drift characterization value exceeds this threshold, it indicates that its change has exceeded the range of normal measurement noise and operating condition perturbation, requiring a reduction in the waste heat recovery ratio and an extension of the heat preservation correction time. When the firing boundary drift characterization value shows an upward trend during the heat preservation stage, the controller sends an opening reduction command to the reflux valve to adjust the valve position towards the closed direction to further reduce the waste heat recovery ratio. At the same time, within the preset upper limit of the heat preservation correction time, it sends a heat preservation delay command to the kiln propulsion control unit to extend the heat preservation time. When the firing boundary drift characterization value tends to stabilize or decrease during the heat preservation stage, the controller maintains the current output state of the reflux valve, bypass valve, fresh air valve, air supply fan, and burner power regulation unit unchanged. The control objective of this stage is to suppress insufficient densification or premature abnormal liquid phase formation, ensuring that the product's bulk density and linear shrinkage rate meet the target requirements.

[0118] In the linkage control, the control commands issued by the controller to each actuator are based on the absolute value of the standardized directional deviation of the corresponding atmosphere parameter as the basis for calculating the adjustment range. The definition of each control command and the method for determining its adjustment range are as follows.

[0119] The opening increase command refers to the controller outputting a control signal to the valve actuator to increase the valve opening by an increment based on the current value. The opening decrease command refers to the controller outputting a control signal to the valve actuator to decrease the valve opening by a decrement based on the current value. The specific value of the increment or decrement is equal to the preset valve position adjustment gain multiplied by the absolute value of the standardized directional deviation of the atmosphere parameter that triggered the command. The product is the valve position change amount for this adjustment. The preset valve position adjustment gain is obtained by statistically analyzing the valve position adjustment range required to restore the corresponding atmosphere parameter to the process target value under different standardized directional deviation levels based on historical operating data of the same firing equipment. The valve position adjustment range should be obtained by linear regression fitting for each unit change in the standardized directional deviation. When the calculated valve position change amount causes the adjusted valve opening to exceed the allowable opening range of the valve, the adjusted valve opening is truncated to the boundary value of the allowable opening range. The allowable opening range is determined according to the mechanical stroke of the valve and the safety operation requirements of the firing system. The above opening increase and opening decrease commands are applicable to fresh air valves, return valves, and bypass valves.

[0120] The frequency increase command refers to the controller outputting a control signal to the frequency converter of the blower to increase the output frequency of the frequency converter by an increment based on the current value. The frequency decrease command refers to the controller outputting a control signal to the frequency converter of the blower to decrease the output frequency of the frequency converter by a decrement based on the current value. The specific value of the increment or decrement is equal to the preset frequency adjustment gain multiplied by the absolute value of the standardized directional deviation of the atmosphere parameter that triggered the command. The product is the frequency change amount of this adjustment. The preset frequency adjustment gain is obtained by statistically analyzing the blower frequency adjustment range required to restore the oxygen partial pressure in the kiln to the process target value under different standardized directional deviation levels based on historical operating data of the same firing equipment. The adjustment range is obtained by linear regression fitting for the fan frequency adjustment range that should be adjusted for each unit change in the standardized directional deviation. When the calculated frequency change causes the adjusted frequency converter output frequency to exceed the allowable operating frequency range of the blower, the adjusted frequency is truncated to the boundary value of the allowable operating frequency range. The allowable operating frequency range is determined based on the rated parameters of the blower motor and the mechanical safe speed.

[0121] The power increase command refers to the controller outputting a control signal to the burner power adjustment unit or auxiliary heating power adjustment unit to increase the output power by an increment based on the current value. The power decrease command refers to the controller outputting a control signal to the burner power adjustment unit or auxiliary heating power adjustment unit to decrease the output power by a reduction based on the current value. The specific value of the increment or reduction is equal to the preset power adjustment gain multiplied by the absolute value of the standardized directional deviation of the atmosphere parameter that triggered the command. The product is the power change amount of this adjustment. The preset power adjustment gain is obtained by statistically analyzing the power adjustment range required to restore the corresponding atmosphere parameter or kiln temperature to the process target value under different standardized directional deviation levels based on historical operating data of the same firing equipment. The power adjustment range should be obtained by linear regression fitting for each unit change in the standardized directional deviation. When the calculated power change amount causes the adjusted output power to exceed the allowable power range of the power adjustment unit, the adjusted power is truncated to the boundary value of the allowable power range. The allowable power range is determined based on the rated power and minimum stable combustion power of the burner or auxiliary heating device.

