High-precision self-adaptive control method and system for floating sealing device of lithium extraction rotary kiln

Abstract

The invention discloses a high-precision self-adaptive control method and system for a floating sealing device of a lithium extraction rotary kiln, and relates to the technical field of intelligent control, and the method comprises the steps: collecting the multi-modal characteristic parameters of the operation state of the rotary kiln, inputting the multi-modal characteristic parameters into a physical information neural network, solving a thermoelastic mechanical partial differential equation, and obtaining the high-precision self-adaptive control of the floating sealing device of the lithium extraction rotary kiln. The method comprises the steps of obtaining a deformation field prediction matrix, outputting a deformation field prediction matrix, solving a multi-target optimization model with minimum leakage rate and wear energy consumption as a joint target based on the deformation field prediction matrix, outputting a sealing compensation instruction vector, and driving a magneto-rheological-piezoelectric composite execution mechanism to complete instruction control of a sealing ring based on the sealing compensation instruction vector. Collecting feedback result data controlled by the instruction, adjusting the weight of each weight coefficient in the multi-objective optimization function on line by adopting a reinforcement learning algorithm based on the feedback result data, and feeding back the adjusted weight to the multi-objective optimization model; and the control stability of the rotary kiln under complex working conditions is ensured.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a high-precision adaptive control method and system for a floating sealing device of a lithium extraction rotary kiln. Background Art

[0002] In the lithium extraction process, the rotary kiln, as a core piece of equipment, needs to roast spodumene at high temperatures to achieve α-β crystal transformation. The kiln body undergoes dynamic deformation due to thermal expansion, mechanical vibration, and material impact, causing fluctuations in the gap between the sealing ring and the kiln body. Traditional sealing devices rely on fixed parameter control and are unable to adapt to deformation and changes in operating conditions in real time. This can easily lead to problems such as high-temperature flue gas leakage, increased energy consumption, and increased equipment wear, affecting lithium extraction efficiency and production safety. Therefore, intelligent dynamic sealing control technology is urgently needed to achieve multi-objective coordinated optimization of leakage rate, energy consumption, and equipment life.

[0003] Traditional PID or empirical formulas rely on fixed parameters and cannot respond to the dynamic deformation of the kiln body and thermal stress fluctuations, resulting in delayed adjustment of the sealing gap and difficulty in balancing leakage rate and wear. Factors such as equipment aging and changes in material properties cause system parameter drift. Traditional control strategies lack online self-healing capabilities and require frequent shutdowns for calibration.

[0004] To this end, the present invention proposes a high-precision adaptive control method and system for a floating sealing device of a lithium extraction rotary kiln. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a high-precision adaptive control method and system for a floating seal device in a lithium extraction rotary kiln, ensuring the control stability of the rotary kiln under complex operating conditions.

[0006] To achieve the above objectives, a high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln is proposed, comprising the following steps:

[0007] Step 1: Collect multi-modal characteristic parameters of the rotary kiln's operating status in real time through the sensor system;

[0008] Step 2: Input the multimodal feature parameters into a pre-trained physical information neural network, solve the partial differential equation of thermoelasticity, and output the deformation field prediction matrix of the future preset time slot;

[0009] Step 3: Based on the deformation field prediction matrix, solve a multi-objective optimization model with the joint goal of minimizing leakage rate and wear energy consumption, and output a sealing compensation instruction vector;

[0010] Step 4: Based on the sealing compensation instruction vector, drive the magnetorheological-piezoelectric composite actuator to complete the instruction control of the sealing ring;

[0011] Step 5: Collect feedback result data of the command control, and based on the feedback result data, use a reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function, and feed the adjusted weights back to step 3;

[0012] The real-time acquisition of multi-modal characteristic parameters of the rotary kiln operating state by the sensor system comprises the following steps:

[0013] Step 11: Deploy a multimodal sensor network at pre-set key locations of the rotary kiln to obtain the dynamic coupling parameters of the sealing system;

[0014] The method of deploying the multimodal sensor network to obtain the dynamic coupling parameters of the sealing system is as follows:

[0015] A laser displacement sensor is installed on the axial surface of the rotary kiln body to capture the axial expansion and contraction and radial runout of the kiln body in real time through the principle of laser triangulation reflection.

[0016] A high-temperature piezoelectric surface acoustic wave sensor is embedded in the circumference of the sealing ring end face, and the contact pressure distribution in the rotary kiln is inverted by the frequency shift of the radio frequency signal.

[0017] A distributed fiber Bragg grating sensor network is spirally laid along the surface of the kiln body. The wavelength offset is analyzed in real time by a demodulator to separate the temperature field distribution and strain field distribution on the kiln body of the rotary kiln.

[0018] The axial expansion and contraction, radial runout, contact pressure distribution, temperature field distribution and strain field distribution constitute dynamic coupling parameters;

[0019] Step 12: performing data synchronization transmission and adaptive preprocessing on the dynamic coupling parameters;

[0020] Step 13: performing spatiotemporal structured storage on the dynamic coupling parameters after data synchronous transmission and adaptive preprocessing to obtain the multimodal characteristic parameters;

[0021] Inputting the multimodal characteristic parameters into a pre-trained physical information neural network, solving the thermoelastic partial differential equation, and outputting a deformation field prediction matrix for a future preset time slot includes the following steps:

[0022] Step 21: Using a multi-input branch fusion structure, construct a physical information neural network architecture that includes the partial differential equations of thermoelasticity;

[0023] The physical information neural network architecture is constructed as follows:

[0024] The physical information neural network architecture is used to embed the partial differential equations of thermoelasticity and realize deformation field prediction;

[0025] The physical information neural network architecture consists of a feature encoder, a physical constraint layer, and a multi-task decoder;

