Lithium ore lithium extraction rotary kiln ladder air duct heat transfer structure optimization method and system

Abstract

The application discloses a lithium ore lithium extraction rotary kiln ladder air duct heat transfer structure optimization method and system, relates to the intelligent control technical field, through real-time collection of multi-source sensing data in the rotary kiln ladder air duct, combining the geometric parameters of the ladder air duct, using the principal component analysis method to extract the coupling characteristics from the multi-source sensing data, and based on the coupling characteristics and the multi-source sensing data, a dynamic heat transfer resistance coefficient matrix is constructed, the coupling characteristics and the dynamic heat transfer resistance coefficient matrix are input into the pre-trained gradient boosting regression tree model, and the real-time effective heat transfer area proportion of each air duct level is output, based on the real-time effective heat transfer area proportion, the adaptive particle swarm optimization algorithm is used to optimize the ladder air duct parameters, the optimized ladder air duct parameters are used for real-time adjustment of the ladder air duct parameters in the rotary kiln through the PID controller, and a closed loop control is formed; the heat transfer efficiency in the calcination process can be greatly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to a method and system for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln. Background Technology

[0002] Rotary kilns are key equipment in the lithium extraction process from lithium ore. They use high temperatures to heat the ore, causing a chemical reaction and extracting the lithium. However, due to the uneven distribution of airflow and heat transfer within the rotary kiln, some areas are often too hot while others are too cold, resulting in uneven heat transfer and affecting the calcination efficiency of the lithium ore.

[0003] Therefore, stepped air duct technology can generally be introduced to assist in airflow guidance and heat transfer, and its geometry directly affects the flow of air and the efficiency of heat transfer. However, the stepped air ducts of traditional rotary kilns are usually fixed in design, lacking flexibility, which makes it impossible to adjust in a timely manner according to changes in airflow during actual operation; that is, the stepped air ducts of current rotary kilns usually adopt a fixed structural design and are not dynamically adjusted according to real-time operating conditions.

[0004] Therefore, this invention proposes a method and system for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a method and system for optimizing the heat transfer structure of the stepped air duct in a rotary kiln for lithium extraction, which can greatly improve the heat transfer efficiency during the calcination process.

[0006] To achieve the above objectives, an optimization method for the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln is proposed, including the following steps:

[0007] Step 1: Real-time acquisition of multi-source sensor data within the rotary kiln's stepped air duct;

[0008] Step 2: Combining the geometric parameters of the stepped air duct, principal component analysis is used to extract coupling features from the multi-source sensor data, and based on the coupling features and multi-source sensor data, a dynamic heat transfer resistance coefficient matrix is ​​constructed.

[0009] Step 3: Input the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, and output the real-time effective heat transfer area ratio of each duct level.

[0010] Step 4: Based on the real-time effective heat transfer area ratio, the parameters of the stepped air duct are optimized using an adaptive particle swarm optimization algorithm;

[0011] Step 5: Adjust the optimized stepped air duct parameters in the rotary kiln in real time using a PID controller to form a closed-loop control.

[0012] The real-time acquisition of multi-source sensor data within the stepped air duct of the rotary kiln includes the following steps:

[0013] Step 11: Install physical sensors on each step to collect various process parameters and material filling rates;

[0014] Step 12: All physical sensors are connected to the database backend through the industrial IoT communication network, and collect the corresponding process parameters or material filling rate according to a unified preset sampling frequency, and then send the process parameters or material filling rate to the database backend.

[0015] Step 13: During the process of collecting process parameters and material filling rate in the database background, a unified timestamp is attached to each physical sensor data; the process parameters, material filling rate and geometric parameters of each timestamp are uniformly aggregated into a multi-dimensional data matrix, and all multi-dimensional data matrices are structured and stored as multi-source sensor data in chronological order.

[0016] The method of uniformly aggregating data into a multi-dimensional data matrix and then structuring and storing all multi-dimensional data matrices in chronological order as multi-source sensor data is as follows:

[0017] Process parameters and material filling rates from various physical sensors are arranged in a time series to form a multidimensional data matrix. Each row in the multidimensional data matrix represents a physical sensor, and each column represents the duct level of that physical sensor in the stepped air duct. mark; This represents the measurement value of the i-th physical sensor at time t in the j-th duct layer;

[0018] The multidimensional data matrix under each timestamp is stored in the database backend in chronological order.

[0019] The extraction of coupling features from multi-source sensor data using principal component analysis, incorporating the geometric parameters of the stepped air duct, includes the following steps:

[0020] Step 211: Read the geometric parameters of the stepped air duct from the calcination control backend in advance, and construct a geometric parameter vector from all geometric parameters;

[0021] Step 212: Represent the combination of each duct level and each physical sensor as a candidate parameter; for each candidate parameter and geometric parameter vector, construct an extended feature vector for each time step; and assemble all geometric parameters into a key parameter set.

[0022] Step 213: Combine the extended feature vectors of all candidate parameters at each time step into an extended feature matrix;

[0023] Step 214: Standardize each row of the extended feature matrix Y(t) to obtain the standardized extended feature matrix, labeled as... ;

[0024] Step 215: Collect standardized extended feature matrices from multiple time points during the historical operation of the stepped air duct. This forms a three-dimensional historical matrix; where N is the number of times data was collected.

[0025] Step 216: Calculate the global mean and covariance matrix based on the historical 3D matrix;

[0026] Step 217: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues. to and the corresponding feature vectors to All eigenvectors constitute the principal component matrix. ;

[0027] Step 218: For the standardized extended feature matrix at time t The data is projected using a fixed global model, and the projection result of each candidate parameter reflects its wind tunnel level in the global principal component space.

