A method for on-line detection of ingredients of continuously conveyed materials and optimization of batching

Abstract

The application provides a kind of component on-line detection and batching optimization method for continuous conveying material, it is related to material conveying technical field, the method comprises: forming raw material identity fingerprint library;Obtain material component variation sequence, construct monitoring index variation sequence based on material component variation sequence;Based on the trend inertia parameter obtained by monitoring index variation sequence, calculate effective deviation according to trend inertia parameter, and judge the current state of transportation system based on effective deviation;When the transportation system is in the trend deviation state, the finished product mixed fingerprint vector of the current material is constructed according to the material component variation sequence;The finished product mixed fingerprint vector is matched with the raw material identity fingerprint library to calculate the similarity, identify the deviation source of the current material;According to the deviation source, generate adjustment instruction. The technical problems of existing technology, such as lag response, unknown deviation source, single compensation strategy, difficulty in adapting to complex working conditions such as frequent fluctuations in raw materials and changes in equipment state are solved.

Description

Technical Field

[0001] This application relates to the field of material conveying technology, and in particular to a method for online component detection and batching optimization of continuously conveyed materials. Background Technology

[0002] With the continuous improvement of industrial automation, industries such as cement, metallurgy, and chemicals have placed higher demands on the control precision and stability of continuous batching processes. Taking cement production as an example, the raw meal preparation section, as a pre-process for clinker calcination, directly determines the mineral composition of the clinker, the thermal stability of the kiln system, and the final product quality through its batching quality. Currently, large-scale cement production lines generally use online component analyzers to monitor the raw meal on the conveyor belt in real time and feed the results back to the batching control system. By adjusting the feed rate of each raw material feeder, closed-loop control of the raw meal composition is achieved. This model has gradually replaced traditional manual sampling and laboratory testing methods, becoming the mainstream technical solution in the industry.

[0003] However, existing continuous batching control technologies still have the following limitations. First, existing systems typically adjust based on the instantaneous deviation between the current detected value and the target value. However, continuous conveying processes have significant time lags, and the current detected value reflects the batching status at a historical moment. Direct adjustment based on instantaneous deviation easily leads to control oscillations of "overshoot-correction-re-overshoot," causing the batching system to be in a state of fluctuation for a long time. Second, existing systems cannot effectively identify the source of deviation. When abnormalities occur in the composition, the system can only blindly adjust the raw material ratio according to the direction of the deviation, unable to distinguish whether the abnormality is caused by changes in the quality of the raw materials themselves, fluctuations in the flow rate of the feeding equipment, or interference from the online detection device. This leads to frequent misadjustments, which not only affect the batching accuracy but may also cause frequent equipment operation and increased energy consumption. Third, existing control strategies are mostly single-mode, lacking differentiated compensation mechanisms for different sources of deviation, making it difficult to simultaneously ensure the coordinated stability of multiple quality indicators.

[0004] In summary, existing continuous batching control technologies suffer from technical problems such as delayed response, unclear sources of deviation, and simplistic compensation strategies, making them ill-suited to complex operating conditions including frequent fluctuations in raw materials and changes in equipment status. Therefore, a solution to these problems is urgently needed. Summary of the Invention

[0005] This disclosure provides an online component detection and batching optimization method for continuously conveyed materials, which solves the technical problems in the prior art, such as delayed response, unclear sources of deviation, single compensation strategy, and difficulty in adapting to complex working conditions such as frequent fluctuations in raw materials and changes in equipment status.

[0006] According to a first aspect of this disclosure, a method for online component detection and batching optimization of continuously conveyed materials is provided, comprising: Each batch of raw materials entering the factory is tested and analyzed to obtain raw material baseline component data. Based on the raw material baseline component data, a raw material identity fingerprint vector is constructed. All raw material baseline component data and raw material identity fingerprint vectors are integrated to form a raw material identity fingerprint database. An online component analysis device is installed on the main conveyor belt to periodically monitor the chemical composition of the continuously conveyed material, obtain the material composition change sequence, and construct a monitoring index change sequence based on the material composition change sequence; Trend inertia parameters are obtained based on the change sequence of monitoring indicators, effective deviation is calculated based on the trend inertia parameters, and the current state of the transportation system is determined based on the effective deviation. The trend inertia parameters include trend inertia intensity and trend acceleration. When the transportation system is in a state of trend deviation, construct the finished product mixture fingerprint vector of the current material based on the material composition change sequence; The similarity matching calculation between the finished product's mixed fingerprint vector and the raw material identity fingerprint database is performed to identify the source of deviation for the current material; Adjustment instructions are generated based on the source of deviation. These adjustment instructions include proportion adjustment instructions, equipment control instructions, data substitution and conservative mode instructions.

[0007] One or more technical solutions provided in this disclosure have at least the following technical effects or advantages: Each batch of incoming raw materials is analyzed to obtain raw material baseline component data; a raw material identity fingerprint vector is constructed based on the raw material baseline component data; all raw material baseline component data and raw material identity fingerprint vectors are integrated to form a raw material identity fingerprint database; an online component analysis device is deployed on the main conveyor belt to periodically monitor the chemical composition of continuously conveyed materials, obtain material component change sequences, and construct monitoring index change sequences based on the material component change sequences; trend inertia parameters are obtained based on the monitoring index change sequences; effective deviations are calculated based on the trend inertia parameters; and the current state of the transportation system is determined based on the effective deviations, wherein the trend inertia parameters include trend inertia strength and trend acceleration; when the transportation system is in a trend deviation state, a finished product mixed fingerprint vector of the current material is constructed based on the material component change sequence; the finished product mixed fingerprint vector is matched with the raw material identity fingerprint database to identify the source of deviation of the current material; and adjustment instructions are generated based on the source of deviation, wherein the adjustment instructions include proportion adjustment instructions, equipment control instructions, data substitution and conservative mode instructions. This invention solves the technical problems of existing technologies, such as delayed response, unclear sources of deviation, and simplistic compensation strategies, making it difficult to adapt to complex operating conditions such as frequent fluctuations in raw materials and changes in equipment status. It achieves significant improvements in the stability, robustness, and intelligence level of the batching system.

[0008] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating an online component detection and batching optimization method for continuously conveyed materials, as provided in an embodiment of this application. Detailed Implementation

[0011] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0012] Example 1: This disclosure provides a method for online component detection and batching optimization of continuously conveyed materials, which is referred to below. Figure 1 The methods include: S1: Conduct laboratory analysis on each batch of raw materials entering the factory to obtain raw material baseline component data. Construct raw material identity fingerprint vectors based on raw material baseline component data. Integrate all raw material baseline component data and raw material identity fingerprint vectors to form a raw material identity fingerprint database. Furthermore, step S1 also includes: Each batch of incoming raw materials is tested and analyzed using X-ray fluorescence analyzer or wet chemical analysis method to determine the content of major oxides and obtain the raw material reference composition data for each raw material. Four sets of element ratio characteristics were constructed based on the raw material baseline composition data. The specific calculation method is as follows: ; in, This represents the characteristic elemental ratio between the calcium oxide content and the silicon dioxide content in the j-th raw material. This represents the characteristic elemental ratio between the calcium oxide content and the aluminum oxide content in the j-th raw material. This represents the characteristic elemental ratio between the content of silicon dioxide and the content of aluminum oxide of the j-th type. This represents the characteristic elemental ratio between the content of the j-th type of iron oxide and the content of aluminum oxide. Statistical processing is performed on the four element ratio characteristics of multiple samples of the same raw material to obtain the raw material identification fingerprint vector, the specific expression of which is as follows: ; in, This represents the raw material identification fingerprint vector for the j-th raw material, where j is the raw material number. The statistical average of the ratio characteristics of the four elements corresponding to the j-th raw material; Integrate all raw material identity fingerprint vectors with the corresponding raw material baseline component data to form a raw material identity fingerprint database.