[0026] The feature encoder is used to receive the coupled tensor of deformation-pressure-temperature-strain in the multimodal feature parameters as input, and extract multi-scale spatiotemporal features through a three-layer convolutional neural network; the convolution kernel size of the three-layer convolutional neural network is 3×3, the step size is 1, and the number of channels is 64, 128, and 256 respectively;

[0027] The physical constraint layer is used to embed the thermoelasticity partial differential equation in the hidden layer;

[0028] The multi-task decoder is used to output a deformation field prediction matrix for a future preset time slot;

[0029] Step 22: using the collected multimodal feature parameters as training samples for the physical information neural network to pre-train the physical information neural network;

[0030] The method of pre-training the physical information neural network is:

[0031] Randomly extract multiple sets of complete working cycle data from all stored multimodal characteristic parameters to form a real data set;

[0032] Each set of operating cycle data in the real data set is simulated using a simulation tool to generate a deformation field label corresponding to the operating cycle data; the deformation field label is in the form of a three-dimensional matrix, the first dimension corresponds to the axial dimension, the second dimension corresponds to the circumferential dimension, and the third dimension includes the axial displacement increment and the radial displacement increment, wherein the axial displacement increment reflects the thermal expansion trend of the kiln body, and the radial displacement increment represents the eccentric vibration amplitude. The deformation degree of the rotary kiln surface is obtained through the axial displacement increment and the radial displacement increment;

[0033] Constructing a loss function for the physical information neural network architecture;

[0034] Using the Adam optimizer, iteratively optimizing the physical information neural network architecture to complete the training process of the physical information neural network architecture;

[0035] Step 23: When the physical information neural network architecture is actually running, the newly acquired multimodal characteristic parameters are collected, and the partial differential equation of thermoelasticity is discretized on the three-dimensional grid of the kiln body; the second-order time derivative of the displacement field output by the physical information neural network architecture is discretized using the central difference scheme, and the stress term is discretized using the finite volume method;

[0036] Step 24: Preset the time slot, input the latest collected multimodal feature parameters into the discretized physical information neural network architecture, obtain the preliminary predicted deformation field prediction matrix through forward propagation, and then perform three Newton-Raphson iterative corrections. After calculating the residual error in each iteration, the final future time slot is generated and the deformation field prediction matrix of the rotary kiln is obtained.

[0037] Solving a multi-objective optimization model based on the deformation field prediction matrix with the joint objective of minimizing leakage rate and wear energy consumption, and outputting a sealing compensation instruction vector comprises the following steps:

[0038] Step 31: Define various controllable parameters as parameter variables, and based on the parameter variables and the deformation field prediction matrix, define a multi-objective optimization function with minimizing the sealing system leakage rate and wear energy consumption as the joint optimization goal;

[0039] Step 32: Design a corresponding set of physical and process constraints based on the physical laws and process requirements of the lithium extraction process of the rotary kiln floating seal device;

[0040] The set of physical and process constraints includes:

[0041] Axial compensation is smaller than the preset axial threshold, and the radial compensation amount is smaller than the preset radial threshold;

[0042] The pressure adjustment amount of each hydraulic bag is less than the preset pressure adjustment threshold, and the total pressure adjustment amount is less than the total pressure adjustment threshold to prevent overload of the sealing surface;

[0043] The compensation speed of each hydraulic bag is less than a preset compensation speed threshold;

[0044] The gap after compensation is smaller than the preset maximum gap;

[0045] Step 33: Taking minimizing the multi-objective optimization function as the optimization objective and the set of physical and process constraints as the set of constraints, a convex optimization model of the multi-objective optimization model is constructed;

[0046] Step 34: Solve the multi-objective optimization model using the interior point method to obtain the solution value of each parameter variable. The solution values ​​of all parameter variables constitute the sealing compensation instruction vector;

[0047] The method of driving the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector is:

[0048] The sealing compensation instruction vector is transmitted to the execution control unit via the CAN bus. The execution control unit drives the control mechanism of the corresponding controllable parameter according to the solution value of the parameter variable of each controllable parameter, and adjusts the corresponding controllable parameter to the corresponding solution value;

[0049] The adjusting the corresponding controllable parameters to the corresponding solution values ​​includes:

[0050] The execution control unit drives the magnetorheological fluid damper to execute the axial displacement instruction, and sets the axial displacement of the magnetorheological fluid damper to the solution value of the axial displacement compensation amount;

[0051] The execution control unit drives the electric stack actuator to execute the radial displacement instruction, and sets the radial displacement of the electric stack actuator to the solution value of the radial displacement compensation amount;

[0052] The execution control unit drives the pressure adjustment value of each hydraulic bag to be the solution value of the corresponding hydraulic bag pressure adjustment amount;

[0053] The method of adjusting the weights of various weight coefficients in the multi-objective optimization function online based on the feedback result data by using the reinforcement learning algorithm includes the following steps:

[0054] Step 41: Use the Actor-Critic framework as the framework and network model of the reinforcement learning algorithm;

[0055] The Actor network in the Actor-Critic framework is a fully connected layer. The input received is the actual execution effect data, the activation function is Tanh, and the output is the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption in the multi-objective optimization function.

[0056] The Critic network structure in the Actor-Critic framework is a fully connected layer with a ReLU activation function. Its input is the action output by the Actor network and the actual execution effect data, and the output is the state value estimation.

[0057] Step 42: Design the state space, action space, and reward function for the reinforcement learning algorithm.

[0058] Step 43: Load multiple sets of historical control cycle data from all collected historical data of multimodal feature parameters as offline datasets; input this offline dataset into the Actor model, update the Actor network using the PPO algorithm, and train and update the Critic network by minimizing the temporal difference error until the reward curve converges;

[0059] Step 44: Input the feedback result data into the Actor network to obtain the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption output by the Actor network, and feed the output adjustment value back to the multi-objective optimization function to dynamically adjust the proportional coefficient in the multi-objective optimization function.