[0122] The heat preservation delay command refers to the controller outputting a control signal to the kiln propulsion control unit to increase the heat preservation time of the high-temperature heat preservation stage by an increment based on the current heat preservation correction time. The specific value of the increment is equal to the preset heat preservation delay gain multiplied by the absolute value of the slope of the linear fitting of the firing boundary drift characterization value within the heat preservation stage. The product is the amount of heat preservation time extension adjusted in this instance. The preset heat preservation delay gain is obtained by statistically analyzing the relationship between the rate of increase of the firing boundary drift characterization value within the heat preservation stage and the amount of heat preservation time extension required to maintain the product's mineral phase transformation and densification, based on historical operating data of the same firing equipment. The magnitude of the heat preservation time extension should be obtained by linear regression fitting for each unit change in the fitting slope. When the calculated heat preservation time extension causes the extended heat preservation time to exceed the preset upper limit of the heat preservation correction time, the extended heat preservation time is truncated to the upper limit.

[0123] All of the above preset adjustment gains are positive numbers. Their physical meaning is the adjustment range that the actuator should make when the corresponding standardized directional deviation deviates by one standard deviation unit. Each adjustment gain is written into the parameter storage area of ​​the industrial controller during system initialization and can be calibrated and corrected in subsequent operation based on the actual control effect.

[0124] The differentiated control in the above three stages all take the atmosphere offset vector and atmosphere offset amplitude updated in real time in step S2 and the firing boundary drift characterization value updated in real time in step S3 as inputs. The execution framework is based on the three control outputs determined in this step. The main contradictions of each stage are subject to key constraints, and the adjustment direction of the actuator is determined according to the direction information of each component in the atmosphere offset vector, so as to realize the direction-sensitive closed-loop correction control.

[0125] Step S5 involves collecting quality feedback information and constructing a comprehensive quality deviation after the current batch of products exits the kiln, and updating the calibration coefficients based on the comprehensive quality deviation. The technical effect of this step is that by introducing quality feedback information from the products exiting the kiln, the calibration coefficients in the calculation model of the firing boundary drift characterization value are corrected batch by batch. This allows the prediction accuracy of the firing boundary drift characterization value for the actual firing quality deviation to continuously improve with the accumulation of batches, thereby making the determination of the control mode and the calculation of the control output increasingly closer to the actual process requirements. The reason for using quality feedback correction instead of relying solely on the initial calibration coefficients is that the initial calibration coefficients are obtained based on historical data through multiple linear regression. Their applicability is limited to the range of raw material characteristics and atmosphere conditions covered by the historical data. When encountering new working conditions not covered by historical data in actual production, the initial calibration coefficients may not accurately predict the firing quality deviation. However, through quality feedback correction, the calibration coefficients can gradually adapt to the new working conditions, improving the system's adaptability. The preset number of stable judgment batches is determined according to the production cycle and statistical confidence requirements of the firing equipment, requiring that the number of consecutive batches selected is sufficient to reflect the stability trend of the process parameters.

[0126] Quality inspection is conducted on the products exiting the kiln, and quality feedback information is collected. This feedback information uses commonly used quality evaluation indicators in the refractory materials field, including linear shrinkage after firing, water absorption, bulk density, and refractoriness. These indicators can be measured according to relevant national standards, industry standards, or enterprise internal control testing procedures. The measured values ​​of each quality evaluation indicator are compared with their corresponding target values ​​to construct the comprehensive quality deviation for the current batch. The comprehensive quality deviation is a comprehensive measure of the deviations of each quality evaluation indicator. It is constructed by calculating the absolute value of the difference between the measured value and the target value of each quality evaluation indicator, dividing it by its respective target value for normalization, and then weighting and summing the results using preset weights for each quality evaluation indicator to obtain a value reflecting the current batch's overall quality deviation. The overall quality deviation value is the degree of deviation of the overall quality of the previous batch of products from the target. The preset weights of each quality evaluation index are determined according to the importance of each quality evaluation index to the product's performance. They are obtained by process technicians assigning values ​​based on the priority of each quality evaluation index in the product standard and the engineering experience of the degree of influence of each index deviation on the final performance of the product. The sum of each preset weight is one. For quality evaluation indices with unilateral qualification requirements, the deviation amplitude is calculated only when the measured value falls in the non-qualified direction. For quality evaluation indices with bilateral target ranges, the deviation of the index is zero when the measured value is within the target range, and the deviation amplitude relative to the nearest boundary is taken when the measured value exceeds the target range.