[0013] Specifically, in a continuous material feeding system, the raw materials involved in the feeding process typically include various minerals such as limestone, sandstone, clay, and iron powder. Due to differences in their mineral composition and geological origin, the proportions of their main oxides exhibit relatively stable characteristics within a certain timeframe. Therefore, this step involves constructing raw material identification fingerprint vectors to quantitatively characterize the chemical composition of various raw materials.

[0014] First, the baseline composition data of each raw material is obtained. Before the raw materials enter the production system, each batch of incoming raw materials undergoes laboratory analysis to determine the content of its main oxide components. The analysis can be performed using X-ray fluorescence analysis or wet chemical analysis methods, and the results are expressed as mass percentages.

[0015] Let the reference composition data of the j-th raw material obtained after testing be: ,in This represents the mass percentage of calcium oxide (CaO) in the j-th raw material. Indicates the silicon dioxide content in the j-th raw material The percentage of mass content, Indicates the amount of alumina in the j-th raw material. The percentage of mass content, Indicates the iron oxide content in the j-th raw material The percentage content of each component is derived from laboratory test data. It should be noted that this example only selects four types of raw materials—limestone, sandstone, clay, and iron powder—as the analytical objects. In reality, it refers to the content of the main oxides in the four types of raw materials.

[0016] After obtaining the aforementioned oxide content, the chemical structural characteristics of the raw materials were characterized by constructing elemental ratio features. Specifically, four representative sets of elemental ratio features were selected, and their calculation method was defined as follows: ; in, This represents the characteristic elemental ratio between the calcium oxide content and the silicon dioxide content in the j-th raw material, used to reflect the relative proportion between alkaline minerals and siliceous minerals in the raw material. This represents the characteristic elemental ratio between the calcium oxide content and the aluminum oxide content in the j-th raw material, used to reflect the proportional characteristics between the calcareous components and the aluminum components. This represents the elemental ratio characteristic between the content of silicon dioxide and the content of aluminum oxide of the j-th type, used to characterize the structural features of silica-alumina minerals. This represents the elemental ratio characteristic between the content of the j-th type of iron oxide and the content of aluminum oxide, used to describe the relative enrichment of iron minerals in the raw material.

[0017] The reason for using the above ratio relationship instead of directly using the absolute content of oxides as the characteristic parameter is that in a continuous batching system, the amount of each raw material fed may be affected by factors such as fluctuations in the feeding equipment and changes in the material accumulation state during actual operation, resulting in short-term disturbances. If the absolute content is directly used as the characteristic parameter, it may be affected by instantaneous flow rate changes, thereby reducing the stability of deviation diagnosis. However, using the ratio relationship between elements as the characteristic parameter can eliminate the influence of flow rate fluctuations to a certain extent, making the constructed feature vector more stable in reflecting the chemical composition and structure of the raw material itself.

[0018] After obtaining the elemental characteristic ratios of a single sample, to improve the stability of the raw material identification fingerprint vector, statistical processing is performed on multiple samples of the same raw material under the same mineral source conditions. Let N be the total number of samples collected for the j-th raw material. j For each sample, its raw material identity fingerprint vector is defined as: ; in, This represents the raw material identification fingerprint vector for the j-th raw material, where j is the raw material number. The statistical average of the ratios of the four elements corresponding to the j-th raw material is obtained by analyzing N. j The average value of N is calculated from the samples. j The quantity is usually between 10 and 20, and can be set to 10 here.

[0019] After constructing the identity fingerprint vectors of various raw materials, all raw material identity fingerprint vectors and corresponding raw material benchmark composition data are synchronously stored in the raw material identity fingerprint database. The raw material benchmark composition data represents the average chemical composition of the raw material during the stable production stage and is used as a benchmark reference in subsequent batching compensation calculations.

[0020] S2: Deploy an online component analysis device on the main conveyor belt to periodically monitor the chemical composition of the continuously conveyed material, obtain the material composition change sequence, and construct a monitoring index change sequence based on the material composition change sequence; Furthermore, step S2 also includes: An online component analysis device is installed in the stable detection area of ​​the main conveyor belt that continuously conveys materials. The online component analysis device is a PGNAA online analyzer. The online component analysis device continuously monitors the composition of the transported material at a fixed sampling period, obtains the material composition vector, retains the material composition vectors of the most recent M sampling periods, and constructs a material composition change sequence. The monitoring indicators are calculated based on the material composition change sequence to form a monitoring indicator change sequence. The monitoring indicators include lime saturation ratio, silicon content, and aluminum content.

[0021] Specifically, after establishing the raw material identification fingerprint database in step S1, in order to realize real-time monitoring and dynamic analysis of the continuous batching process, this step constructs an online composition monitoring system for continuously conveyed materials to obtain material composition data reflecting the current production status, and further constructs monitoring indicators for subsequent trend analysis and control decisions based on this data.

[0022] First, an online PGNAA analyzer based on the principle of transient gamma neutron activation analysis was used as an online component analysis device to monitor the conveyed materials. The PGNAA online analyzer was installed in the stable detection area of ​​the main conveyor belt that continuously conveys materials. This area is preferably located in the section of the main belt after the convergence of each raw material, and meets the conditions of stable material layer thickness, uniform lateral distribution of materials, and minimal vibration interference. Specifically, the neutron source and gamma ray detector are arranged above and below the conveyor belt to form a through-beam or surround detection structure. The neutron source emits fast neutrons into the material layer, which penetrate the material layer. Each element in the material generates characteristic gamma rays under the excitation of the neutrons, and the detector collects the gamma ray signals generated after penetrating the material. During the detection process, the PGNAA online analyzer first obtains the content information of each major element in the material through energy spectrum analysis. Then, based on the pre-calibrated element-oxide conversion relationship, the element content is converted into the mass percentage content of the corresponding oxide, thereby obtaining the chemical composition data of the material. To ensure the stability of the test results, the thickness of the material layer on the conveyor belt is preferably controlled within the range of 200mm to 400mm, and the material is evenly distributed in the cross-sectional direction by a belt shaping device.

[0023] In actual monitoring, the PGNAA online analyzer performs continuous monitoring with a fixed sampling period, preferably 60 seconds, meaning a complete component analysis result is output every 60 seconds. Let the material component vector obtained at sampling time t be... ,in Let be the mass percentage of calcium oxide in the material at time t. Let t be the mass percentage of silica in the material at time t. Let t be the mass percentage of alumina in the material. This represents the mass percentage of iron oxide in the material at time t. All the above data are calculated by the internal data processing unit of the PGNAA device and transmitted in real time to the upper control system or data processing server via the industrial communication interface.

[0024] To support subsequent trend analysis, the material composition vector is cached, retaining historical data from the most recent M sampling periods to construct a material composition change sequence. It should be noted that the value of M should match the material's transport lag time from the batching point to the monitoring point, typically ranging from 15 to 30, corresponding to a production process timescale of approximately 15 to 30 minutes, thus ensuring physical consistency for subsequent trend analysis.

[0025] After obtaining the material composition change sequence, in order to further reflect the quality status of the material, it is necessary to calculate three commonly used monitoring indicators in cement production based on the above oxide content, including lime saturation ratio KH(t), silica ratio SM(t), and alumina ratio AM(t). The specific calculation method is as follows: Lime saturation ratio: ; The numerator represents the calcium oxide content, the denominator represents the theoretical equivalent amount of acidic oxides required to react with calcium oxide, and the lime saturation ratio can be used to reflect the relative excess or deficiency of limestone components.