[0028] Step 219: In the projection results of each candidate parameter, analyze the loading of the first component of the corresponding extended vector in each principal component, and preset the loading threshold. If the absolute value of the loading in more than a preset number of principal components is greater than the loading threshold, then the candidate parameter is added to the key parameter set; the final key parameter set constitutes the coupling feature.

[0029] The construction of the dynamic heat transfer resistance coefficient matrix based on coupling characteristics and multi-source sensor data includes the following steps:

[0030] Step 221: Set the dynamic heat transfer resistance coefficient matrix to R(t);

[0031] Step 222: For the j-th duct level, read the type of physical sensor contained in the j-th duct level from the key parameter set of coupling features, and then combine it with the material filling rate of the j-th duct level and the geometric parameters of the stepped duct to form the key parameters.

[0032] Step 223: Describe the motion characteristics of airflow in the stepped duct using fluid dynamics equations, and then combine the heat conduction model to describe the heat transfer based on the temperature distribution, thereby obtaining the heat transfer resistance coefficients for any two duct levels.

[0033] Step 224: Collect the actual heat transfer resistance coefficients between each duct level during the historical operation of the stepped duct, and use regression analysis or neural networks to fit the parameters in the calculation equation of the heat transfer resistance coefficient.

[0034] The process of inputting the coupling characteristics and dynamic heat transfer resistance coefficient matrix into a pre-trained gradient boosting regression tree model and outputting the real-time effective heat transfer area ratio of each duct level includes the following steps:

[0035] Step 31: Collect the coupling feature vector and heat transfer resistance coefficient matrix at each moment during the operation of the historical stepped air duct to form the training input sample, and the real-time effective heat transfer area ratio of each air duct level to form the training label.

[0036] Step 32: Construct a gradient boosting regression tree model containing multiple decision trees;

[0037] The gradient boosting regression tree model is composed of multiple decision trees.

[0038] Each decision tree consists of multiple branch nodes and leaf nodes. Each branch node splits according to the dimension of the input features. The goal is to make the mean of each split subset as similar as possible in the output label, i.e. the effective heat transfer area ratio. Each leaf node contains the final regression value, i.e. the model's prediction value for the input features.

[0039] Each decision tree reduces prediction error by fitting the residual of the previous decision tree, and the output of each decision tree is weighted and summed with the output of the previous decision tree to obtain the final prediction result.

[0040] The input to the gradient boosting regression tree model is a combination of coupled feature vectors and dynamic heat transfer resistance coefficient matrix. The output of the gradient boosting regression tree model is a continuous value, representing the real-time effective heat transfer area ratio of each duct level.

[0041] Step 33: Train the gradient boosting regression tree model using the training input samples and training labels;

[0042] The method for training the gradient boosting regression tree model is as follows:

[0043] The gradient boosting regression tree model is initialized by setting it to a constant value, using the average value of the effective heat transfer area ratio in the training data as the initial value.

[0044] For each training input sample, calculate the current residual of the gradient boosting regression tree model. The residual is the difference between the actual value and the predicted value.

[0045] Each round of training involves building a new decision tree to fit the current residual. The new decision tree optimizes the residual, making the output value of the gradient boosting regression tree model closer to the actual value of the training label.

[0046] The training of each decision tree depends on the error of the previous training round. The decision tree finds the best split point by dividing the training input samples, so that the output error on each branch is minimized.

[0047] The final output of the gradient boosting regression tree model is a weighted sum of all decision trees.

[0048] Step 34: After the gradient boosting regression tree model is trained, real-time target input samples are collected by collecting real-time coupling features and dynamic heat transfer resistance coefficient matrix. The target input samples are then input into the gradient boosting regression tree model to obtain the real-time effective heat transfer area ratio of each duct level output by the gradient boosting regression tree model.

[0049] The optimization of the stepped air duct parameters based on the real-time effective heat transfer area ratio using the adaptive particle swarm optimization algorithm includes the following steps:

[0050] Step 41: Construct the target optimization function based on the real-time effective heat transfer area ratio, and select the optimized stepped air duct parameters;

[0051] The optimized stepped air duct parameters include the gas injection valve opening G and the guide vane angle θ;

[0052] Step 42: Design the fitness function based on the objective optimization function;

[0053] Step 43: Initialize the particle swarm for the adaptive particle swarm algorithm;

[0054] The method for initializing the particle swarm in the adaptive particle swarm algorithm is as follows:

[0055] Generate a set of random particles, each particle representing a set of stepped air duct parameters;

[0056] A motion range is preset for each parameter of each particle so that the particle is confined to the motion range during each generation update;

[0057] The position of each particle is updated using the particle update formula.

[0058] Step 44: Apply the adaptive particle swarm optimization algorithm to optimize the position of the particle swarm, gradually approach the optimal solution, and obtain the optimized stepped air duct parameters.

[0059] The method of adjusting the optimized stepped air duct parameters in real time within the rotary kiln using a PID controller to form a closed-loop control is as follows:

[0060] The optimized stepped air duct parameters are input into the PID controller, which uses these parameters as the target adjustment value. The PID controller then adjusts the stepped air duct parameters in the rotary kiln to the target adjustment value based on this target value.

[0061] A heat transfer structure optimization system for the stepped air duct of a lithium extraction rotary kiln is proposed, which includes a multi-source data collection module, a feature extraction module, a heat transfer area ratio output module, and an air duct parameter optimization module; the modules are connected to each other electrically.

[0062] The multi-source data collection module collects multi-source sensor data in the stepped air duct of the rotary kiln in real time and sends the multi-source sensor data to the feature extraction module.

[0063] The feature extraction module, in combination with the geometric parameters of the stepped air duct, uses principal component analysis to extract coupling features from multi-source sensor data, and constructs a dynamic heat transfer resistance coefficient matrix based on the coupling features and multi-source sensor data, and sends the coupling features and dynamic heat transfer resistance coefficient matrix to the heat transfer area ratio output module.