[0127] Since the overall quality deviation is a dimensionless deviation value constructed based on the quality evaluation index of the kiln-exit products, while the firing boundary drift characterization value is a dimensionless risk index constructed based on the raw material sensitivity index and atmosphere offset amplitude, although both are dimensionless quantities, their numerical ranges and construction bases are different, and they cannot be directly compared arithmetically. To make them comparable, the overall quality deviation is first subjected to scale mapping processing before calculating the prediction error. The reason for introducing scale mapping processing is that the calibration coefficient update method driven by prediction error requires the predicted value and the true value to be on the same numerical scale. Otherwise, the magnitude of the prediction error will not be able to truly reflect the degree of deviation in prediction accuracy, which may lead to the update step size of the calibration coefficient being too large or too small, affecting the convergence speed and stability. The scale mapping processing method is to multiply the overall quality deviation of the current batch by a preset scale mapping coefficient to obtain the mapped overall quality deviation value, so that its numerical range is on the same order of magnitude as the numerical range of the firing boundary drift characterization value.

[0128] The preset scale mapping coefficient is obtained by collecting the comprehensive quality deviation values ​​accumulated by the same firing equipment in historical batches and calculating the mean. The initial value of the scale mapping coefficient is obtained by dividing the mean of the representative values ​​of the firing boundary drift characterization values ​​of multiple historical batches by the mean of the comprehensive quality deviation values. When the mean of the comprehensive quality deviation values ​​of historical batches is zero or lower than the preset lower limit of the mean quality deviation, the lower limit of the mean quality deviation is used as the denominator. The lower limit of the mean quality deviation is determined according to the quality inspection resolution, the allowable deviation of the product standard, and the enterprise's internal control inspection procedures. The representative value of the firing boundary drift characterization value is the time-weighted average of the firing boundary drift characterization values ​​during the firing process of historical batches. Specifically, it is calculated by taking the average of the representative values ​​of the firing boundary drift characterization values ​​during the firing process of a certain batch. The calculated firing boundary drift characterization value is combined using a time-weighted method to obtain a single value, which is used to summarize the overall level of the firing boundary drift characterization value of the entire batch during the firing process as a scalar value. The time-weighted method is as follows: the weight assigned to the firing boundary drift characterization value at each sampling time during the high-temperature holding stage is greater than the weight assigned to the firing boundary drift characterization value at each sampling time during the heating and phase change transition stages; the weight assigned to the firing boundary drift characterization value at each sampling time during the heating and phase change transition stages is greater than the weight assigned to the firing boundary drift characterization value at each sampling time during the preheating stage; the weight of each stage is determined according to the degree of influence of that stage on the final product quality, so as to reflect the difference in the degree of influence of different firing stages on the final product quality.

[0129] The reason for adopting the above-mentioned stage-differentiated time-weighted method is that the high-temperature holding stage is the main completion stage of the mullite reaction and densification process. The firing boundary drift characterization value of this stage has the most direct and significant impact on the final product quality, so it should be given the highest weight. The heating and phase transformation transition stage is the initiation and transition stage of mineral phase transformation. The atmosphere shift in this stage will affect the selection of mineral phase transformation path and has a moderate impact on the final product quality. The preheating stage mainly completes the drying and preheating of the billet. The atmosphere shift in this stage has a relatively small impact on the final product quality. If each stage is given the same weight, the firing boundary drift characterization value of the preheating stage will occupy a proportion in the representative value that is disproportionate to its actual impact. This may lead to the representative value not being able to accurately reflect the risk level of the stage that has the most critical impact on product quality during the firing process. The representative value of the firing boundary drift characterization value of this batch is obtained by multiplying the firing boundary drift characterization value of each sampling time by the corresponding normalized time weight and summing them.