[0060] A high-precision adaptive control system for the floating seal device of a lithium extraction rotary kiln is proposed, which includes a characteristic parameter collection module, a deformation matrix generation module, a control instruction generation module, and a dynamic weight update module. The various models are electrically connected.

[0061] The characteristic parameter collection module collects the multimodal characteristic parameters of the rotary kiln's operating status in real time through the sensor system and sends the multimodal characteristic parameters to the deformation matrix generation module;

[0062] A deformation matrix generation module inputs the multimodal feature parameters into a pre-trained physical information neural network, solves the partial differential equation of thermoelasticity, outputs a deformation field prediction matrix for a future preset time slot, and sends the deformation field prediction matrix to a control instruction generation module;

[0063] a control instruction generation module, which solves a multi-objective optimization model with the joint objective of minimizing leakage rate and wear energy consumption based on the deformation field prediction matrix, outputs a sealing compensation instruction vector, and sends the sealing compensation instruction vector to the dynamic weight update module;

[0064] The dynamic weight update module drives the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector, collects the feedback result data of the command control, and uses the reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function based on the feedback result data, and sends the adjusted weights to the control command generation module.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] The present invention collects the deformation, temperature and pressure data of the rotary kiln in real time through a multimodal sensing network, combines it with a physical information neural network to predict the deformation field trend, and adopts reinforcement learning to dynamically optimize the leakage rate and energy consumption weights to drive the composite actuator to accurately adjust the sealing gap. The physical information neural network uses the thermoelastic mechanics equation as an internal constraint to achieve physical consistency in deformation field prediction and avoid the generalization risk of a pure data-driven model. The dynamic weight adjustment mechanism based on reinforcement learning, through online learning of the game relationship between leakage rate and energy consumption, autonomously balances multi-objective conflicts, overcomes the static defects of the fixed weight strategy, and thus achieves the coordination of large-scale deformation compensation and high-precision fine-tuning, ensuring control stability under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of a high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln in Example 1 of the present invention;

[0068] Figure 2 This is a network structure diagram of the physical information neural network in Example 1 of the present invention;

[0069] Figure 3 This is a module connection diagram of the high-precision adaptive control system of the floating seal device of the lithium extraction rotary kiln in Example 2 of the present invention. DETAILED DESCRIPTION

[0070] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1 As shown, the high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln includes the following steps:

[0073] Step 1: Collect multi-modal characteristic parameters of the rotary kiln's operating status in real time through the sensor system;

[0074] Step 2: Input the multimodal feature parameters into a pre-trained physical information neural network, solve the partial differential equation of thermoelasticity, and output the deformation field prediction matrix of the future preset time slot;

[0075] Step 3: Based on the deformation field prediction matrix, solve a multi-objective optimization model with the joint goal of minimizing leakage rate and wear energy consumption, and output a sealing compensation instruction vector;

[0076] Step 4: Based on the sealing compensation instruction vector, drive the magnetorheological-piezoelectric composite actuator to complete the instruction control of the sealing ring;

[0077] Step 5: Collect feedback result data of the command control, and based on the feedback result data, use a reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function, and feed the adjusted weights back to step 3;

[0078] To achieve all-round perception of the operating status of the rotary kiln sealing device, this embodiment deploys a multi-type industrial-grade sensor network in key areas of the kiln body and sealing ring to collect multimodal characteristic parameters such as deformation, pressure and temperature in real time.

[0079] In an embodiment of the present invention, the real-time acquisition of multimodal characteristic parameters of the rotary kiln operating state by the sensor system includes the following steps:

[0080] Step 11: Deploy a multimodal sensor network at pre-set key locations of the rotary kiln to obtain the dynamic coupling parameters of the sealing system;

[0081] Specifically, the method of deploying the multimodal sensing network to obtain the dynamic coupling parameters of the sealing system is as follows:

[0082] In this embodiment, a set of high-precision laser displacement sensors is installed on the axial surface of the rotary kiln body at intervals of 50 cm, with a total of 20 sets of sensors deployed. These sensors use the principle of laser triangulation to capture the axial expansion and contraction and radial runout of the kiln body in real time. The sampling frequency of these high-precision laser displacement sensors is set to 100 Hz, with an accuracy of ±0.01 mm and a tolerance for kiln surface temperatures up to 350°C. The axial expansion and contraction ΔL is used to quantify the length change caused by thermal expansion of the kiln body, and the radial runout ΔR reflects the amplitude of the mechanical vibration during the kiln body rotation.

[0083] Furthermore, 16 high-temperature piezoelectric surface acoustic wave sensors were embedded circumferentially on the end face of the sealing ring, with a spacing of 22.5°. The contact pressure distribution in the rotary kiln was inverted by the frequency shift of the radio frequency signal. The measurement range was set to 0MPa-5MPa and the resolution was set to 0.5kPa.

[0084] At the same time, a distributed fiber Bragg grating sensor network is spirally laid along the surface of the kiln body, with a grating measurement point set every 10 cm, for a total of 1,200 measurement points. The wavelength offset is analyzed in real time by a demodulator to separate the temperature field distribution and strain field distribution on the kiln body of the rotary kiln.

[0085] The axial expansion and contraction, radial runout, contact pressure distribution, temperature field distribution and strain field distribution constitute dynamic coupling parameters;

[0086] Step 12: performing data synchronization transmission and adaptive preprocessing on the dynamic coupling parameters;

[0087] Specifically, the method of performing data synchronization transmission and adaptive preprocessing on the dynamic coupling parameters is:

[0088] All sensor data from each sensor is connected to the edge computing gateway via armored twisted pair shielded cables, and the clocks of each sensor are synchronized in advance to ensure timestamp alignment;

[0089] In order to address the interference of high dust and strong vibration environment in the kiln body on signal quality, a hierarchical noise reduction process of dynamic coupling parameters is implemented. Specifically:

[0090] Since the laser deformation signal collected by the high-precision laser displacement sensor is significantly affected by mechanical vibration, a 5-layer wavelet packet decomposition such as the wavelet base DB4 is used to perform hard threshold filtering on the high-frequency noise, retaining the effective frequency band of 0-50Hz.