[0064] The heat transfer area ratio output module inputs the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, outputs the real-time effective heat transfer area ratio of each duct level, and sends the real-time effective heat transfer area ratio to the duct parameter optimization module.

[0065] The air duct parameter optimization module optimizes the stepped air duct parameters based on the real-time effective heat transfer area ratio and uses an adaptive particle swarm optimization algorithm. The optimized stepped air duct parameters are then adjusted in real time by a PID controller to form a closed-loop control.

[0066] Compared with the prior art, the beneficial effects of the present invention are:

[0067] This invention acquires multi-source sensor data in real time within the stepped air duct of a rotary kiln. Combining this data with the geometric parameters of the stepped air duct, principal component analysis (PCA) is used to extract coupling features from the multi-source sensor data. Based on these coupling features and the multi-source sensor data, a dynamic heat transfer resistance coefficient matrix is ​​constructed. This matrix is ​​then input into a pre-trained gradient boosting regression tree model, outputting the real-time effective heat transfer area ratio for each air duct level. Based on this ratio, an adaptive particle swarm optimization (PSO) algorithm is used to optimize the stepped air duct parameters. The optimized parameters are then used in real-time by a PID controller to adjust the stepped air duct parameters within the rotary kiln, forming a closed-loop control. By extracting coupling features from the multi-source sensor data using PCA and combining this with the dynamic heat transfer resistance coefficient matrix, a feature model related to the air duct parameters is constructed. The adaptive particle swarm optimization algorithm, based on real-time feedback data, automatically adjusts the stepped air duct parameters within the rotary kiln to optimize heat transfer efficiency, thereby ensuring more uniform airflow distribution and heat transfer, and improving overall heat transfer efficiency. Attached Figure Description

[0068] Figure 1 This is a diagram of an internal heat transfer rotary kiln and its internal structure.

[0069] Figure 2 This is a flowchart of the method for optimizing the heat transfer structure of the stepped air duct in the lithium extraction rotary kiln of the present invention, in Embodiment 1 of the present invention;

[0070] Figure 3 This is a diagram showing the module connection relationship of the optimized heat transfer structure system of the lithium extraction rotary kiln stepped air duct in Embodiment 2 of the present invention. Detailed Implementation

[0071] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] In the traditional process of lithium extraction rotary kilns, external heating indirectly transfers heat through the outer cylinder of the kiln. Heat must penetrate steel plates typically over 50mm thick to reach the internal material, resulting in a long heat transfer path, high thermal resistance, high energy consumption, and bulky equipment with high manufacturing and maintenance costs. To address this, a technology was introduced that incorporates multi-layered stepped airflow channels on the inner wall of the kiln. This allows high-temperature flue gas to flow directly through the 8-12mm thick channel walls for heat exchange with the material, transforming the traditional "external wall heat conduction" mode into "embedded radiation-convection composite heat transfer." This structure increases the effective heat transfer area by 50%-100% and enhances the uniformity of gas-solid heat exchange through turbulence created by the stepped drop, improving heat exchange efficiency by over 50%. Simultaneously, it reduces steel plate thickness, achieving lightweight equipment and overall energy savings of over 40%. Figure 1 The diagram shows the internal heat transfer rotary kiln and its internal structure; the diagram shows the cylinder of the stepped air duct of the internal heat transfer rotary kiln, as well as example cross-sectional views of the inlet and outlet of the stepped air duct.

[0073] However, the geometric parameters of stepped ducts, such as their inclination angle, spacing, and wall thickness, are closely coupled with process parameters such as airflow pressure, flue gas velocity, local temperature, oxygen concentration, carbon dioxide concentration, and material filling rate. Improper parameter matching can lead to uneven airflow distribution, localized thermal stress concentration, and even duct structure deformation. For example, excessively large step spacing weakens turbulence, while excessively small inclination angles easily cause ash accumulation and blockage. Furthermore, the thin-walled structure's sensitivity to temperature fluctuations requires real-time dynamic adjustment of flue gas flow to balance heat transfer efficiency and structural reliability. Therefore, maximizing energy utilization efficiency while ensuring long-term stable operation of the equipment requires multi-parameter synergistic optimization, combined with heat transfer efficiency prediction and structural mechanics analysis. This is the core challenge for the industrial application of stepped duct technology.

[0074] Example 1

[0075] like Figure 2 As shown, the method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln includes the following steps:

[0076] Step 1: Real-time acquisition of multi-source sensor data within the rotary kiln's stepped air duct;

[0077] Step 2: Combining the geometric parameters of the stepped air duct, principal component analysis is used to extract coupling features from the multi-source sensor data, and based on the coupling features and multi-source sensor data, a dynamic heat transfer resistance coefficient matrix is ​​constructed.

[0078] Step 3: Input the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, and output the real-time effective heat transfer area ratio of each duct level.

[0079] Step 4: Based on the real-time effective heat transfer area ratio, the parameters of the stepped air duct are optimized using an adaptive particle swarm optimization algorithm;

[0080] Step 5: Adjust the optimized stepped air duct parameters in the rotary kiln in real time using a PID controller to form a closed-loop control.

[0081] In an embodiment of the present invention, the real-time acquisition of multi-source sensor data within the stepped air duct of the rotary kiln includes the following steps:

[0082] Step 11: Install physical sensors on each step to collect various process parameters and material filling rates;

[0083] Specifically, the material filling rate refers to the ratio between the volume occupied by materials such as ore particles in the rotary kiln and the available space volume in the kiln.