[0026] Silicon content: ; Among them, the silicon ratio can be used to reflect the proportional relationship between silicon components and iron and aluminum components.

[0027] Aluminum ratio: ; Among them, the aluminum ratio is used to reflect the proportional relationship between the aluminum component and the iron component.

[0028] Through the above calculations, three corresponding monitoring indicators can be obtained for each sampling period, thus forming a sequence of monitoring indicator changes. It should be noted that in actual industrial applications, the KH value is usually controlled within the range of 1.00 to 1.05, SM within the range of 2.3 to 2.8, and AM within the range of 1.3 to 1.7. The specific value range can be adjusted according to the clinker type and process requirements.

[0029] Furthermore, to improve data reliability, outlier handling is required before data is included in subsequent calculations. When a significant anomaly occurs in the detected value within a sampling period, such as a negative oxide content or a change in value between adjacent periods exceeding 10%–20% of the previous value, the data is identified as an outlier. For outliers, a linear interpolation method based on adjacent time points is preferred for correction, thereby ensuring the continuity and stability of the time series.

[0030] S3: Obtain trend inertia parameters based on the change sequence of monitoring indicators, calculate the effective deviation based on the trend inertia parameters, and determine the current state of the transportation system based on the effective deviation. The trend inertia parameters include trend inertia strength and trend acceleration. Furthermore, step S3 also includes: Obtain the lime saturation ratio data from the monitoring indicator change sequence, and calculate the trend inertia strength. The specific expression is as follows: ; Where I(t) represents the trend inertia intensity at time t, KH(t) represents the lime saturation ratio at the current time, KH(tk) represents the lime saturation ratio in the kth historical sampling period, and n represents the inertia window length. α represents the weighting coefficient, and α represents the inertia decay coefficient. The formula for calculating the trend acceleration based on strong inertia is as follows: ; Where A(t) represents the trend acceleration, I(t) represents the trend inertia strength at the current moment, and I(t-1) represents the trend inertia strength of the previous sampling period; The adaptive inertial compensation coefficient is calculated based on trend acceleration. The specific calculation formula is as follows: ; Where β represents the adaptive inertia compensation coefficient, Here, γ is the basic compensation coefficient, η is the acceleration influence amplitude coefficient, tanh is the hyperbolic tangent function, sign represents the sign function, and KH(t) represents the lime saturation ratio at the current time t. The target set value representing the lime saturation ratio; The effective deviation is calculated based on the trend inertia strength, trend acceleration, and adaptive inertia compensation coefficient. The specific expression is as follows: ; in, K represents the effective deviation after inertia correction, and KH(t) represents the lime saturation ratio at the current time t. β represents the target setpoint for lime saturation ratio, I(t) represents the adaptive inertia compensation coefficient, μ represents the trend inertia intensity, A(t) represents the acceleration compensation coefficient, and sign represents the sign function. The dynamic intervention threshold is set, and the specific calculation formula is as follows: ; Where δ(t) represents the dynamic intervention threshold at time t, ρ is the basic intervention threshold, κ is the threshold dynamic adjustment amplitude coefficient, and κ is the threshold sensitivity coefficient. The state of the transportation system is determined based on the quantitative relationship between effective deviation and dynamic intervention threshold. At that time, it is determined that the current transportation system is in a state of trend deviation.

[0031] Specifically, after obtaining the monitoring index change sequence under continuous time scale in step S2, this step obtains the trend inertia parameter based on the sequence, quantitatively characterizes the trend direction and change intensity of the current material composition fluctuation, and thus determines whether the current system is in a trend deviation state that requires batching adjustment.

[0032] In continuous batching production, there is a significant time lag between the material feeding points and the online detection location, typically ranging from 10 to 20 minutes. Therefore, the monitoring index value at the current monitoring time t actually reflects the batching status at a historical moment. Directly adjusting based on the current detection deviation would cause the adjustment to act on a state that has already occurred, leading to system oscillations in a cycle of "overshoot—reversal—re-overshoot." To avoid this problem, this step compensates for historical trends through trend inertia modeling, transforming control decisions from instantaneous responses to trend-aware responses.

[0033] In the specific implementation process, the lime saturation ratio KH(t) is first selected as the sole trend analysis object from the monitoring indicator change sequence constructed in step S2. The reason for choosing the lime saturation ratio as the core indicator is that it directly determines the balance between limestone and acid oxides in the raw meal and is the dominant indicator affecting the mineral composition of clinker. When the lime saturation ratio remains stable, the silica and alumina ratios are usually within a reasonable range. If trend criteria are established independently for multiple indicators at the same time, it will lead to conflicts between multiple adjustment signals, making it difficult to form a unique control decision. Therefore, this step uniformly uses the lime saturation ratio as the basic variable for trend analysis.

[0034] After obtaining the lime saturation ratio data, the trend inertia strength is calculated to quantify the direction and intensity of the current trend. The specific expression is as follows: ; Where I(t) represents the trend inertia strength at time t, and its physical meaning is the trend momentum of the current indicator change. When I(t) > 0, it indicates that the current indicator is in an upward trend, and when I(t) < 0, it indicates that the current indicator is in a downward trend. The magnitude of the value reflects the strength of the trend; a larger value indicates a stronger trend. KH(t) represents the lime saturation ratio at the current moment, and KH(tk) represents the lime saturation ratio in the kth historical sampling period. Both are calculated in step S2. n represents the inertial window length, indicating the number of historical periods involved in the calculation. Its value is determined based on the material lag time, and here it can be taken as 20, corresponding to approximately 20 minutes of historical data. These represent weighting coefficients, used to adjust the contribution levels of different data points. The goal is to ensure that the sum of all weighting coefficients equals 1; therefore, an equal-weighting approach is preferred. α represents the inertia decay coefficient, used to control the degree of influence of historical data on the current trend. Its value ranges from 0.05 to 0.2. In this embodiment, α = 0.1 is used. The exponential term... This is used to attenuate historical data from a distant time period, so that the older the data, the smaller its contribution to the current trend. The denominator k is used to normalize the time interval, so that the difference between more distant time points contributes less to the trend proportionally, thereby ensuring the stability of the trend calculation.

[0035] After obtaining the inertial intensity I(t), to further characterize the dynamic properties of the trend change, a trend acceleration is introduced, which is defined as: ; Where A(t) represents the trend acceleration, I(t) represents the trend inertia strength at the current moment, and I(t-1) represents the trend inertia strength in the previous sampling period. When A(t) is greater than 0, it indicates that the current trend is strengthening, that is, the speed of deviation from the target is accelerating. When A(t) is less than 0, it indicates that the current trend is weakening, that is, the system may enter the natural regression stage. This parameter is used to identify whether the trend is in the accelerated deviation or natural convergence stage.