[0130] The scale mapping coefficient is synchronously corrected in subsequent batches as the calibration coefficient is updated. The correction criterion is the mean ratio between the mapped comprehensive quality deviation value and the representative value of the firing boundary drift characterization value. The mean ratio is calculated as follows: after each batch updates the calibration coefficient, the arithmetic mean of the mapped comprehensive quality deviation values ​​of the most recently preset mean ratio statistical batches is taken as the numerator, and the arithmetic mean of the representative values ​​of the firing boundary drift characterization values ​​of the most recently preset mean ratio statistical batches is taken as the denominator. The two are then divided to obtain the mean ratio. The number of preset mean ratio statistical batches is determined according to the statistical smoothing requirements and production cycle. When the mean ratio is equal to one, it indicates that the mapped comprehensive quality deviation and the representative value of the firing boundary drift characterization value are completely matched on the numerical scale, and no correction is required. When the mean ratio deviates from one, it indicates that there is a deviation in the numerical scale between the two, and the scale mapping coefficient needs to be corrected.

[0131] The preset mapping deviation tolerance is a positive number used to define the acceptable range of mean ratio deviation from one. It is obtained by calculating the standard deviation of the sequence of representative values ​​of the mapped comprehensive quality deviation value and the firing boundary drift characterization value accumulated in historical batches of the same firing equipment under normal operating conditions, and using a preset multiple of the standard deviation as the value of the mapping deviation tolerance. The preset multiple is determined according to the balance between the scale mapping accuracy requirements and the correction frequency.

[0132] When the absolute value of the difference between the mean ratio and one does not exceed the preset mapping deviation tolerance, the controller maintains the current scale mapping coefficient unchanged; when the absolute value of the difference between the mean ratio and one exceeds the preset mapping deviation tolerance, the controller performs a small-step correction on the scale mapping coefficient. The correction direction and method are as follows: when the mean ratio is greater than one, it indicates that the mean of the overall quality deviation after mapping is greater than the mean of the representative value of the firing boundary drift characterization value, that is, the current scale mapping coefficient is too large, resulting in a high value of the overall quality deviation after mapping. At this time, the controller subtracts the preset mapping coefficient correction step size from the current scale mapping coefficient to obtain the corrected scale mapping coefficient, so that the value of the overall quality deviation after mapping decreases, thereby causing the mean ratio to fall back towards one; when the mean ratio is less than one, it indicates that the overall quality deviation after mapping is... If the mean deviation is less than the mean of the representative value of the firing boundary drift characterization value, meaning the current scale mapping coefficient is too small, resulting in a low overall quality deviation value after mapping, the controller adds a preset mapping coefficient correction step size to the current scale mapping coefficient to obtain the corrected scale mapping coefficient, thereby increasing the overall quality deviation value after mapping and causing the mean ratio to rise back towards one. The preset mapping coefficient correction step size is a positive number, obtained by multiplying the current value of the scale mapping coefficient by a preset mapping coefficient correction ratio. The preset mapping coefficient correction ratio is determined based on the convergence speed and stability requirements of the scale mapping coefficient. The corrected scale mapping coefficient is written into the parameter storage area of ​​the industrial controller for use in the next batch of scale mapping processing.

[0133] Based on the mapped comprehensive quality deviation, the preset first calibration coefficient, preset second calibration coefficient, and preset third calibration coefficient used in step S3 to calculate the firing boundary drift characterization value are updated in small steps. The reason for using small-step updates instead of a one-time refit of the calibration coefficients is that a one-time refit requires accumulating enough new batch data to ensure the statistical reliability of the regression results. During the data accumulation period, the calibration coefficients cannot be updated. The small-step update method can use the quality feedback information of each batch after it leaves the kiln to incrementally correct the calibration coefficients, so that the calibration coefficients can continuously track the slow drift of the process conditions and have better real-time adaptability. The small-step update is based on a prediction error-driven correction method. Its core idea is to update the firing boundary drift table... The eigenvalue is considered as the predicted value of the overall quality deviation during the firing process. The overall quality deviation after scale mapping after exiting the kiln is regarded as the true value. The difference between the prediction error (i.e., the mapped overall quality deviation) and the representative value of the firing boundary drift characterization value is used as the signal driving the calibration coefficient update. The reason for adopting the prediction error-driven correction method is that this method is mathematically equivalent to the stochastic gradient descent optimization process with mean square error as the loss function. This helps to improve the stability of the calibration coefficient update process. Under the condition of reasonable learning rate selection, the calibration coefficients can converge to a stable value that minimizes the prediction error batch by batch. At the same time, this method only requires a set of data of the current batch for each update, without the need to store and process a large amount of historical data. The computational and storage overhead is minimal, making it suitable for implementation in industrial controllers.