[0091] The surface acoustic wave pressure signal collected by the high-temperature piezoelectric surface acoustic wave sensor is interfered by transient impact, so a sliding median filter with a window length of 100 milliseconds is implemented to suppress abnormal spikes;

[0092] The fiber temperature data collected by the distributed fiber Bragg grating sensor network is based on the spatial correlation of adjacent measurement points, and the outliers are corrected by the Kriging interpolation method. The interpolation weight is calculated based on the inverse square of the distance.

[0093] Step 13: performing spatiotemporal structured storage on the dynamic coupling parameters after data synchronous transmission and adaptive preprocessing to obtain the multimodal characteristic parameters;

[0094] Specifically, the spatiotemporal structured storage method is:

[0095] A spatiotemporal fusion engine is built within the edge server. Dynamic coupling parameters, after synchronous data transmission and adaptive preprocessing, are structured and stored in the form of a three-dimensional spatial grid model of the kiln body. This 3D grid is sector-shaped with an axial resolution of 5 cm and a circumferential resolution of 1°. Using a 10-millisecond time window, the sensor data streams are aligned using cubic spline interpolation to generate a coupled tensor of deformation, pressure, temperature, and strain. This tensor has 200 grids in the axial dimension and 360 grids in the circumferential dimension. The third dimension corresponds to four channels of data: the deformation vector (composed of axial expansion and radial runout), the pressure value in the contact pressure distribution, the temperature value in the temperature field distribution, and the strain value in the strain field distribution. The storage layer adopts a time-series database partitioning strategy, generating independent data blocks every hour. Deformation and pressure data are compressed using Delta encoding, while temperature and deformation data are compressed using lossy floating-point compression.

[0096] In a further preferred embodiment, to ensure long-term data reliability, 30 days of historical data can be loaded every morning to train a linear regression model to dynamically compensate for sensor baseline drift. When the difference between the laser deformation data and the estimated fiber strain value exceeds 0.05 mm, the pre-trained LSTM anomaly diagnosis model is triggered to reallocate sensor weight coefficients and isolate and mark the faulty node. Ultimately, a multimodal feature parameter set is generated.

[0097] In an embodiment of the present invention, inputting the multimodal feature parameters into a pre-trained physical information neural network, solving the partial differential equation of thermoelasticity, and outputting a deformation field prediction matrix for a future preset time slot includes the following steps:

[0098] Step 21: Using a multi-input branch fusion structure, construct a physical information neural network architecture that includes the partial differential equations of thermoelasticity;

[0099] Specifically, the physical information neural network architecture is constructed as follows:

[0100] The physical information neural network architecture is used to embed the partial differential equations of thermoelasticity and realize deformation field prediction;

[0101] The physical information neural network architecture consists of a feature encoder, a physical constraint layer, and a multi-task decoder;

[0102] The feature encoder is used to receive the coupled tensor of deformation-pressure-temperature-strain in the multimodal feature parameters as input, and extract multi-scale spatiotemporal features through a three-layer convolutional neural network; the convolution kernel size of the three-layer convolutional neural network is 3×3, the step size is 1, and the number of channels is 64, 128, and 256 respectively;

[0103] The physical constraint layer is used to embed the partial differential equation of thermoelasticity in the hidden layer;

[0104] The multi-task decoder is used to output a deformation field prediction matrix for a future preset time slot;

[0105] Specifically, the physical constraint layer embeds the thermoelastic partial differential equation in the hidden layer as follows: ;

[0106] in, is the displacement field, t is the time, σ is the stress tensor, x and y are the horizontal and vertical axes respectively, and The stress tensors for the x-axis and y-axis, ρ=7850kg / m 3 is the density of the kiln material, ∇ represents the gradient operator, and ∇·σ represents the net force generated by the stress gradient per unit volume, where:

[0107] ;

[0108] G is the shear modulus, u and v are the axial displacement and radial displacement respectively; E=200GPa is the elastic modulus;

[0109] , represents the thermal stress term; is the sealing contact force vector derived from the pressure field; is the coefficient of thermal expansion, Temperature change value;

[0110] The calculation process of the sealing contact force calculated from the pressure field is as follows:

[0111] The contact surface of the rotary kiln is modeled as an annular area with an inner diameter of r1 = 1.5m, an outer diameter of r2 = 1.6m, and a circumferential angle θ of 0 to 360 degrees. The annular surface is discretized into 360 10 grids, each with a circumferential resolution of 1° and a radial resolution of 1 cm;

[0112] Collecting the discrete pressure values ​​measured by all deployed high-temperature piezoelectric surface acoustic wave sensors, constructing the circumferential continuous pressure distribution of the rotary kiln using a cubic spline interpolation method, and then generating the full annular pressure field P(r,θ,t) through radial linear interpolation;

[0113] Calculate the normal contact force of each mesh element and calculate the sum of the normal contact forces Ft of all mesh elements;

[0114] Specifically, the calculation formula of the normal contact force of each grid unit is P(r,θ,t)×ΔA, where ΔA is the area of ​​each grid unit; then the sum of the normal contact forces is the sum of the normal contact forces of all grid units;

[0115] According to the sum of the normal contact force, the axial component, radial component and circumferential component are calculated respectively, and finally the sealing contact force vector is formed. , the sealing contact force vector consists of axial component, radial component and circumferential component;