[0036] After obtaining the trend acceleration, the adaptive inertia compensation coefficient is calculated based on the trend acceleration. The specific calculation formula is as follows: ; Where β represents the adaptive inertia compensation coefficient, The basic compensation coefficient serves as the benchmark for inertial compensation. Its value must ensure that, within typical production fluctuations, the inertial compensation term does not exceed 50% of the current deviation to avoid over-compensation leading to control reversal oscillations. Based on PID control engineering tuning experience, the basic compensation coefficient typically ranges from 0.2 to 0.4. This embodiment uses [a specific value]. γ is the acceleration influence amplitude coefficient, used to control the maximum adjustment range of the trend acceleration on the compensation coefficient. Its value should ensure that β is always within the range of [0,1] to avoid compensation amplification or sign reversal. Based on the mathematical properties of the hyperbolic tangent function and its application in control engineering, the specific value range of the acceleration influence amplitude coefficient is 0 to 0.5. In this embodiment, γ = 0.3 is taken. η is the acceleration sensitivity coefficient, used to adjust the sensitivity of the trend acceleration to the compensation coefficient. Its value should match the statistical distribution range of the trend acceleration, so that... Within the typical fluctuation range, the compensation coefficient falls within the linear response range of the tanh function, thus ensuring smooth changes in the compensation coefficient. The specific value range is 5 to 20; in this embodiment, η = 10 is used. The acceleration influence amplitude coefficient and acceleration sensitivity coefficient adjust the dynamic characteristics of the compensation coefficient from two dimensions: amplitude range and response speed, respectively. `sign` represents the sign function, and `KH(t)` represents the lime saturation ratio at the current time `t`. The target setpoint represents the lime saturation ratio, and tanh is the hyperbolic tangent function used to map the acceleration to the interval [-1, 1], ensuring that the compensation coefficient is always within a reasonable range. The physical meaning of this formula is that when the trend acceleration differs from the current deviation sign, the system is undergoing natural regression, and β needs to be less than [a certain value]. To reduce the compensation intensity and avoid interfering with the natural regression process; when the trend acceleration has the same sign as the current deviation, the deviation accelerates, and β needs to be greater than 1. This is to enhance compensation and encourage the system to intervene earlier.

[0037] In traditional control methods, the deviation is usually the difference between the current value and the target setpoint. However, this deviation does not consider trend factors and is prone to misadjustment during the inertia regression phase. Therefore, this step introduces a trend inertia correction mechanism to construct an effective deviation, the specific expression of which is: ; in, K represents the effective deviation after inertia correction, and KH(t) represents the lime saturation ratio at the current time t. The target set value representing the lime saturation ratio is determined by process requirements; in this embodiment, it is taken as... β represents the adaptive inertia compensation coefficient, I(t) represents the trend inertia intensity, and μ represents the acceleration compensation coefficient, the value of which must be significantly smaller than the basic compensation coefficient. To ensure that the acceleration term only participates in the calculation as a trend correction term and avoids dominating the control decision, the specific value range is 0 to 0.5. In this embodiment, μ = 0.2 is used. A(t) represents the trend acceleration, and sign represents the sign function. When the value is greater than 0, the function outputs 1; when... When the value is zero, the function outputs 0; when... When the value is less than zero, the function outputs -1.

[0038] The above inertia compensation item The purpose of this compensation term is to determine whether the current deviation direction aligns with the trend direction, indicating that the system is moving further away from the target. In this case, the compensation term is positive, increasing the effective deviation and triggering adjustment earlier. Conversely, if the current deviation direction is opposite to the trend direction, indicating that the system has entered a natural regression process, the compensation term is negative, reducing the effective deviation and avoiding unnecessary adjustment.

[0039] The above acceleration compensation term The function of this is that when the trend acceleration is greater than 0 and has the same sign as the deviation, the acceleration term is positive, further reducing the effective deviation and strengthening the judgment of not intervening during natural regression. When the trend acceleration is greater than 0 and has the same sign as the deviation, the acceleration term is negative, further increasing the effective deviation and strengthening the judgment of early intervention when the deviation is accelerated. When the trend acceleration is less than 0, the acceleration term has the opposite effect, reflecting the impact of trend deceleration on decision-making.

[0040] It should be noted that the trend acceleration participates in the effective deviation calculation in this step in two forms: indirect modulation and direct compensation. The influence of the trend inertia term is modulated by the adaptive inertia compensation coefficient, which is an adjustment of the amplitude of the first-order trend. Simultaneously, it directly reflects the impact of the trend change rate on the evolution of the system deviation as an independent compensation term, which is a dynamic correction of the second-order trend. Since both act at different physical levels, they do not constitute redundant calculations.

[0041] After obtaining the effective deviation, a dynamic intervention threshold is set, and the specific calculation formula is as follows: ; Where δ(t) represents the dynamic intervention threshold at time t, The basic intervention threshold is determined according to the allowable deviation range of the process in the "Cement Plant Design Code". For the lime saturation ratio index, it is usually taken as 0.02~0.05. In this embodiment, it is taken as... ρ is the threshold dynamic adjustment amplitude coefficient, ranging from 0 to 0.5, used to control the maximum adjustment amplitude of acceleration on the threshold. In this embodiment, ρ = 0.3 is used. κ is the threshold sensitivity coefficient, ranging from 5 to 20. In this embodiment, κ = 10 is used. tanh is the hyperbolic tangent function. The physical meaning of this formula is that when... When the value is large, the trend changes drastically, δ(t) decreases, making the system more susceptible to intervention to respond to rapid changes. When the value is small, the trend is gradual, and δ(t) is close to... Maintain normal intervention sensitivity, when When the value approaches 0, the trend stabilizes. Restore the basic threshold.

[0042] After determining the effective deviation and dynamic intervention threshold, a comprehensive intervention judgment condition is constructed. When the condition is met... If the current transportation system is determined to be in a state of trend deviation, it is necessary to proceed to the subsequent deviation source tracing and adjustment steps; otherwise, it is assumed that the current deviation can be eliminated naturally through system inertia, and no adjustment operation is performed.

[0043] S4: When the transportation system is in a state of trend deviation, construct the finished product mixture fingerprint vector of the current material based on the material composition change sequence; Specifically, after determining in step S3 that the current system is in a trend deviation state, it indicates that the fluctuation of the current material composition has exceeded the range within which the system can naturally converge, and it is necessary to further identify the source of the deviation. Since the materials in the continuous batching system are formed by mixing multiple raw materials in a certain proportion, it is impossible to determine which raw material is causing the anomaly by the deviation of a single monitoring indicator. Therefore, it is necessary to quantitatively characterize the overall chemical structure characteristics of the current mixture to provide characteristic input data for subsequent deviation tracing.

[0044] In the specific implementation process, the material composition change sequence obtained in step S2 is first called. This sequence contains the mass percentage content of calcium oxide, silicon dioxide, aluminum oxide, and iron oxide detected by the online component analysis device at the current moment. After obtaining the above oxide content, following the construction method of the raw material identity fingerprint vector in step S1, the element ratio calculation is performed on the current finished product material to construct the finished product mixed fingerprint vector, the specific expression of which is: ,in, The calculation method is the same as in step S1. The calculation methods are the same, both involving element ratio characteristics, which together constitute a characteristic description of the chemical structure of the current mixture.

[0045] It should be noted that the reason for using elemental ratios rather than absolute content to construct the fingerprint vector is that significant flow rate disturbances exist in the system during continuous feeding and batching. When the amount of a raw material fed fluctuates, even if the chemical composition of that raw material itself remains unchanged, the mass percentage of each oxide in the finished product will change accordingly due to the overall change in proportion, thus producing a dilution effect. For example, when the amount of limestone fed decreases, the CaO content in the finished product decreases, and simultaneously... , When the relative proportions of certain components increase, directly using absolute content for matching might misjudge a change in limestone grade. However, when using ratio characteristics, if only quantitative changes (i.e., changes in proportion) occur, the component contents change proportionally, and the ratio remains relatively stable. But when qualitative changes occur (i.e., changes in raw material composition), the ratio will shift significantly. Therefore, ratio characteristics can effectively distinguish between apparent changes caused by flow disturbances and fundamental changes caused by changes in raw material grade, thus significantly improving the accuracy of subsequent deviation tracing.