[0134] The specific update method is as follows: First, calculate the prediction error, which is the comprehensive quality deviation after mapping the current batch minus the representative value of the firing boundary drift characterization value calculated during the firing process of the current batch; then, add a calibration correction amount to a calibration coefficient value used in the current batch to obtain the updated calibration coefficient value for the next batch; where the calibration correction amount is the product of the preset learning rate, the prediction error, and the input amount corresponding to the calibration coefficient. The preset learning rate is a positive number less than one, and it is obtained based on the process adjustment inertia of the firing system and the quality fluctuation amplitude between batches; the selected learning rate is... The reason for a positive number less than one is that an excessively large learning rate will cause the calibration coefficients to update too much between adjacent batches, resulting in the calibration coefficients oscillating around the optimal value and failing to converge, or even causing the calibration coefficients to diverge; an excessively small learning rate will result in a slow convergence speed of the calibration coefficients, requiring too many batches to adapt to the new operating conditions; the specific value of the learning rate needs to strike a balance between convergence speed and stability, and is usually determined based on the process adjustment inertia of the firing system; the input quantity corresponding to the preset first calibration coefficient is the raw material sensitivity index of this batch calculated in step S1, and the input quantity corresponding to the preset second calibration coefficient is... The input quantity is the representative value of the atmosphere shift amplitude during the firing process of this batch, calculated in step S2. The input quantity corresponding to the preset third calibration coefficient is the product of the raw material sensitivity index and the representative value of the atmosphere shift amplitude. The representative value of the atmosphere shift amplitude is the time-weighted average of the atmosphere shift amplitude at each sampling time during the firing process of this batch. Its time weighting method is the same as the time weighting method of the representative value of the firing boundary drift characterization value, that is, the weight assigned to the sampling value in the high-temperature holding stage is greater than the weight assigned to the sampling value in the heating and phase change transition stage, and the weight assigned to the sampling value in the heating and phase change transition stage is greater than the weight assigned to the sampling value in the heating and phase change transition stage. The weights assigned to the sampled values ​​during the preheating stage are determined based on the degree of impact of each stage on the final product quality, reflecting the differences in the degree of impact of different firing stages on the final product quality. All inputs involved in updating the calibration coefficients are first subjected to amplitude limiting processing. The raw material sensitivity index, the representative value of the atmosphere offset amplitude, and their product are each limited to the preset allowable input range in the process boundary table. After the update is completed, the controller performs projection processing on the first, second, and third calibration coefficients to keep them within their respective allowable value ranges. Values ​​exceeding the allowable range are not included in the calculation of the next batch.

[0135] The principle behind the above update method is as follows: when the prediction error is positive, meaning the mapped overall quality deviation is greater than the representative value of the firing boundary drift characterization value, it indicates that the current calibration coefficient underestimates the firing risk. The calibration coefficient needs to be increased to make the firing boundary drift characterization value calculated in the next batch larger, thus triggering a more conservative control mode. When the prediction error is negative, it indicates that the current calibration coefficient overestimates the firing risk. The calibration coefficient needs to be decreased to avoid excessive conservatism leading to an unnecessary reduction in waste heat utilization. Mathematically, this update method is equivalent to using the representative value of the firing boundary drift characterization value... The mean square error between the value and the overall quality deviation after mapping is the loss function, and the stochastic gradient descent process with each calibration coefficient as the parameter to be optimized ensures that the calibration coefficients converge batch by batch to a stable value that minimizes the prediction error, provided that the learning rate is reasonably selected. To further ensure the stability of the update process, a range constraint is imposed on each updated calibration coefficient. That is, if an updated calibration coefficient exceeds its preset allowable range, it is truncated to the boundary value of that range. The allowable range is determined based on the physical meaning of the calibration coefficient and the statistics of its historical values.

[0136] Through small-step corrections batch by batch, the calibration coefficients gradually converge to the optimal values ​​that match the current firing equipment, raw material sources, and process conditions. The updated first, second, and third calibration coefficients are written into the parameter storage area of ​​the industrial controller for use in step S3 during the next batch of firing.