[0084] During the operation of a rotary kiln, material continuously enters from the feed inlet and moves within the kiln. By the time it reaches the outlet, some of the material has been converted or extracted into components such as lithium, while the remaining ore is discharged. Changes in the feed rate and output lead to variations in the material filling rate, which is therefore constantly changing.

[0085] The sensor that collects the material filling rate can acquire data in real time using a magnetic sensor. The principle behind this magnetic sensor is that the material inside the rotary kiln generally contains magnetic components. The magnetic sensor can detect the distribution of these magnetic components, and when the material passes through the sensor, the sensor's signal strength changes. By analyzing the signal changes, the material's accumulation and filling degree can be determined.

[0086] The physical sensors and their applications include:

[0087] Temperature sensors are used to measure local temperature, pressure sensors are used to monitor airflow pressure, airflow velocity sensors are used to detect combustion gas flow rate, and gas concentration sensors are set in some key areas to detect the concentration of gases such as oxygen and carbon dioxide to reflect the combustion state.

[0088] Furthermore, all physical sensors are industrial-grade high-temperature anti-interference equipment with high sampling frequencies, such as once per second or faster, to meet real-time monitoring requirements. The air duct level and number of each physical sensor are recorded in the system, forming a sensor set.

[0089] Step 12: All physical sensors are connected to the database backend through the industrial IoT communication network, and collect the corresponding process parameters or material filling rate according to a unified preset sampling frequency, and then send the process parameters or material filling rate to the database backend.

[0090] In a preferred embodiment, for real-time process parameters or material filling rate, preliminary filtering and noise reduction preprocessing can be performed in advance by using embedded multi-dimensional filtering algorithms such as Kalman filtering or low-pass filtering, and by adjusting the parameters adaptively according to the predefined process stage.

[0091] Step 13: During the process of collecting process parameters and material filling rate in the database background, a unified timestamp is attached to each physical sensor data; the process parameters, material filling rate and geometric parameters of each timestamp are uniformly aggregated into a multi-dimensional data matrix, and all multi-dimensional data matrices are structured and stored as multi-source sensor data in chronological order.

[0092] Specifically, the method of uniformly aggregating data into a multi-dimensional data matrix and then structuring and storing all multi-dimensional data matrices in chronological order as multi-source sensor data is as follows:

[0093] Process parameters and material filling rates from various physical sensors are arranged in a time series to form a multidimensional data matrix. Each row in the multidimensional data matrix represents a physical sensor, and each column represents the duct level of that physical sensor in the stepped air duct. mark; This represents the measurement value of the i-th physical sensor at time t in the j-th duct layer;

[0094] The multidimensional data matrix under each timestamp is stored in the database backend in chronological order.

[0095] In a further preferred embodiment, in order to ensure data accuracy and long-term stability, it is also necessary to periodically calibrate the process parameters and material filling rate. Specifically, this can be achieved by using correction models such as linear regression or neural network correction algorithms built from historical and real-time data to correct sensor drift and abrupt changes in data.

[0096] Furthermore, the extraction of coupling features from multi-source sensor data using principal component analysis, in conjunction with the geometric parameters of the stepped air duct, includes the following steps:

[0097] Step 211: Read the geometric parameters of the stepped air duct from the calcination control backend in advance, and construct a geometric parameter vector from all geometric parameters;

[0098] The geometric parameters include the tilt angle that affects the airflow direction and heat transfer efficiency, the distance between the stepped layers in the duct, and the thickness of the duct wall that affects the heat conduction effect and the thermal stability of the equipment.

[0099] Step 212: Represent the combination of each duct level and each physical sensor as a candidate parameter; for each candidate parameter and geometric parameter vector, construct an extended feature vector for each time step; and assemble all geometric parameters into a key parameter set.

[0100] Specifically, the method for constructing the extended feature vector is as follows:

[0101] Label the geometric parameter vector as ;

[0102] Then the extended feature vector corresponding to the candidate parameters formed by the combination of the i-th physical sensor at the j-th duct level. for This allows us to consider not only the fluctuations of the candidate parameters themselves in subsequent analyses, but also to introduce geometric conditions as contextual information.

[0103] Step 213: Combine the extended feature vectors of all candidate parameters at each time step into an extended feature matrix;

[0104] The extended feature matrix Y(t) is in the form of:

[0105] ,in, Let K represent the number of all geometric parameters, M be the total number of all candidate parameters, and K be the denoting value of K. ;

[0106] Step 214: Standardize each row of the extended feature matrix Y(t) to obtain the standardized extended feature matrix, labeled as... ;

[0107] Step 215: Collect standardized extended feature matrices from multiple time points during the historical operation of the stepped air duct. This forms a three-dimensional historical matrix; where N is the number of times data was collected.

[0108] Step 216: Calculate the global mean and covariance matrix based on the historical 3D matrix;

[0109] Specifically, the global mean The calculation formula is:

[0110] ;

[0111] The formula for calculating the covariance matrix C is as follows:

[0112] ;

[0113] Step 217: Perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues. to and the corresponding feature vectors to All eigenvectors constitute the principal component matrix. ;

[0114] Step 218: For the standardized extended feature matrix at time t The data is projected using a fixed global model, and the projection result of each candidate parameter reflects its wind tunnel level in the global principal component space.

[0115] Specifically, the global model is as follows:

[0116] ;

[0117] in, yes By multiplying with a vector of all 1 rows, and then copying the resulting mean vector M times, the result is obtained... A matrix of the same size is used to align each candidate parameter vector with the global mean to eliminate the influence of the mean; thus, Z(t) is formed by first decentering the extended feature vector of each candidate parameter, and then using a fixed principal component matrix. These decentralized vectors are projected onto the principal component space, where T represents the transpose; each column of the obtained Z(t) is a low-dimensional representation of the candidate parameters in the principal component space, thereby capturing the main coupling features of the candidate parameters under fixed geometric conditions and ensuring a unified and stable semantic at different time points;

[0118] Step 219: In the projection results of each candidate parameter, analyze the loading of the first component of the corresponding extended vector in each principal component, and preset the loading threshold. If the absolute value of the loading in more than a preset number of principal components is greater than the loading threshold, then the candidate parameter is added to the key parameter set; the final key parameter set constitutes the coupling feature.