[0046] S5: Perform similarity matching calculations between the finished product's mixed fingerprint vector and the raw material identity fingerprint database to identify the source of deviation for the current material; Furthermore, step S5 also includes: The similarity calculation is performed between the finished product's hybrid fingerprint vector and the raw material identity fingerprint database. The specific calculation formula is as follows: ; in, The similarity between the finished product's mixed fingerprint vector and the raw material feature fingerprint vector of the j-th raw material at time t is represented. The finished product mixture fingerprint vector representing the current material. The raw material identity fingerprint vector representing the j-th raw material; The maximum similarity is selected based on the similarity calculation results. Raw materials with dominant deviation ; The real-time flow data of each raw material is obtained from the DCS system or the feed scale control system, where the set flow rate of the j-th raw material is denoted as... The actual measured flow rate is recorded as The flow deviation rate is calculated based on the two types of data, and the specific calculation formula is as follows: ; in, This represents the flow rate deviation rate of the raw material at time t; Determine the flow deviation threshold based on equipment accuracy. Simultaneously, a similarity threshold is introduced to construct classification rules, wherein the similarity threshold includes a high matching threshold. With medium matching threshold : When satisfied Furthermore, if there are recent changes in test data or batch switching records for the corresponding raw materials, the current deviation is determined to be a change in the grade of the raw materials. When satisfied And the absolute value of the flow deviation rate of the corresponding raw materials. The current deviation is determined to be caused by fluctuations in the material feed rate; When satisfied If a fingerprint combination that does not conform to physical laws is detected, the current deviation is determined to be due to abnormal interference.

[0047] Specifically, after obtaining the finished product mixed fingerprint vector at the current moment, in order to identify the specific source of raw materials causing the component deviation, this step performs similarity matching calculation between the finished product mixed fingerprint vector and the raw material identity fingerprint database constructed in step S1, thereby realizing quantitative source tracing analysis of the current deviation source.

[0048] First, the raw material identification fingerprint database stored in step S1 is accessed. This fingerprint database contains the raw material feature fingerprint vectors of various raw materials. The resulting hybrid fingerprint vector obtained in step S4 is then... Similarity is calculated between the fingerprint vectors of each raw material and the fingerprint vectors. To eliminate the influence of dimensions and ensure computational stability, cosine similarity is used as the matching index. The specific calculation formula is as follows: ; in, The similarity between the finished product's mixed fingerprint vector and the raw material feature fingerprint vector of the j-th raw material at time t is denoted by , with a value ranging from [0,1]. The closer the value is to 1, the more similar the structural features of the two are. The finished product mixture fingerprint vector representing the current material. The numerator is the vector representing the identity fingerprint of the j-th raw material. The numerator is the vector dot product, which reflects the consistency of the two fingerprints in each feature dimension. The denominator is the Euclidean norm product of the two vectors, which is used for normalization to eliminate the influence of different magnitudes on the calculation results.

[0049] By performing the above calculations on all raw materials, the similarity vector at the current moment can be obtained. ,in This represents the similarity to the raw material fingerprint of limestone. This represents the similarity to the raw material identity fingerprint of sandstone. This represents the similarity to the raw material fingerprint of iron powder. This represents the similarity to the raw material fingerprint of clay.

[0050] Since the materials being transported are actually mixtures of multiple raw materials, their fingerprint features are a comprehensive reflection of the superposition of features from each raw material. Therefore, the similarity vector not only reflects the matching degree of a single raw material but also the result of the coupling effect of multiple raw materials. When one similarity is significantly higher than others, it indicates that the structural features of the current finished product are mainly dominated by that raw material. When multiple similarities are close and all are high, it indicates the existence of multi-source coupling bias. When all similarities are low, it indicates that the current fingerprint features cannot be explained by existing raw material fingerprints, and there may be detection anomalies or unknown perturbations.

[0051] To implement the executable decision logic for the project, the maximum value in the similarity vector S(t) is defined as the maximum similarity. At the same time, the corresponding raw materials are recorded as the dominant deviation raw materials. .

[0052] Furthermore, to distinguish different types of deviation sources, a similarity threshold and a traffic deviation threshold are introduced to construct classification rules. The similarity threshold includes a high-match threshold. With medium matching threshold .

[0053] High matching threshold Used to identify changes in raw material composition, its value ranges from 0.85 to 0.95, and in this embodiment, it is taken as... =0.90; Medium matching threshold Used to identify traffic disturbances, its value ranges from 0.50 to 0.70, and in this embodiment, it is taken as... =0.60.

[0054] Simultaneously, real-time flow data of each raw material is obtained from the DCS system or the feed scale control system, where the set flow rate of the j-th raw material is denoted as... The actual measured flow rate is recorded as The flow deviation rate is calculated based on the two types of data, and the specific calculation formula is as follows: ; in, This represents the flow deviation rate of the raw material at time t, and the flow deviation threshold. The value is determined based on the equipment's accuracy, typically ranging from 0.05 to 0.10, which is 5% to 10% of the set flow rate. In this embodiment, it is taken as... =0.05.

[0055] Based on the above parameters, the logic for classifying and determining the source of deviation is constructed as follows: When satisfied Furthermore, if there are recent changes in the test data or batch switching records of the corresponding raw materials, the current deviation is determined to be due to a change in the grade of the raw materials. Physically, this means that while the finished product's fingerprint is highly similar to that of a certain raw material, it deviates from the target state overall, indicating that the chemical composition of the raw material itself has changed.

[0056] When satisfied And the absolute value of the flow deviation rate of the corresponding raw materials. The current deviation is determined to be caused by fluctuations in the feed rate of the raw material. Physically, this means that while the finished product retains the fingerprint characteristics of the raw material, the overall proportions are unbalanced due to the actual feed rate deviating from the set value.

[0057] When satisfied If a fingerprint combination that does not conform to physical laws is detected, the current deviation is determined to be due to abnormal interference. A combination that does not conform to physical laws refers to a fingerprint ratio that exceeds the reasonable chemical composition range of cement raw materials, for example, when... and If the threshold value is not met, it can be determined that the detection signal of a certain element is abnormal. The above threshold values ​​are derived from the statistical range of long-term production experience of cement raw materials.

[0058] S6: Generate adjustment instructions based on the source of deviation. The adjustment instructions include proportion adjustment instructions, equipment control instructions, data substitution and conservative mode instructions.

[0059] Furthermore, step S6 also includes: Identify the source of the deviation and generate adjustment instructions based on the source of the deviation: When the deviation type is determined to be a change in raw material grade, a ratio adjustment instruction is generated to adjust the ratio of the raw material with the dominant deviation, while coordinating the ratio of other raw materials to maintain the total amount balance. When the deviation type is determined to be a fluctuation in the material flow rate, a device control command is generated to directly correct the output of the material feeding device and trigger an abnormal alarm. When the deviation type is determined to be detection interference, a data substitution and conservative mode command is generated, and the predicted value is used to replace the measured data in subsequent calculations. At the same time, the control system is switched to conservative mode.