[0137] After updating the calibration coefficients, the controller performs stability assessment on the quality feedback information of several consecutive preset stability assessment batches. The stability conditions require all of the following conditions to be met simultaneously: the linear shrinkage rate fluctuation of the several consecutive preset stability assessment batches does not exceed the preset linear shrinkage rate fluctuation threshold; the water absorption rate and bulk density of the several consecutive preset stability assessment batches are both stable within their respective preset target ranges; and the refractoriness and high-temperature performance indicators specified in the process boundary table of the several consecutive preset stability assessment batches all meet the requirements of the corresponding product standards or enterprise internal control standards. The types of high-temperature performance indicators are defined in the product grade settings. The parameters are immediately written into the process boundary table; parameters not written into the process boundary table are not used as equivalent substitute parameters. The unit product energy consumption of several consecutive preset stable batches is lower than the preset unit product energy consumption upper limit under the benchmark firing regime. The preset linear shrinkage rate fluctuation threshold is determined based on the consistency requirements for linear shrinkage rate in the product standard, specifically the allowable deviation range of linear shrinkage rate in the product standard. The preset target range is determined based on the qualified judgment intervals for water absorption rate and bulk density in the product standard. The preset unit product energy consumption upper limit is determined based on the standard energy consumption level under the benchmark firing regime, used to verify that the optimized process scheme is not inferior to the benchmark scheme in terms of energy saving.

[0138] The steps S1 to S5 described above constitute a complete waste heat firing control process for low-grade hard clay. Step S1 is executed once before each batch is loaded into the kiln; steps S2 to S4 are executed in real-time or periodically during firing; and step S5 is executed once after each batch exits the kiln. The entire process, through the organic integration of raw material sensitivity pre-coding, waste heat atmosphere shift identification, firing boundary drift characterization value calculation and mode determination, linkage control output and phased differentiated control, and quality feedback correction, achieves the technical goal of stably controlling refractoriness, densification degree, and linear shrinkage rate while ensuring waste heat utilization, thus reducing the risks of under-firing, over-firing, and performance dispersion. The comprehensive technical effects of the above method are reflected in the following aspects: Regarding waste heat utilization, through the dynamic switching of the three-level control mode, high recovery is prioritized under conditions of low raw material sensitivity and small atmosphere shift. The model maximizes waste heat utilization, gradually reducing the waste heat recovery ratio only when the risk of firing boundary drift increases. This avoids the waste heat caused by the conservative waste heat recovery strategy adopted in all traditional methods to ensure firing quality. In terms of firing quality stability, through the coupled calculation of raw material sensitivity index and atmosphere deviation amplitude and staged differentiated control, the system can adaptively adjust process parameters for different raw material batches and different atmosphere conditions, thereby reducing the batch-to-batch fluctuation of key quality indicators such as refractoriness, bulk density, water absorption and linear shrinkage. In terms of process adaptability, through the batch-by-batch update mechanism of calibration coefficient driven by prediction error in step S5, the system can continuously track the impact of slow drift factors such as changes in raw material source, equipment aging and seasonal climate change on the firing process, maintaining control accuracy without frequent manual intervention.

[0139] Example 2:

[0140] See Figure 3 As shown, a waste heat firing system for low-grade hard clay is provided for performing the aforementioned waste heat firing method for low-grade hard clay. The system includes:

[0141] Before loading the kiln, the raw material sensitive module 101 collects and normalizes the detection information of the current batch of raw materials, and then calculates the weighted sum of the normalized values ​​to obtain the raw material sensitive index.

[0142] The atmosphere offset module 102 collects the atmosphere parameters of the recycled gas in the waste heat recovery loop during the firing process, calculates the standardized directional deviation of each atmosphere parameter, and calculates the atmosphere offset amplitude based on each standardized directional deviation.

[0143] The boundary drift module 103 calculates the firing boundary drift characterization value by the raw material sensitivity index, atmosphere offset amplitude and preset calibration coefficient, and determines the current control mode based on the comparison result of the firing boundary drift characterization value and the preset threshold.

[0144] The linkage control module 104 generates a control output based on the firing boundary drift characterization value, and performs linkage control during the firing process based on the control output, control mode and each standardized directional deviation.

[0145] The feedback update module 105 collects quality feedback information and constructs a comprehensive quality deviation after the current batch of products leaves the kiln, and updates the calibration coefficient based on the comprehensive quality deviation.

[0146] The embodiments of this application have been described above, but these embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments based on the guidance of these embodiments, and all of them are within the protection scope of these embodiments.