[0116] Specifically, the calculation formula of the axial component is: ,in, is the inclination angle of the sealing surface of the rotary kiln;

[0117] The calculation formula of the radial component is: , where m1 is the mass of the sealing ring, is the angular velocity of the kiln body, is the average radius of the kiln body; Indicates the centrifugal effect of the kiln body;

[0118] The calculation formula of the circumferential component is: ;in, is the friction coefficient of the rotary kiln surface;

[0119] A module structure diagram of the physical information neural network is as follows Figure 2 As shown;

[0120] Step 22: using the collected multimodal feature parameters as training samples for the physical information neural network to pre-train the physical information neural network;

[0121] Specifically, the method of pre-training the physical information neural network is:

[0122] Randomly extract several groups of complete working cycle data from all stored multimodal characteristic parameters to form a real data set;

[0123] Each set of operating cycle data in the real data set is simulated using a simulation tool to generate a deformation field label corresponding to the operating cycle data; the deformation field label is in the form of a three-dimensional matrix, the first dimension corresponds to the axial dimension, the second dimension corresponds to the circumferential dimension, and the third dimension includes the axial displacement increment and the radial displacement increment, wherein the axial displacement increment reflects the thermal expansion trend of the kiln body, and the radial displacement increment represents the eccentric vibration amplitude, thereby expressing the degree of deformation of the rotary kiln surface through the axial displacement increment and the radial displacement increment;

[0124] Constructing a loss function for the physical information neural network architecture;

[0125] Specifically, the loss function of the physical information neural network architecture is the sum of the mean square error between the deformation field prediction matrix output by the task decoder and the deformation field label and the physical conservation loss error;

[0126] The physical conservation loss error is the difference between the left and right forms of the thermoelastic partial differential equation in the physical constraint layer, and is used to theoretically constrain the fitting process of the physical constraint layer of the physical information neural network architecture to ensure the authenticity of the physical information neural network architecture's simulation process of the actual rotary kiln lithium extraction process;

[0127] Finally, the Adam optimizer is used to iteratively optimize the physical information neural network architecture to complete the model training process;

[0128] Step 23: When the physical information neural network architecture is actually running, the newly acquired multimodal characteristic parameters are collected, and the partial differential equation of thermoelasticity is discretized on the three-dimensional grid of the kiln body; the second-order time derivative of the displacement field output by the physical information neural network architecture is discretized using the central difference scheme, and the stress term is discretized using the finite volume method;

[0129] Step 24: Preset the time slot, input the latest collected multimodal feature parameters into the discretized physical information neural network architecture, obtain the preliminary predicted deformation field prediction matrix through forward propagation, and then perform three Newton-Raphson iterative corrections. After calculating the residual error in each iteration, the final future time slot is generated and the deformation field prediction matrix of the rotary kiln is obtained.

[0130] It can be understood that the physical information neural network integrates multiple characteristics related to the deformation of the rotary kiln, such as axial expansion and contraction, radial runout, contact pressure distribution, temperature field distribution, and strain field distribution, to achieve deformation prediction of the rotary kiln. The prediction result can directly drive the generation of sealing compensation instructions. For example, the compensation amount of the magnetorheological fluid damper is dynamically adjusted by the axial displacement increment, the displacement compensation of the piezoelectric actuator is controlled by the radial displacement increment, and the pressure ratio of the sealing ring hydraulic bladder array is optimized in combination with the contact pressure distribution, thereby achieving the coordinated optimization of leakage rate and wear energy consumption.

[0131] Furthermore, solving a multi-objective optimization model with the joint objective of minimizing leakage rate and wear energy consumption based on the deformation field prediction matrix and outputting a sealing compensation instruction vector includes the following steps:

[0132] Step 31: Define various controllable parameters as parameter variables, and based on the parameter variables and the deformation field prediction matrix, define a multi-objective optimization function with minimizing the sealing system leakage rate and wear energy consumption as the joint optimization goal;

[0133] Specifically, the construction process of the multi-objective optimization function includes:

[0134] The defined parameter variables include but are not limited to axial displacement compensation , radial displacement compensation , Hydraulic bladder pressure adjustment ,in, is the pressure adjustment of the Nth hydraulic bladder, N is the number of all hydraulic bladders, and all parameter variables constitute the sealing compensation instruction vector [ ];

[0135] Mark the sealing system leakage rate as , wear energy consumption is marked as ;

[0136] The calculation formula for the sealing system leakage rate is:

[0137] ;

[0138] Among them, k is the preset proportional coefficient, is the predicted sealing gap of the sealing device, is the initial gap, is the deformation field prediction matrix, then is the deformation field gradient vector obtained by spatial difference calculation, d is the direction vector, that is, the unit direction vector of the sealing surface normal, and the deformation field gradient vector is determined by the geometric installation angle of the sealing ring; therefore It represents the projection of the deformation gradient on the normal direction of the sealing surface, and then superimposes the initial gap The predicted gap of the sealing device after deformation can be obtained; The greater the pressure difference, the stronger the driving force for fluid leakage, and the higher the leakage rate. The kinematic viscosity of the object in the sealing system reflects the internal friction resistance of the object when it flows. The greater the kinematic viscosity, the greater the flow resistance, and the lower the leakage rate.