[0060] Furthermore, step S6 also includes: Based on the type of raw material with dominant deviation, the composition change vector of this raw material is defined as: ; in, This represents the vector of compositional changes for the j-th raw material. This is the baseline composition of the raw material under its historical stable state. The latest composition detected at the current moment; Based on the component change vector, the impact of component changes on the overall composition of the mixture is calculated. The specific calculation formula is as follows: ; in, This represents the impact of changes in raw materials on the overall composition of the mixture. This represents the proportion of the j-th raw material at the current moment. The vector representing the compositional change of the j-th raw material; A proportion adjustment amount is set, and the impact of the proportion adjustment on the overall composition of the mixture is quantitatively expressed as follows: ; in, This represents the impact of changes in the proportions on the overall composition of the mixture. This represents the amount of adjustment in the proportion. This is the baseline composition of the raw material under its historical stable state; Based on the impact of raw material changes on the overall composition of the mixture and the impact of proportion changes on the overall composition of the mixture, a complete compensation condition is established. Furthermore, a proportion compensation calculation formula is constructed by correlating monitoring indicators. The specific expression is as follows: ; in, Let j be the amount of material to be adjusted in the proportion of the j-th raw material. Here, W is the compensation gain coefficient, W is the weight vector, and J is the sensitivity matrix. This represents the vector of compositional changes for the j-th raw material. This is the baseline composition of the raw material under its historical stable state. This represents the ratio of the j-th raw material at the current moment; Based on the calculated adjustment amount, a new ingredient setting value is generated. The specific calculation formula is as follows: ; in, This represents the new ingredient settings. For the current raw material ratio, This represents the adjustment amount for the proportion of the j-th raw material; Meanwhile, to ensure that the overall proportions meet the conservation constraints and that the sum of all raw material proportions is 1, the proportions of other unchanged raw materials are corrected by scaling them proportionally.

[0061] Furthermore, step S6 also includes: Read the flow deviation rate of the j-th raw material at time t, and use an incremental adjustment algorithm with integral compensation to obtain the equipment control output ratio based on the current flow deviation rate. The specific calculation formula is as follows: ; in, The current device control output ratio. The ratio of the device control output at the previous moment. The current flow deviation rate. This represents the flow deviation rate at time td. The proportional compensation gain is ξ, where ξ represents the integral compensation gain, d is the index value, and D represents the length of the integral window. Issue a corresponding equipment malfunction alarm based on the direction of deviation. If the value is greater than 0, a blockage alarm is triggered. If the value is less than 0, an alarm for uncontrolled material feeding is triggered.

[0062] Specifically, after identifying the sources of deviation, the deviation diagnosis results need to be transformed into executable adjustment instructions. The core of this process lies in defining different adjustment methods for various sources of deviation, thereby achieving quality control of transported materials.

[0063] The data sources involved in this step have a clear hierarchical relationship. Among them, the current material composition vector... The online monitoring data from step S2, the dominant deviation raw materials and the source of deviation are obtained from the similarity analysis results in step S5, and the raw material reference composition data for each raw material. The data originates from the raw material baseline composition data synchronously stored during step S1 when establishing the raw material identity fingerprint database, and the current proportions of each raw material. Flow data is obtained directly from the DCS system or the batching control system. , The signal originates from the feeder scale in the DCS system.

[0064] Based on the above data, the first step is to generate a ratio adjustment instruction in response to changes in raw material grade. The meaning of this instruction is to offset the comprehensive impact of changes in the composition of the dominant deviation raw material on the monitoring indicators by adjusting the ratio of the dominant deviation raw material, while coordinating the ratio of other raw materials to maintain the total amount balance.

[0065] Specifically, a multi-component coupled compensation model is first constructed. When the j-th raw material is identified as the dominant deviation raw material, the component change vector of this raw material is defined as: ; in, The vector representing the compositional change of the j-th raw material is specifically expressed as follows: This indicates the amount of variation of the raw material across the four oxide dimensions. This is the baseline composition of the raw material under its historical stable state, derived from the raw material baseline composition data in step S1. The latest composition detected at the current moment is derived from the material composition vector in step S2.

[0066] In a continuous batching process, the overall composition of the mixture is determined by the composition and proportion of each raw material. Let the proportion of the j-th raw material at the current moment be... When the composition of the j-th raw material changes but its proportion has not been adjusted, the impact of this change on the overall composition of the mixture can be quantified as follows: ; in, This represents the impact of changes in raw materials on the overall composition of the mixture. This represents the proportion of the j-th raw material at the current moment. This represents the vector of compositional changes for the j-th raw material.

[0067] To eliminate the impact of this component change on the quality of the finished product, the proportion of the j-th raw material needs to be adjusted. Let the adjustment amount be... Then the raw material composition remains at the baseline composition. Under these conditions, the impact of proportion adjustment on the overall composition of the mixture can be quantified as follows: ; in, This represents the impact of changes in the proportions on the overall composition of the mixture. This represents the amount of adjustment in the proportion. This is the baseline composition of the raw material under its historical stable state.

[0068] Therefore, to fully compensate for changes in composition, the effects of composition changes and proportion adjustments on the composition of the mixture should cancel each other out, meaning their sum should be zero. The specific expression is: This formula is a vector equation, indicating that the sum of the effects of compositional changes and proportion adjustments is zero across the four oxide dimensions.

[0069] Furthermore, the changes in the mixture composition are correlated with monitoring indicators. In cement production, the monitoring indicators, lime saturation coefficient KH, silica ratio SM, and alumina ratio AM, are nonlinear functions of oxide content. To simplify calculations, this nonlinear relationship is linearized to first order under small perturbation conditions. Let the target monitoring indicator vector be Y(t)=[KH(t),SM(t),AM(t)], and the change in the monitoring indicators when the mixture composition changes can be approximately expressed as... Where ΔY represents the change in the monitoring indicator. This represents the overall change in the composition of the mixture, which is the sum of the impact of changes in raw materials on the overall composition of the mixture and the impact of changes in the mix proportions on the overall composition of the mixture. Specifically, it is expressed as... J is the sensitivity matrix.

[0070] It should be noted that the sensitivity matrix J has a dimension of 3×4 and is used to describe the degree of influence of changes in the content of the four oxides on the three monitoring indicators, including the lime saturation coefficient KH, silicon content SM, and aluminum content AM. Its specific form is as follows: ; The first row represents the effect of changes in the content of each oxide on the lime saturation coefficient KH, the second row represents the effect of changes in the content of each oxide on the silicon content SM, and the third row represents the effect of changes in the content of each oxide on the aluminum content AM.

[0071] This matrix is ​​obtained by collecting historical production data from the factory and establishing a multiple linear regression model to solve for the regression coefficient matrix. In this embodiment, based on typical cement raw material conditions, with CaO content of 42.0%, SiO2 content of 13.5%, Al2O3 content of 3.2%, and Fe2O3 content of 2.1%, the specific values ​​of the sensitivity matrix calculated using the mechanistic model are as follows: ; Each row in the matrix corresponds to a monitoring indicator, and each column corresponds to a type of oxide. The matrix elements represent the change in the monitoring indicator when the content of the oxide changes by 1%.

[0072] Substituting the compensation conditions into the linearization model yields... Since our goal is to eliminate the impact of component changes through ratio adjustments, the change in the adjusted monitoring index should be zero, i.e., ΔY=0. Therefore, we can obtain... .

[0073] Based on this, the impact of changes in a single raw material component on the overall monitoring indicators is expressed in a weighted manner, and a formula for calculating the proportion compensation is constructed. The specific expression is as follows: ; in, Let j be the amount of material to be adjusted in the proportion of the j-th raw material. The gain coefficient is used to compensate for over-adjustment and is typically set to 0.5~1.0. In this embodiment, it is set to 0.8 to suppress overshoot. w is a weight vector used to characterize the importance of different oxides to the target quality index, and J is a sensitivity matrix used to represent the influence coefficient of each oxide on the monitoring index. This represents the vector of compositional changes for the j-th raw material. This is the baseline composition of the raw material under its historical stable state. Represents the proportion of the j-th raw material at the current moment, where... This indicates the comprehensive impact of changes in the raw material composition on the monitoring indicators in a weighted sense. This is a normalization factor used to map the proportion adjustment amount to a reasonable range.