[0119] The analysis of the loading of the first component of the corresponding extended vector in each principal component refers to examining the absolute value of each component in each eigenvector. A candidate parameter with a larger loading indicates that its value contributes significantly to the overall variation in that principal component, and thus it is considered that the candidate parameter has high importance under the current geometric conditions.

[0120] For example, if the screening results during operation show that the temperature sensor with candidate parameters (i1,j1) at the air duct level j1 and the airflow velocity sensor with candidate parameters (i2,j2) at the air duct level j2 both have high loads, then a set of key parameters can be output. ;

[0121] This allows for the selection of the combination of air duct level and physical sensor most affected by fixed geometric parameters from numerous data acquisition air duct levels and various types of physical sensors, reducing the difficulty of subsequent data processing.

[0122] In an embodiment of the present invention, constructing a dynamic heat transfer resistance coefficient matrix based on coupling characteristics and multi-source sensor data includes the following steps:

[0123] Step 221: Set the dynamic heat transfer resistance coefficient matrix as R(t), where R(t) is a matrix related to time t, reflecting the heat transfer resistance of different air duct levels in the rotary kiln;

[0124] Specifically, the elements of matrix R(t) This represents the heat transfer resistance coefficient between the j1st and j2nd duct levels at time t.

[0125] Step 222: For the j-th duct level, read the type of physical sensor contained in the j-th duct level from the key parameter set of coupling features, and then combine it with the material filling rate of the j-th duct level and the geometric parameters of the stepped duct to form the key parameters.

[0126] Step 223: Describe the motion characteristics of airflow in the stepped duct using fluid dynamics equations, and then combine the heat conduction model to describe the heat transfer based on the temperature distribution, thereby obtaining the heat transfer resistance coefficients for any two duct levels.

[0127] Specifically, the heat conduction model is such as the Fourier heat conduction equation;

[0128] The heat transfer resistance coefficient of any two air duct levels It can be represented as:

[0129] ,in, It incorporates the effects of fluid dynamics and heat conduction, and the specific forms and parameters are fitted based on historical data; These represent the key parameters of the j1st and j2nd air duct levels, respectively. These are the correction factors that need to be fitted;

[0130] Step 224: Collect the actual heat transfer resistance coefficients between each duct level during the historical operation of the stepped duct, and use regression analysis, neural networks or other optimization algorithms to fit the parameters in the calculation equation of the heat transfer resistance coefficient.

[0131] In a further preferred embodiment, since the operation of the rotary kiln is dynamic, the heat transfer resistance coefficient matrix needs to be dynamically adjusted over time. Therefore, a new heat transfer resistance coefficient can be dynamically calculated based on the process parameters collected by various physical sensors at each air duct level, so that the heat transfer resistance coefficient matrix can be corrected according to the current working conditions. This correction method can also be based on historical data and statistical fitting using regression analysis, neural networks, or other optimization algorithms.

[0132] Furthermore, the step of inputting the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model and outputting the real-time effective heat transfer area ratio of each duct level includes the following steps:

[0133] Step 31: Collect the coupling feature vector and heat transfer resistance coefficient matrix at each moment during the operation of the historical stepped air duct to form the training input sample, and the real-time effective heat transfer area ratio of each air duct level to form the training label.

[0134] Specifically, the coupling feature vector is the parameter value of the key parameters of each air duct level at each moment during the historical operation period, extracted through the above-mentioned process of extracting coupling features.

[0135] The heat transfer resistance coefficient matrix is ​​obtained through the above-mentioned heat transfer resistance coefficient matrix extraction process.

[0136] The real-time effective heat transfer area ratio represents the heat transfer capacity of each air duct level at that point in time. The actual value of this effective heat transfer area ratio is generally obtained through experimental data or numerical simulation.

[0137] Step 32: Construct a gradient boosting regression tree model containing multiple decision trees;

[0138] Specifically, the gradient boosting regression tree model is composed of multiple decision trees;

[0139] Each decision tree consists of multiple branch nodes and leaf nodes. Each branch node splits according to a certain dimension of the input feature. The goal is to make the mean of each split subset as similar as possible in the output label, i.e. the effective heat transfer area ratio. Each leaf node contains the final regression value, i.e. the model's prediction value for a specific input feature.

[0140] Each decision tree gradually reduces prediction error by fitting the residual of the previous decision tree. The output of each decision tree is weighted and summed with the output of the previous decision tree to obtain the final prediction result.

[0141] The input to the gradient boosting regression tree model is a combination of coupled feature vectors and dynamic heat transfer resistance coefficient matrix, and the output of the gradient boosting regression tree model is a continuous value, representing the real-time effective heat transfer area ratio of each duct level.

[0142] Step 33: Train the gradient boosting regression tree model using the training input samples and training labels;

[0143] Specifically, the method for training the gradient boosting regression tree model is as follows:

[0144] The gradient boosting regression tree model is initialized by setting it to a constant value, using the average value of the effective heat transfer area ratio in the training data as the initial value.

[0145] For each training input sample, calculate the current residual of the gradient boosting regression tree model. The residual is the difference between the actual value and the predicted value.

[0146] Each round of training involves building a new decision tree to fit the current residual. The new decision tree optimizes the residual, making the output value of the gradient boosting regression tree model closer to the actual value of the training label.