[0139] The calculation formula of the wear energy consumption is: ,in, is the friction coefficient of the nth hydraulic bladder, is the contact pressure of the nth hydraulic bladder, is the compensation speed of the nth hydraulic bag, where It is the preset control period, usually set to 5 seconds. is the local displacement of the sealing ring generated in the nth hydraulic bag, The calculation formula is ,in, is the stiffness of the nth hydraulic bladder, is the pressure adjustment amount of the nth hydraulic bladder;

[0140] Then the expression of the multi-objective optimization function is: ;in, and All are preset proportional coefficients;

[0141] Step 32: Design a corresponding set of physical and process constraints based on the physical laws and process requirements of the lithium extraction process of the rotary kiln floating seal device;

[0142] Specifically, the set of physical and process constraints includes:

[0143] Axial compensation is smaller than the preset axial threshold, and the radial compensation amount is smaller than the preset radial threshold;

[0144] The pressure adjustment of each hydraulic bag is less than the preset pressure adjustment threshold, and the total pressure adjustment should be less than the total pressure adjustment threshold to prevent overload of the sealing surface;

[0145] The compensation speed of each hydraulic bag is less than a preset compensation speed threshold;

[0146] Gap after compensation The clearance must be smaller than the preset maximum clearance to ensure the minimum safety clearance requirement;

[0147] Step 33: Taking minimizing the multi-objective optimization function as the optimization objective and the set of physical and process constraints as the set of constraints, a convex optimization model of the multi-objective optimization model is constructed;

[0148] Step 34: Solve the multi-objective optimization model using the interior point method to obtain the solution value of each parameter variable. The solution values ​​of all parameter variables constitute the sealing compensation instruction vector;

[0149] Furthermore, the method of driving the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector is:

[0150] The sealing compensation instruction vector is transmitted to the execution control unit via the CAN bus. The execution control unit drives the control mechanism of the corresponding controllable parameter according to the solution value of the parameter variable of each controllable parameter, and adjusts the corresponding controllable parameter to the corresponding solution value;

[0151] Specifically, adjusting the corresponding controllable parameters to the corresponding solution values ​​includes:

[0152] The execution control unit drives the magnetorheological fluid damper to execute the axial displacement instruction, and sets the axial displacement of the magnetorheological fluid damper to the solution value of the axial displacement compensation amount;

[0153] The execution control unit drives the electric stack actuator to execute the radial displacement instruction, and sets the radial displacement of the electric stack actuator to the solution value of the radial displacement compensation amount;

[0154] The execution control unit drives the pressure adjustment value of each hydraulic bag to be the solution value of the corresponding hydraulic bag pressure adjustment amount;

[0155] Furthermore, the method of collecting the feedback result data of the instruction control is:

[0156] Through a multi-source sensor network, the actual execution effect data of the dynamic coupling parameters after the seal compensation control is collected in real time, forming a closed-loop feedback to optimize system performance;

[0157] Specifically, the actual execution effect data is collected in the following manner:

[0158] Axial and radial displacement compensation is monitored by high-precision displacement sensors, while the contact pressure distribution of the sealing surface is captured by an embedded pressure sensor array. Meanwhile, vibration sensors record the mechanical vibration spectrum of the actuator. All sensor data is synchronized with a unified clock, filtered for noise reduction, and formatted for standardization to generate a feedback dataset containing displacement deviation, pressure distribution, and vibration characteristics. This feedback dataset serves as actual execution performance data.

[0159] Furthermore, the method of adjusting the weights of various weight coefficients in the multi-objective optimization function online using a reinforcement learning algorithm based on the feedback result data includes the following steps:

[0160] Step 41: Use the Actor-Critic framework as the framework and network model of the reinforcement learning algorithm;

[0161] Specifically, the Actor network in the Actor-Critic framework is a fully connected layer, the input received is the actual execution effect data, the activation function is Tanh, and the output is the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption in the multi-objective optimization function; it is understandable that due to sudden changes in feed volume, temperature fluctuations, etc., the deformation characteristics of the sealing gap will change. Therefore, if the weight between the sealing system leakage rate and wear energy consumption is fixed, it will not be able to adapt to the rapidly changing leakage rate or energy consumption requirements, making it difficult to achieve global optimization under different working conditions; therefore, it is generally necessary to automatically increase the proportional weight of the sealing system leakage rate when there is a high risk of leakage, such as when a sudden temperature rise causes an increase in deformation, and implement a balanced weight under stable working conditions;

[0162] The Critic network structure in the Actor-Critic framework is a fully connected layer with a ReLU activation function. Its input is the action output by the Actor network and the actual execution effect data, and the output is the state value estimation.

[0163] Step 42: Design the state space, action space, and reward function for the reinforcement learning algorithm.

[0164] Specifically, the state space includes dynamic coupling parameters collected by various sensors in the actual execution effect data;

[0165] The action space includes adjustment values ​​of proportionality coefficients of sealing system leakage rate and wear energy consumption;

[0166] In an embodiment of the present invention, the reward function The definition can be set as:

[0167] ;

[0168] Among them, punishment for violations of safety constraints, is the local stress state of the kiln surface, and the training objective is to maximize the discounted cumulative reward ,in, is the discount factor, which is generally set to 0.95. The farther the time node is from the current time, the smaller its impact on the reward of the current time; I( ) is an indicator function, which is 1 if the condition is met, otherwise it is 0. Δδ is the displacement offset. Therefore, when the absolute value of the second-order norm of Δδ exceeds 0.15, 20 is deducted to penalize the change. Similarly, when the stress When it exceeds 250, 15 points will be deducted to punish high stress;

[0169] Step 43: Load several sets of historical control cycle data from all collected historical data of multimodal feature parameters as offline datasets; input this offline dataset into the Actor model, update the Actor network using the PPO algorithm, and train and update the Critic network by minimizing the temporal difference error until the reward curve converges;

[0170] Step 44: Input the feedback result data into the Actor network to obtain the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption output by the Actor network, and feed the output adjustment value back to the multi-objective optimization function to dynamically adjust the proportional coefficient in the multi-objective optimization function.