[0074] It should be noted that the introduced weight vector w performs unified scalarization processing on multiple monitoring indicators. (Details to follow) Each weight represents the importance of the corresponding monitoring indicator, satisfying the following conditions: When KH is the primary monitoring target, w can be set to [0.6, 0.25, 0.15].

[0075] The physical meaning of this formula is that it measures the comprehensive impact of changes in raw material composition on the overall monitoring indicators by using a multi-component weighted method, and then adjusts the ratio in the opposite direction to offset the impact, thereby achieving coordinated and stable control of multiple indicators.

[0076] Based on the calculated adjustment amount, a new ingredient setting value is generated. The specific calculation formula is as follows: ; in, This represents the new ingredient settings. For the current raw material ratio, This represents the adjustment amount for the proportion of the j-th raw material; Meanwhile, to ensure that the overall proportions meet the conservation constraints and that the sum of all raw material proportions is 1, the proportions of other unchanged raw materials need to be normalized. This is usually done by scaling up proportionally, and the new ingredient settings for the remaining raw materials are... This processing method ensures that while the main raw materials are adjusted, the remaining raw materials are changed in proportion, thereby maintaining the stability of the overall ingredient structure.

[0077] When the deviation type is determined to be fluctuation in the feed flow rate, a device control command is generated to directly correct the output of the feed device, thereby eliminating the deviation between the set flow rate and the actual flow rate, and triggering an abnormal alarm.

[0078] Specifically, in this case, component compensation is no longer used; instead, control correction is directly based on the equipment's flow deviation. First, the flow deviation rate of the j-th raw material at time t is read. Then, an incremental adjustment algorithm with integral compensation is used to obtain the equipment control output ratio based on the current flow deviation rate. The specific calculation formula is as follows: ; in, This represents the current control output ratio of the device, indicating the proportion of the current output relative to the device's maximum output capacity. In practical applications, multiplying this value by the device's maximum output capacity, such as the maximum frequency of the frequency converter or the maximum speed of the feeder, yields the specific control parameters. The ratio of the device control output at the previous moment. The current flow deviation rate. This represents the flow deviation rate at time td. The proportional compensation gain is used to quickly adjust according to the current deviation. ξ represents the integral compensation gain, which is used to eliminate the persistent steady-state deviation. d is the index value, and D represents the length of the integral window, which is usually 3 to 5 cycles. In this embodiment, D=3 is used. This value matches the material conveying lag time and the detection cycle, ensuring that the integral term can cover 3 to 5 minutes of historical deviation information, thereby effectively eliminating the steady-state deviation caused by changes in equipment characteristics.

[0079] It should be clarified that the proportional compensation gain can be determined with reference to PID control engineering experience and equipment response characteristics. When the inertia is small, the adjustment is smooth and the response is slow, making it suitable for feeding equipment with high inertia, such as belt scales. When the inertia is large, the response is fast, but it is prone to oscillation. It is suitable for feeding equipment with low inertia, such as screw feeders and rotor scales. In this embodiment, we take... =0.8. The integral compensation gain can be determined with reference to the PID integral time constant tuning experience, and is usually taken as 0.8. The value of ξ is 1 / 5 to 1 / 10. If ξ is too large, it will cause integral saturation. If it is too small, the steady-state error will be eliminated slowly. In this embodiment, ξ is taken as 0.1.

[0080] At the same time, an alarm for the corresponding equipment abnormality is issued based on the direction of the deviation. A value greater than 0 indicates insufficient actual material feeding, and this occurs for multiple consecutive cycles, such as three cycles where the absolute value of the flow deviation rate is greater than 0. This triggers a material blockage alarm, indicating potential issues such as material bridging, feeder jamming, or an empty hopper. The value is less than 0, indicating that the actual material feed is excessive, and the absolute value of the flow deviation rate is greater than 0 for multiple consecutive cycles. This triggers a feeding out-of-control alarm, indicating possible problems such as valve not closing properly, feeder speed overshoot, or frequency converter malfunction.

[0081] When the deviation type is determined to be detection interference, a data substitution and conservative mode command is generated, and the predicted value is used to replace the measured data in subsequent calculations. At the same time, the control system is switched to conservative mode.

[0082] Specifically, control calculations based on current monitoring data will be suspended. Instead, a predictive model will be used to generate alternative data, and an exponential smoothing model will be used to calculate the predicted values. The specific calculation formula is as follows: ; in, The predicted values ​​are x(t-1) and x(t-2), which are the monitoring values ​​from the previous two periods, derived from the material composition change sequence. λ is the smoothing coefficient, typically between 0.6 and 0.8. The calculated predicted values ​​will replace the current actual monitoring data in subsequent calculations. This is because if detection interference occurs, the current measured data will be unreliable and must be replaced. Simultaneously, the control system is switched to conservative mode, reducing the control gain to 50% of its original value and increasing the intervention threshold to 1.5 times its original value to avoid erroneous adjustments. If detection interference is still detected for several consecutive periods, such as three periods, a maintenance alarm for the detection equipment is triggered, prompting the operator to check the online analyzer.

[0083] After completing the above differential compensation calculation, the new ratio settings or control commands for each raw material are sent to the industrial control system to realize closed-loop regulation of the continuous batching process and enter the real-time monitoring and decision-making cycle of the next cycle.

[0084] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for online component detection and batching optimization of continuously conveyed materials, characterized in that, The method includes: Each batch of raw materials entering the factory is tested and analyzed to obtain raw material baseline component data. Based on the raw material baseline component data, a raw material identity fingerprint vector is constructed. All raw material baseline component data and raw material identity fingerprint vectors are integrated to form a raw material identity fingerprint database. An online component analysis device is installed on the main conveyor belt to periodically monitor the chemical composition of the continuously conveyed material, obtain the material composition change sequence, and construct a monitoring index change sequence based on the material composition change sequence; Trend inertia parameters are obtained based on the change sequence of monitoring indicators, effective deviation is calculated based on the trend inertia parameters, and the current state of the transportation system is determined based on the effective deviation. The trend inertia parameters include trend inertia intensity and trend acceleration. When the transportation system is in a state of trend deviation, construct the finished product mixture fingerprint vector of the current material based on the material composition change sequence; The similarity matching calculation between the finished product's mixed fingerprint vector and the raw material identity fingerprint database is performed to identify the source of deviation for the current material; Adjustment instructions are generated based on the source of deviation. These adjustment instructions include proportion adjustment instructions, equipment control instructions, data substitution and conservative mode instructions.

2. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 1, characterized in that, The raw material identification fingerprint database is formed, including: Each batch of incoming raw materials is tested and analyzed using X-ray fluorescence analyzer or wet chemical analysis method to determine the content of major oxides and obtain the raw material reference composition data for each raw material. Four sets of element ratio characteristics were constructed based on the raw material baseline composition data. The specific calculation method is as follows: ； in, This represents the characteristic elemental ratio between the calcium oxide content and the silicon dioxide content in the j-th raw material. This represents the characteristic elemental ratio between the calcium oxide content and the aluminum oxide content in the j-th raw material. This represents the characteristic elemental ratio between the content of silicon dioxide and the content of aluminum oxide of the j-th type. This represents the characteristic elemental ratio between the content of the j-th type of iron oxide and the content of aluminum oxide. Statistical processing is performed on the four element ratio characteristics of multiple samples of the same raw material to obtain the raw material identification fingerprint vector, the specific expression of which is as follows: ； in, This represents the raw material identification fingerprint vector for the j-th raw material, where j is the raw material number. The statistical average of the ratio characteristics of the four elements corresponding to the j-th raw material; Integrate all raw material identity fingerprint vectors with the corresponding raw material baseline component data to form a raw material identity fingerprint database.

3. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 1, characterized in that, Obtain the material composition change sequence, and construct the monitoring indicator change sequence based on the material composition change sequence, including: An online component analysis device is installed in the stable detection area of ​​the main conveyor belt that continuously conveys materials. The online component analysis device is a PGNAA online analyzer. The online component analysis device continuously monitors the composition of the transported material at a fixed sampling period, obtains the material composition vector, retains the material composition vectors of the most recent M sampling periods, and constructs a material composition change sequence. The monitoring indicators are calculated based on the material composition change sequence to form a monitoring indicator change sequence. The monitoring indicators include lime saturation ratio, silicon content, and aluminum content.

4. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 1, characterized in that, Trend inertia parameters are obtained based on the change sequence of monitoring indicators. Effective deviations are calculated based on these trend inertia parameters, and the current state of the transportation system is determined based on the effective deviations, including: Obtain the lime saturation ratio data from the monitoring indicator change sequence, and calculate the trend inertia strength. The specific expression is as follows: ； Where I(t) represents the trend inertia intensity at time t, KH(t) represents the lime saturation ratio at the current time, KH(tk) represents the lime saturation ratio in the kth historical sampling period, and n represents the inertia window length. α represents the weighting coefficient, and α represents the inertia decay coefficient. The formula for calculating the trend acceleration based on strong inertia is as follows: ； Where A(t) represents the trend acceleration, I(t) represents the trend inertia strength at the current moment, and I(t-1) represents the trend inertia strength of the previous sampling period; The adaptive inertial compensation coefficient is calculated based on trend acceleration. The specific calculation formula is as follows: ； Where β represents the adaptive inertia compensation coefficient, The basic compensation coefficient, γ is the acceleration influence amplitude coefficient, η is the acceleration sensitivity coefficient, tanh is the hyperbolic tangent function, sign represents the sign function, and KH(t) represents the lime saturation ratio at the current time t. The target set value representing the lime saturation ratio; The effective deviation is calculated based on the trend inertia strength, trend acceleration, and adaptive inertia compensation coefficient. The specific expression is as follows: ； in, K represents the effective deviation after inertia correction, and KH(t) represents the lime saturation ratio at the current time t. β represents the target setpoint for lime saturation ratio, I(t) represents the adaptive inertia compensation coefficient, μ represents the trend inertia intensity, A(t) represents the acceleration compensation coefficient, and sign represents the sign function. The dynamic intervention threshold is set, and the specific calculation formula is as follows: ； in, The dynamic intervention threshold representing time t. ρ is the basic intervention threshold, κ is the threshold dynamic adjustment amplitude coefficient, and κ is the threshold sensitivity coefficient. The state of the transportation system is determined based on the quantitative relationship between effective deviation and dynamic intervention threshold. At that time, it is determined that the current transportation system is in a state of trend deviation.

5. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 1, characterized in that, Identify the sources of deviation for the current material, including: The similarity calculation is performed between the finished product's hybrid fingerprint vector and the raw material identity fingerprint database. The specific calculation formula is as follows: ； in, The similarity between the finished product's mixed fingerprint vector and the raw material feature fingerprint vector of the j-th raw material at time t is represented. The finished product mixture fingerprint vector representing the current material. The raw material identity fingerprint vector representing the j-th raw material; The maximum similarity is selected based on the similarity calculation results. Raw materials with dominant deviation ; The real-time flow data of each raw material is obtained from the DCS system or the feed scale control system, where the set flow rate of the j-th raw material is denoted as... The actual measured flow rate is recorded as The flow deviation rate is calculated based on the two types of data, and the specific calculation formula is as follows: ； in, This represents the flow rate deviation rate of the raw material at time t; Determine the flow deviation threshold based on equipment accuracy. Simultaneously, a similarity threshold is introduced to construct classification rules, wherein the similarity threshold includes a high matching threshold. With medium matching threshold : When satisfied Furthermore, if there are recent changes in test data or batch switching records for the corresponding raw materials, the current deviation is determined to be a change in the grade of the raw materials. When satisfied And the absolute value of the flow deviation rate of the corresponding raw materials. The current deviation is determined to be caused by fluctuations in the material feed rate; When satisfied If a fingerprint combination that does not conform to physical laws is detected, the source of the current deviation is determined to be abnormal interference.

6. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 1, characterized in that, Adjustment instructions are generated based on the source of the deviation, including: Identify the source of the deviation and generate adjustment instructions based on the source of the deviation: When the deviation type is determined to be a change in raw material grade, a ratio adjustment instruction is generated to adjust the ratio of the raw material with the dominant deviation, while coordinating the ratio of other raw materials to maintain the total amount balance. When the deviation type is determined to be a fluctuation in the material flow rate, a device control command is generated to directly correct the output of the material feeding device and trigger an abnormal alarm. When the deviation type is determined to be detection interference, a data substitution and conservative mode command is generated, and the predicted value is used to replace the measured data in subsequent calculations. At the same time, the control system is switched to conservative mode.

7. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 6, characterized in that, Generate proportion adjustment instructions, including: Based on the type of raw material with dominant deviation, the composition change vector of this raw material is defined as: ； in, This represents the vector of compositional changes for the j-th raw material. This is the baseline composition of the raw material under its historical stable state. The latest composition detected at the current moment; Based on the component change vector, the impact of component changes on the overall composition of the mixture is calculated. The specific calculation formula is as follows: ； in, This represents the impact of changes in raw materials on the overall composition of the mixture. This represents the proportion of the j-th raw material at the current moment. The vector representing the compositional change of the j-th raw material; A proportion adjustment amount is set, and the impact of the proportion adjustment on the overall composition of the mixture is quantitatively expressed as follows: ； in, This represents the impact of changes in the proportions on the overall composition of the mixture. This represents the amount of adjustment in the proportion. This is the baseline composition of the raw material under its historical stable state; Based on the impact of raw material changes on the overall composition of the mixture and the impact of proportion changes on the overall composition of the mixture, a complete compensation condition is established. Furthermore, a proportion compensation calculation formula is constructed by linking monitoring indicators. The specific expression is as follows: ； in, Let j be the amount of material to be adjusted in the proportion of the j-th raw material. Here, w is the compensation gain coefficient, w is the weight vector, and J is the sensitivity matrix. This represents the vector of compositional changes for the j-th raw material. This is the baseline composition of the raw material under its historical stable state. This represents the ratio of the j-th raw material at the current moment; Based on the calculated adjustment amount, a new ingredient setting value is generated. The specific calculation formula is as follows: ； in, This represents the new ingredient settings. For the current raw material ratio, This represents the adjustment amount for the proportion of the j-th raw material; Meanwhile, to ensure that the overall proportions meet the conservation constraints and that the sum of all raw material proportions is 1, the proportions of other unchanged raw materials are corrected by scaling them proportionally.

8. The method for online component detection and batching optimization of continuously conveyed materials as described in claim 6, characterized in that, Generate device control commands, including: Read the flow deviation rate of the j-th raw material at time t, and use an incremental adjustment algorithm with integral compensation to obtain the equipment control output ratio based on the current flow deviation rate. The specific calculation formula is as follows: ； in, The current device control output ratio. The ratio of the device control output at the previous moment. The current flow deviation rate. This represents the flow deviation rate at time td. The proportional compensation gain is ξ, where ξ represents the integral compensation gain, d is the index value, and D represents the length of the integral window. Issue a corresponding equipment malfunction alarm based on the direction of deviation. If the value is greater than 0, a blockage alarm is triggered. If the value is less than 0, an alarm for uncontrolled material feeding is triggered.

Images