[0147] The training of each decision tree depends on the error of the previous training round. The decision tree finds the best split point by dividing the training input samples, so that the output error on each branch is minimized.

[0148] The final output of the gradient boosting regression tree model is a weighted sum of all decision trees.

[0149] Step 34: After the gradient boosting regression tree model is trained, real-time target input samples are formed by collecting real-time coupling features and dynamic heat transfer resistance coefficient matrix. The target input samples are then input into the gradient boosting regression tree model to obtain the real-time effective heat transfer area ratio of each duct level output by the gradient boosting regression tree model.

[0150] It should be noted that the first extracted coupling feature vector can compress highly correlated multidimensional features into a few important principal components, greatly reducing the dimensionality of the input data. Secondly, the dynamic heat transfer resistance coefficient matrix integrates the influence of duct geometry and physical characteristics on heat transfer, and can comprehensively consider the heat transfer relationship between different duct levels without the need to input detailed parameters of each physical characteristic of each duct level separately. By combining the coupling feature vector and the dynamic heat transfer resistance coefficient matrix, the complex coupling relationship inside the system can be captured. That is, the coupling feature vector captures the main changing trend of sensor data, while the dynamic heat transfer resistance coefficient matrix can combine the geometric structure, airflow and heat transfer state of the duct level, thereby improving the prediction accuracy of the real-time effective heat transfer area ratio.

[0151] Furthermore, the optimization of the stepped air duct parameters based on the real-time effective heat transfer area ratio using an adaptive particle swarm optimization algorithm includes the following steps:

[0152] Step 41: Construct the target optimization function based on the real-time effective heat transfer area ratio, and select the optimized stepped air duct parameters;

[0153] In an embodiment of the present invention, the expression of the objective optimization function is: Where J represents the number of all air duct levels; The effective heat transfer area ratio of the j-th duct level is predicted by the gradient boosting regression tree model; the optimization objective of the objective optimization function is to maximize the function value of the objective optimization function.

[0154] The optimized stepped air duct parameters in the embodiments of the present invention include the gas injection valve opening degree G and the guide vane angle θ.

[0155] The opening degree of the gas injection valve directly affects the flow rate and distribution of the gas inside the kiln, thereby affecting airflow dynamics and heat transfer.

[0156] The angle of the guide plate affects the direction of airflow, which in turn affects the distribution and transfer of heat within the kiln.

[0157] Step 42: Design the fitness function based on the objective optimization function;

[0158] In an embodiment of the present invention, the fitness function is designed as follows:

[0159] fitness function It can be represented in the following form:

[0160] ,in, and These represent the initial gas injection valve opening and the guide vane angle, respectively, indicating the initial or reference state of the rotary kiln. 'This is a penalty coefficient used to control the magnitude of changes in the gas injection valve opening and guide vane angle during the optimization process, to avoid over-adjustment;

[0161] The fitness function is used in the particle swarm optimization algorithm to evaluate the quality of the solution represented by each particle. Therefore, the fitness function aims to maximize the effective heat transfer area ratio while limiting the changes in the gas injection valve opening and the guide vane angle, so that the parameter adjustment during the optimization process is not too drastic and the rotary kiln is kept running stably.

[0162] Step 43: Initialize the particle swarm for the adaptive particle swarm algorithm;

[0163] In an embodiment of the present invention, the method for initializing the particle swarm of the adaptive particle swarm algorithm is as follows:

[0164] Generate a set of random particles, each particle representing a set of stepped air duct parameters, that is, each particle represents a possible solution;

[0165] A motion range is preset for each parameter of each particle so that the particle is limited to a reasonable motion range during each generation update. In this embodiment, the gas injection valve opening is set to a reasonable range, such as between 0% and 100%; the guide vane angle is set to a range between 0 and 90 degrees.

[0166] The position of each particle is updated using the following particle update formula;

[0167] ,in, and represents the gas injection valve opening at time t+1 and time t, respectively, while represents This represents the updated velocity value of the particles at the opening of the gas injection valve. and These represent the deflector angles at time t+1 and time t, respectively. This indicates the speed update value in the direction of the deflector angle.

[0168] Step 44: Apply the adaptive particle swarm optimization algorithm to optimize the position of the particle swarm, gradually approach the optimal solution, and obtain the optimized stepped air duct parameters;

[0169] In an embodiment of the present invention, the process of applying the adaptive particle swarm optimization algorithm to optimize the position of the particle swarm and gradually approach the optimal solution is as follows:

[0170] The velocity and position of each particle will be updated based on the current fitness function, the individual best position, and the global best position according to the following formula;

[0171] The speed update formula is:

[0172]

[0173] ,

[0174] in, Inertial weights control the effect of the particle's own velocity. and These are the acceleration constants, which determine the degree to which a particle approaches its individual optimal position and its global optimal position. The optimal position for the individual gas injection valve opening. The optimal position for the angle of the individual air deflector; and These are randomly generated numbers to ensure the randomness of the search; and These are the globally optimal positions for the opening of the gas injection valve and the globally optimal positions for the guide vane angle, respectively.

[0175] It should be noted that the individual best position is the best solution found by each particle in the optimization history, that is, the position with the best fitness that the particle has experienced; while the global best position is the position of the best solution in the particle swarm, which guides the search process of the particle swarm based on the overall experience of the particle swarm. The global best position is updated as follows: in each iteration, each particle in the particle swarm compares the individual best position with the individual best position of other particles. If the fitness value of a particle's individual best position is better than the current global best fitness, then the global best position is updated to the individual best position of that particle.