[0171] Example 2

[0172] like Figure 3 As shown, the high-precision adaptive control system of the floating seal device of the lithium extraction rotary kiln includes a characteristic parameter collection module, a deformation matrix generation module, a control instruction generation module and a dynamic weight update module; wherein each model is electrically connected;

[0173] The characteristic parameter collection module collects the multimodal characteristic parameters of the rotary kiln's operating status in real time through the sensor system and sends the multimodal characteristic parameters to the deformation matrix generation module;

[0174] A deformation matrix generation module inputs the multimodal feature parameters into a pre-trained physical information neural network, solves the partial differential equation of thermoelasticity, outputs a deformation field prediction matrix for a future preset time slot, and sends the deformation field prediction matrix to a control instruction generation module;

[0175] a control instruction generation module, which solves a multi-objective optimization model with the joint objective of minimizing leakage rate and wear energy consumption based on the deformation field prediction matrix, outputs a sealing compensation instruction vector, and sends the sealing compensation instruction vector to the dynamic weight update module;

[0176] The dynamic weight update module drives the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector, collects the feedback result data of the command control, and uses the reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function based on the feedback result data, and sends the adjusted weights to the control command generation module.

[0177] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0178] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0179] The above preset parameters or preset thresholds are all set by those skilled in the art according to actual conditions or obtained through large amounts of data simulation.

[0180] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln, characterized in that: The following steps are involved: Step 1: Collect multi-modal characteristic parameters of the rotary kiln's operating status in real time through the sensor system; Step 2: Input the multimodal feature parameters into a pre-trained physical information neural network, solve the partial differential equation of thermoelasticity, and output the deformation field prediction matrix of the future preset time slot; Step 3: Based on the deformation field prediction matrix, solve a multi-objective optimization model with the joint goal of minimizing leakage rate and wear energy consumption, and output a sealing compensation instruction vector; Step 4: Based on the sealing compensation instruction vector, drive the magnetorheological-piezoelectric composite actuator to complete the instruction control of the sealing ring; Step 5: Collect the feedback result data of the command control, and based on the feedback result data, use the reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function, and feed the adjusted weights back to step 3.

2. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 1 is characterized in that: The real-time acquisition of multi-modal characteristic parameters of the rotary kiln operating state by the sensor system comprises the following steps: Step 11: Deploy a multimodal sensor network at pre-set key locations of the rotary kiln to obtain the dynamic coupling parameters of the sealing system; Step 12: performing data synchronization transmission and adaptive preprocessing on the dynamic coupling parameters; Step 13: Perform spatiotemporal structured storage on the dynamic coupling parameters after data synchronous transmission and adaptive preprocessing to obtain the multimodal characteristic parameters.

3. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 2 is characterized in that: The method of deploying the multimodal sensor network to obtain the dynamic coupling parameters of the sealing system is as follows: A laser displacement sensor is installed on the axial surface of the rotary kiln body to capture the axial expansion and contraction and radial runout of the kiln body in real time through the principle of laser triangulation reflection. A high-temperature piezoelectric surface acoustic wave sensor is embedded in the circumference of the sealing ring end face, and the contact pressure distribution in the rotary kiln is inverted by the frequency shift of the radio frequency signal. A distributed fiber Bragg grating sensor network is spirally laid along the surface of the kiln body. The wavelength offset is analyzed in real time by a demodulator to separate the temperature field distribution and strain field distribution on the kiln body of the rotary kiln. The axial expansion and contraction amount, radial runout amount, contact pressure distribution, temperature field distribution and strain field distribution constitute dynamic coupling parameters.

4. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 3 is characterized in that: Inputting the multimodal characteristic parameters into a pre-trained physical information neural network, solving the thermoelastic partial differential equation, and outputting a deformation field prediction matrix for a future preset time slot includes the following steps: Step 21: Using a multi-input branch fusion structure, construct a physical information neural network architecture that includes the partial differential equations of thermoelasticity; Step 22: using the collected multimodal feature parameters as training samples for the physical information neural network to pre-train the physical information neural network; Step 23: When the physical information neural network architecture is actually running, the newly acquired multimodal characteristic parameters are collected, and the partial differential equation of thermoelasticity is discretized on the three-dimensional grid of the kiln body; the second-order time derivative of the displacement field output by the physical information neural network architecture is discretized using the central difference scheme, and the stress term is discretized using the finite volume method; Step 24: Pre-set the time slot, input the latest collected multimodal feature parameters into the discretized physical information neural network architecture, obtain the preliminary predicted deformation field prediction matrix through forward propagation, and then perform three Newton-Raphson iterative corrections. Calculate the residual for each iteration, generate the final future time slot, and obtain the deformation field prediction matrix of the rotary kiln.

5. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 4 is characterized in that: The physical information neural network architecture is constructed as follows: The physical information neural network architecture is used to embed the partial differential equations of thermoelasticity and realize deformation field prediction; The physical information neural network architecture consists of a feature encoder, a physical constraint layer, and a multi-task decoder; The feature encoder is used to receive the coupled tensor of deformation-pressure-temperature-strain in the multimodal feature parameters as input, and extract multi-scale spatiotemporal features through a three-layer convolutional neural network; the convolution kernel size of the three-layer convolutional neural network is 3×3, the step size is 1, and the number of channels is 64, 128, and 256 respectively; The physical constraint layer is used to embed the thermoelasticity partial differential equation in the hidden layer; The multi-task decoder is used to output a deformation field prediction matrix of a future preset time slot.

6. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 5 is characterized in that: The method of pre-training the physical information neural network is: Randomly extract multiple sets of complete working cycle data from all stored multimodal characteristic parameters to form a real data set; Each set of operating cycle data in the real data set is simulated using a simulation tool to generate a deformation field label corresponding to the operating cycle data; the deformation field label is in the form of a three-dimensional matrix, the first dimension corresponds to the axial dimension, the second dimension corresponds to the circumferential dimension, and the third dimension includes the axial displacement increment and the radial displacement increment, wherein the axial displacement increment reflects the thermal expansion trend of the kiln body, and the radial displacement increment represents the eccentric vibration amplitude. The deformation degree of the rotary kiln surface is obtained through the axial displacement increment and the radial displacement increment; Constructing a loss function for the physical information neural network architecture; The Adam optimizer is used to iteratively optimize the physical information neural network architecture to complete the training process of the physical information neural network architecture.