Claims

1. A method for firing low-grade hard clay using waste heat, applied in firing equipment with a waste heat recovery loop, characterized in that, Includes the following steps: Before loading the kiln, the detection information of the current batch of raw materials is collected and normalized, and the normalized values ​​are weighted and summed to obtain the raw material sensitivity index. During the firing process, the atmospheric parameters of the recycled gas in the waste heat recovery loop are collected, the standardized directional deviation of each atmospheric parameter is calculated, and the atmospheric offset amplitude is calculated based on each standardized directional deviation. The firing boundary drift characterization value is calculated by the raw material sensitivity index, atmosphere offset amplitude and preset calibration coefficient. The current control mode is determined based on the comparison result of the firing boundary drift characterization value and the preset threshold. A control output is generated based on the firing boundary drift characterization value. Based on the control output, control mode and each standardized directional deviation, linkage control is executed during the firing process. After the current batch of products leaves the kiln, quality feedback information is collected and a comprehensive quality deviation is constructed. The calibration coefficient is then updated based on the comprehensive quality deviation.

2. The method for waste heat firing of low-grade hard clay according to claim 1, characterized in that, Methods for calculating the atmospheric offset amplitude include: Atmospheric parameters include recycled gas temperature, oxygen partial pressure, water vapor volume fraction, carbon monoxide concentration, and carbon dioxide concentration; After applying sliding window mean filtering to the atmosphere parameters, the standardized directed biases of the recycled gas temperature, oxygen partial pressure, and water vapor volume fraction relative to the process target values ​​are calculated; the standardized directed biases of the logarithmic ratios are calculated based on the carbon monoxide concentration and carbon dioxide concentration. The atmosphere offset amplitude is the weighted sum of the absolute values ​​of the directed deviations of the reuse gas temperature standardization, oxygen partial pressure standardization, water vapor volume fraction standardization, and logarithmic ratio standardization, and the corresponding preset atmosphere weight coefficients.

3. The method for waste heat firing of low-grade hard clay according to claim 2, characterized in that, Methods for calculating the standardized directed bias of the log-ratio include: The measured logarithmic ratio was calculated based on the measured values ​​of carbon monoxide and carbon dioxide concentrations. Calculate the target logarithmic ratio based on the process target values ​​of carbon monoxide concentration and carbon dioxide concentration; The logarithmic ratio is calculated by adding the carbon monoxide concentration to a preset positive bias concentration as the numerator, adding the carbon dioxide concentration to a preset positive bias concentration as the denominator, and taking the natural logarithm of the fraction. Calculate the difference between the measured logarithmic ratio and the target logarithmic ratio, and then divide it by the standard deviation of the logarithmic ratio under normal operating conditions to obtain the standardized directional bias of the logarithmic ratio.

4. The method for waste heat firing of low-grade hard clay according to claim 1, characterized in that, Methods for determining the current control mode include: The first term is obtained by multiplying the raw material sensitivity index by a preset first calibration coefficient, the second term is obtained by multiplying the atmosphere offset amplitude by a preset second calibration coefficient, the third term is obtained by multiplying the raw material sensitivity index by the atmosphere offset amplitude and then by a preset third calibration coefficient, and the first, second and third terms are added together to obtain the firing boundary drift characterization value. When the firing boundary drift characterization value is less than or equal to the preset first threshold, the control mode is determined to be high recovery mode; When the firing boundary drift characterization value is greater than the preset first threshold and less than or equal to the preset second threshold, the control mode is determined to be the balance mode. When the value of the firing boundary drift characterization is greater than the preset second threshold, the control mode is determined to be steady-state protection mode.

5. The method for waste heat firing of low-grade hard clay according to claim 4, characterized in that, Methods for generating control output quantities include: Control output includes waste heat recovery ratio, insulation correction time, and air supply path selection; The waste heat recovery ratio is obtained by subtracting the ratio term from the preset baseline waste heat recovery ratio and then truncating it. The ratio term is the preset recovery ratio adjustment coefficient multiplied by the firing boundary drift characterization value. The heat preservation correction time is obtained by multiplying the preset heat preservation time adjustment coefficient by the firing boundary drift characterization value, adding the preset reference heat preservation time, and then subjecting it to truncation and rate of change limitation. The air supply path is selected as one of the following: preheating priority path, combustion priority path, or bypass gas stabilization path.