[0176] It is easy to understand that in the above speed update formula, the first term, i.e. and Both terms are inertial terms, which can be seen as particles continuing to move forward based on their historical velocity "inertia," maintaining their original motion trend and preventing them from stopping completely or losing their momentum; while the second term, namely... and All of these demonstrate how particles move towards their optimal individual positions, while the third formula, namely... and All of these demonstrate how particles move toward the global optimal position of the entire group;

[0177] Furthermore, the optimized stepped air duct parameters are adjusted in real time by a PID controller to form a closed-loop control method:

[0178] The optimized stepped air duct parameters are input into the PID controller, which uses these parameters as the target adjustment value. The PID controller then adjusts the stepped air duct parameters in the rotary kiln to the target adjustment value based on this target value.

[0179] Example 2

[0180] like Figure 3 As shown, the lithium extraction rotary kiln stepped air duct heat transfer structure optimization system includes a multi-source data collection module, a feature extraction module, a heat transfer area ratio output module, and an air duct parameter optimization module; wherein, the modules are connected to each other by electrical means.

[0181] The multi-source data collection module collects multi-source sensor data in the stepped air duct of the rotary kiln in real time and sends the multi-source sensor data to the feature extraction module.

[0182] The feature extraction module, in combination with the geometric parameters of the stepped air duct, uses principal component analysis to extract coupling features from multi-source sensor data, and constructs a dynamic heat transfer resistance coefficient matrix based on the coupling features and multi-source sensor data, and sends the coupling features and dynamic heat transfer resistance coefficient matrix to the heat transfer area ratio output module.

[0183] The heat transfer area ratio output module inputs the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, outputs the real-time effective heat transfer area ratio of each duct level, and sends the real-time effective heat transfer area ratio to the duct parameter optimization module.

[0184] The air duct parameter optimization module optimizes the stepped air duct parameters based on the real-time effective heat transfer area ratio and uses an adaptive particle swarm optimization algorithm. The optimized stepped air duct parameters are then adjusted in real time by a PID controller to form a closed-loop control.

[0185] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0186] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0187] The preset parameters or preset thresholds mentioned above are all set by those skilled in the art based on actual conditions or obtained through large-scale data simulation.

[0188] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for optimizing the heat transfer structure of a stepped air duct in a rotary kiln for lithium extraction, characterized in that, Includes the following steps: Step 1: Real-time acquisition of multi-source sensor data within the rotary kiln's stepped air duct; Step 2: Combining the geometric parameters of the stepped air duct, principal component analysis is used to extract coupling features from the multi-source sensor data, and based on the coupling features and multi-source sensor data, a dynamic heat transfer resistance coefficient matrix is ​​constructed. Step 3: Input the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, and output the real-time effective heat transfer area ratio of each duct level. Step 4: Based on the real-time effective heat transfer area ratio, the parameters of the stepped air duct are optimized using an adaptive particle swarm optimization algorithm; Step 5: The optimized stepped air duct parameters are adjusted in real time using a PID controller to form a closed-loop control; the real-time acquisition of multi-source sensor data in the rotary kiln stepped air duct includes the following steps: Step 11: Install physical sensors on each step to collect various process parameters and material filling rates; Step 12: All physical sensors are connected to the database backend through the industrial IoT communication network, and collect the corresponding process parameters or material filling rate according to a unified preset sampling frequency, and then send the process parameters or material filling rate to the database backend. Step 13: During the process of collecting process parameters and material filling rate in the database background, a unified timestamp is attached to each physical sensor data; the process parameters, material filling rate and geometric parameters of each timestamp are uniformly aggregated into a multi-dimensional data matrix, and all multi-dimensional data matrices are structured and stored as multi-source sensor data in chronological order. The construction of the dynamic heat transfer resistance coefficient matrix based on coupling characteristics and multi-source sensor data includes the following steps: Step 221: Set the dynamic heat transfer resistance coefficient matrix to R(t); Step 222: For the j-th duct level, read the type of physical sensor contained in the j-th duct level from the key parameter set of coupling features, and then combine it with the material filling rate of the j-th duct level and the geometric parameters of the stepped duct to form the key parameters. Step 223: Describe the motion characteristics of airflow in the stepped duct using fluid dynamics equations, and then combine the heat conduction model to describe the heat transfer based on the temperature distribution, thereby obtaining the heat transfer resistance coefficients for any two duct levels. Step 224: Collect the actual heat transfer resistance coefficients between each duct level during the historical operation of the stepped duct, and use regression analysis or neural networks to fit the parameters in the calculation equation of the heat transfer resistance coefficient. The optimization of the stepped air duct parameters based on the real-time effective heat transfer area ratio using the adaptive particle swarm optimization algorithm includes the following steps: Step 41: Construct the target optimization function based on the real-time effective heat transfer area ratio, and select the optimized stepped air duct parameters; The optimized stepped air duct parameters include the gas injection valve opening G and the guide vane angle θ; Step 42: Design the fitness function based on the objective optimization function; Step 43: Initialize the particle swarm for the adaptive particle swarm algorithm; Step 44: Apply the adaptive particle swarm optimization algorithm to optimize the position of the particle swarm and obtain the optimized stepped air duct parameters.

2. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 1, characterized in that, The method of uniformly aggregating data into a multi-dimensional data matrix and then structuring and storing all multi-dimensional data matrices in chronological order as multi-source sensor data is as follows: Process parameters and material filling rates from various physical sensors are arranged in a time series to form a multidimensional data matrix. Each row in the multidimensional data matrix represents a physical sensor, and each column represents the duct level of that physical sensor in the stepped air duct. mark; This represents the measurement value of the i-th physical sensor at time t in the j-th duct layer; The multidimensional data matrix under each timestamp is stored in the database backend in chronological order.

3. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 2, characterized in that, The extraction of coupling features from multi-source sensor data using principal component analysis, incorporating the geometric parameters of the stepped air duct, includes the following steps: Step 211: Read the geometric parameters of the stepped air duct from the calcination control backend in advance, and construct a geometric parameter vector from all geometric parameters; Step 212: Represent the combination of each duct level and each physical sensor as a candidate parameter; for each candidate parameter and geometric parameter vector, construct an extended feature vector for each time step; and assemble all geometric parameters into a key parameter set. Step 213: Combine the extended feature vectors of all candidate parameters at each time step into an extended feature matrix; Step 214: Standardize each row of the extended feature matrix Y(t) to obtain the standardized extended feature matrix, labeled as... ; Step 215: Collect standardized extended feature matrices from multiple time points during the historical operation of the stepped air duct. This forms a three-dimensional historical matrix; where N is the number of times data was collected. Step 216: Calculate the global mean and covariance matrix based on the historical 3D matrix; Step 217: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues. to and the corresponding feature vectors to All eigenvectors constitute the principal component matrix. ; Step 218: For the standardized extended feature matrix at time t The data is projected using a fixed global model, and the projection result of each candidate parameter reflects its wind tunnel level in the global principal component space. Step 219: In the projection results of each candidate parameter, analyze the loading of the first component of the corresponding extended vector in each principal component, and preset the loading threshold. If the absolute value of the loading in more than a preset number of principal components is greater than the loading threshold, then the candidate parameter is added to the key parameter set; the final key parameter set constitutes the coupling feature.

4. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 3, characterized in that, The process of inputting the coupling characteristics and dynamic heat transfer resistance coefficient matrix into a pre-trained gradient boosting regression tree model and outputting the real-time effective heat transfer area ratio of each duct level includes the following steps: Step 31: Collect the coupling feature vector and heat transfer resistance coefficient matrix at each moment during the operation of the historical stepped air duct to form the training input sample, and the real-time effective heat transfer area ratio of each air duct level to form the training label. Step 32: Construct a gradient boosting regression tree model containing multiple decision trees; Step 33: Train the gradient boosting regression tree model using the training input samples and training labels; Step 34: After the gradient boosting regression tree model is trained, real-time target input samples are collected by collecting real-time coupling features and dynamic heat transfer resistance coefficient matrix. The target input samples are then input into the gradient boosting regression tree model to obtain the real-time effective heat transfer area ratio of each duct level output by the gradient boosting regression tree model.

5. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 4, characterized in that, The gradient boosting regression tree model is composed of multiple decision trees. Each decision tree consists of multiple branch nodes and leaf nodes. Each branch node splits according to the dimension of the input features, and each leaf node contains the final regression value, which is the model's prediction of the input features. Each decision tree reduces prediction error by fitting the residual of the previous decision tree, and the output of each decision tree is weighted and summed with the output of the previous decision tree to obtain the final prediction result. The input to the gradient boosting regression tree model is a combination of coupled feature vectors and dynamic heat transfer resistance coefficient matrix. The output of the gradient boosting regression tree model is a continuous value, representing the real-time effective heat transfer area ratio of each duct level.

6. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 5, characterized in that, The method for training the gradient boosting regression tree model is as follows: The gradient boosting regression tree model is initialized by setting it to a constant value, using the average value of the effective heat transfer area ratio in the training data as the initial value. For each training input sample, calculate the current residual of the gradient boosting regression tree model. The residual is the difference between the actual value and the predicted value. Each training round involves building a new decision tree to fit the current residual, and the new decision tree optimizes the residual. The training of each decision tree depends on the error of the previous training round. The decision tree finds the best split point by dividing the training input samples, so that the output error on each branch is minimized. The final output of the gradient boosting regression tree model is a weighted sum of all decision trees.

7. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 6, characterized in that, The method for initializing the particle swarm in the adaptive particle swarm algorithm is as follows: Generate a set of random particles, each particle representing a set of stepped air duct parameters; A motion range is preset for each parameter of each particle so that the particle is confined to the motion range during each generation update; The position of each particle is updated using the particle update formula.

8. The method for optimizing the heat transfer structure of the stepped air duct in a lithium extraction rotary kiln according to claim 7, characterized in that, The method of adjusting the optimized stepped air duct parameters in real time within the rotary kiln using a PID controller to form a closed-loop control is as follows: The optimized stepped air duct parameters are input into the PID controller, which uses these parameters as the target adjustment value. The PID controller then adjusts the stepped air duct parameters in the rotary kiln to the target adjustment value based on this target value.

9. A system for optimizing the heat transfer structure of a stepped air duct in a rotary kiln for lithium extraction, used to implement the method for optimizing the heat transfer structure of a stepped air duct in a rotary kiln for lithium extraction as described in any one of claims 1-8, characterized in that... It includes a multi-source data collection module, a feature extraction module, a heat transfer area ratio output module, and a duct parameter optimization module; the modules are connected electrically. The multi-source data collection module collects multi-source sensor data in the stepped air duct of the rotary kiln in real time and sends the multi-source sensor data to the feature extraction module. The feature extraction module, in combination with the geometric parameters of the stepped air duct, uses principal component analysis to extract coupling features from multi-source sensor data, and constructs a dynamic heat transfer resistance coefficient matrix based on the coupling features and multi-source sensor data, and sends the coupling features and dynamic heat transfer resistance coefficient matrix to the heat transfer area ratio output module. The heat transfer area ratio output module inputs the coupling features and dynamic heat transfer resistance coefficient matrix into the pre-trained gradient boosting regression tree model, outputs the real-time effective heat transfer area ratio of each duct level, and sends the real-time effective heat transfer area ratio to the duct parameter optimization module. The air duct parameter optimization module optimizes the stepped air duct parameters based on the real-time effective heat transfer area ratio and uses an adaptive particle swarm optimization algorithm. The optimized stepped air duct parameters are then adjusted in real time by a PID controller to form a closed-loop control.

Images