7. The high-precision adaptive control method for the floating seal device of the lithium extraction rotary kiln according to claim 6 is characterized in that: Solving a multi-objective optimization model based on the deformation field prediction matrix with the joint objective of minimizing leakage rate and wear energy consumption, and outputting a sealing compensation instruction vector comprises the following steps: Step 31: Define various controllable parameters as parameter variables, and based on the parameter variables and the deformation field prediction matrix, define a multi-objective optimization function with minimizing the sealing system leakage rate and wear energy consumption as the joint optimization goal; Step 32: Design a corresponding set of physical and process constraints based on the physical laws and process requirements of the lithium extraction process of the rotary kiln floating seal device; Step 33: Taking minimizing the multi-objective optimization function as the optimization objective and the set of physical and process constraints as the set of constraints, a convex optimization model of the multi-objective optimization model is constructed; Step 34: Use the interior point method to solve the multi-objective optimization model to obtain the solution value of each parameter variable. The solution values ​​of all parameter variables constitute the sealing compensation instruction vector.

8. The high-precision adaptive control method for the floating seal device of a lithium extraction rotary kiln according to claim 7 is characterized in that: The set of physical and process constraints includes: Axial compensation is smaller than the preset axial threshold, and the radial compensation amount is smaller than the preset radial threshold; The pressure adjustment amount of each hydraulic bladder is less than a preset pressure adjustment threshold, and the total pressure adjustment amount is less than the total pressure adjustment threshold; The compensation speed of each hydraulic bag is less than a preset compensation speed threshold; The gap after compensation is smaller than the preset maximum gap.

9. The high-precision adaptive control method for the floating seal device of a lithium extraction rotary kiln according to claim 8, characterized in that: The method of driving the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector is: The sealing compensation instruction vector is transmitted to the execution control unit through the CAN bus. The execution control unit drives the control mechanism of the corresponding controllable parameters according to the solution values ​​of the parameter variables of each controllable parameter, and adjusts the corresponding controllable parameters to the corresponding solution values.

10. The high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln according to claim 9, characterized in that: The adjusting the corresponding controllable parameters to the corresponding solution values ​​includes: The execution control unit drives the magnetorheological fluid damper to execute the axial displacement instruction, and sets the axial displacement of the magnetorheological fluid damper to the solution value of the axial displacement compensation amount; The execution control unit drives the electric stack actuator to execute the radial displacement instruction, and sets the radial displacement of the electric stack actuator to the solution value of the radial displacement compensation amount; The execution control unit drives the pressure adjustment value of each hydraulic bag to be the solution value of the corresponding hydraulic bag pressure adjustment amount.

11. The high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln according to claim 9, characterized in that: The method of adjusting the weights of various weight coefficients in the multi-objective optimization function online based on the feedback result data by using the reinforcement learning algorithm includes the following steps: Step 41: Use the Actor-Critic framework as the framework and network model of the reinforcement learning algorithm; Step 42: Design the state space, action space, and reward function for the reinforcement learning algorithm. Step 43: Load multiple sets of historical control cycle data from all collected historical data of multimodal feature parameters as offline datasets; input this offline dataset into the Actor model, update the Actor network using the PPO algorithm, and train and update the Critic network by minimizing the temporal difference error until the reward curve converges; Step 44: Input the feedback result data into the Actor network to obtain the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption output by the Actor network, and feed the output adjustment value back to the multi-objective optimization function to dynamically adjust the proportional coefficient in the multi-objective optimization function.

12. The high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln according to claim 11, characterized in that: The Actor network in the Actor-Critic framework is a fully connected layer. The input received is the actual execution effect data, the activation function is Tanh, and the output is the adjustment value of the proportional coefficient of the sealing system leakage rate and wear energy consumption in the multi-objective optimization function. The Critic network structure in the Actor-Critic framework is a fully connected layer with a ReLU activation function. Its input is the action output by the Actor network and the actual execution effect data, and its output is a state value estimate.

13. A high-precision adaptive control system for a floating seal device of a lithium extraction rotary kiln, which is used to implement the high-precision adaptive control method for a floating seal device of a lithium extraction rotary kiln according to any one of claims 1 to 12, characterized in that: It includes a feature parameter collection module, a deformation matrix generation module, a control instruction generation module and a dynamic weight update module; wherein each model is electrically connected; The characteristic parameter collection module collects the multimodal characteristic parameters of the rotary kiln's operating status in real time through the sensor system and sends the multimodal characteristic parameters to the deformation matrix generation module; A deformation matrix generation module inputs the multimodal feature parameters into a pre-trained physical information neural network, solves the partial differential equation of thermoelasticity, outputs a deformation field prediction matrix for a future preset time slot, and sends the deformation field prediction matrix to a control instruction generation module; a control instruction generation module, which solves a multi-objective optimization model with the joint objective of minimizing leakage rate and wear energy consumption based on the deformation field prediction matrix, outputs a sealing compensation instruction vector, and sends the sealing compensation instruction vector to the dynamic weight update module; The dynamic weight update module drives the magnetorheological-piezoelectric composite actuator to complete the command control of the sealing ring based on the sealing compensation command vector, collects the feedback result data of the command control, and uses the reinforcement learning algorithm to online adjust the weights of each weight coefficient in the multi-objective optimization function based on the feedback result data, and sends the adjusted weights to the control command generation module.

Images