6. The method for waste heat firing of low-grade hard clay according to claim 5, characterized in that, Methods for implementing linkage control include: Based on the decision-based control mode, the control output quantity is subjected to linkage control, wherein: In high recovery mode, the cutoff range of the waste heat recovery ratio is between the preset upper limit of the waste heat recovery ratio and the preset benchmark waste heat recovery ratio. The air supply path is selected as either the preheating priority path or the combustion priority path, and the heat preservation correction time is set to the preset benchmark heat preservation time. In balance mode, the cutoff range of the waste heat recovery ratio is between the preset benchmark waste heat recovery ratio and the preset lower limit of the waste heat recovery ratio, and the air supply path is selected as the combustion priority path. In steady-state protection mode, the waste heat recovery ratio is set to the preset lower limit of the waste heat recovery ratio, the air supply path is switched to the bypass stable gas path, and the insulation correction time is set to the preset upper limit of the insulation correction time.

7. The method for waste heat firing of low-grade hard clay according to claim 6, characterized in that, Based on the determined control output, linkage control is executed at different stages of the firing process, specifically including: The firing process includes a preheating stage, a heating and phase change transition stage, and a high-temperature holding stage; During the preheating stage, the standardized directional deviations of water vapor volume fraction and recycled gas temperature are used as constraints, and the fresh air valve, return valve, or auxiliary heating power is adjusted according to the deviation direction of the standardized directional deviations. During the heating and phase change transition phase, the oxygen partial pressure standardization directional deviation and the logarithmic ratio of carbon monoxide concentration to carbon dioxide concentration standardization directional deviation are used as constraints. The blower, waste heat recovery ratio or burner power are adjusted according to the deviation direction of the corresponding standardization directional deviation. During the high-temperature heat preservation stage, the proportion of waste heat recovery and the heat preservation correction time are finely adjusted based on the firing boundary drift characterization value.

8. The method for waste heat firing of low-grade hard clay according to claim 7, characterized in that, Methods for fine-tuning the waste heat recovery ratio and insulation correction time include: Linear fitting is performed on the firing boundary drift characterization value of several control cycles within a continuously preset trend fitting period during the heat preservation stage. When the slope obtained by fitting is greater than the preset trend slope threshold, the firing boundary drift characterization value is determined to be in an upward trend; when the slope obtained by fitting is not greater than the preset trend slope threshold, the firing boundary drift characterization value is determined to be in a stable or downward trend. When an upward trend is detected, a command to reduce the opening of the reflux valve is sent to reduce the proportion of waste heat recovery. Within the preset upper limit of the allowable heat preservation correction time, a heat preservation delay command is sent to the firing equipment to extend the heat preservation correction time. When the trend is determined to be stable or declining, the current waste heat recovery ratio and insulation correction time should be maintained unchanged.

9. A method for firing low-grade hard clay using residual heat according to claim 1 or 4, characterized in that, The method for updating the calibration coefficients includes: Multiply the overall quality deviation by the preset scale mapping coefficient to obtain the mapped overall quality deviation; The prediction error is obtained by subtracting the representative value of the firing boundary drift characterization value of the current batch from the overall quality deviation after mapping of the current batch. The representative value of the firing boundary drift characterization value is the time-weighted average value of the firing boundary drift characterization value of the current batch. The calibration correction is obtained by multiplying the preset learning rate and prediction error by the corresponding input. The current calibration coefficient is added to the calibration correction to obtain the updated calibration coefficient, and a range constraint is applied to the updated calibration coefficient. The first, second, and third calibration coefficients correspond to the raw material sensitivity index, the atmosphere offset amplitude, and the product of the raw material sensitivity index and the atmosphere offset amplitude, respectively.

10. A waste heat firing system for low-grade hard clay, characterized in that, A system for performing a waste heat firing method for low-grade hard clay as described in any one of claims 1-9; the system comprises: The raw material sensitivity module collects and normalizes the detection information of the current batch of raw materials before loading the kiln, and then calculates the weighted sum of the normalized values ​​to obtain the raw material sensitivity index. The atmosphere offset module collects the atmosphere parameters of the recycled gas in the waste heat recovery loop during the firing process, calculates the standardized directional deviation of each atmosphere parameter, and calculates the atmosphere offset amplitude based on each standardized directional deviation. The boundary drift module calculates the firing boundary drift characterization value by combining the raw material sensitivity index, atmosphere offset amplitude, and preset calibration coefficients, and determines the current control mode based on the comparison result between the firing boundary drift characterization value and the preset threshold. The linkage control module generates a control output based on the firing boundary drift characterization value, and performs linkage control during the firing process based on the control output, control mode and each standardized directional deviation. The feedback update module collects quality feedback information and constructs a comprehensive quality deviation after the current batch of products leaves the kiln, and updates the calibration coefficients based on the comprehensive quality deviation.

Images